Add Ctrl+C quit handling to REPL

- Add _quit_requested flag to TerminalDisplay
- Add request_quit() method to TerminalDisplay
- Handle 'ctrl_c' key in REPL input loops in both pipeline_runner.py and main.py
- When Ctrl+C is pressed, request_quit() is called which sets the flag
- The main loop checks is_quit_requested() and raises KeyboardInterrupt

.gitignore

@@ -1,4 +1,17 @@
 __pycache__/
 *.pyc
 .mainline_venv/
+.venv/
+uv.lock
 .mainline_cache_*.json
+.DS_Store
+htmlcov/
+.coverage
+.pytest_cache/
+*.egg-info/
+coverage.xml
+*.dot
+*.png
+test-reports/
+.opencode/
+tests/comparison_output/

.opencode/skills/mainline-architecture/SKILL.md

@@ -0,0 +1,97 @@
+---
+name: mainline-architecture
+description: Pipeline stages, capability resolution, and core architecture patterns
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers Mainline's pipeline architecture - the Stage-based system for dependency resolution, data flow, and component composition.
+
+## Key Concepts
+
+### Stage Class (engine/pipeline/core.py)
+
+The `Stage` ABC is the foundation. All pipeline components inherit from it:
+
+```python
+class Stage(ABC):
+    name: str
+    category: str  # "source", "effect", "overlay", "display", "camera"
+    optional: bool = False
+    
+    @property
+    def capabilities(self) -> set[str]:
+        """What this stage provides (e.g., 'source.headlines')"""
+        return set()
+    
+    @property
+    def dependencies(self) -> set[str]:
+        """What this stage needs (e.g., {'source'})"""
+        return set()
+```
+
+### Capability-Based Dependencies
+
+The Pipeline resolves dependencies using **prefix matching**:
+- `"source"` matches `"source.headlines"`, `"source.poetry"`, etc.
+- `"camera.state"` matches the camera state capability
+- This allows flexible composition without hardcoding specific stage names
+
+### Minimum Capabilities
+
+The pipeline requires these minimum capabilities to function:
+- `"source"` - Data source capability
+- `"render.output"` - Rendered content capability
+- `"display.output"` - Display output capability
+- `"camera.state"` - Camera state for viewport filtering
+
+These are automatically injected if missing (auto-injection).
+
+### DataType Enum
+
+PureData-style data types for inlet/outlet validation:
+- `SOURCE_ITEMS`: List[SourceItem] - raw items from sources
+- `ITEM_TUPLES`: List[tuple] - (title, source, timestamp) tuples  
+- `TEXT_BUFFER`: List[str] - rendered ANSI buffer
+- `RAW_TEXT`: str - raw text strings
+- `PIL_IMAGE`: PIL Image object
+
+### Pipeline Execution
+
+The Pipeline (engine/pipeline/controller.py):
+1. Collects all stages from StageRegistry
+2. Resolves dependencies using prefix matching
+3. Executes stages in dependency order
+4. Handles errors for non-optional stages
+
+### Canvas & Camera
+
+- **Canvas** (`engine/canvas.py`): 2D rendering surface with dirty region tracking
+- **Camera** (`engine/camera.py`): Viewport controller for scrolling content
+
+Canvas tracks dirty regions automatically when content is written via `put_region`, `put_text`, `fill`, enabling partial buffer updates.
+
+## Adding New Stages
+
+1. Create a class inheriting from `Stage`
+2. Define `capabilities` and `dependencies` properties
+3. Implement required abstract methods
+4. Register in StageRegistry or use as adapter
+
+## Common Patterns
+
+- Use adapters (engine/pipeline/adapters.py) to wrap existing components as stages
+- Set `optional=True` for stages that can fail gracefully
+- Use `stage_type` and `render_order` for execution ordering
+- Clock stages update state independently of data flow
+
+## Sources
+
+- engine/pipeline/core.py - Stage base class
+- engine/pipeline/controller.py - Pipeline implementation
+- engine/pipeline/adapters/ - Stage adapters
+- docs/PIPELINE.md - Pipeline documentation

.opencode/skills/mainline-display/SKILL.md

@@ -0,0 +1,163 @@
+---
+name: mainline-display
+description: Display backend implementation and the Display protocol
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers Mainline's display backend system - how to implement new display backends and how the Display protocol works.
+
+## Key Concepts
+
+### Display Protocol
+
+All backends implement a common Display protocol (in `engine/display/__init__.py`):
+
+```python
+class Display(Protocol):
+    width: int
+    height: int
+    
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize the display"""
+        ...
+    
+    def show(self, buf: list[str], border: bool = False) -> None:
+        """Display the buffer"""
+        ...
+    
+    def clear(self) -> None:
+        """Clear the display"""
+        ...
+    
+    def cleanup(self) -> None:
+        """Clean up resources"""
+        ...
+    
+    def get_dimensions(self) -> tuple[int, int]:
+        """Return (width, height)"""
+        ...
+```
+
+### DisplayRegistry
+
+Discovers and manages backends:
+
+```python
+from engine.display import DisplayRegistry
+display = DisplayRegistry.create("terminal")  # or "websocket", "null", "multi"
+```
+
+### Available Backends
+
+| Backend | File | Description |
+|---------|------|-------------|
+| terminal | backends/terminal.py | ANSI terminal output |
+| websocket | backends/websocket.py | Web browser via WebSocket |
+| null | backends/null.py | Headless for testing |
+| multi | backends/multi.py | Forwards to multiple displays |
+| moderngl | backends/moderngl.py | GPU-accelerated OpenGL rendering (optional) |
+
+### WebSocket Backend
+
+- WebSocket server: port 8765
+- HTTP server: port 8766 (serves client/index.html)
+- Client has ANSI color parsing and fullscreen support
+
+### Multi Backend
+
+Forwards to multiple displays simultaneously - useful for `terminal + websocket`.
+
+## Adding a New Backend
+
+1. Create `engine/display/backends/my_backend.py`
+2. Implement the Display protocol methods
+3. Register in `engine/display/__init__.py`'s `DisplayRegistry`
+
+Required methods:
+- `init(width: int, height: int, reuse: bool = False)` - Initialize display
+- `show(buf: list[str], border: bool = False)` - Display buffer
+- `clear()` - Clear screen
+- `cleanup()` - Clean up resources
+- `get_dimensions() -> tuple[int, int]` - Get terminal dimensions
+
+Optional methods:
+- `title(text: str)` - Set window title
+- `cursor(show: bool)` - Control cursor
+
+## Usage
+
+```bash
+python mainline.py --display terminal  # default
+python mainline.py --display websocket
+python mainline.py --display moderngl  # GPU-accelerated (requires moderngl)
+```
+
+## Common Bugs and Patterns
+
+### BorderMode.OFF Enum Bug
+
+**Problem**: `BorderMode.OFF` has enum value `1` (not `0`), and Python enums are always truthy.
+
+**Incorrect Code**:
+```python
+if border:
+    buffer = render_border(buffer, width, height, fps, frame_time)
+```
+
+**Correct Code**:
+```python
+from engine.display import BorderMode
+if border and border != BorderMode.OFF:
+    buffer = render_border(buffer, width, height, fps, frame_time)
+```
+
+**Why**: Checking `if border:` evaluates to `True` even when `border == BorderMode.OFF` because enum members are always truthy in Python.
+
+### Context Type Mismatch
+
+**Problem**: `PipelineContext` and `EffectContext` have different APIs for storing data.
+
+- `PipelineContext`: Uses `set()`/`get()` for services
+- `EffectContext`: Uses `set_state()`/`get_state()` for state
+
+**Pattern for Passing Data**:
+```python
+# In pipeline setup (uses PipelineContext)
+ctx.set("pipeline_order", pipeline.execution_order)
+
+# In EffectPluginStage (must copy to EffectContext)
+effect_ctx.set_state("pipeline_order", ctx.get("pipeline_order"))
+```
+
+### Terminal Display ANSI Patterns
+
+**Screen Clearing**:
+```python
+output = "\033[H\033[J" + "".join(buffer)
+```
+
+**Cursor Positioning** (used by HUD effect):
+- `\033[row;colH` - Move cursor to row, column
+- Example: `\033[1;1H` - Move to row 1, column 1
+
+**Key Insight**: Terminal display joins buffer lines WITHOUT newlines, relying on ANSI cursor positioning codes to move the cursor to the correct location for each line.
+
+### EffectPluginStage Context Copying
+
+**Problem**: When effects need access to pipeline services (like `pipeline_order`), they must be copied from `PipelineContext` to `EffectContext`.
+
+**Pattern**:
+```python
+# In EffectPluginStage.process()
+# Copy pipeline_order from PipelineContext services to EffectContext state
+pipeline_order = ctx.get("pipeline_order")
+if pipeline_order:
+    effect_ctx.set_state("pipeline_order", pipeline_order)
+```
+
+This ensures effects can access `ctx.get_state("pipeline_order")` in their process method.

.opencode/skills/mainline-effects/SKILL.md

@@ -0,0 +1,113 @@
+---
+name: mainline-effects
+description: How to add new effect plugins to Mainline's effect system
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers Mainline's effect plugin system - how to create, configure, and integrate visual effects into the pipeline.
+
+## Key Concepts
+
+### EffectPlugin ABC (engine/effects/types.py)
+
+All effects must inherit from `EffectPlugin` and implement:
+
+```python
+class EffectPlugin(ABC):
+    name: str
+    config: EffectConfig
+    param_bindings: dict[str, dict[str, str | float]] = {}
+    supports_partial_updates: bool = False
+    
+    @abstractmethod
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Process buffer with effect applied"""
+        ...
+    
+    @abstractmethod
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect"""
+        ...
+```
+
+### EffectContext
+
+Passed to every effect's process method:
+
+```python
+@dataclass
+class EffectContext:
+    terminal_width: int
+    terminal_height: int
+    scroll_cam: int
+    ticker_height: int
+    camera_x: int = 0
+    mic_excess: float = 0.0
+    grad_offset: float = 0.0
+    frame_number: int = 0
+    has_message: bool = False
+    items: list = field(default_factory=list)
+    _state: dict[str, Any] = field(default_factory=dict)
+```
+
+Access sensor values via `ctx.get_sensor_value("sensor_name")`.
+
+### EffectConfig
+
+Configuration dataclass:
+
+```python
+@dataclass
+class EffectConfig:
+    enabled: bool = True
+    intensity: float = 1.0
+    params: dict[str, Any] = field(default_factory=dict)
+```
+
+### Partial Updates
+
+For performance optimization, set `supports_partial_updates = True` and implement `process_partial`:
+
+```python
+class MyEffect(EffectPlugin):
+    supports_partial_updates = True
+    
+    def process_partial(self, buf, ctx, partial: PartialUpdate) -> list[str]:
+        # Only process changed regions
+        ...
+```
+
+## Adding a New Effect
+
+1. Create file in `effects_plugins/my_effect.py`
+2. Inherit from `EffectPlugin`
+3. Implement `process()` and `configure()`
+4. Add to `effects_plugins/__init__.py` (runtime discovery via issubclass checks)
+
+## Param Bindings
+
+Declarative sensor-to-param mappings:
+
+```python
+param_bindings = {
+    "intensity": {"sensor": "mic", "transform": "linear"},
+    "rate": {"sensor": "oscillator", "transform": "exponential"},
+}
+```
+
+Transforms: `linear`, `exponential`, `threshold`
+
+## Effect Chain
+
+Effects are chained via `engine/effects/chain.py` - processes each effect in order, passing output to next.
+
+## Existing Effects
+
+See `effects_plugins/`:
+- noise.py, fade.py, glitch.py, firehose.py
+- border.py, crop.py, tint.py, hud.py

.opencode/skills/mainline-presets/SKILL.md

@@ -0,0 +1,103 @@
+---
+name: mainline-presets
+description: Creating pipeline presets in TOML format for Mainline
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers how to create pipeline presets in TOML format for Mainline's rendering pipeline.
+
+## Key Concepts
+
+### Preset Loading Order
+
+Presets are loaded from multiple locations (later overrides earlier):
+1. Built-in: `engine/presets.toml`
+2. User config: `~/.config/mainline/presets.toml`
+3. Local override: `./presets.toml`
+
+### PipelinePreset Dataclass
+
+```python
+@dataclass
+class PipelinePreset:
+    name: str
+    description: str = ""
+    source: str = "headlines"      # Data source
+    display: str = "terminal"      # Display backend
+    camera: str = "scroll"          # Camera mode
+    effects: list[str] = field(default_factory=list)
+    border: bool = False
+```
+
+### TOML Format
+
+```toml
+[presets.my-preset]
+description = "My custom pipeline"
+source = "headlines"
+display = "terminal"
+camera = "scroll"
+effects = ["noise", "fade"]
+border = true
+```
+
+## Creating a Preset
+
+### Option 1: User Config
+
+Create/edit `~/.config/mainline/presets.toml`:
+
+```toml
+[presets.my-cool-preset]
+description = "Noise and glitch effects"
+source = "headlines"
+display = "terminal"
+effects = ["noise", "glitch"]
+```
+
+### Option 2: Local Override
+
+Create `./presets.toml` in project root:
+
+```toml
+[presets.dev-inspect]
+description = "Pipeline introspection for development"
+source = "headlines"
+display = "terminal"
+effects = ["hud"]
+```
+
+### Option 3: Built-in
+
+Edit `engine/presets.toml` (requires PR to repository).
+
+## Available Sources
+
+- `headlines` - RSS news feeds
+- `poetry` - Literature mode
+- `pipeline-inspect` - Live DAG visualization
+
+## Available Displays
+
+- `terminal` - ANSI terminal
+- `websocket` - Web browser
+- `null` - Headless
+- `moderngl` - GPU-accelerated (optional)
+
+## Available Effects
+
+See `effects_plugins/`:
+- noise, fade, glitch, firehose
+- border, crop, tint, hud
+
+## Validation Functions
+
+Use these from `engine/pipeline/presets.py`:
+- `validate_preset()` - Validate preset structure
+- `validate_signal_path()` - Detect circular dependencies
+- `generate_preset_toml()` - Generate skeleton preset

.opencode/skills/mainline-sensors/SKILL.md

@@ -0,0 +1,136 @@
+---
+name: mainline-sensors
+description: Sensor framework for real-time input in Mainline
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers Mainline's sensor framework - how to use, create, and integrate sensors for real-time input.
+
+## Key Concepts
+
+### Sensor Base Class (engine/sensors/__init__.py)
+
+```python
+class Sensor(ABC):
+    name: str
+    unit: str = ""
+    
+    @property
+    def available(self) -> bool:
+        """Whether sensor is currently available"""
+        return True
+    
+    @abstractmethod
+    def read(self) -> SensorValue | None:
+        """Read current sensor value"""
+        ...
+    
+    def start(self) -> None:
+        """Initialize sensor (optional)"""
+        pass
+    
+    def stop(self) -> None:
+        """Clean up sensor (optional)"""
+        pass
+```
+
+### SensorValue Dataclass
+
+```python
+@dataclass
+class SensorValue:
+    sensor_name: str
+    value: float
+    timestamp: float
+    unit: str = ""
+```
+
+### SensorRegistry
+
+Discovers and manages sensors globally:
+
+```python
+from engine.sensors import SensorRegistry
+registry = SensorRegistry()
+sensor = registry.get("mic")
+```
+
+### SensorStage
+
+Pipeline adapter that provides sensor values to effects:
+
+```python
+from engine.pipeline.adapters import SensorStage
+stage = SensorStage(sensor_name="mic")
+```
+
+## Built-in Sensors
+
+| Sensor | File | Description |
+|--------|------|-------------|
+| MicSensor | sensors/mic.py | Microphone input (RMS dB) |
+| OscillatorSensor | sensors/oscillator.py | Test sine wave generator |
+| PipelineMetricsSensor | sensors/pipeline_metrics.py | FPS, frame time, etc. |
+
+## Param Bindings
+
+Effects declare sensor-to-param mappings:
+
+```python
+class GlitchEffect(EffectPlugin):
+    param_bindings = {
+        "intensity": {"sensor": "mic", "transform": "linear"},
+    }
+```
+
+### Transform Functions
+
+- `linear` - Direct mapping to param range
+- `exponential` - Exponential scaling
+- `threshold` - Binary on/off
+
+## Adding a New Sensor
+
+1. Create `engine/sensors/my_sensor.py`
+2. Inherit from `Sensor` ABC
+3. Implement required methods
+4. Register in `SensorRegistry`
+
+Example:
+```python
+class MySensor(Sensor):
+    name = "my-sensor"
+    unit = "units"
+    
+    def read(self) -> SensorValue | None:
+        return SensorValue(
+            sensor_name=self.name,
+            value=self._read_hardware(),
+            timestamp=time.time(),
+            unit=self.unit
+        )
+```
+
+## Using Sensors in Effects
+
+Access sensor values via EffectContext:
+
+```python
+def process(self, buf, ctx):
+    mic_level = ctx.get_sensor_value("mic")
+    if mic_level and mic_level > 0.5:
+        # Apply intense effect
+        ...
+```
+
+Or via param_bindings (automatic):
+
+```python
+# If intensity is bound to "mic", it's automatically 
+# available in self.config.intensity
+```

.opencode/skills/mainline-sources/SKILL.md

@@ -0,0 +1,87 @@
+---
+name: mainline-sources
+description: Adding new RSS feeds and data sources to Mainline
+compatibility: opencode
+metadata:
+  audience: developers
+  source_type: codebase
+---
+
+## What This Skill Covers
+
+This skill covers how to add new data sources (RSS feeds, poetry) to Mainline.
+
+## Key Concepts
+
+### Feeds Dictionary (engine/sources.py)
+
+All feeds are defined in a simple dictionary:
+
+```python
+FEEDS = {
+    "Feed Name": "https://example.com/feed.xml",
+    # Category comments help organize:
+    # Science & Technology
+    # Economics & Business
+    # World & Politics
+    # Culture & Ideas
+}
+```
+
+### Poetry Sources
+
+Project Gutenberg URLs for public domain literature:
+
+```python
+POETRY_SOURCES = {
+    "Author Name": "https://www.gutenberg.org/cache/epub/1234/pg1234.txt",
+}
+```
+
+### Language & Script Mapping
+
+The sources.py also contains language/script detection mappings used for auto-translation and font selection.
+
+## Adding a New RSS Feed
+
+1. Edit `engine/sources.py`
+2. Add entry to `FEEDS` dict under appropriate category:
+   ```python
+   "My Feed": "https://example.com/feed.xml",
+   ```
+3. The feed will be automatically discovered on next run
+
+### Feed Requirements
+
+- Must be valid RSS or Atom XML
+- Should have `<title>` elements for items
+- Must be HTTP/HTTPS accessible
+
+## Adding Poetry Sources
+
+1. Edit `engine/sources.py`
+2. Add to `POETRY_SOURCES` dict:
+   ```python
+   "Author": "https://www.gutenberg.org/cache/epub/XXXX/pgXXXX.txt",
+   ```
+
+### Poetry Requirements
+
+- Plain text (UTF-8)
+- Project Gutenberg format preferred
+- No DRM-protected sources
+
+## Data Flow
+
+Feeds are fetched via `engine/fetch.py`:
+- `fetch_feed(url)` - Fetches and parses RSS/Atom
+- Results cached for fast restarts
+- Filtered via `engine/filter.py` for content cleaning
+
+## Categories
+
+Organize new feeds by category using comments:
+- Science & Technology
+- Economics & Business  
+- World & Politics
+- Culture & Ideas

.python-version

@@ -0,0 +1 @@
+3.12

AGENTS.md

@@ -0,0 +1,465 @@
+# Agent Development Guide
+
+## Development Environment
+
+This project uses:
+- **mise** (mise.jdx.dev) - tool version manager and task runner
+- **uv** - fast Python package installer
+- **ruff** - linter and formatter (line-length 88, target Python 3.10)
+- **pytest** - test runner with strict marker enforcement
+
+### Setup
+
+```bash
+mise run install          # Install dependencies
+# Or: uv sync --all-extras   # includes mic, websocket support
+```
+
+### Available Commands
+
+```bash
+# Testing
+mise run test            # Run all tests
+mise run test-cov       # Run tests with coverage report
+pytest tests/test_foo.py::TestClass::test_method  # Run single test
+
+# Linting & Formatting
+mise run lint            # Run ruff linter
+mise run lint-fix       # Run ruff with auto-fix
+mise run format         # Run ruff formatter
+
+# CI
+mise run ci             # Full CI pipeline (topics-init + lint + test-cov)
+```
+
+### Running a Single Test
+
+```bash
+# Run a specific test function
+pytest tests/test_eventbus.py::TestEventBusInit::test_init_creates_empty_subscribers
+
+# Run all tests in a file
+pytest tests/test_eventbus.py
+
+# Run tests matching a pattern
+pytest -k "test_subscribe"
+```
+
+### Git Hooks
+
+Install hooks at start of session:
+```bash
+ls -la .git/hooks/pre-commit  # Verify installed
+hk init --mise                # Install if missing
+mise run pre-commit           # Run manually
+```
+
+## Code Style Guidelines
+
+### Imports (three sections, alphabetical within each)
+
+```python
+# 1. Standard library
+import os
+import threading
+from collections import defaultdict
+from collections.abc import Callable
+from dataclasses import dataclass, field
+from typing import Any
+
+# 2. Third-party
+from abc import ABC, abstractmethod
+
+# 3. Local project
+from engine.events import EventType
+```
+
+### Type Hints
+
+- Use type hints for all function signatures (parameters and return)
+- Use `|` for unions (Python 3.10+): `EventType | None`
+- Use `dict[K, V]`, `list[V]` (generic syntax): `dict[str, list[int]]`
+- Use `Callable[[ArgType], ReturnType]` for callbacks
+
+```python
+def subscribe(self, event_type: EventType, callback: Callable[[Any], None]) -> None:
+    ...
+
+def get_sensor_value(self, sensor_name: str) -> float | None:
+    return self._state.get(f"sensor.{sensor_name}")
+```
+
+### Naming Conventions
+
+- **Classes**: `PascalCase` (e.g., `EventBus`, `EffectPlugin`)
+- **Functions/methods**: `snake_case` (e.g., `get_event_bus`, `process_partial`)
+- **Constants**: `SCREAMING_SNAKE_CASE` (e.g., `CURSOR_OFF`)
+- **Private methods**: `_snake_case` prefix (e.g., `_initialize`)
+- **Type variables**: `PascalCase` (e.g., `T`, `EffectT`)
+
+### Dataclasses
+
+Use `@dataclass` for simple data containers:
+
+```python
+@dataclass
+class EffectContext:
+    terminal_width: int
+    terminal_height: int
+    scroll_cam: int
+    ticker_height: int = 0
+    _state: dict[str, Any] = field(default_factory=dict, repr=False)
+```
+
+### Abstract Base Classes
+
+Use ABC for interface enforcement:
+
+```python
+class EffectPlugin(ABC):
+    name: str
+    config: EffectConfig
+    
+    @abstractmethod
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        ...
+    
+    @abstractmethod
+    def configure(self, config: EffectConfig) -> None:
+        ...
+```
+
+### Error Handling
+
+- Catch specific exceptions, not bare `Exception`
+- Use `try/except` with fallbacks for optional features
+- Silent pass in event callbacks to prevent one handler from breaking others
+
+```python
+# Good: specific exception
+try:
+    term_size = os.get_terminal_size()
+except OSError:
+    term_width = 80
+
+# Good: silent pass in callbacks
+for callback in callbacks:
+    try:
+        callback(event)
+    except Exception:
+        pass
+```
+
+### Thread Safety
+
+Use locks for shared state:
+
+```python
+class EventBus:
+    def __init__(self):
+        self._lock = threading.Lock()
+    
+    def publish(self, event_type: EventType, event: Any = None) -> None:
+        with self._lock:
+            callbacks = list(self._subscribers.get(event_type, []))
+```
+
+### Comments
+
+- **DO NOT ADD comments** unless explicitly required
+- Let code be self-documenting with good naming
+- Use docstrings only for public APIs or complex logic
+
+### Testing Patterns
+
+Follow pytest conventions:
+
+```python
+class TestEventBusSubscribe:
+    """Tests for EventBus.subscribe method."""
+    
+    def test_subscribe_adds_callback(self):
+        """subscribe() adds a callback for an event type."""
+        bus = EventBus()
+        def callback(e):
+            return None
+        bus.subscribe(EventType.NTFY_MESSAGE, callback)
+        assert bus.subscriber_count(EventType.NTFY_MESSAGE) == 1
+```
+
+- Use classes to group related tests (`Test<ClassName>`, `Test<method_name>`)
+- Test docstrings follow `"<method>() <action>"` pattern
+- Use descriptive assertion messages via pytest behavior
+
+## Workflow Rules
+
+### Before Committing
+
+1. Run tests: `mise run test`
+2. Run linter: `mise run lint`
+3. Review changes: `git diff`
+
+### On Failing Tests
+
+- **Out-of-date test**: Update test to match new expected behavior
+- **Correctly failing test**: Fix implementation, not the test
+
+**Never** modify a test to make it pass without understanding why it failed.
+
+## Testing
+
+Tests live in `tests/` and follow the pattern `test_*.py`.
+
+Run all tests:
+```bash
+mise run test
+```
+
+Run with coverage:
+```bash
+mise run test-cov
+```
+
+The project uses pytest with strict marker enforcement. Test configuration is in `pyproject.toml` under `[tool.pytest.ini_options]`.
+
+### Test Coverage Strategy
+
+Current coverage: 56% (463 tests)
+
+Key areas with lower coverage (acceptable for now):
+- **app.py** (8%): Main entry point - integration heavy, requires terminal
+- **scroll.py** (10%): Terminal-dependent rendering logic (unused)
+
+Key areas with good coverage:
+- **display/backends/null.py** (95%): Easy to test headlessly
+- **display/backends/terminal.py** (96%): Uses mocking
+- **display/backends/multi.py** (100%): Simple forwarding logic
+- **effects/performance.py** (99%): Pure Python logic
+- **eventbus.py** (96%): Simple event system
+- **effects/controller.py** (95%): Effects command handling
+
+Areas needing more tests:
+- **websocket.py** (48%): Network I/O, hard to test in CI
+- **ntfy.py** (50%): Network I/O, hard to test in CI
+- **mic.py** (61%): Audio I/O, hard to test in CI
+
+Note: Terminal-dependent modules (scroll, layers render) are harder to test in CI.
+Performance regression tests are in `tests/test_benchmark.py` with `@pytest.mark.benchmark`.
+
+## Architecture Notes
+
+- **ntfy.py** - standalone notification poller with zero internal dependencies
+- **sensors/** - Sensor framework (MicSensor, OscillatorSensor) for real-time input
+- **eventbus.py** provides thread-safe event publishing for decoupled communication
+- **effects/** - plugin architecture with performance monitoring
+- The new pipeline architecture: source → render → effects → display
+
+#### Canvas & Camera
+
+- **Canvas** (`engine/canvas.py`): 2D rendering surface with dirty region tracking
+- **Camera** (`engine/camera.py`): Viewport controller for scrolling content
+
+The Canvas tracks dirty regions automatically when content is written (via `put_region`, `put_text`, `fill`), enabling partial buffer updates for optimized effect processing.
+
+### Pipeline Architecture
+
+The new Stage-based pipeline architecture provides capability-based dependency resolution:
+
+- **Stage** (`engine/pipeline/core.py`): Base class for pipeline stages
+- **Pipeline** (`engine/pipeline/controller.py`): Executes stages with capability-based dependency resolution
+- **PipelineConfig** (`engine/pipeline/controller.py`): Configuration for pipeline instance
+- **StageRegistry** (`engine/pipeline/registry.py`): Discovers and registers stages
+- **Stage Adapters** (`engine/pipeline/adapters.py`): Wraps existing components as stages
+
+#### Pipeline Configuration
+
+The `PipelineConfig` dataclass configures pipeline behavior:
+
+```python
+@dataclass
+class PipelineConfig:
+    source: str = "headlines"      # Data source identifier
+    display: str = "terminal"      # Display backend identifier
+    camera: str = "vertical"       # Camera mode identifier
+    effects: list[str] = field(default_factory=list)  # List of effect names
+    enable_metrics: bool = True    # Enable performance metrics
+```
+
+**Available sources**: `headlines`, `poetry`, `empty`, `list`, `image`, `metrics`, `cached`, `transform`, `composite`, `pipeline-inspect`
+**Available displays**: `terminal`, `null`, `replay`, `websocket`, `pygame`, `moderngl`, `multi`
+**Available camera modes**: `FEED`, `SCROLL`, `HORIZONTAL`, `OMNI`, `FLOATING`, `BOUNCE`, `RADIAL`
+
+#### Capability-Based Dependencies
+
+Stages declare capabilities (what they provide) and dependencies (what they need). The Pipeline resolves dependencies using prefix matching:
+- `"source"` matches `"source.headlines"`, `"source.poetry"`, etc.
+- `"camera.state"` matches the camera state capability
+- This allows flexible composition without hardcoding specific stage names
+
+#### Minimum Capabilities
+
+The pipeline requires these minimum capabilities to function:
+- `"source"` - Data source capability
+- `"render.output"` - Rendered content capability
+- `"display.output"` - Display output capability
+- `"camera.state"` - Camera state for viewport filtering
+
+These are automatically injected if missing by the `ensure_minimum_capabilities()` method.
+
+#### Sensor Framework
+
+- **Sensor** (`engine/sensors/__init__.py`): Base class for real-time input sensors
+- **SensorRegistry**: Discovers available sensors
+- **SensorStage**: Pipeline adapter that provides sensor values to effects
+- **MicSensor** (`engine/sensors/mic.py`): Self-contained microphone input
+- **OscillatorSensor** (`engine/sensors/oscillator.py`): Test sensor for development
+- **PipelineMetricsSensor** (`engine/sensors/pipeline_metrics.py`): Exposes pipeline metrics as sensor values
+
+Sensors support param bindings to drive effect parameters in real-time.
+
+#### Pipeline Introspection
+
+- **PipelineIntrospectionSource** (`engine/data_sources/pipeline_introspection.py`): Renders live ASCII visualization of pipeline DAG with metrics
+- **PipelineIntrospectionDemo** (`engine/pipeline/pipeline_introspection_demo.py`): 3-phase demo controller for effect animation
+
+Preset: `pipeline-inspect` - Live pipeline introspection with DAG and performance metrics
+
+#### Partial Update Support
+
+Effect plugins can opt-in to partial buffer updates for performance optimization:
+- Set `supports_partial_updates = True` on the effect class
+- Implement `process_partial(buf, ctx, partial)` method
+- The `PartialUpdate` dataclass indicates which regions changed
+
+### Preset System
+
+Presets use TOML format (no external dependencies):
+
+- Built-in: `engine/presets.toml`
+- User config: `~/.config/mainline/presets.toml`
+- Local override: `./presets.toml`
+
+- **Preset loader** (`engine/pipeline/preset_loader.py`): Loads and validates presets
+- **PipelinePreset** (`engine/pipeline/presets.py`): Dataclass for preset configuration
+
+Functions:
+- `validate_preset()` - Validate preset structure
+- `validate_signal_path()` - Detect circular dependencies
+- `generate_preset_toml()` - Generate skeleton preset
+
+### Display System
+
+- **Display abstraction** (`engine/display/`): swap display backends via the Display protocol
+  - `display/backends/terminal.py` - ANSI terminal output
+  - `display/backends/websocket.py` - broadcasts to web clients via WebSocket
+  - `display/backends/null.py` - headless display for testing
+  - `display/backends/multi.py` - forwards to multiple displays simultaneously
+  - `display/backends/moderngl.py` - GPU-accelerated OpenGL rendering (optional)
+  - `display/__init__.py` - DisplayRegistry for backend discovery
+
+- **WebSocket display** (`engine/display/backends/websocket.py`): real-time frame broadcasting to web browsers
+  - WebSocket server on port 8765
+  - HTTP server on port 8766 (serves HTML client)
+  - Client at `client/index.html` with ANSI color parsing and fullscreen support
+
+- **Display modes** (`--display` flag):
+  - `terminal` - Default ANSI terminal output
+  - `websocket` - Web browser display (requires websockets package)
+  - `moderngl` - GPU-accelerated rendering (requires moderngl package)
+
+### Effect Plugin System
+
+- **EffectPlugin ABC** (`engine/effects/types.py`): abstract base class for effects
+  - All effects must inherit from EffectPlugin and implement `process()` and `configure()`
+  - Runtime discovery via `effects_plugins/__init__.py` using `issubclass()` checks
+  
+- **EffectRegistry** (`engine/effects/registry.py`): manages registered effects
+- **EffectChain** (`engine/effects/chain.py`): chains effects in pipeline order
+
+### Command & Control
+
+- C&C uses separate ntfy topics for commands and responses
+- `NTFY_CC_CMD_TOPIC` - commands from cmdline.py
+- `NTFY_CC_RESP_TOPIC` - responses back to cmdline.py
+- Effects controller handles `/effects` commands (list, on/off, intensity, reorder, stats)
+
+### Pipeline Documentation
+
+The rendering pipeline is documented in `docs/PIPELINE.md` using Mermaid diagrams.
+
+**IMPORTANT**: When making significant architectural changes to the rendering pipeline (new layers, effects, display backends), update `docs/PIPELINE.md` to reflect the changes:
+1. Edit `docs/PIPELINE.md` with the new architecture
+2. If adding new SVG diagrams, render them manually using an external tool (e.g., Mermaid Live Editor)
+3. Commit both the markdown and any new diagram files
+
+### Pipeline Mutation API
+
+The Pipeline class supports dynamic mutation during runtime via the mutation API:
+
+**Core Methods:**
+- `add_stage(name, stage, initialize=True)` - Add a stage to the pipeline
+- `remove_stage(name, cleanup=True)` - Remove a stage and rebuild execution order
+- `replace_stage(name, new_stage, preserve_state=True)` - Replace a stage with another
+- `swap_stages(name1, name2)` - Swap two stages
+- `move_stage(name, after=None, before=None)` - Move a stage in execution order
+- `enable_stage(name)` - Enable a stage
+- `disable_stage(name)` - Disable a stage
+
+**New Methods (Issue #35):**
+- `cleanup_stage(name)` - Clean up specific stage without removing it
+- `remove_stage_safe(name, cleanup=True)` - Alias for remove_stage that explicitly rebuilds
+- `can_hot_swap(name)` - Check if a stage can be safely hot-swapped
+  - Returns False for stages that provide minimum capabilities as sole provider
+  - Returns True for swappable stages
+
+**WebSocket Commands:**
+Commands can be sent via WebSocket to mutate the pipeline at runtime:
+```json
+{"action": "remove_stage", "stage": "stage_name"}
+{"action": "swap_stages", "stage1": "name1", "stage2": "name2"}
+{"action": "enable_stage", "stage": "stage_name"}
+{"action": "disable_stage", "stage": "stage_name"}
+{"action": "cleanup_stage", "stage": "stage_name"}
+{"action": "can_hot_swap", "stage": "stage_name"}
+```
+
+**Implementation Files:**
+- `engine/pipeline/controller.py` - Pipeline class with mutation methods
+- `engine/app/pipeline_runner.py` - `_handle_pipeline_mutation()` function
+- `engine/pipeline/ui.py` - execute_command() with docstrings
+- `tests/test_pipeline_mutation_commands.py` - Integration tests
+
+## Skills Library
+
+A skills library MCP server (`skills`) is available for capturing and tracking learned knowledge. Skills are stored in `~/.skills/`.
+
+### Workflow
+
+**Before starting work:**
+1. Run `local_skills_list_skills` to see available skills
+2. Use `local_skills_peek_skill({name: "skill-name"})` to preview relevant skills
+3. Use `local_skills_skill_slice({name: "skill-name", query: "your question"})` to get relevant sections
+
+**While working:**
+- If a skill was wrong or incomplete: `local_skills_update_skill` → `local_skills_record_assessment` → `local_skills_report_outcome({quality: 1})`
+- If a skill worked correctly: `local_skills_report_outcome({quality: 4})` (normal) or `quality: 5` (perfect)
+
+**End of session:**
+- Run `local_skills_reflect_on_session({context_summary: "what you did"})` to identify new skills to capture
+- Use `local_skills_create_skill` to add new skills
+- Use `local_skills_record_assessment` to score them
+
+### Useful Tools
+- `local_skills_review_stale_skills()` - Skills due for review (negative days_until_due)
+- `local_skills_skills_report()` - Overview of entire collection
+- `local_skills_validate_skill({name: "skill-name"})` - Load skill for review with sources
+
+### Agent Skills
+
+This project also has Agent Skills (SKILL.md files) in `.opencode/skills/`. Use the `skill` tool to load them:
+- `skill({name: "mainline-architecture"})` - Pipeline stages, capability resolution
+- `skill({name: "mainline-effects"})` - How to add new effect plugins
+- `skill({name: "mainline-display"})` - Display backend implementation
+- `skill({name: "mainline-sources"})` - Adding new RSS feeds
+- `skill({name: "mainline-presets"})` - Creating pipeline presets
+- `skill({name: "mainline-sensors"})` - Sensor framework usage

README.md

@@ -2,79 +2,258 @@
 
 > *Digital consciousness stream. Matrix aesthetic · THX-1138 hue.*
 
-A full-screen terminal news ticker that renders live global headlines in large OTF-font block characters with a white-hot → deep green gradient. Headlines auto-translate into the native script of their subject region. Ambient mic input warps the glitch rate in real time. A `--poetry` mode replaces the feed with public-domain literary passages.
+A full-screen terminal news ticker that renders live global headlines in large OTF-font block characters with a white-hot → deep green gradient. Headlines auto-translate into the native script of their subject region. Ambient mic input warps the glitch rate in real time. A `--poetry` mode replaces the feed with public-domain literary passages. Live messages can be pushed to the display over [ntfy.sh](https://ntfy.sh).
 
 ---
 
-## Run
+## Using
+
+### Run
 
 ```bash
-python3 mainline.py            # news stream
-python3 mainline.py --poetry   # literary consciousness mode
-python3 mainline.py -p         # same
+python3 mainline.py                      # news stream
+python3 mainline.py --poetry             # literary consciousness mode
+python3 mainline.py -p                   # same
+python3 mainline.py --firehose           # dense rapid-fire headline mode
+python3 mainline.py --display websocket  # web browser display only
+python3 mainline.py --no-font-picker     # skip interactive font picker
+python3 mainline.py --font-file path.otf # use a specific font file
+python3 mainline.py --font-dir ~/fonts   # scan a different font folder
+python3 mainline.py --font-index 1       # select face index within a collection
 ```
 
-First run bootstraps a local `.mainline_venv/` and installs deps (`feedparser`, `Pillow`, `sounddevice`, `numpy`). Subsequent runs start immediately.
+Or with uv:
 
----
+```bash
+uv run mainline.py
+```
 
-## Config
+First run bootstraps dependencies. Use `uv sync --all-extras` for mic support.
 
-At the top of `mainline.py`:
+### Command & Control (C&C)
+
+Control mainline remotely using `cmdline.py`:
+
+```bash
+uv run cmdline.py                    # Interactive TUI
+uv run cmdline.py /effects list      # List all effects
+uv run cmdline.py /effects stats     # Show performance stats
+uv run cmdline.py -w /effects stats  # Watch mode (auto-refresh)
+```
+
+Commands are sent via ntfy.sh topics - useful for controlling a daemonized mainline instance.
+
+### Config
+
+All constants live in `engine/config.py`:
 
 | Constant | Default | What it does |
 |---|---|---|
 | `HEADLINE_LIMIT` | `1000` | Total headlines per session |
+| `FEED_TIMEOUT` | `10` | Per-feed HTTP timeout (seconds) |
 | `MIC_THRESHOLD_DB` | `50` | dB floor above which glitches spike |
-| `_FONT_PATH` | hardcoded path | Path to your OTF/TTF display font |
-| `_FONT_SZ` | `60` | Font render size (affects block density) |
-| `_RENDER_H` | `8` | Terminal rows per headline line |
+| `NTFY_TOPIC` | klubhaus URL | ntfy.sh JSON stream for messages |
+| `NTFY_CC_CMD_TOPIC` | klubhaus URL | ntfy.sh topic for C&C commands |
+| `NTFY_CC_RESP_TOPIC` | klubhaus URL | ntfy.sh topic for C&C responses |
+| `NTFY_RECONNECT_DELAY` | `5` | Seconds before reconnecting after dropped SSE |
+| `MESSAGE_DISPLAY_SECS` | `30` | How long an ntfy message holds the screen |
+| `FONT_DIR` | `fonts/` | Folder scanned for `.otf`, `.ttf`, `.ttc` files |
+| `FONT_PATH` | first file in `FONT_DIR` | Active display font |
+| `FONT_PICKER` | `True` | Show interactive font picker at boot |
+| `FONT_SZ` | `60` | Font render size (affects block density) |
+| `RENDER_H` | `8` | Terminal rows per headline line |
+| `SSAA` | `4` | Super-sampling factor |
+| `SCROLL_DUR` | `5.625` | Seconds per headline |
+| `FRAME_DT` | `0.05` | Frame interval in seconds (20 FPS) |
+| `FIREHOSE_H` | `12` | Firehose zone height (terminal rows) |
+| `GRAD_SPEED` | `0.08` | Gradient sweep speed |
 
-**Font:** `_FONT_PATH` is hardcoded to a local path. Update it to point to whatever display font you want — anything with strong contrast and wide letterforms works well.
+### Display Modes
+
+Mainline supports multiple display backends:
+
+- **Terminal** (`--display terminal`): ANSI terminal output (default)
+- **WebSocket** (`--display websocket`): Stream to web browser clients
+- **ModernGL** (`--display moderngl`): GPU-accelerated rendering (optional)
+
+WebSocket mode serves a web client at http://localhost:8766 with ANSI color support and fullscreen mode.
+
+### Feeds
+
+~25 sources across four categories: **Science & Technology**, **Economics & Business**, **World & Politics**, **Culture & Ideas**. Add or swap feeds in `engine/sources.py` → `FEEDS`.
+
+**Poetry mode** pulls from Project Gutenberg: Whitman, Dickinson, Thoreau, Emerson. Sources are in `engine/sources.py` → `POETRY_SOURCES`.
+
+### Fonts
+
+A `fonts/` directory is bundled with demo faces. On startup, an interactive picker lists all discovered faces with a live half-block preview.
+
+Navigation: `↑`/`↓` or `j`/`k` to move, `Enter` or `q` to select.
+
+To add your own fonts, drop `.otf`, `.ttf`, or `.ttc` files into `fonts/`.
+
+### ntfy.sh
+
+Mainline polls a configurable ntfy.sh topic in the background. When a message arrives, the scroll pauses and the message renders full-screen.
+
+To push a message:
+
+```bash
+curl -d "Body text" -H "Title: Alert title" https://ntfy.sh/your_topic
+```
 
 ---
 
-## How it works
+## Internals
 
-- Feeds are fetched and filtered on startup (sports and vapid content stripped)
-- Headlines are rasterized via Pillow into half-block characters (`▀▄█ `) at the configured font size
-- A left-to-right ANSI gradient colors each character: white-hot leading edge trails off to near-black
-- Subject-region detection runs a regex pass on each headline; matches trigger a Google Translate call and font swap to the appropriate script (CJK, Arabic, Devanagari, etc.) using macOS system fonts
-- The mic stream runs in a background thread, feeding RMS dB into the glitch probability calculation each frame
-- The viewport scrolls through a virtual canvas of pre-rendered blocks; fade zones at top and bottom dissolve characters probabilistically
+### How it works
+
+- On launch, the font picker scans `fonts/` and presents a live-rendered TUI for face selection
+- Feeds are fetched and filtered on startup; results are cached for fast restarts
+- Headlines are rasterized via Pillow with 4× SSAA into half-block characters
+- The ticker uses a sweeping white-hot → deep green gradient
+- Subject-region detection triggers Google Translate and font swap for non-Latin scripts
+- The mic stream runs in a background thread, feeding RMS dB into glitch probability
+- The viewport scrolls through pre-rendered blocks with fade zones
+- An ntfy.sh SSE stream runs in a background thread for messages and C&C commands
+
+### Architecture
+
+```
+engine/
+  __init__.py           package marker
+  app.py                main(), font picker TUI, boot sequence, C&C poller
+  config.py             constants, CLI flags, glyph tables
+  sources.py            FEEDS, POETRY_SOURCES, language/script maps
+  terminal.py           ANSI codes, tw/th, type_out, boot_ln
+  filter.py             HTML stripping, content filter
+  translate.py          Google Translate wrapper + region detection
+  render.py             OTF → half-block pipeline (SSAA, gradient)
+  effects/              plugin architecture for visual effects
+    types.py            EffectPlugin ABC, EffectConfig, EffectContext
+    registry.py         effect registration and lookup
+    chain.py            effect pipeline chaining
+    controller.py       handles /effects commands
+    performance.py      performance monitoring
+    legacy.py           legacy functional effects
+  effects_plugins/      effect plugin implementations
+    noise.py            noise effect
+    fade.py             fade effect
+    glitch.py           glitch effect
+    firehose.py         firehose effect
+  fetch.py              RSS/Gutenberg fetching + cache
+  ntfy.py               NtfyPoller — standalone, zero internal deps
+  mic.py                MicMonitor — standalone, graceful fallback
+  scroll.py             stream() frame loop + message rendering
+  viewport.py           terminal dimension tracking
+  frame.py              scroll step calculation, timing
+  layers.py             ticker zone, firehose, message overlay
+  eventbus.py           thread-safe event publishing
+  events.py             event types and definitions
+  controller.py         coordinates ntfy/mic monitoring
+  emitters.py           background emitters
+  types.py              type definitions
+  display/              Display backend system
+    __init__.py         DisplayRegistry, get_monitor
+    backends/
+      terminal.py       ANSI terminal display
+      websocket.py      WebSocket server for browser clients
+      null.py          headless display for testing
+      multi.py         forwards to multiple displays
+      moderngl.py       GPU-accelerated OpenGL rendering
+  benchmark.py          performance benchmarking tool
+```
 
 ---
 
-## Feeds
+## Development
 
-~25 sources across four categories: **Science & Technology**, **Economics & Business**, **World & Politics**, **Culture & Ideas**. Add or swap in `FEEDS`.
+### Setup
 
-**Poetry mode** pulls from Project Gutenberg: Whitman, Dickinson, Thoreau, Emerson.
+Requires Python 3.10+ and [uv](https://docs.astral.sh/uv/).
+
+```bash
+uv sync                              # minimal (no mic)
+uv sync --all-extras                 # with mic support
+uv sync --all-extras --group dev     # full dev environment
+```
+
+### Tasks
+
+With [mise](https://mise.jdx.dev/):
+
+```bash
+mise run test            # run test suite
+mise run test-cov       # run with coverage report
+
+mise run lint           # ruff check
+mise run lint-fix       # ruff check --fix
+mise run format         # ruff format
+
+mise run run            # terminal display
+mise run run-websocket  # web display only
+mise run run-client     # terminal + web
+
+mise run cmd            # C&C command interface
+mise run cmd-stats      # watch effects stats
+
+mise run benchmark      # run performance benchmarks
+mise run benchmark-json # save as JSON
+
+mise run topics-init    # initialize ntfy topics
+```
+
+### Testing
+
+```bash
+uv run pytest
+uv run pytest --cov=engine --cov-report=term-missing
+
+# Run with mise
+mise run test
+mise run test-cov
+
+# Run performance benchmarks
+mise run benchmark
+mise run benchmark-json
+
+# Run benchmark hook mode (for CI)
+uv run python -m engine.benchmark --hook
+```
+
+Performance regression tests are in `tests/test_benchmark.py` marked with `@pytest.mark.benchmark`.
+
+### Linting
+
+```bash
+uv run ruff check engine/ mainline.py
+uv run ruff format engine/ mainline.py
+```
+
+Pre-commit hooks run lint automatically via `hk`.
 
 ---
 
-## Ideas / Future
+## Roadmap
 
 ### Performance
-- **Concurrent feed fetching** — startup currently blocks sequentially on ~25 HTTP requests; `concurrent.futures.ThreadPoolExecutor` would cut load time to the slowest single feed
-- **Background refresh** — re-fetch feeds in a daemon thread so a long session stays current without restart
-- **Translation pre-fetch** — run translate calls concurrently during the boot sequence rather than on first render
+- Concurrent feed fetching with ThreadPoolExecutor
+- Background feed refresh daemon
+- Translation pre-fetch during boot
 
 ### Graphics
-- **Matrix rain underlay** — katakana column rain rendered at low opacity beneath the scrolling blocks as a background layer
-- **Animated gradient** — shift the white-hot leading edge left/right each frame for a pulse/comet effect
-- **CRT simulation** — subtle dim scanlines every N rows, occasional brightness ripple across the full screen
-- **Sixel / iTerm2 inline images** — bypass half-blocks entirely and stream actual bitmap frames for true resolution; would require a capable terminal
-- **Parallax secondary column** — a second, dimmer, faster-scrolling stream of ambient text at reduced opacity on one side
+- Matrix rain katakana underlay
+- CRT scanline simulation
+- Sixel/iTerm2 inline images
+- Parallax secondary column
 
 ### Cyberpunk Vibes
-- **Keyword watch list** — highlight or strobe any headline matching tracked terms (names, topics, tickers)
-- **Breaking interrupt** — full-screen flash + synthesized blip when a high-priority keyword hits
-- **Live data overlay** — secondary ticker strip at screen edge: BTC price, ISS position, geomagnetic index
-- **Theme switcher** — `--amber` (phosphor), `--ice` (electric cyan), `--red` (alert state) palette modes via CLI flag
-- **Persona modes** — `--surveillance`, `--oracle`, `--underground` as feed presets with matching color themes and boot copy
-- **Synthesized audio** — short static bursts tied to glitch events, independent of mic input
+- Keyword watch list with strobe effects
+- Breaking interrupt with synthesized audio
+- Live data overlay (BTC, ISS position)
+- Theme switcher (amber, ice, red)
+- Persona modes (surveillance, oracle, underground)
 
 ---
 
-*macOS only (system font paths hardcoded). Python 3.9+.*
+*Python 3.10+. Primary display font is user-selectable via bundled `fonts/` picker.*

REPL_USAGE.md

@@ -0,0 +1,116 @@
+# REPL Usage Guide
+
+The REPL (Read-Eval-Print Loop) effect provides an interactive command-line interface for controlling Mainline's pipeline in real-time.
+
+## How to Access the REPL
+
+### Method 1: Using CLI Arguments (Recommended)
+
+Run Mainline with the `repl` effect added to the effects list:
+
+```bash
+# With empty source (for testing)
+python mainline.py --pipeline-source empty --pipeline-effects repl
+
+# With headlines source (requires network)
+python mainline.py --pipeline-source headlines --pipeline-effects repl
+
+# With poetry source
+python mainline.py --pipeline-source poetry --pipeline-effects repl
+```
+
+### Method 2: Using a Preset
+
+Add a preset to your `~/.config/mainline/presets.toml` or `./presets.toml`:
+
+```toml
+[presets.repl]
+description = "Interactive REPL control"
+source = "headlines"
+display = "terminal"
+effects = ["repl"]
+viewport_width = 80
+viewport_height = 24
+```
+
+Then run:
+```bash
+python mainline.py --preset repl
+```
+
+### Method 3: Using Graph Config
+
+Create a TOML file (e.g., `repl_config.toml`):
+
+```toml
+source = "empty"
+display = "terminal"
+effects = ["repl"]
+```
+
+Then run:
+```bash
+python mainline.py --graph-config repl_config.toml
+```
+
+## REPL Commands
+
+Once the REPL is active, you can type commands:
+
+- **help** - Show available commands
+- **status** - Show pipeline status and metrics
+- **effects** - List all effects in the pipeline
+- **effect \<name\> \<on|off\>** - Toggle an effect
+- **param \<effect\> \<param\> \<value\>** - Set effect parameter
+- **pipeline** - Show current pipeline order
+- **clear** - Clear output buffer
+- **quit/exit** - Show exit message (use Ctrl+C to actually exit)
+
+## Keyboard Controls
+
+- **Enter** - Execute command
+- **Up/Down arrows** - Navigate command history
+- **Backspace** - Delete last character
+- **Ctrl+C** - Exit Mainline
+
+## Visual Features
+
+The REPL displays:
+- **HUD header** (top 3 lines): Shows FPS, frame time, command count, and output buffer size
+- **Content area**: Main content from the data source
+- **Separator line**: Visual divider
+- **REPL area**: Output buffer and input prompt
+
+## Example Session
+
+```
+MAINLINE REPL | FPS: 60.0 | 12.5ms
+COMMANDS: 3 | [2/3]
+OUTPUT: 5 lines
+────────────────────────────────────────
+Content from source appears here...
+More content...
+────────────────────────────────────────
+> help
+Available commands:
+  help - Show this help
+  status - Show pipeline status
+  effects - List all effects
+  effect <name> <on|off> - Toggle effect
+  param <effect> <param> <value> - Set parameter
+  pipeline - Show current pipeline order
+  clear - Clear output buffer
+  quit - Show exit message
+> effects
+Pipeline effects:
+  1. repl
+> effect repl off
+Effect 'repl' set to off
+```
+
+## Notes
+
+- The REPL effect needs a content source to overlay on (e.g., headlines, poetry, empty)
+- The REPL uses terminal display with raw input mode
+- Command history is preserved across sessions (up to 50 commands)
+- Pipeline mutations (enabling/disabling effects) are handled automatically

TODO.md

@@ -0,0 +1,63 @@
+# Tasks
+
+## Documentation Updates
+- [x] Remove references to removed display backends (sixel, kitty) from all documentation
+- [x] Remove references to deprecated "both" display mode
+- [x] Update AGENTS.md to reflect current architecture and remove merge conflicts
+- [x] Update Agent Skills (.opencode/skills/) to match current codebase
+- [x] Update docs/ARCHITECTURE.md to remove SixelDisplay references
+- [x] Verify ModernGL backend is properly documented and registered
+- [ ] Update docs/PIPELINE.md to reflect Stage-based architecture (outdated legacy flowchart) [#41](https://git.notsosm.art/david/Mainline/issues/41)
+
+## Code & Features
+- [ ] Check if luminance implementation exists for shade/tint effects (see [#26](https://git.notsosm.art/david/Mainline/issues/26) related: need to verify render/blocks.py has luminance calculation)
+- [x] Add entropy/chaos score metadata to effects for auto-categorization and intensity control [#32](https://git.notsosm.art/david/Mainline/issues/32) (closed - completed)
+- [ ] Finish ModernGL display backend: integrate window system, implement glyph caching, add event handling, and support border modes [#42](https://git.notsosm.art/david/Mainline/issues/42)
+- [x] Integrate UIPanel with pipeline: register stages, link parameter schemas, handle events, implement hot-reload.
+- [x] Move cached fixture headlines to engine/fixtures/headlines.json and update default source to use fixture.
+- [x] Add interactive UI panel for pipeline configuration (right-side panel) with stage toggles and param sliders.
+- [x] Enumerate all effect plugin parameters automatically for UI control (intensity, decay, etc.)
+- [ ] Implement pipeline hot-rebuild when stage toggles or params change, preserving camera and display state [#43](https://git.notsosm.art/david/Mainline/issues/43)
+
+## Test Suite Cleanup & Feature Implementation
+### Phase 1: Test Suite Cleanup (In Progress)
+- [x] Port figment feature to modern pipeline architecture
+- [x] Create `engine/effects/plugins/figment.py` (full port)
+- [x] Add `figment.py` to `engine/effects/plugins/`
+- [x] Copy SVG files to `figments/` directory
+- [x] Update `pyproject.toml` with figment extra
+- [x] Add `test-figment` preset to `presets.toml`
+- [x] Update pipeline adapters for overlay effects
+- [x] Clean up `test_adapters.py` (removed 18 mock-only tests)
+- [x] Verify all tests pass (652 passing, 20 skipped, 58% coverage)
+- [ ] Review remaining mock-heavy tests in `test_pipeline.py`
+- [ ] Review `test_effects.py` for implementation detail tests
+- [ ] Identify additional tests to remove/consolidate
+- [ ] Target: ~600 tests total
+
+### Phase 2: Acceptance Test Expansion (Planned)
+- [ ] Create `test_message_overlay.py` for message rendering
+- [ ] Create `test_firehose.py` for firehose rendering
+- [ ] Create `test_pipeline_order.py` for execution order verification
+- [ ] Expand `test_figment_effect.py` for animation phases
+- [ ] Target: 10-15 new acceptance tests
+
+### Phase 3: Post-Branch Features (Planned)
+- [ ] Port message overlay system from `upstream_layers.py`
+- [ ] Port firehose rendering from `upstream_layers.py`
+- [ ] Create `MessageOverlayStage` for pipeline integration
+- [ ] Verify figment renders in correct order (effects → figment → messages → display)
+
+### Phase 4: Visual Quality Improvements (Planned)
+- [ ] Compare upstream vs current pipeline output
+- [ ] Implement easing functions for figment animations
+- [ ] Add animated gradient shifts
+- [ ] Improve strobe effect patterns
+- [ ] Use introspection to match visual style
+
+## Gitea Issues Tracking
+- [#37](https://git.notsosm.art/david/Mainline/issues/37): Refactor app.py and adapter.py for better maintainability
+- [#35](https://git.notsosm.art/david/Mainline/issues/35): Epic: Pipeline Mutation API for Stage Hot-Swapping
+- [#34](https://git.notsosm.art/david/Mainline/issues/34): Improve benchmarking system and performance tests
+- [#33](https://git.notsosm.art/david/Mainline/issues/33): Add web-based pipeline editor UI
+- [#26](https://git.notsosm.art/david/Mainline/issues/26): Add Streaming display backend

analysis/visual_output_comparison.md

@@ -0,0 +1,158 @@
+# Visual Output Comparison: Upstream/Main vs Sideline
+
+## Summary
+
+A comprehensive comparison of visual output between `upstream/main` and the sideline branch (`feature/capability-based-deps`) reveals fundamental architectural differences in how content is rendered and displayed.
+
+## Captured Outputs
+
+### Sideline (Pipeline Architecture)
+- **File**: `output/sideline_demo.json`
+- **Format**: Plain text lines without ANSI cursor positioning
+- **Content**: Readable headlines with gradient colors applied
+
+### Upstream/Main (Monolithic Architecture)
+- **File**: `output/upstream_demo.json`
+- **Format**: Lines with explicit ANSI cursor positioning codes
+- **Content**: Cursor positioning codes + block characters + ANSI colors
+
+## Key Architectural Differences
+
+### 1. Buffer Content Structure
+
+**Sideline Pipeline:**
+```python
+# Each line is plain text with ANSI colors
+buffer = [
+    "The Download: OpenAI is building...",
+    "OpenAI is throwing everything...",
+    # ... more lines
+]
+```
+
+**Upstream Monolithic:**
+```python
+# Each line includes cursor positioning
+buffer = [
+    "\033[10;1H       \033[2;38;5;238mﾕ\033[0m \033[2;38;5;37mﾓ\033[0m ...",
+    "\033[11;1H\033[K",  # Clear line 11
+    # ... more lines with positioning
+]
+```
+
+### 2. Rendering Approach
+
+**Sideline (Pipeline Architecture):**
+- Stages produce plain text buffers
+- Display backend handles cursor positioning
+- `TerminalDisplay.show()` prepends `\033[H\033[J` (home + clear)
+- Lines are appended sequentially
+
+**Upstream (Monolithic Architecture):**
+- `render_ticker_zone()` produces buffers with explicit positioning
+- Each line includes `\033[{row};1H` to position cursor
+- Display backend writes buffer directly to stdout
+- Lines are positioned explicitly in the buffer
+
+### 3. Content Rendering
+
+**Sideline:**
+- Headlines rendered as plain text
+- Gradient colors applied via ANSI codes
+- Ticker effect via camera/viewport filtering
+
+**Upstream:**
+- Headlines rendered as block characters (▀,  ▄, █, etc.)
+- Japanese katakana glyphs used for glitch effect
+- Explicit row positioning for each line
+
+## Visual Output Analysis
+
+### Sideline Frame 0 (First 5 lines):
+```
+Line 0: 'The Download: OpenAI is building a fully automated researcher...'
+Line 1: 'OpenAI is throwing everything into building a fully automated...'
+Line 2: 'Mind-altering substances are (still) falling short in clinical...'
+Line 3: 'The Download: Quantum computing for health...'
+Line 4: 'Can quantum computers now solve health care problems...'
+```
+
+### Upstream Frame 0 (First 5 lines):
+```
+Line 0: ''
+Line 1: '\x1b[2;1H\x1b[K'
+Line 2: '\x1b[3;1H\x1b[K'
+Line 3: '\x1b[4;1H\x1b[2;38;5;238m \x1b[0m  \x1b[2;38;5;238mﾘ\x1b[0m ...'
+Line 4: '\x1b[5;1H\x1b[K'
+```
+
+## Implications for Visual Comparison
+
+### Challenges with Direct Comparison
+1. **Different buffer formats**: Plain text vs. positioned ANSI codes
+2. **Different rendering pipelines**: Pipeline stages vs. monolithic functions
+3. **Different content generation**: Headlines vs. block characters
+
+### Approaches for Visual Verification
+
+#### Option 1: Render and Compare Terminal Output
+- Run both branches with `TerminalDisplay`
+- Capture terminal output (not buffer)
+- Compare visual rendering
+- **Challenge**: Requires actual terminal rendering
+
+#### Option 2: Normalize Buffers for Comparison
+- Convert upstream positioned buffers to plain text
+- Strip ANSI cursor positioning codes
+- Compare normalized content
+- **Challenge**: Loses positioning information
+
+#### Option 3: Functional Equivalence Testing
+- Verify features work the same way
+- Test message overlay rendering
+- Test effect application
+- **Challenge**: Doesn't verify exact visual match
+
+## Recommendations
+
+### For Exact Visual Match
+1. **Update sideline to match upstream architecture**:
+   - Change `MessageOverlayStage` to return positioned buffers
+   - Update terminal display to handle positioned buffers
+   - This requires significant refactoring
+
+2. **Accept architectural differences**:
+   - The sideline pipeline architecture is fundamentally different
+   - Visual differences are expected and acceptable
+   - Focus on functional equivalence
+
+### For Functional Verification
+1. **Test message overlay rendering**:
+   - Verify message appears in correct position
+   - Verify gradient colors are applied
+   - Verify metadata bar is displayed
+
+2. **Test effect rendering**:
+   - Verify glitch effect applies block characters
+   - Verify firehose effect renders correctly
+   - Verify figment effect integrates properly
+
+3. **Test pipeline execution**:
+   - Verify stage execution order
+   - Verify capability resolution
+   - Verify dependency injection
+
+## Conclusion
+
+The visual output comparison reveals that `sideline` and `upstream/main` use fundamentally different rendering architectures:
+
+- **Upstream**: Explicit cursor positioning in buffer, monolithic rendering
+- **Sideline**: Plain text buffer, display handles positioning, pipeline rendering
+
+These differences are **architectural**, not bugs. The sideline branch has successfully adapted the upstream features to a new pipeline architecture.
+
+### Next Steps
+1. ✅ Document architectural differences (this file)
+2. ⏳ Create functional tests for visual verification
+3. ⏳ Update Gitea issue #50 with findings
+4. ⏳ Consider whether to adapt sideline to match upstream rendering style

client/editor.html

@@ -0,0 +1,313 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Mainline Pipeline Editor</title>
+    <style>
+        * { margin: 0; padding: 0; box-sizing: border-box; }
+        body {
+            font-family: 'Fira Code', 'Cascadia Code', 'Consolas', monospace;
+            background: #1a1a1a;
+            color: #eee;
+            display: flex;
+            min-height: 100vh;
+        }
+        #sidebar {
+            width: 300px;
+            background: #222;
+            padding: 15px;
+            border-right: 1px solid #333;
+            overflow-y: auto;
+        }
+        #main {
+            flex: 1;
+            padding: 20px;
+            overflow-y: auto;
+        }
+        h2 {
+            font-size: 14px;
+            color: #888;
+            margin-bottom: 10px;
+            text-transform: uppercase;
+        }
+        .section {
+            margin-bottom: 20px;
+        }
+        .stage-list {
+            list-style: none;
+        }
+        .stage-item {
+            display: flex;
+            align-items: center;
+            padding: 6px 8px;
+            background: #333;
+            margin-bottom: 2px;
+            cursor: pointer;
+            border-radius: 4px;
+        }
+        .stage-item:hover { background: #444; }
+        .stage-item.selected { background: #0066cc; }
+        .stage-item input[type="checkbox"] {
+            margin-right: 8px;
+            transform: scale(1.2);
+        }
+        .stage-name {
+            flex: 1;
+            font-size: 13px;
+        }
+        .param-group {
+            background: #2a2a2a;
+            padding: 10px;
+            border-radius: 4px;
+        }
+        .param-row {
+            display: flex;
+            align-items: center;
+            margin-bottom: 8px;
+            font-size: 12px;
+        }
+        .param-name {
+            width: 100px;
+            color: #aaa;
+        }
+        .param-slider {
+            flex: 1;
+            margin: 0 10px;
+        }
+        .param-value {
+            width: 50px;
+            text-align: right;
+            color: #4f4;
+        }
+        .preset-list {
+            display: flex;
+            flex-wrap: wrap;
+            gap: 6px;
+        }
+        .preset-btn {
+            background: #333;
+            border: 1px solid #444;
+            color: #ccc;
+            padding: 4px 10px;
+            border-radius: 3px;
+            cursor: pointer;
+            font-size: 11px;
+        }
+        .preset-btn:hover { background: #444; }
+        .preset-btn.active { background: #0066cc; border-color: #0077ff; color: #fff; }
+        button.action-btn {
+            background: #0066cc;
+            border: none;
+            color: white;
+            padding: 8px 12px;
+            border-radius: 4px;
+            cursor: pointer;
+            font-size: 12px;
+            margin-right: 5px;
+            margin-bottom: 5px;
+        }
+        button.action-btn:hover { background: #0077ee; }
+        #status {
+            position: fixed;
+            bottom: 10px;
+            left: 10px;
+            font-size: 11px;
+            color: #666;
+        }
+        #status.connected { color: #4f4; }
+        #status.disconnected { color: #f44; }
+        #pipeline-view {
+            margin-top: 10px;
+        }
+        .pipeline-node {
+            display: inline-block;
+            padding: 4px 8px;
+            margin: 2px;
+            background: #333;
+            border-radius: 3px;
+            font-size: 11px;
+        }
+        .pipeline-node.enabled { border-left: 3px solid #4f4; }
+        .pipeline-node.disabled { border-left: 3px solid #666; opacity: 0.6; }
+    </style>
+</head>
+<body>
+    <div id="sidebar">
+        <div class="section">
+            <h2>Preset</h2>
+            <div id="preset-list" class="preset-list"></div>
+        </div>
+        <div class="section">
+            <h2>Stages</h2>
+            <ul id="stage-list" class="stage-list"></ul>
+        </div>
+        <div class="section">
+            <h2>Parameters</h2>
+            <div id="param-editor" class="param-group"></div>
+        </div>
+    </div>
+    <div id="main">
+        <h2>Pipeline</h2>
+        <div id="pipeline-view"></div>
+        <div style="margin-top: 20px;">
+            <button class="action-btn" onclick="sendCommand({action: 'cycle_preset', direction: 1})">Next Preset</button>
+            <button class="action-btn" onclick="sendCommand({action: 'cycle_preset', direction: -1})">Prev Preset</button>
+        </div>
+    </div>
+    <div id="status">Disconnected</div>
+
+    <script>
+        const ws = new WebSocket(`ws://${location.hostname}:8765`);
+        let state = { stages: {}, preset: '', presets: [], selected_stage: null };
+
+        function updateStatus(connected) {
+            const status = document.getElementById('status');
+            status.textContent = connected ? 'Connected' : 'Disconnected';
+            status.className = connected ? 'connected' : 'disconnected';
+        }
+
+        function connect() {
+            ws.onopen = () => {
+                updateStatus(true);
+                // Request initial state
+                ws.send(JSON.stringify({ type: 'state_request' }));
+            };
+            ws.onclose = () => {
+                updateStatus(false);
+                setTimeout(connect, 2000);
+            };
+            ws.onerror = () => {
+                updateStatus(false);
+            };
+            ws.onmessage = (event) => {
+                try {
+                    const data = JSON.parse(event.data);
+                    if (data.type === 'state') {
+                        state = data.state;
+                        render();
+                    }
+                } catch (e) {
+                    console.error('Parse error:', e);
+                }
+            };
+        }
+
+        function sendCommand(command) {
+            ws.send(JSON.stringify({ type: 'command', command }));
+        }
+
+        function render() {
+            renderPresets();
+            renderStageList();
+            renderPipeline();
+            renderParams();
+        }
+
+        function renderPresets() {
+            const container = document.getElementById('preset-list');
+            container.innerHTML = '';
+            (state.presets || []).forEach(preset => {
+                const btn = document.createElement('button');
+                btn.className = 'preset-btn' + (preset === state.preset ? ' active' : '');
+                btn.textContent = preset;
+                btn.onclick = () => sendCommand({ action: 'change_preset', preset });
+                container.appendChild(btn);
+            });
+        }
+
+        function renderStageList() {
+            const list = document.getElementById('stage-list');
+            list.innerHTML = '';
+            Object.entries(state.stages || {}).forEach(([name, info]) => {
+                const li = document.createElement('li');
+                li.className = 'stage-item' + (name === state.selected_stage ? ' selected' : '');
+                li.innerHTML = `
+                    <input type="checkbox" ${info.enabled ? 'checked' : ''} 
+                           onchange="sendCommand({action: 'toggle_stage', stage: '${name}'})">
+                    <span class="stage-name">${name}</span>
+                `;
+                li.onclick = (e) => {
+                    if (e.target.type !== 'checkbox') {
+                        sendCommand({ action: 'select_stage', stage: name });
+                    }
+                };
+                list.appendChild(li);
+            });
+        }
+
+        function renderPipeline() {
+            const view = document.getElementById('pipeline-view');
+            view.innerHTML = '';
+            const stages = Object.entries(state.stages || {});
+            if (stages.length === 0) {
+                view.textContent = '(No stages)';
+                return;
+            }
+            stages.forEach(([name, info]) => {
+                const span = document.createElement('span');
+                span.className = 'pipeline-node' + (info.enabled ? ' enabled' : ' disabled');
+                span.textContent = name;
+                view.appendChild(span);
+            });
+        }
+
+        function renderParams() {
+            const container = document.getElementById('param-editor');
+            container.innerHTML = '';
+            const selected = state.selected_stage;
+            if (!selected || !state.stages[selected]) {
+                container.innerHTML = '<div style="color:#666;font-size:11px;">(select a stage)</div>';
+                return;
+            }
+            const stage = state.stages[selected];
+            if (!stage.params || Object.keys(stage.params).length === 0) {
+                container.innerHTML = '<div style="color:#666;font-size:11px;">(no params)</div>';
+                return;
+            }
+            Object.entries(stage.params).forEach(([key, value]) => {
+                const row = document.createElement('div');
+                row.className = 'param-row';
+                // Infer min/max/step from typical ranges
+                let min = 0, max = 1, step = 0.1;
+                if (typeof value === 'number') {
+                    if (value > 1) { max = value * 2; step = 1; }
+                    else { max = 1; step = 0.1; }
+                }
+                row.innerHTML = `
+                    <div class="param-name">${key}</div>
+                    <input type="range" class="param-slider" min="${min}" max="${max}" step="${step}" 
+                           value="${value}"
+                           oninput="adjustParam('${key}', this.value)">
+                    <div class="param-value">${typeof value === 'number' ? Number(value).toFixed(2) : value}</div>
+                `;
+                container.appendChild(row);
+            });
+        }
+
+        function adjustParam(param, newValue) {
+            const selected = state.selected_stage;
+            if (!selected) return;
+            // Update display immediately for responsiveness
+            const num = parseFloat(newValue);
+            if (!isNaN(num)) {
+                // Show updated value
+                document.querySelectorAll('.param-value').forEach(el => {
+                    if (el.parentElement.querySelector('.param-name').textContent === param) {
+                        el.textContent = num.toFixed(2);
+                    }
+                });
+            }
+            // Send command
+            sendCommand({
+                action: 'adjust_param',
+                stage: selected,
+                param: param,
+                delta: num - (state.stages[selected].params[param] || 0)
+            });
+        }
+
+        connect();
+    </script>
+</body>
+</html>

client/index.html

@@ -0,0 +1,369 @@
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>Mainline Terminal</title>
+    <style>
+        * {
+            margin: 0;
+            padding: 0;
+            box-sizing: border-box;
+        }
+        body {
+            background: #0a0a0a;
+            color: #ccc;
+            font-family: 'Fira Code', 'Cascadia Code', 'Consolas', monospace;
+            display: flex;
+            flex-direction: column;
+            align-items: center;
+            justify-content: center;
+            min-height: 100vh;
+            padding: 20px;
+        }
+        body.fullscreen {
+            padding: 0;
+        }
+        body.fullscreen #controls {
+            display: none;
+        }
+        #container {
+            position: relative;
+        }
+        canvas {
+            background: #000;
+            border: 1px solid #333;
+            image-rendering: pixelated;
+            image-rendering: crisp-edges;
+        }
+        body.fullscreen canvas {
+            border: none;
+            width: 100vw;
+            height: 100vh;
+            max-width: 100vw;
+            max-height: 100vh;
+        }
+        #controls {
+            display: flex;
+            gap: 10px;
+            margin-top: 10px;
+            align-items: center;
+        }
+        #controls button {
+            background: #333;
+            color: #ccc;
+            border: 1px solid #555;
+            padding: 5px 12px;
+            cursor: pointer;
+            font-family: inherit;
+            font-size: 12px;
+        }
+        #controls button:hover {
+            background: #444;
+        }
+        #controls input {
+            width: 60px;
+            background: #222;
+            color: #ccc;
+            border: 1px solid #444;
+            padding: 4px 8px;
+            font-family: inherit;
+            text-align: center;
+        }
+        #status {
+            margin-top: 10px;
+            font-size: 12px;
+            color: #666;
+        }
+        #status.connected {
+            color: #4f4;
+        }
+        #status.disconnected {
+            color: #f44;
+        }
+    </style>
+</head>
+<body>
+    <div id="container">
+        <canvas id="terminal"></canvas>
+    </div>
+    <div id="controls">
+        <label>Cols: <input type="number" id="cols" value="80" min="20" max="200"></label>
+        <label>Rows: <input type="number" id="rows" value="24" min="10" max="60"></label>
+        <button id="apply">Apply</button>
+        <button id="fullscreen">Fullscreen</button>
+    </div>
+    <div id="status" class="disconnected">Connecting...</div>
+
+    <script>
+        const canvas = document.getElementById('terminal');
+        const ctx = canvas.getContext('2d');
+        const status = document.getElementById('status');
+        const colsInput = document.getElementById('cols');
+        const rowsInput = document.getElementById('rows');
+        const applyBtn = document.getElementById('apply');
+        const fullscreenBtn = document.getElementById('fullscreen');
+
+        const CHAR_WIDTH = 9;
+        const CHAR_HEIGHT = 16;
+
+        const ANSI_COLORS = {
+            0: '#000000', 1: '#cd3131', 2: '#0dbc79', 3: '#e5e510',
+            4: '#2472c8', 5: '#bc3fbc', 6: '#11a8cd', 7: '#e5e5e5',
+            8: '#666666', 9: '#f14c4c', 10: '#23d18b', 11: '#f5f543',
+            12: '#3b8eea', 13: '#d670d6', 14: '#29b8db', 15: '#ffffff',
+        };
+
+        let cols = 80;
+        let rows = 24;
+        let ws = null;
+
+        function resizeCanvas() {
+            canvas.width = cols * CHAR_WIDTH;
+            canvas.height = rows * CHAR_HEIGHT;
+        }
+
+        function parseAnsi(text) {
+            if (!text) return [];
+            
+            const tokens = [];
+            let currentText = '';
+            let fg = '#cccccc';
+            let bg = '#000000';
+            let bold = false;
+            let i = 0;
+            let inEscape = false;
+            let escapeCode = '';
+
+            while (i < text.length) {
+                const char = text[i];
+                
+                if (inEscape) {
+                    if (char >= '0' && char <= '9' || char === ';' || char === '[') {
+                        escapeCode += char;
+                    }
+                    
+                    if (char === 'm') {
+                        const codes = escapeCode.replace('\x1b[', '').split(';');
+                        
+                        for (const code of codes) {
+                            const num = parseInt(code) || 0;
+                            
+                            if (num === 0) {
+                                fg = '#cccccc';
+                                bg = '#000000';
+                                bold = false;
+                            } else if (num === 1) {
+                                bold = true;
+                            } else if (num === 22) {
+                                bold = false;
+                            } else if (num === 39) {
+                                fg = '#cccccc';
+                            } else if (num === 49) {
+                                bg = '#000000';
+                            } else if (num >= 30 && num <= 37) {
+                                fg = ANSI_COLORS[num - 30 + (bold ? 8 : 0)] || '#cccccc';
+                            } else if (num >= 40 && num <= 47) {
+                                bg = ANSI_COLORS[num - 40] || '#000000';
+                            } else if (num >= 90 && num <= 97) {
+                                fg = ANSI_COLORS[num - 90 + 8] || '#cccccc';
+                            } else if (num >= 100 && num <= 107) {
+                                bg = ANSI_COLORS[num - 100 + 8] || '#000000';
+                            } else if (num >= 1 && num <= 256) {
+                                // 256 colors
+                                if (num < 16) {
+                                    fg = ANSI_COLORS[num] || '#cccccc';
+                                } else if (num < 232) {
+                                    const c = num - 16;
+                                    const r = Math.floor(c / 36) * 51;
+                                    const g = Math.floor((c % 36) / 6) * 51;
+                                    const b = (c % 6) * 51;
+                                    fg = `#${r.toString(16).padStart(2,'0')}${g.toString(16).padStart(2,'0')}${b.toString(16).padStart(2,'0')}`;
+                                } else {
+                                    const gray = (num - 232) * 10 + 8;
+                                    fg = `#${gray.toString(16).repeat(2)}`;
+                                }
+                            }
+                        }
+                        
+                        if (currentText) {
+                            tokens.push({ text: currentText, fg, bg, bold });
+                            currentText = '';
+                        }
+                        inEscape = false;
+                        escapeCode = '';
+                    }
+                } else if (char === '\x1b' && text[i + 1] === '[') {
+                    if (currentText) {
+                        tokens.push({ text: currentText, fg, bg, bold });
+                        currentText = '';
+                    }
+                    inEscape = true;
+                    escapeCode = '';
+                    i++;
+                } else {
+                    currentText += char;
+                }
+                i++;
+            }
+
+            if (currentText) {
+                tokens.push({ text: currentText, fg, bg, bold });
+            }
+
+            return tokens;
+        }
+
+        function renderLine(text, x, y, lineHeight) {
+            const tokens = parseAnsi(text);
+            let xOffset = x;
+
+            for (const token of tokens) {
+                if (token.text) {
+                    if (token.bold) {
+                        ctx.font = 'bold 16px monospace';
+                    } else {
+                        ctx.font = '16px monospace';
+                    }
+
+                    const metrics = ctx.measureText(token.text);
+                    
+                    if (token.bg !== '#000000') {
+                        ctx.fillStyle = token.bg;
+                        ctx.fillRect(xOffset, y - 2, metrics.width + 1, lineHeight);
+                    }
+                    
+                    ctx.fillStyle = token.fg;
+                    ctx.fillText(token.text, xOffset, y);
+                    xOffset += metrics.width;
+                }
+            }
+        }
+
+        function connect() {
+            const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
+            const wsUrl = `${protocol}//${window.location.hostname}:8765`;
+            
+            ws = new WebSocket(wsUrl);
+
+            ws.onopen = () => {
+                status.textContent = 'Connected';
+                status.className = 'connected';
+                sendSize();
+            };
+
+            ws.onclose = () => {
+                status.textContent = 'Disconnected - Reconnecting...';
+                status.className = 'disconnected';
+                setTimeout(connect, 1000);
+            };
+
+            ws.onerror = () => {
+                status.textContent = 'Connection error';
+                status.className = 'disconnected';
+            };
+
+            ws.onmessage = (event) => {
+                try {
+                    const data = JSON.parse(event.data);
+                    
+                    if (data.type === 'frame') {
+                        cols = data.width || 80;
+                        rows = data.height || 24;
+                        colsInput.value = cols;
+                        rowsInput.value = rows;
+                        resizeCanvas();
+                        render(data.lines || []);
+                    } else if (data.type === 'clear') {
+                        ctx.fillStyle = '#000';
+                        ctx.fillRect(0, 0, canvas.width, canvas.height);
+                    } else if (data.type === 'state') {
+                        // Log state updates for debugging (can be extended for UI)
+                        console.log('State update:', data.state);
+                    }
+                } catch (e) {
+                    console.error('Failed to parse message:', e);
+                }
+            };
+        }
+
+        function sendSize() {
+            if (ws && ws.readyState === WebSocket.OPEN) {
+                ws.send(JSON.stringify({
+                    type: 'resize',
+                    width: parseInt(colsInput.value),
+                    height: parseInt(rowsInput.value)
+                }));
+            }
+        }
+
+        function render(lines) {
+            ctx.fillStyle = '#000';
+            ctx.fillRect(0, 0, canvas.width, canvas.height);
+
+            ctx.font = '16px monospace';
+            ctx.textBaseline = 'top';
+
+            const lineHeight = CHAR_HEIGHT;
+            const maxLines = Math.min(lines.length, rows);
+
+            for (let i = 0; i < maxLines; i++) {
+                const line = lines[i] || '';
+                renderLine(line, 0, i * lineHeight, lineHeight);
+            }
+        }
+
+        function calculateViewportSize() {
+            const isFullscreen = document.fullscreenElement !== null;
+            const padding = isFullscreen ? 0 : 40;
+            const controlsHeight = isFullscreen ? 0 : 60;
+            const availableWidth = window.innerWidth - padding;
+            const availableHeight = window.innerHeight - controlsHeight;
+            cols = Math.max(20, Math.floor(availableWidth / CHAR_WIDTH));
+            rows = Math.max(10, Math.floor(availableHeight / CHAR_HEIGHT));
+            colsInput.value = cols;
+            rowsInput.value = rows;
+            resizeCanvas();
+            console.log('Fullscreen:', isFullscreen, 'Size:', cols, 'x', rows);
+            sendSize();
+        }
+
+        applyBtn.addEventListener('click', () => {
+            cols = parseInt(colsInput.value);
+            rows = parseInt(rowsInput.value);
+            resizeCanvas();
+            sendSize();
+        });
+
+        fullscreenBtn.addEventListener('click', () => {
+            if (!document.fullscreenElement) {
+                document.body.classList.add('fullscreen');
+                document.documentElement.requestFullscreen().then(() => {
+                    calculateViewportSize();
+                });
+            } else {
+                document.exitFullscreen().then(() => {
+                    calculateViewportSize();
+                });
+            }
+        });
+
+        document.addEventListener('fullscreenchange', () => {
+            if (!document.fullscreenElement) {
+                document.body.classList.remove('fullscreen');
+                calculateViewportSize();
+            }
+        });
+
+        window.addEventListener('resize', () => {
+            if (document.fullscreenElement) {
+                calculateViewportSize();
+            }
+        });
+
+        // Initial setup
+        resizeCanvas();
+        connect();
+    </script>
+</body>
+</html>

cmdline.py

@@ -0,0 +1,256 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+"""
+Command-line utility for interacting with mainline via ntfy.
+
+Usage:
+    python cmdline.py                     # Interactive TUI mode
+    python cmdline.py --help              # Show help
+    python cmdline.py /effects list       # Send single command via ntfy
+    python cmdline.py /effects stats      # Get performance stats via ntfy
+    python cmdline.py -w /effects stats   # Watch mode (polls for stats)
+
+The TUI mode provides:
+    - Arrow keys to navigate command history
+    - Tab completion for commands
+    - Auto-refresh for performance stats
+
+C&C works like a serial port:
+    1. Send command to ntfy_cc_topic
+    2. Mainline receives, processes, responds to same topic
+    3. Cmdline polls for response
+"""
+
+import os
+
+os.environ["FORCE_COLOR"] = "1"
+os.environ["TERM"] = "xterm-256color"
+
+import argparse
+import json
+import sys
+import time
+import threading
+import urllib.request
+from pathlib import Path
+
+from engine import config
+from engine.terminal import CLR, CURSOR_OFF, CURSOR_ON, G_DIM, G_HI, RST, W_GHOST
+
+try:
+    CC_CMD_TOPIC = config.NTFY_CC_CMD_TOPIC
+    CC_RESP_TOPIC = config.NTFY_CC_RESP_TOPIC
+except AttributeError:
+    CC_CMD_TOPIC = "https://ntfy.sh/klubhaus_terminal_mainline_cc_cmd/json"
+    CC_RESP_TOPIC = "https://ntfy.sh/klubhaus_terminal_mainline_cc_resp/json"
+
+
+class NtfyResponsePoller:
+    """Polls ntfy for command responses."""
+
+    def __init__(self, cmd_topic: str, resp_topic: str, timeout: float = 10.0):
+        self.cmd_topic = cmd_topic
+        self.resp_topic = resp_topic
+        self.timeout = timeout
+        self._last_id = None
+        self._lock = threading.Lock()
+
+    def _build_url(self) -> str:
+        from urllib.parse import parse_qs, urlencode, urlparse, urlunparse
+
+        parsed = urlparse(self.resp_topic)
+        params = parse_qs(parsed.query, keep_blank_values=True)
+        params["since"] = [self._last_id if self._last_id else "20s"]
+        new_query = urlencode({k: v[0] for k, v in params.items()})
+        return urlunparse(parsed._replace(query=new_query))
+
+    def send_and_wait(self, cmd: str) -> str:
+        """Send command and wait for response."""
+        url = self.cmd_topic.replace("/json", "")
+        data = cmd.encode("utf-8")
+
+        req = urllib.request.Request(
+            url,
+            data=data,
+            headers={
+                "User-Agent": "mainline-cmdline/0.1",
+                "Content-Type": "text/plain",
+            },
+            method="POST",
+        )
+
+        try:
+            urllib.request.urlopen(req, timeout=5)
+        except Exception as e:
+            return f"Error sending command: {e}"
+
+        return self._wait_for_response(cmd)
+
+    def _wait_for_response(self, expected_cmd: str = "") -> str:
+        """Poll for response message."""
+        start = time.time()
+        while time.time() - start < self.timeout:
+            try:
+                url = self._build_url()
+                req = urllib.request.Request(
+                    url, headers={"User-Agent": "mainline-cmdline/0.1"}
+                )
+                with urllib.request.urlopen(req, timeout=10) as resp:
+                    for line in resp:
+                        try:
+                            data = json.loads(line.decode("utf-8", errors="replace"))
+                        except json.JSONDecodeError:
+                            continue
+                        if data.get("event") == "message":
+                            self._last_id = data.get("id")
+                            msg = data.get("message", "")
+                            if msg:
+                                return msg
+            except Exception:
+                pass
+            time.sleep(0.5)
+        return "Timeout waiting for response"
+
+
+AVAILABLE_COMMANDS = """Available commands:
+  /effects list          - List all effects and status
+  /effects <name> on    - Enable an effect
+  /effects <name> off   - Disable an effect
+  /effects <name> intensity <0.0-1.0> - Set effect intensity
+  /effects reorder <name1>,<name2>,... - Reorder pipeline
+  /effects stats        - Show performance statistics
+  /help                 - Show this help
+  /quit                 - Exit
+"""
+
+
+def print_header():
+    w = 60
+    print(CLR, end="")
+    print(CURSOR_OFF, end="")
+    print(f"\033[1;1H", end="")
+    print(f"  \033[1;38;5;231m╔{'═' * (w - 6)}╗\033[0m")
+    print(
+        f"  \033[1;38;5;231m║\033[0m \033[1;38;5;82mMAINLINE\033[0m \033[3;38;5;245mCommand Center\033[0m \033[1;38;5;231m  ║\033[0m"
+    )
+    print(f"  \033[1;38;5;231m╚{'═' * (w - 6)}╝\033[0m")
+    print(f"  \033[2;38;5;37mCMD: {CC_CMD_TOPIC.split('/')[-2]}\033[0m")
+    print(f"  \033[2;38;5;37mRESP: {CC_RESP_TOPIC.split('/')[-2]}\033[0m")
+    print()
+
+
+def print_response(response: str, is_error: bool = False) -> None:
+    """Print response with nice formatting."""
+    print()
+    if is_error:
+        print(f"  \033[1;38;5;196m✗ Error\033[0m")
+        print(f"  \033[38;5;196m{'─' * 40}\033[0m")
+    else:
+        print(f"  \033[1;38;5;82m✓ Response\033[0m")
+        print(f"  \033[38;5;37m{'─' * 40}\033[0m")
+
+    for line in response.split("\n"):
+        print(f"  {line}")
+    print()
+
+
+def interactive_mode():
+    """Interactive TUI for sending commands."""
+    import readline
+
+    print_header()
+    poller = NtfyResponsePoller(CC_CMD_TOPIC, CC_RESP_TOPIC)
+
+    print(f"  \033[38;5;245mType /help for commands, /quit to exit\033[0m")
+    print()
+
+    while True:
+        try:
+            cmd = input(f"  \033[1;38;5;82m❯\033[0m {G_HI}").strip()
+        except (EOFError, KeyboardInterrupt):
+            print()
+            break
+
+        if not cmd:
+            continue
+
+        if cmd.startswith("/"):
+            if cmd == "/quit" or cmd == "/exit":
+                print(f"\n  \033[1;38;5;245mGoodbye!{RST}\n")
+                break
+
+            if cmd == "/help":
+                print(f"\n{AVAILABLE_COMMANDS}\n")
+                continue
+
+            print(f"  \033[38;5;245m⟳ Sending to mainline...{RST}")
+            result = poller.send_and_wait(cmd)
+            print_response(result, is_error=result.startswith("Error"))
+        else:
+            print(f"\n  \033[1;38;5;196m⚠ Commands must start with /{RST}\n")
+
+    print(CURSOR_ON, end="")
+    return 0
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Mainline command-line interface",
+        formatter_class=argparse.RawDescriptionHelpFormatter,
+        epilog=AVAILABLE_COMMANDS,
+    )
+    parser.add_argument(
+        "command",
+        nargs="?",
+        default=None,
+        help="Command to send (e.g., /effects list)",
+    )
+    parser.add_argument(
+        "--watch",
+        "-w",
+        action="store_true",
+        help="Watch mode: continuously poll for stats (Ctrl+C to exit)",
+    )
+
+    args = parser.parse_args()
+
+    if args.command is None:
+        return interactive_mode()
+
+    poller = NtfyResponsePoller(CC_CMD_TOPIC, CC_RESP_TOPIC)
+
+    if args.watch and "/effects stats" in args.command:
+        import signal
+
+        def handle_sigterm(*_):
+            print(f"\n  \033[1;38;5;245mStopped watching{RST}")
+            print(CURSOR_ON, end="")
+            sys.exit(0)
+
+        signal.signal(signal.SIGTERM, handle_sigterm)
+
+        print_header()
+        print(f"  \033[38;5;245mWatching /effects stats (Ctrl+C to exit)...{RST}\n")
+        try:
+            while True:
+                result = poller.send_and_wait(args.command)
+                print(f"\033[2J\033[1;1H", end="")
+                print(
+                    f"  \033[1;38;5;82m❯\033[0m Performance Stats - \033[1;38;5;245m{time.strftime('%H:%M:%S')}{RST}"
+                )
+                print(f"  \033[38;5;37m{'─' * 44}{RST}")
+                for line in result.split("\n"):
+                    print(f"  {line}")
+                time.sleep(2)
+        except KeyboardInterrupt:
+            print(f"\n  \033[1;38;5;245mStopped watching{RST}")
+            return 0
+        return 0
+
+    result = poller.send_and_wait(args.command)
+    print(result)
+    return 0
+
+
+if __name__ == "__main__":
+    main()

completion/mainline-completion.bash

@@ -0,0 +1,106 @@
+# Mainline bash completion script
+#
+# To install:
+#   source /path/to/completion/mainline-completion.bash
+#
+# Or add to ~/.bashrc:
+#   source /path/to/completion/mainline-completion.bash
+
+_mainline_completion() {
+    local cur prev words cword
+    _init_completion || return
+
+    # Get current word and previous word
+    cur="${COMP_WORDS[COMP_CWORD]}"
+    prev="${COMP_WORDS[COMP_CWORD-1]}"
+
+    # Completion options based on previous word
+    case "${prev}" in
+        --display)
+            # Display backends
+            COMPREPLY=($(compgen -W "terminal null replay websocket pygame moderngl" -- "${cur}"))
+            return
+            ;;
+
+        --pipeline-source)
+            # Available sources
+            COMPREPLY=($(compgen -W "headlines poetry empty fixture pipeline-inspect" -- "${cur}"))
+            return
+            ;;
+
+        --pipeline-effects)
+            # Available effects (comma-separated)
+            local effects="afterimage border crop fade firehose glitch hud motionblur noise tint"
+            COMPREPLY=($(compgen -W "${effects}" -- "${cur}"))
+            return
+            ;;
+
+        --pipeline-camera)
+            # Camera modes
+            COMPREPLY=($(compgen -W "feed scroll horizontal omni floating bounce radial" -- "${cur}"))
+            return
+            ;;
+
+        --pipeline-border)
+            # Border modes
+            COMPREPLY=($(compgen -W "off simple ui" -- "${cur}"))
+            return
+            ;;
+
+        --pipeline-display)
+            # Display backends (same as --display)
+            COMPREPLY=($(compgen -W "terminal null replay websocket pygame moderngl" -- "${cur}"))
+            return
+            ;;
+
+        --theme)
+            # Theme colors
+            COMPREPLY=($(compgen -W "green orange purple blue red" -- "${cur}"))
+            return
+            ;;
+
+        --viewport)
+            # Viewport size suggestions
+            COMPREPLY=($(compgen -W "80x24 100x30 120x40 60x20" -- "${cur}"))
+            return
+            ;;
+
+        --preset)
+            # Presets (would need to query available presets)
+            COMPREPLY=($(compgen -W "demo demo-base demo-pygame demo-camera-showcase poetry headlines empty test-basic test-border test-scroll-camera" -- "${cur}"))
+            return
+            ;;
+
+        --positioning)
+            # Positioning modes
+            COMPREPLY=($(compgen -W "absolute relative mixed" -- "${cur}"))
+            return
+            ;;
+    esac
+
+    # Flag completion (start with --)
+    if [[ "${cur}" == -* ]]; then
+        COMPREPLY=($(compgen -W "
+            --display
+            --pipeline-source
+            --pipeline-effects
+            --pipeline-camera
+            --pipeline-display
+            --pipeline-ui
+            --pipeline-border
+            --viewport
+            --preset
+            --theme
+            --positioning
+            --websocket
+            --websocket-port
+            --allow-unsafe
+            --help
+        " -- "${cur}"))
+        return
+    fi
+}
+
+complete -F _mainline_completion mainline.py
+complete -F _mainline_completion python\ -m\ engine.app
+complete -F _mainline_completion python\ -m\ mainline

completion/mainline-completion.fish

@@ -0,0 +1,81 @@
+# Fish completion script for Mainline
+#
+# To install:
+#   source /path/to/completion/mainline-completion.fish
+#
+# Or copy to ~/.config/fish/completions/mainline.fish
+
+# Define display backends
+set -l display_backends terminal null replay websocket pygame moderngl
+
+# Define sources
+set -l sources headlines poetry empty fixture pipeline-inspect
+
+# Define effects
+set -l effects afterimage border crop fade firehose glitch hud motionblur noise tint
+
+# Define camera modes
+set -l cameras feed scroll horizontal omni floating bounce radial
+
+# Define border modes
+set -l borders off simple ui
+
+# Define themes
+set -l themes green orange purple blue red
+
+# Define presets
+set -l presets demo demo-base demo-pygame demo-camera-showcase poetry headlines empty test-basic test-border test-scroll-camera test-figment test-message-overlay
+
+# Main completion function
+function __mainline_complete
+    set -l cmd (commandline -po)
+    set -l token (commandline -t)
+
+    # Complete display backends
+    complete -c mainline.py -n '__fish_seen_argument --display' -a "$display_backends" -d 'Display backend'
+
+    # Complete sources
+    complete -c mainline.py -n '__fish_seen_argument --pipeline-source' -a "$sources" -d 'Data source'
+
+    # Complete effects
+    complete -c mainline.py -n '__fish_seen_argument --pipeline-effects' -a "$effects" -d 'Effect plugin'
+
+    # Complete camera modes
+    complete -c mainline.py -n '__fish_seen_argument --pipeline-camera' -a "$cameras" -d 'Camera mode'
+
+    # Complete display backends (pipeline)
+    complete -c mainline.py -n '__fish_seen_argument --pipeline-display' -a "$display_backends" -d 'Display backend'
+
+    # Complete border modes
+    complete -c mainline.py -n '__fish_seen_argument --pipeline-border' -a "$borders" -d 'Border mode'
+
+    # Complete themes
+    complete -c mainline.py -n '__fish_seen_argument --theme' -a "$themes" -d 'Color theme'
+
+    # Complete presets
+    complete -c mainline.py -n '__fish_seen_argument --preset' -a "$presets" -d 'Preset name'
+
+    # Complete viewport sizes
+    complete -c mainline.py -n '__fish_seen_argument --viewport' -a '80x24 100x30 120x40 60x20' -d 'Viewport size (WxH)'
+
+    # Complete flag options
+    complete -c mainline.py -n 'not __fish_seen_argument --display' -l display -d 'Display backend' -a "$display_backends"
+    complete -c mainline.py -n 'not __fish_seen_argument --preset' -l preset -d 'Preset to use' -a "$presets"
+    complete -c mainline.py -n 'not __fish_seen_argument --viewport' -l viewport -d 'Viewport size (WxH)' -a '80x24 100x30 120x40 60x20'
+    complete -c mainline.py -n 'not __fish_seen_argument --theme' -l theme -d 'Color theme' -a "$themes"
+    complete -c mainline.py -l websocket -d 'Enable WebSocket server'
+    complete -c mainline.py -n 'not __fish_seen_argument --websocket-port' -l websocket-port -d 'WebSocket port' -a '8765'
+    complete -c mainline.py -l allow-unsafe -d 'Allow unsafe pipeline configuration'
+    complete -c mainline.py -n 'not __fish_seen_argument --help' -l help -d 'Show help'
+
+    # Pipeline-specific flags
+    complete -c mainline.py -n 'not __fish_seen_argument --pipeline-source' -l pipeline-source -d 'Data source' -a "$sources"
+    complete -c mainline.py -n 'not __fish_seen_argument --pipeline-effects' -l pipeline-effects -d 'Effect plugins (comma-separated)' -a "$effects"
+    complete -c mainline.py -n 'not __fish_seen_argument --pipeline-camera' -l pipeline-camera -d 'Camera mode' -a "$cameras"
+    complete -c mainline.py -n 'not __fish_seen_argument --pipeline-display' -l pipeline-display -d 'Display backend' -a "$display_backends"
+    complete -c mainline.py -l pipeline-ui -d 'Enable UI panel'
+    complete -c mainline.py -n 'not __fish_seen_argument --pipeline-border' -l pipeline-border -d 'Border mode' -a "$borders"
+end
+
+# Register the completion function
+__mainline_complete

completion/mainline-completion.zsh

@@ -0,0 +1,48 @@
+#compdef mainline.py
+
+# Mainline zsh completion script
+#
+# To install:
+#   source /path/to/completion/mainline-completion.zsh
+#
+# Or add to ~/.zshrc:
+#   source /path/to/completion/mainline-completion.zsh
+
+# Define completion function
+_mainline() {
+    local -a commands
+    local curcontext="$curcontext" state line
+    typeset -A opt_args
+
+    _arguments -C \
+        '(-h --help)'{-h,--help}'[Show help]' \
+        '--display=[Display backend]:backend:(terminal null replay websocket pygame moderngl)' \
+        '--preset=[Preset to use]:preset:(demo demo-base demo-pygame demo-camera-showcase poetry headlines empty test-basic test-border test-scroll-camera test-figment test-message-overlay)' \
+        '--viewport=[Viewport size]:size:(80x24 100x30 120x40 60x20)' \
+        '--theme=[Color theme]:theme:(green orange purple blue red)' \
+        '--websocket[Enable WebSocket server]' \
+        '--websocket-port=[WebSocket port]:port:' \
+        '--allow-unsafe[Allow unsafe pipeline configuration]' \
+        '(-)*: :{_files}' \
+        && ret=0
+
+    # Handle --pipeline-* arguments
+    if [[ -n ${words[*]} ]]; then
+        _arguments -C \
+            '--pipeline-source=[Data source]:source:(headlines poetry empty fixture pipeline-inspect)' \
+            '--pipeline-effects=[Effect plugins]:effects:(afterimage border crop fade firehose glitch hud motionblur noise tint)' \
+            '--pipeline-camera=[Camera mode]:camera:(feed scroll horizontal omni floating bounce radial)' \
+            '--pipeline-display=[Display backend]:backend:(terminal null replay websocket pygame moderngl)' \
+            '--pipeline-ui[Enable UI panel]' \
+            '--pipeline-border=[Border mode]:mode:(off simple ui)' \
+            '--viewport=[Viewport size]:size:(80x24 100x30 120x40 60x20)' \
+            && ret=0
+    fi
+
+    return ret
+}
+
+# Register completion function
+compdef _mainline mainline.py
+compdef _mainline "python -m engine.app"
+compdef _mainline "python -m mainline"

docs/ARCHITECTURE.md

@@ -0,0 +1,153 @@
+# Mainline Architecture Diagrams
+
+> These diagrams use Mermaid. Render with: `npx @mermaid-js/mermaid-cli -i ARCHITECTURE.md` or view in GitHub/GitLab/Notion.
+
+## Class Hierarchy (Mermaid)
+
+```mermaid
+classDiagram
+    class Stage {
+        <<abstract>>
+        +str name
+        +set[str] capabilities
+        +set[str] dependencies
+        +process(data, ctx) Any
+    }
+    
+    Stage <|-- DataSourceStage
+    Stage <|-- CameraStage
+    Stage <|-- FontStage
+    Stage <|-- ViewportFilterStage
+    Stage <|-- EffectPluginStage
+    Stage <|-- DisplayStage
+    Stage <|-- SourceItemsToBufferStage
+    Stage <|-- PassthroughStage
+    Stage <|-- ImageToTextStage
+    Stage <|-- CanvasStage
+    
+    class EffectPlugin {
+        <<abstract>>
+        +str name
+        +EffectConfig config
+        +process(buf, ctx) list[str]
+        +configure(config) None
+    }
+    
+    EffectPlugin <|-- NoiseEffect
+    EffectPlugin <|-- FadeEffect
+    EffectPlugin <|-- GlitchEffect
+    EffectPlugin <|-- FirehoseEffect
+    EffectPlugin <|-- CropEffect
+    EffectPlugin <|-- TintEffect
+    
+    class Display {
+        <<protocol>>
+        +int width
+        +int height
+        +init(width, height, reuse)
+        +show(buffer, border)
+        +clear() None
+        +cleanup() None
+    }
+    
+    Display <|.. TerminalDisplay
+    Display <|.. NullDisplay
+    Display <|.. PygameDisplay
+    Display <|.. WebSocketDisplay
+    
+    class Camera {
+        +int viewport_width
+        +int viewport_height
+        +CameraMode mode
+        +apply(buffer, width, height) list[str]
+    }
+    
+    class Pipeline {
+        +dict[str, Stage] stages
+        +PipelineContext context
+        +execute(data) StageResult
+    }
+    
+    Pipeline --> Stage
+    Stage --> Display
+```
+
+## Data Flow (Mermaid)
+
+```mermaid
+flowchart LR
+    DataSource[Data Source] --> DataSourceStage
+    DataSourceStage --> FontStage
+    FontStage --> CameraStage
+    CameraStage --> EffectStages
+    EffectStages --> DisplayStage
+    DisplayStage --> TerminalDisplay
+    DisplayStage --> BrowserWebSocket
+    DisplayStage --> SixelDisplay
+    DisplayStage --> NullDisplay
+```
+
+## Effect Chain (Mermaid)
+
+```mermaid
+flowchart LR
+    InputBuffer --> NoiseEffect
+    NoiseEffect --> FadeEffect
+    FadeEffect --> GlitchEffect
+    GlitchEffect --> FirehoseEffect
+    FirehoseEffect --> Output
+```
+
+> **Note:** Each effect must preserve buffer dimensions (line count and visible width).
+
+## Stage Capabilities
+
+```mermaid
+flowchart TB
+    subgraph "Capability Resolution"
+        D[DataSource<br/>provides: source.*]
+        C[Camera<br/>provides: render.output]
+        E[Effects<br/>provides: render.effect]
+        DIS[Display<br/>provides: display.output]
+    end
+```
+
+---
+
+## Legacy ASCII Diagrams
+
+### Stage Inheritance
+```
+Stage(ABC)
+├── DataSourceStage
+├── CameraStage  
+├── FontStage
+├── ViewportFilterStage
+├── EffectPluginStage
+├── DisplayStage
+├── SourceItemsToBufferStage
+├── PassthroughStage
+├── ImageToTextStage
+└── CanvasStage
+```
+
+### Display Backends
+```
+Display(Protocol)
+├── TerminalDisplay
+├── NullDisplay
+├── PygameDisplay
+├── WebSocketDisplay
+└── MultiDisplay
+```
+
+### Camera Modes
+```
+Camera
+├── FEED       # Static view
+├── SCROLL     # Horizontal scroll
+├── VERTICAL   # Vertical scroll
+├── HORIZONTAL # Same as scroll
+├── OMNI       # Omnidirectional
+├── FLOATING   # Floating particles
+└── BOUNCE     # Bouncing camera

docs/GRAPH_SYSTEM_SUMMARY.md

@@ -0,0 +1,178 @@
+# Graph-Based Pipeline System - Implementation Summary
+
+## Overview
+
+Implemented a graph-based scripting language to replace the verbose `XYZStage` naming convention in Mainline's pipeline architecture. The new system represents pipelines as nodes and connections, providing a more intuitive way to define, configure, and orchestrate pipelines.
+
+## Files Created
+
+### Core Graph System
+- `engine/pipeline/graph.py` - Core graph abstraction (Node, Connection, Graph classes)
+- `engine/pipeline/graph_adapter.py` - Adapter to convert Graph to Pipeline with existing Stage classes
+- `engine/pipeline/graph_toml.py` - TOML-based graph configuration loader
+
+### Tests
+- `tests/test_graph_pipeline.py` - Comprehensive test suite (17 tests, all passing)
+- `examples/graph_dsl_demo.py` - Demo script showing the new DSL
+- `examples/test_graph_integration.py` - Integration test verifying pipeline execution
+- `examples/pipeline_graph.toml` - Example TOML configuration file
+
+### Documentation
+- `docs/graph-dsl.md` - Complete DSL documentation with examples
+- `docs/GRAPH_SYSTEM_SUMMARY.md` - This summary document
+
+## Key Features
+
+### 1. Graph Abstraction
+- **Node Types**: `source`, `camera`, `effect`, `position`, `display`, `render`, `overlay`
+- **Connections**: Directed edges between nodes with automatic dependency resolution
+- **Validation**: Cycle detection and disconnected node warnings
+
+### 2. DSL Syntax Options
+
+#### TOML Configuration
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.camera]
+type = "camera"
+mode = "scroll"
+
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.5
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> camera -> noise -> display"]
+```
+
+#### Python API
+```python
+from engine.pipeline.graph import Graph, NodeType
+from engine.pipeline.graph_adapter import graph_to_pipeline
+
+graph = Graph()
+graph.node("source", NodeType.SOURCE, source="headlines")
+graph.node("camera", NodeType.CAMERA, mode="scroll")
+graph.node("noise", NodeType.EFFECT, effect="noise", intensity=0.5)
+graph.node("display", NodeType.DISPLAY, backend="terminal")
+graph.chain("source", "camera", "noise", "display")
+
+pipeline = graph_to_pipeline(graph)
+```
+
+#### Dictionary/JSON Input
+```python
+from engine.pipeline.graph_adapter import dict_to_pipeline
+
+data = {
+    "nodes": {
+        "source": "headlines",
+        "noise": {"type": "effect", "effect": "noise", "intensity": 0.5},
+        "display": {"type": "display", "backend": "terminal"}
+    },
+    "connections": ["source -> noise -> display"]
+}
+
+pipeline = dict_to_pipeline(data)
+```
+
+### 3. Pipeline Integration
+
+The graph system integrates with the existing pipeline architecture:
+
+- **Auto-injection**: Pipeline automatically injects required stages (camera_update, render, etc.)
+- **Capability Resolution**: Uses existing capability-based dependency system
+- **Type Safety**: Validates data flow between stages (TEXT_BUFFER, SOURCE_ITEMS, etc.)
+- **Backward Compatible**: Works alongside existing preset system
+
+### 4. Node Configuration
+
+| Node Type | Config Options | Example |
+|-----------|----------------|---------|
+| `source` | `source`: "headlines", "poetry", "empty" | `{"type": "source", "source": "headlines"}` |
+| `camera` | `mode`: "scroll", "feed", "horizontal", etc.<br>`speed`: float | `{"type": "camera", "mode": "scroll", "speed": 1.0}` |
+| `effect` | `effect`: effect name<br>`intensity`: 0.0-1.0 | `{"type": "effect", "effect": "noise", "intensity": 0.5}` |
+| `position` | `mode`: "absolute", "relative", "mixed" | `{"type": "position", "mode": "mixed"}` |
+| `display` | `backend`: "terminal", "null", "websocket" | `{"type": "display", "backend": "terminal"}` |
+
+## Implementation Details
+
+### Graph Adapter Logic
+
+1. **Node Mapping**: Converts graph nodes to appropriate Stage classes
+2. **Effect Intensity**: Sets effect intensity globally (consistent with existing architecture)
+3. **Camera Creation**: Maps mode strings to Camera factory methods
+4. **Dependencies**: Effects automatically depend on `render.output`
+5. **Type Flow**: Ensures TEXT_BUFFER flow between render and effects
+
+### Validation
+
+- **Disconnected Nodes**: Warns about nodes without connections
+- **Cycle Detection**: Detects circular dependencies using DFS
+- **Type Validation**: Pipeline validates inlet/outlet type compatibility
+
+## Files Modified
+
+### Core Pipeline
+- `engine/pipeline/controller.py` - Pipeline class (no changes needed, uses existing architecture)
+- `engine/pipeline/graph_adapter.py` - Added effect intensity setting, fixed PositionStage creation
+- `engine/app/pipeline_runner.py` - Added graph config support
+
+### Documentation
+- `AGENTS.md` - Updated with task tracking
+
+## Test Results
+
+```
+17 tests passed in 0.23s
+- Graph creation and manipulation
+- Connection handling and validation
+- TOML loading and parsing
+- Pipeline conversion and execution
+- Effect intensity configuration
+- Camera mode mapping
+- Positioning mode support
+```
+
+## Usage Examples
+
+### Running with Graph Config
+```bash
+python -c "
+from engine.effects.plugins import discover_plugins
+from engine.pipeline.graph_toml import load_pipeline_from_toml
+
+discover_plugins()
+pipeline = load_pipeline_from_toml('examples/pipeline_graph.toml')
+"
+```
+
+### Integration with Pipeline Runner
+```bash
+# The pipeline runner now supports graph configs
+# (Implementation in progress)
+```
+
+## Benefits
+
+1. **Simplified Configuration**: No need to manually create Stage instances
+2. **Visual Representation**: Graph structure is easier to understand than class hierarchy
+3. **Automatic Dependency Resolution**: Pipeline handles stage ordering automatically
+4. **Flexible Composition**: Easy to add/remove/modify pipeline stages
+5. **Backward Compatible**: Existing presets and stages continue to work
+
+## Future Enhancements
+
+1. **CLI Integration**: Add `--graph-config` flag to mainline command
+2. **Visual Builder**: Web-based drag-and-drop pipeline editor
+3. **Script Execution**: Support for loops, conditionals, and timing in graph scripts
+4. **Parameter Binding**: Real-time sensor-to-parameter bindings in graph config
+5. **Pipeline Inspection**: Visual DAG representation with metrics

docs/Mainline Renderer + ntfy Message Queue for ESP32.md

@@ -3,29 +3,29 @@
 mainline\.py does heavy work unsuitable for ESP32: 25\+ HTTPS/TLS RSS feeds, OTF font rasterization via Pillow, Google Translate API calls, and complex text layout\. Simultaneously, messages arriving on `ntfy.sh/klubhaus_terminal_mainline` need to interrupt the news ticker on the same device\.
 ## Architecture: Server \+ Thin Client
 Split the system into two halves that are designed together\.
-**Server \(mainline\.py `--serve` mode, runs on any always\-on machine\)**
+**Server (mainline\.py `--serve` mode, runs on any always\-on machine)**
 * Reuses existing feed fetching, caching, content filtering, translation, and Pillow font rendering pipeline — no duplication\.
-* Pre\-renders each headline into a 1\-bit bitmap strip \(the OTF→half\-block pipeline already produces this as an intermediate step in `_render_line()`\)\.
+* Pre\-renders each headline into a 1\-bit bitmap strip (the OTF→half\-block pipeline already produces this as an intermediate step in `_render_line()`)\.
 * Exposes a lightweight HTTP API the ESP32 polls\.
-**ESP32 thin client \(Arduino sketch\)**
-* Polls the mainline server for pre\-rendered headline bitmaps over plain HTTP \(no TLS needed if on the same LAN\)\.
-* Polls `ntfy.sh/klubhaus_terminal_mainline` directly for messages, reusing the proven `NetManager::httpGet()` \+ JSON parsing pattern from DoorbellLogic \(`DoorbellLogic.cpp:155-192`\)\.
-* Manages scrolling, gradient coloring, and glitch effects locally \(cheap per\-frame GPU work\)\.
+**ESP32 thin client (Arduino sketch)**
+* Polls the mainline server for pre\-rendered headline bitmaps over plain HTTP (no TLS needed if on the same LAN)\.
+* Polls `ntfy.sh/klubhaus_terminal_mainline` directly for messages, reusing the proven `NetManager::httpGet()` \+ JSON parsing pattern from DoorbellLogic (`DoorbellLogic.cpp:155-192`)\.
+* Manages scrolling, gradient coloring, and glitch effects locally (cheap per\-frame GPU work)\.
 * When an ntfy message arrives, the scroll is paused and the message takes over the display — same interrupt pattern as the doorbell's ALERT→DASHBOARD flow\.
-## Server API \(mainline repo\)
-New file: `serve.py` \(or `--serve` mode in mainline\.py\)\.
+## Server API (mainline repo)
+New file: `serve.py` (or `--serve` mode in mainline\.py)\.
 Endpoints:
 * `GET /api/headlines` — returns JSON array of headline metadata: `[{"id": 0, "src": "Nature", "ts": "14:30", "width": 280, "height": 16, "bitmap": "<base64 1-bit packed>"}]`\. Bitmaps are 1\-bit\-per\-pixel, row\-major, packed 8px/byte\. The ESP32 applies gradient color locally\.
 * `GET /api/config` — returns `{"count": 120, "version": "...", "mode": "news"}` so the ESP32 knows what it's getting\.
 * `GET /api/health` — `{"ok": true, "last_fetch": "...", "headline_count": 120}`
-The server renders at a configurable target width \(e\.g\. 800px for Board 3, 320px for Boards 1/2\) via a `--width` flag or query parameter\. Height is fixed per headline by the font size\.
-The server refreshes feeds on a timer \(reusing `_SCROLL_DUR` cadence or a longer interval\), re\-renders, and serves the latest set\. The ESP32 polls `/api/headlines` periodically \(e\.g\. every 60s\) and swaps in the new set\.
-## Render pipeline \(server side\)
+The server renders at a configurable target width (e\.g\. 800px for Board 3, 320px for Boards 1/2) via a `--width` flag or query parameter\. Height is fixed per headline by the font size\.
+The server refreshes feeds on a timer (reusing `_SCROLL_DUR` cadence or a longer interval), re\-renders, and serves the latest set\. The ESP32 polls `/api/headlines` periodically (e\.g\. every 60s) and swaps in the new set\.
+## Render pipeline (server side)
 The existing `_render_line()` in mainline\.py already does:
 1. `ImageFont.truetype()` → `ImageDraw.text()` → grayscale `Image`
 2. Resize to target height
-3. Threshold to 1\-bit \(the `thr = 80` step\)
-For the server, we stop at step 3 and pack the 1\-bit data into bytes instead of converting to half\-block Unicode\. This is the exact same pipeline, just with a different output format\. The `_big_wrap()` and `_lr_gradient()` logic stays on the server for layout; gradient *coloring* moves to the ESP32 \(it's just an index lookup per pixel column\)\.
+3. Threshold to 1\-bit (the `thr = 80` step)
+For the server, we stop at step 3 and pack the 1\-bit data into bytes instead of converting to half\-block Unicode\. This is the exact same pipeline, just with a different output format\. The `_big_wrap()` and `_lr_gradient()` logic stays on the server for layout; gradient *coloring* moves to the ESP32 (it's just an index lookup per pixel column)\.
 ## ESP32 client
 ### State machine
 ```warp-runnable-command
@@ -35,19 +35,19 @@ BOOT → SCROLL ⇄ MESSAGE
 * **BOOT** — WiFi connect, initial headline fetch from server\.
 * **SCROLL** — Vertical scroll through pre\-rendered headlines with local gradient \+ glitch\. Polls server for new headlines periodically\. Polls ntfy every 15s\.
 * **MESSAGE** — ntfy message arrived\. Scroll paused, message displayed\. Auto\-dismiss after timeout or touch\-dismiss\. Returns to SCROLL\.
-* **OFF** — Backlight off after inactivity \(polling continues in background\)\.
+* **OFF** — Backlight off after inactivity (polling continues in background)\.
 ### ntfy integration
 The ESP32 polls `https://ntfy.sh/klubhaus_terminal_mainline/json?since=20s&poll=1` on the same 15s interval as the doorbell polls its topics\. When a message event arrives:
 1. Parse JSON: `{"event": "message", "title": "...", "message": "..."}`
 2. Save current scroll position\.
 3. Transition to MESSAGE state\.
-4. Render message text using the display library's built\-in fonts \(messages are short, no custom font needed\)\.
-5. After `MESSAGE_TIMEOUT_MS` \(e\.g\. 30s\) or touch, restore scroll position and resume\.
-This is architecturally identical to `DoorbellLogic::onAlert()` → `dismissAlert()`, just with different content\. The ntfy polling runs independently of the server connection, so messages work even if the mainline server is offline \(the device just shows the last cached headlines\)\.
+4. Render message text using the display library's built\-in fonts (messages are short, no custom font needed)\.
+5. After `MESSAGE_TIMEOUT_MS` (e\.g\. 30s) or touch, restore scroll position and resume\.
+This is architecturally identical to `DoorbellLogic::onAlert()` → `dismissAlert()`, just with different content\. The ntfy polling runs independently of the server connection, so messages work even if the mainline server is offline (the device just shows the last cached headlines)\.
 ### Headline storage
-* Board 3 \(8 MB PSRAM\): store all ~120 headline bitmaps in PSRAM\. At 800px × 16px × 1 bit = 1\.6 KB each → ~192 KB total\. Trivial\.
-* Boards 1/2 \(PSRAM TBD\): at 320px × 16px = 640 bytes each → ~77 KB for 120 headlines\. Fits if PSRAM is present\. Without PSRAM, keep ~20 headlines in a ring buffer \(~13 KB\)\.
-### Gradient coloring \(local\)
+* Board 3 (8 MB PSRAM): store all ~120 headline bitmaps in PSRAM\. At 800px × 16px × 1 bit = 1\.6 KB each → ~192 KB total\. Trivial\.
+* Boards 1/2 (PSRAM TBD): at 320px × 16px = 640 bytes each → ~77 KB for 120 headlines\. Fits if PSRAM is present\. Without PSRAM, keep ~20 headlines in a ring buffer (~13 KB)\.
+### Gradient coloring (local)
 The 12\-step ANSI gradient in mainline\.py maps to 12 RGB565 values:
 ```warp-runnable-command
 const uint16_t GRADIENT[] = {
@@ -68,8 +68,8 @@ mainline.py          (existing, unchanged)
 serve.py             (new — HTTP server, imports mainline rendering functions)
 klubhaus-doorbell-hardware.md  (existing)
 ```
-`serve.py` imports the rendering functions from mainline\.py \(after refactoring them into importable form — they're currently top\-level but not wrapped in `if __name__`\)\.
-### klubhaus\-doorbell repo \(or mainline repo under firmware/\)
+`serve.py` imports the rendering functions from mainline\.py (after refactoring them into importable form — they're currently top\-level but not wrapped in `if __name__`)\.
+### klubhaus\-doorbell repo (or mainline repo under firmware/)
 ```warp-runnable-command
 boards/esp32-mainline/
 ├── esp32-mainline.ino    Main sketch
@@ -79,31 +79,31 @@ boards/esp32-mainline/
 ├── HeadlineStore.h/.cpp  Bitmap ring buffer in PSRAM
 └── NtfyPoller.h/.cpp     ntfy.sh polling (extracted from DoorbellLogic pattern)
 ```
-The display driver is reused from the target board \(e\.g\. `DisplayDriverGFX` for Board 3\)\. `MainlineLogic` replaces `DoorbellLogic` as the state machine but follows the same patterns\.
+The display driver is reused from the target board (e\.g\. `DisplayDriverGFX` for Board 3)\. `MainlineLogic` replaces `DoorbellLogic` as the state machine but follows the same patterns\.
 ## Branch strategy recommendation
 The work spans two repos and has clear dependency ordering\.
-### Phase 1 — Finish current branch \(mainline repo\)
-**Branch:** `feat/arduino` \(current\)
+### Phase 1 — Finish current branch (mainline repo)
+**Branch:** `feat/arduino` (current)
 **Content:** Hardware spec doc\. Already done\.
 **Action:** Merge to main when ready\.
-### Phase 2 — Server renderer \(mainline repo\)
-**Branch:** `feat/renderer` \(branch from main after Phase 1 merges\)
+### Phase 2 — Server renderer (mainline repo)
+**Branch:** `feat/renderer` (branch from main after Phase 1 merges)
 **Content:**
-* Refactor mainline\.py rendering functions to be importable \(extract from `__main__` guard\)
+* Refactor mainline\.py rendering functions to be importable (extract from `__main__` guard)
 * `serve.py` — HTTP server with `/api/headlines`, `/api/config`, `/api/health`
-* Bitmap packing utility \(1\-bit row\-major\)
-**Why a separate branch:** This changes mainline\.py's structure \(refactoring for imports\) and adds a new entry point\. It's a self\-contained, testable unit — you can verify the API with `curl` before touching any Arduino code\.
-### Phase 3 — ESP32 client \(klubhaus\-doorbell repo, or mainline repo\)
+* Bitmap packing utility (1\-bit row\-major)
+**Why a separate branch:** This changes mainline\.py's structure (refactoring for imports) and adds a new entry point\. It's a self\-contained, testable unit — you can verify the API with `curl` before touching any Arduino code\.
+### Phase 3 — ESP32 client (klubhaus\-doorbell repo, or mainline repo)
 **Branch:** `feat/mainline-client` in whichever repo hosts it
 **Content:**
 * `MainlineLogic` state machine
 * `HeadlineStore` bitmap buffer
 * `NtfyPoller` for `klubhaus_terminal_mainline`
 * Board\-specific sketch for the target board
-**Depends on:** Phase 2 \(needs a running server to test against\)
+**Depends on:** Phase 2 (needs a running server to test against)
 **Repo decision:** If you have push access to klubhaus\-doorbell, it fits naturally as a new board target alongside the existing doorbell sketches — it reuses `NetManager`, `IDisplayDriver`, and the vendored display libraries\. If not, put it under `mainline/firmware/` and vendor the shared KlubhausCore library\.
 ### Merge order
-1. `feat/arduino` → main \(hardware spec\)
-2. `feat/renderer` → main \(server\)
-3. `feat/mainline-client` → main in whichever repo \(ESP32 client\)
+1. `feat/arduino` → main (hardware spec)
+2. `feat/renderer` → main (server)
+3. `feat/mainline-client` → main in whichever repo (ESP32 client)
 Each phase is independently testable and doesn't block the other until Phase 3 needs a running server\.

docs/PIPELINE.md

@@ -0,0 +1,223 @@
+# Mainline Pipeline
+
+## Architecture Overview
+
+The Mainline pipeline uses a **Stage-based architecture** with **capability-based dependency resolution**. Stages declare capabilities (what they provide) and dependencies (what they need), and the Pipeline resolves dependencies using prefix matching.
+
+```
+Source Stage → Render Stage → Effect Stages → Display Stage
+      ↓
+Camera Stage (provides camera.state capability)
+```
+
+### Capability-Based Dependency Resolution
+
+Stages declare capabilities and dependencies:
+- **Capabilities**: What the stage provides (e.g., `source`, `render.output`, `display.output`, `camera.state`)
+- **Dependencies**: What the stage needs (e.g., `source`, `render.output`, `camera.state`)
+
+The Pipeline resolves dependencies using **prefix matching**:
+- `"source"` matches `"source.headlines"`, `"source.poetry"`, etc.
+- `"camera.state"` matches the camera state capability provided by `CameraClockStage`
+- This allows flexible composition without hardcoding specific stage names
+
+### Minimum Capabilities
+
+The pipeline requires these minimum capabilities to function:
+- `"source"` - Data source capability (provides raw items)
+- `"render.output"` - Rendered content capability
+- `"display.output"` - Display output capability
+- `"camera.state"` - Camera state for viewport filtering
+
+These are automatically injected if missing by the `ensure_minimum_capabilities()` method.
+
+### Stage Registry
+
+The `StageRegistry` discovers and registers stages automatically:
+- Scans `engine/stages/` for stage implementations
+- Registers stages by their declared capabilities
+- Enables runtime stage discovery and composition
+
+## Stage-Based Pipeline Flow
+
+```mermaid
+flowchart TD
+    subgraph Stages["Stage Pipeline"]
+        subgraph SourceStage["Source Stage (provides: source.*)"]
+            Headlines[HeadlinesSource]
+            Poetry[PoetrySource]
+            Pipeline[PipelineSource]
+        end
+
+        subgraph RenderStage["Render Stage (provides: render.*)"]
+            Render[RenderStage]
+            Canvas[Canvas]
+            Camera[Camera]
+        end
+
+        subgraph EffectStages["Effect Stages (provides: effect.*)"]
+            Noise[NoiseEffect]
+            Fade[FadeEffect]
+            Glitch[GlitchEffect]
+            Firehose[FirehoseEffect]
+            Hud[HudEffect]
+        end
+
+        subgraph DisplayStage["Display Stage (provides: display.*)"]
+            Terminal[TerminalDisplay]
+            Pygame[PygameDisplay]
+            WebSocket[WebSocketDisplay]
+            Null[NullDisplay]
+        end
+    end
+
+    subgraph Capabilities["Capability Map"]
+        SourceCaps["source.headlines<br/>source.poetry<br/>source.pipeline"]
+        RenderCaps["render.output<br/>render.canvas"]
+        EffectCaps["effect.noise<br/>effect.fade<br/>effect.glitch"]
+        DisplayCaps["display.output<br/>display.terminal"]
+    end
+
+    SourceStage --> RenderStage
+    RenderStage --> EffectStages
+    EffectStages --> DisplayStage
+
+    SourceStage --> SourceCaps
+    RenderStage --> RenderCaps
+    EffectStages --> EffectCaps
+    DisplayStage --> DisplayCaps
+
+    style SourceStage fill:#f9f,stroke:#333
+    style RenderStage fill:#bbf,stroke:#333
+    style EffectStages fill:#fbf,stroke:#333
+    style DisplayStage fill:#bfb,stroke:#333
+```
+
+## Stage Adapters
+
+Existing components are wrapped as Stages via adapters:
+
+### Source Stage Adapter
+- Wraps `HeadlinesDataSource`, `PoetryDataSource`, etc.
+- Provides `source.*` capabilities
+- Fetches data and outputs to pipeline buffer
+
+### Render Stage Adapter
+- Wraps `StreamController`, `Camera`, `render_ticker_zone`
+- Provides `render.output` capability
+- Processes content and renders to canvas
+
+### Effect Stage Adapter
+- Wraps `EffectChain` and individual effect plugins
+- Provides `effect.*` capabilities
+- Applies visual effects to rendered content
+
+### Display Stage Adapter
+- Wraps `TerminalDisplay`, `PygameDisplay`, etc.
+- Provides `display.*` capabilities
+- Outputs final buffer to display backend
+
+## Pipeline Mutation API
+
+The Pipeline supports dynamic mutation during runtime:
+
+### Core Methods
+- `add_stage(name, stage, initialize=True)` - Add a stage
+- `remove_stage(name, cleanup=True)` - Remove a stage and rebuild execution order
+- `replace_stage(name, new_stage, preserve_state=True)` - Replace a stage
+- `swap_stages(name1, name2)` - Swap two stages
+- `move_stage(name, after=None, before=None)` - Move a stage in execution order
+- `enable_stage(name)` / `disable_stage(name)` - Enable/disable stages
+
+### Safety Checks
+- `can_hot_swap(name)` - Check if a stage can be safely hot-swapped
+- `cleanup_stage(name)` - Clean up specific stage without removing it
+
+### WebSocket Commands
+The mutation API is accessible via WebSocket for remote control:
+```json
+{"action": "remove_stage", "stage": "stage_name"}
+{"action": "swap_stages", "stage1": "name1", "stage2": "name2"}
+{"action": "enable_stage", "stage": "stage_name"}
+{"action": "cleanup_stage", "stage": "stage_name"}
+```
+
+## Camera Modes
+
+The Camera supports the following modes:
+
+- **FEED**: Single item view (static or rapid cycling)
+- **SCROLL**: Smooth vertical scrolling (movie credits style)
+- **HORIZONTAL**: Left/right movement
+- **OMNI**: Combination of vertical and horizontal
+- **FLOATING**: Sinusoidal/bobbing motion
+- **BOUNCE**: DVD-style bouncing off edges
+- **RADIAL**: Polar coordinate scanning (radar sweep)
+
+Note: Camera state is provided by `CameraClockStage` (capability: `camera.state`) which updates independently of data flow. The `CameraStage` applies viewport transformations (capability: `camera`).
+
+## Animation & Presets
+
+```mermaid
+flowchart LR
+    subgraph Preset["Preset"]
+        PP[PipelineParams]
+        AC[AnimationController]
+    end
+    
+    subgraph AnimationController["AnimationController"]
+        Clock[Clock]
+        Events[Events]
+        Triggers[Triggers]
+    end
+    
+    subgraph Triggers["Trigger Types"]
+        TIME[TIME]
+        FRAME[FRAME]
+        CYCLE[CYCLE]
+        COND[CONDITION]
+        MANUAL[MANUAL]
+    end
+    
+    PP --> AC
+    Clock --> AC
+    Events --> AC
+    Triggers --> Events
+```
+
+## Camera Modes State Diagram
+
+```mermaid
+stateDiagram-v2
+    [*] --> Vertical
+    Vertical --> Horizontal: mode change
+    Horizontal --> Omni: mode change
+    Omni --> Floating: mode change
+    Floating --> Trace: mode change
+    Trace --> Vertical: mode change
+
+    state Vertical {
+        [*] --> ScrollUp
+        ScrollUp --> ScrollUp: +y each frame
+    }
+
+    state Horizontal {
+        [*] --> ScrollLeft
+        ScrollLeft --> ScrollLeft: +x each frame
+    }
+
+    state Omni {
+        [*] --> Diagonal
+        Diagonal --> Diagonal: +x, +y each frame
+    }
+
+    state Floating {
+        [*] --> Bobbing
+        Bobbing --> Bobbing: sin(time) for x,y
+    }
+
+    state Trace {
+        [*] --> FollowPath
+        FollowPath --> FollowPath: node by node
+    }
+```

docs/Refactor mainline.md

@@ -0,0 +1,255 @@
+#
+
+Refactor mainline\.py into modular package
+
+## Problem
+
+`mainline.py` is a single 1085\-line file with ~10 interleaved concerns\. This prevents:
+
+* Reusing the ntfy doorbell interrupt in other visualizers
+* Importing the render pipeline from `serve.py` \(future ESP32 HTTP server\)
+* Testing any concern in isolation
+* Porting individual layers to Rust independently
+
+## Target structure
+
+```warp-runnable-command
+mainline.py              # thin entrypoint: venv bootstrap → engine.app.main()
+engine/
+  __init__.py
+  config.py              # constants, CLI flags, glyph tables
+  sources.py             # FEEDS, POETRY_SOURCES, SOURCE_LANGS, _LOCATION_LANGS
+  terminal.py            # ANSI codes, tw/th, type_out, slow_print, boot_ln
+  filter.py              # HTML stripping, content filter (_SKIP_RE)
+  translate.py           # Google Translate wrapper + location→language detection
+  render.py              # OTF font loading, _render_line, _big_wrap, _lr_gradient, _make_block
+  effects.py             # noise, glitch_bar, _fade_line, _vis_trunc, _firehose_line, _next_headline
+  fetch.py               # RSS/Gutenberg fetching, cache load/save
+  ntfy.py                # NtfyPoller class — standalone, zero internal deps
+  mic.py                 # MicMonitor class — standalone
+  scroll.py              # stream() frame loop + message rendering
+  app.py                 # main(), TITLE art, boot sequence, signal handler
+```
+
+The package is named `engine/` to avoid a naming conflict with the `mainline.py` entrypoint\.
+
+## Module dependency graph
+
+```warp-runnable-command
+config      ← (nothing)
+sources     ← (nothing)
+terminal    ← (nothing)
+filter      ← (nothing)
+translate   ← sources
+render      ← config, terminal, sources
+effects     ← config, terminal, sources
+fetch       ← config, sources, filter, terminal
+ntfy        ← (nothing — stdlib only, fully standalone)
+mic         ← (nothing — sounddevice only)
+scroll      ← config, terminal, render, effects, ntfy, mic
+app         ← everything above
+```
+
+Critical property: **ntfy\.py and mic\.py have zero internal dependencies**, making ntfy reusable by any visualizer\.
+
+## Module details
+
+### mainline\.py \(entrypoint — slimmed down\)
+
+Keeps only the venv bootstrap \(lines 10\-38\) which must run before any third\-party imports\. After bootstrap, delegates to `engine.app.main()`\.
+
+### engine/config\.py
+
+From current mainline\.py:
+
+* `HEADLINE_LIMIT`, `FEED_TIMEOUT`, `MIC_THRESHOLD_DB` \(lines 55\-57\)
+* `MODE`, `FIREHOSE` CLI flag parsing \(lines 58\-59\)
+* `NTFY_TOPIC`, `NTFY_POLL_INTERVAL`, `MESSAGE_DISPLAY_SECS` \(lines 62\-64\)
+* `_FONT_PATH`, `_FONT_SZ`, `_RENDER_H` \(lines 147\-150\)
+* `_SCROLL_DUR`, `_FRAME_DT`, `FIREHOSE_H` \(lines 505\-507\)
+* `GLITCH`, `KATA` glyph tables \(lines 143\-144\)
+
+### engine/sources\.py
+
+Pure data, no logic:
+
+* `FEEDS` dict \(lines 102\-140\)
+* `POETRY_SOURCES` dict \(lines 67\-80\)
+* `SOURCE_LANGS` dict \(lines 258\-266\)
+* `_LOCATION_LANGS` dict \(lines 269\-289\)
+* `_SCRIPT_FONTS` dict \(lines 153\-165\)
+* `_NO_UPPER` set \(line 167\)
+
+### engine/terminal\.py
+
+ANSI primitives and terminal I/O:
+
+* All ANSI constants: `RST`, `BOLD`, `DIM`, `G_HI`, `G_MID`, `G_LO`, `G_DIM`, `W_COOL`, `W_DIM`, `W_GHOST`, `C_DIM`, `CLR`, `CURSOR_OFF`, `CURSOR_ON` \(lines 83\-99\)
+* `tw()`, `th()` \(lines 223\-234\)
+* `type_out()`, `slow_print()`, `boot_ln()` \(lines 355\-386\)
+
+### engine/filter\.py
+
+* `_Strip` HTML parser class \(lines 205\-214\)
+* `strip_tags()` \(lines 217\-220\)
+* `_SKIP_RE` compiled regex \(lines 322\-346\)
+* `_skip()` predicate \(lines 349\-351\)
+
+### engine/translate\.py
+
+* `_TRANSLATE_CACHE` \(line 291\)
+* `_detect_location_language()` \(lines 294\-300\) — imports `_LOCATION_LANGS` from sources
+* `_translate_headline()` \(lines 303\-319\)
+
+### engine/render\.py
+
+The OTF→terminal pipeline\. This is exactly what `serve.py` will import to produce 1\-bit bitmaps for the ESP32\.
+
+* `_GRAD_COLS` gradient table \(lines 169\-182\)
+* `_font()`, `_font_for_lang()` with lazy\-load \+ cache \(lines 185\-202\)
+* `_render_line()` — OTF text → half\-block terminal rows \(lines 567\-605\)
+* `_big_wrap()` — word\-wrap \+ render \(lines 608\-636\)
+* `_lr_gradient()` — apply left→right color gradient \(lines 639\-656\)
+* `_make_block()` — composite: translate → render → colorize a headline \(lines 718\-756\)\. Imports from translate, sources\.
+
+### engine/effects\.py
+
+Visual effects applied during the frame loop:
+
+* `noise()` \(lines 237\-245\)
+* `glitch_bar()` \(lines 248\-252\)
+* `_fade_line()` — probabilistic character dissolve \(lines 659\-680\)
+* `_vis_trunc()` — ANSI\-aware width truncation \(lines 683\-701\)
+* `_firehose_line()` \(lines 759\-801\) — imports config\.MODE, sources\.FEEDS/POETRY\_SOURCES
+* `_next_headline()` — pool management \(lines 704\-715\)
+
+### engine/fetch\.py
+
+* `fetch_feed()` \(lines 390\-396\)
+* `fetch_all()` \(lines 399\-426\) — imports filter\.\_skip, filter\.strip\_tags, terminal\.boot\_ln
+* `_fetch_gutenberg()` \(lines 429\-456\)
+* `fetch_poetry()` \(lines 459\-472\)
+* `_cache_path()`, `_load_cache()`, `_save_cache()` \(lines 476\-501\)
+
+### engine/ntfy\.py — standalone, reusable
+
+Refactored from the current globals \+ thread \(lines 531\-564\) and the message rendering section of `stream()` \(lines 845\-909\) into a class:
+
+```python
+class NtfyPoller:
+    def __init__(self, topic_url, poll_interval=15, display_secs=30):
+        ...
+    def start(self):
+        """Start background polling thread."""
+    def get_active_message(self):
+        """Return (title, body, timestamp) if a message is active and not expired, else None."""
+    def dismiss(self):
+        """Manually dismiss current message."""
+```
+
+Dependencies: `urllib.request`, `json`, `threading`, `time` — all stdlib\. No internal imports\.
+Other visualizers use it like:
+
+```python
+from engine.ntfy import NtfyPoller
+poller = NtfyPoller("https://ntfy.sh/my_topic/json?since=20s&poll=1")
+poller.start()
+# in render loop:
+msg = poller.get_active_message()
+if msg:
+    title, body, ts = msg
+    render_my_message(title, body)  # visualizer-specific
+```
+
+### engine/mic\.py — standalone
+
+Refactored from the current globals \(lines 508\-528\) into a class:
+
+```python
+class MicMonitor:
+    def __init__(self, threshold_db=50):
+        ...
+    def start(self) -> bool:
+        """Start background mic stream. Returns False if unavailable."""
+    def stop(self):
+        ...
+    @property
+    def db(self) -> float:
+        """Current RMS dB level."""
+    @property
+    def excess(self) -> float:
+        """dB above threshold (clamped to 0)."""
+```
+
+Dependencies: `sounddevice`, `numpy` \(both optional — graceful fallback\)\.
+
+### engine/scroll\.py
+
+The `stream()` function \(lines 804\-990\)\. Receives its dependencies via arguments or imports:
+
+* `stream(items, ntfy_poller, mic_monitor, config)` or similar
+* Message rendering \(lines 855\-909\) stays here since it's terminal\-display\-specific — a different visualizer would render messages differently
+
+### engine/app\.py
+
+The orchestrator:
+
+* `TITLE` ASCII art \(lines 994\-1001\)
+* `main()` \(lines 1004\-1084\): CLI handling, signal setup, boot animation, fetch, wire up ntfy/mic/scroll
+
+## Execution order
+
+### Step 1: Create engine/ package skeleton
+
+Create `engine/__init__.py` and all empty module files\.
+
+### Step 2: Extract pure data modules \(zero\-dep\)
+
+Move constants and data dicts into `config.py`, `sources.py`\. These have no logic dependencies\.
+
+### Step 3: Extract terminal\.py
+
+Move ANSI codes and terminal I/O helpers\. No internal deps\.
+
+### Step 4: Extract filter\.py and translate\.py
+
+Both are small, self\-contained\. translate imports from sources\.
+
+### Step 5: Extract render\.py
+
+Font loading \+ the OTF→half\-block pipeline\. Imports from config, terminal, sources\. This is the module `serve.py` will later import\.
+
+### Step 6: Extract effects\.py
+
+Visual effects\. Imports from config, terminal, sources\.
+
+### Step 7: Extract fetch\.py
+
+Feed/Gutenberg fetching \+ caching\. Imports from config, sources, filter, terminal\.
+
+### Step 8: Extract ntfy\.py and mic\.py
+
+Refactor globals\+threads into classes\. Zero internal deps\.
+
+### Step 9: Extract scroll\.py
+
+The frame loop\. Last to extract because it depends on everything above\.
+
+### Step 10: Extract app\.py
+
+The `main()` function, boot sequence, signal handler\. Wire up all modules\.
+
+### Step 11: Slim down mainline\.py
+
+Keep only venv bootstrap \+ `from engine.app import main; main()`\.
+
+### Step 12: Verify
+
+Run `python3 mainline.py`, `python3 mainline.py --poetry`, and `python3 mainline.py --firehose` to confirm identical behavior\. No behavioral changes in this refactor\.
+
+## What this enables
+
+* **serve\.py** \(future\): `from engine.render import _render_line, _big_wrap` \+ `from engine.fetch import fetch_all` — imports the pipeline directly
+* **Other visualizers**: `from engine.ntfy import NtfyPoller` — doorbell feature with no coupling to mainline's scroll engine
+* **Rust port**: Clear boundaries for what to port first \(ntfy client, render pipeline\) vs what stays in Python \(fetching, caching — the server side\)
+* **Testing**: Each module can be unit\-tested in isolation

docs/SUMMARY.md

@@ -0,0 +1,30 @@
+# Mainline Documentation Summary
+
+## Core Architecture
+- [Pipeline Architecture](PIPELINE.md) - Pipeline stages, capability resolution, DAG execution
+- [Graph-Based DSL](graph-dsl.md) - New graph abstraction for pipeline configuration
+
+## Pipeline Configuration
+- [Hybrid Config](hybrid-config.md) - **Recommended**: Preset simplicity + graph flexibility
+- [Graph DSL](graph-dsl.md) - Verbose node-based graph definition
+- [Presets Usage](presets-usage.md) - Creating and using pipeline presets
+
+## Feature Documentation
+- [Positioning Analysis](positioning-analysis.md) - Positioning modes and tradeoffs
+- [Pipeline Introspection](pipeline_introspection.md) - Live pipeline visualization
+
+## Implementation Details
+- [Graph System Summary](GRAPH_SYSTEM_SUMMARY.md) - Complete implementation overview
+
+## Quick Start
+
+**Recommended: Hybrid Configuration**
+```toml
+[pipeline]
+source = "headlines"
+camera = { mode = "scroll" }
+effects = [{ name = "noise", intensity = 0.3 }]
+display = { backend = "terminal" }
+```
+
+See `docs/hybrid-config.md` for details.

docs/analysis_graph_dsl_duplicative.md

@@ -0,0 +1,236 @@
+# Analysis: Graph DSL Duplicative Issue
+
+## Executive Summary
+
+The current Graph DSL implementation in Mainline is **duplicative** because:
+
+1. **Node definitions are repeated**: Every node requires a full `[nodes.name]` block with `type` and specific config, even when the type can often be inferred
+2. **Connections are separate**: The `[connections]` list must manually reference node names that were just defined
+3. **Type specification is redundant**: The `type = "effect"` is always the same as the key name prefix
+4. **No implicit connections**: Even linear pipelines require explicit connection strings
+
+This creates significant verbosity compared to the preset system.
+
+---
+
+## What Makes the Script Feel "Duplicative"
+
+### 1. Type Specification Redundancy
+
+```toml
+[nodes.noise]
+type = "effect"        # ← Redundant: already know it's an effect from context
+effect = "noise"
+intensity = 0.3
+```
+
+**Why it's redundant:**
+- The `[nodes.noise]` section name suggests it's a custom node
+- The `effect = "noise"` key implies it's an effect type
+- The parser could infer the type from the presence of `effect` key
+
+### 2. Connection String Redundancy
+
+```toml
+[connections]
+list = ["source -> camera -> noise -> fade -> glitch -> firehose -> display"]
+```
+
+**Why it's redundant:**
+- All node names were already defined in individual blocks above
+- For linear pipelines, the natural flow is obvious
+- The connection order matches the definition order
+
+### 3. Verbosity Comparison
+
+**Preset System (10 lines):**
+```toml
+[presets.upstream-default]
+source = "headlines"
+display = "terminal"
+camera = "scroll"
+effects = ["noise", "fade", "glitch", "firehose"]
+camera_speed = 1.0
+viewport_width = 80
+viewport_height = 24
+```
+
+**Graph DSL (39 lines):**
+- 3.9x more lines for the same pipeline
+- Each effect requires 4 lines instead of 1 line in preset system
+- Connection string repeats all node names
+
+---
+
+## Syntactic Sugar Options
+
+### Option 1: Type Inference (Immediate)
+
+**Current:**
+```toml
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.3
+```
+
+**Proposed:**
+```toml
+[nodes.noise]
+effect = "noise"        # Type inferred from 'effect' key
+intensity = 0.3
+```
+
+**Implementation:** Modify `graph_toml.py` to infer node type from keys:
+- `effect` key → type = "effect"
+- `backend` key → type = "display"
+- `source` key → type = "source"
+- `mode` key → type = "camera"
+
+### Option 2: Implicit Linear Connections
+
+**Current:**
+```toml
+[connections]
+list = ["source -> camera -> noise -> fade -> display"]
+```
+
+**Proposed:**
+```toml
+[connections]
+implicit = true  # Auto-connect all nodes in definition order
+```
+
+**Implementation:** If `implicit = true`, automatically create connections between consecutive nodes.
+
+### Option 3: Inline Node Definitions
+
+**Current:**
+```toml
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.3
+
+[nodes.fade]
+type = "effect"
+effect = "fade"
+intensity = 0.5
+```
+
+**Proposed:**
+```toml
+[graph]
+nodes = [
+    { name = "source", source = "headlines" },
+    { name = "noise", effect = "noise", intensity = 0.3 },
+    { name = "fade", effect = "fade", intensity = 0.5 },
+    { name = "display", backend = "terminal" }
+]
+connections = ["source -> noise -> fade -> display"]
+```
+
+### Option 4: Hybrid Preset-Graph System
+
+```toml
+[presets.custom]
+source = "headlines"
+display = "terminal"
+camera = "scroll"
+effects = [
+    { name = "noise", intensity = 0.3 },
+    { name = "fade", intensity = 0.5 }
+]
+```
+
+---
+
+## Comparative Analysis: Other Systems
+
+### GitHub Actions
+```yaml
+steps:
+  - uses: actions/checkout@v2
+  - uses: actions/setup-node@v2
+  - run: npm install
+```
+- Steps in order, no explicit connection syntax
+- Type inference from `uses` or `run`
+
+### Apache Airflow
+```python
+task1 = PythonOperator(...)
+task2 = PythonOperator(...)
+task1 >> task2  # Minimal connection syntax
+```
+
+### Jenkins Pipeline
+```groovy
+stages {
+    stage('Build') { steps { sh 'make' } }
+    stage('Test') { steps { sh 'make test' } }
+}
+```
+- Implicit sequential execution
+
+---
+
+## Recommended Improvements
+
+### Immediate (Backward Compatible)
+
+1. **Type Inference** - Make `type` field optional
+2. **Implicit Connections** - Add `implicit = true` option
+3. **Array Format** - Support `nodes = ["a", "b", "c"]` format
+
+### Example: Improved Configuration
+
+**Current (39 lines):**
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.camera]
+type = "camera"
+mode = "scroll"
+speed = 1.0
+
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.3
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> camera -> noise -> display"]
+```
+
+**Improved (13 lines, 67% reduction):**
+```toml
+[graph]
+nodes = [
+    { name = "source", source = "headlines" },
+    { name = "camera", mode = "scroll", speed = 1.0 },
+    { name = "noise", effect = "noise", intensity = 0.3 },
+    { name = "display", backend = "terminal" }
+]
+
+[connections]
+implicit = true  # Auto-connects: source -> camera -> noise -> display
+```
+
+---
+
+## Conclusion
+
+The Graph DSL's duplicative nature stems from:
+1. **Explicit type specification** when it could be inferred
+2. **Separate connection definitions** that repeat node names
+3. **Verbose node definitions** for simple cases
+4. **Lack of implicit defaults** for linear pipelines
+
+The recommended improvements focus on **type inference** and **implicit connections** as immediate wins that reduce verbosity by 50%+ while maintaining full flexibility for complex pipelines.

docs/graph-dsl.md

@@ -0,0 +1,210 @@
+# Graph-Based Pipeline DSL
+
+This document describes the new graph-based DSL for defining pipelines in Mainline.
+
+## Overview
+
+The graph DSL represents pipelines as nodes and connections, replacing the verbose `XYZStage` naming convention with a more intuitive graph abstraction.
+
+## TOML Syntax
+
+### Basic Pipeline
+
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.camera]
+type = "camera"
+mode = "scroll"
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> camera -> display"]
+```
+
+### With Effects
+
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.5
+
+[nodes.fade]
+type = "effect"
+effect = "fade"
+intensity = 0.8
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> noise -> fade -> display"]
+```
+
+### With Positioning
+
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.position]
+type = "position"
+mode = "mixed"
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> position -> display"]
+```
+
+## Python API
+
+### Basic Construction
+
+```python
+from engine.pipeline.graph import Graph, NodeType
+
+graph = Graph()
+graph.node("source", NodeType.SOURCE, source="headlines")
+graph.node("camera", NodeType.CAMERA, mode="scroll")
+graph.node("display", NodeType.DISPLAY, backend="terminal")
+graph.chain("source", "camera", "display")
+
+pipeline = graph_to_pipeline(graph)
+```
+
+### With Effects
+
+```python
+from engine.pipeline.graph import Graph, NodeType
+
+graph = Graph()
+graph.node("source", NodeType.SOURCE, source="headlines")
+graph.node("noise", NodeType.EFFECT, effect="noise", intensity=0.5)
+graph.node("fade", NodeType.EFFECT, effect="fade", intensity=0.8)
+graph.node("display", NodeType.DISPLAY, backend="terminal")
+graph.chain("source", "noise", "fade", "display")
+
+pipeline = graph_to_pipeline(graph)
+```
+
+### Dictionary/JSON Input
+
+```python
+from engine.pipeline.graph_adapter import dict_to_pipeline
+
+data = {
+    "nodes": {
+        "source": "headlines",
+        "noise": {"type": "effect", "effect": "noise", "intensity": 0.5},
+        "display": {"type": "display", "backend": "terminal"}
+    },
+    "connections": ["source -> noise -> display"]
+}
+
+pipeline = dict_to_pipeline(data)
+```
+
+## CLI Usage
+
+### Using Graph Config File
+
+```bash
+mainline --graph-config pipeline.toml
+```
+
+### Inline Graph Definition
+
+```bash
+mainline --graph 'source:headlines -> noise:noise:0.5 -> display:terminal'
+```
+
+### With Preset Override
+
+```bash
+mainline --preset demo --graph-modify 'add:noise:0.5 after:source'
+```
+
+## Node Types
+
+| Type | Description | Config Options |
+|------|-------------|----------------|
+| `source` | Data source | `source`: "headlines", "poetry", "empty", etc. |
+| `camera` | Viewport camera | `mode`: "scroll", "feed", "horizontal", etc. `speed`: float |
+| `effect` | Visual effect | `effect`: effect name, `intensity`: 0.0-1.0 |
+| `position` | Positioning mode | `mode`: "absolute", "relative", "mixed" |
+| `display` | Output backend | `backend`: "terminal", "null", "websocket", etc. |
+| `render` | Text rendering | (auto-injected) |
+| `overlay` | Message overlay | (auto-injected) |
+
+## Advanced Features
+
+### Conditional Connections
+
+```toml
+[connections]
+list = ["source -> camera -> display"]
+# Effects can be conditionally enabled/disabled
+```
+
+### Parameter Binding
+
+```toml
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 1.0
+# intensity can be bound to sensor values at runtime
+```
+
+### Pipeline Inspection
+
+```toml
+[nodes.inspect]
+type = "pipeline-inspect"
+# Renders live pipeline visualization
+```
+
+## Comparison with Stage-Based Approach
+
+### Old (Stage-Based)
+
+```python
+pipeline = Pipeline()
+pipeline.add_stage("source", DataSourceStage(HeadlinesDataSource()))
+pipeline.add_stage("camera", CameraStage(Camera.scroll()))
+pipeline.add_stage("render", FontStage())
+pipeline.add_stage("noise", EffectPluginStage(noise_effect))
+pipeline.add_stage("display", DisplayStage(terminal_display))
+```
+
+### New (Graph-Based)
+
+```python
+graph = Graph()
+graph.node("source", NodeType.SOURCE, source="headlines")
+graph.node("camera", NodeType.CAMERA, mode="scroll")
+graph.node("noise", NodeType.EFFECT, effect="noise")
+graph.node("display", NodeType.DISPLAY, backend="terminal")
+graph.chain("source", "camera", "noise", "display")
+pipeline = graph_to_pipeline(graph)
+```
+
+The graph system automatically:
+- Inserts the render stage between camera and effects
+- Handles capability-based dependency resolution
+- Auto-injects required stages (camera_update, render, etc.)

docs/hybrid-config.md

@@ -0,0 +1,267 @@
+# Hybrid Preset-Graph Configuration
+
+The hybrid configuration format combines the simplicity of presets with the flexibility of graphs, providing a concise way to define pipelines.
+
+## Overview
+
+The hybrid format uses **70% less space** than the verbose node-based DSL while providing the same functionality.
+
+### Comparison
+
+**Verbose Node DSL (39 lines):**
+```toml
+[nodes.source]
+type = "source"
+source = "headlines"
+
+[nodes.camera]
+type = "camera"
+mode = "scroll"
+speed = 1.0
+
+[nodes.noise]
+type = "effect"
+effect = "noise"
+intensity = 0.3
+
+[nodes.display]
+type = "display"
+backend = "terminal"
+
+[connections]
+list = ["source -> camera -> noise -> display"]
+```
+
+**Hybrid Config (20 lines):**
+```toml
+[pipeline]
+source = "headlines"
+
+camera = { mode = "scroll", speed = 1.0 }
+
+effects = [
+    { name = "noise", intensity = 0.3 }
+]
+
+display = { backend = "terminal" }
+```
+
+## Syntax
+
+### Basic Structure
+
+```toml
+[pipeline]
+source = "headlines"
+camera = { mode = "scroll", speed = 1.0 }
+effects = [
+    { name = "noise", intensity = 0.3 },
+    { name = "fade", intensity = 0.5 }
+]
+display = { backend = "terminal", positioning = "mixed" }
+```
+
+### Configuration Options
+
+#### Source
+```toml
+source = "headlines"  # Built-in source: headlines, poetry, empty, etc.
+```
+
+#### Camera
+```toml
+# Inline object notation
+camera = { mode = "scroll", speed = 1.0 }
+
+# Or shorthand (uses defaults)
+camera = "scroll"
+```
+
+Available modes: `scroll`, `feed`, `horizontal`, `omni`, `floating`, `bounce`, `radial`
+
+#### Effects
+```toml
+# Array of effect configurations
+effects = [
+    { name = "noise", intensity = 0.3 },
+    { name = "fade", intensity = 0.5, enabled = true }
+]
+
+# Or shorthand (uses defaults)
+effects = ["noise", "fade"]
+```
+
+Available effects: `noise`, `fade`, `glitch`, `firehose`, `tint`, `hud`, etc.
+
+#### Display
+```toml
+# Inline object notation
+display = { backend = "terminal", positioning = "mixed" }
+
+# Or shorthand
+display = "terminal"
+```
+
+Available backends: `terminal`, `null`, `websocket`, `pygame`
+
+### Viewport Settings
+```toml
+[pipeline]
+viewport_width = 80
+viewport_height = 24
+```
+
+## Usage Examples
+
+### Minimal Configuration
+```toml
+[pipeline]
+source = "headlines"
+display = "terminal"
+```
+
+### With Camera and Effects
+```toml
+[pipeline]
+source = "headlines"
+camera = { mode = "scroll", speed = 1.0 }
+effects = [
+    { name = "noise", intensity = 0.3 },
+    { name = "fade", intensity = 0.5 }
+]
+display = { backend = "terminal", positioning = "mixed" }
+```
+
+### Full Configuration
+```toml
+[pipeline]
+source = "poetry"
+camera = { mode = "scroll", speed = 1.5 }
+effects = [
+    { name = "noise", intensity = 0.2 },
+    { name = "fade", intensity = 0.4 },
+    { name = "glitch", intensity = 0.3 },
+    { name = "firehose", intensity = 0.5 }
+]
+display = { backend = "terminal", positioning = "mixed" }
+viewport_width = 100
+viewport_height = 30
+```
+
+## Python API
+
+### Loading from TOML File
+```python
+from engine.pipeline.hybrid_config import load_hybrid_config
+
+config = load_hybrid_config("examples/hybrid_config.toml")
+pipeline = config.to_pipeline()
+```
+
+### Creating Config Programmatically
+```python
+from engine.pipeline.hybrid_config import (
+    PipelineConfig,
+    CameraConfig,
+    EffectConfig,
+    DisplayConfig,
+)
+
+config = PipelineConfig(
+    source="headlines",
+    camera=CameraConfig(mode="scroll", speed=1.0),
+    effects=[
+        EffectConfig(name="noise", intensity=0.3),
+        EffectConfig(name="fade", intensity=0.5),
+    ],
+    display=DisplayConfig(backend="terminal", positioning="mixed"),
+)
+
+pipeline = config.to_pipeline(viewport_width=80, viewport_height=24)
+```
+
+### Converting to Graph
+```python
+from engine.pipeline.hybrid_config import PipelineConfig
+
+config = PipelineConfig(source="headlines", display={"backend": "terminal"})
+graph = config.to_graph()  # Returns Graph object for further manipulation
+```
+
+## How It Works
+
+The hybrid config system:
+
+1. **Parses TOML** into a `PipelineConfig` dataclass
+2. **Converts to Graph** internally using automatic linear connections
+3. **Reuses existing adapter** to convert graph to pipeline stages
+4. **Maintains backward compatibility** with verbose node DSL
+
+### Automatic Connection Logic
+
+The system automatically creates linear connections:
+```
+source -> camera -> effects[0] -> effects[1] -> ... -> display
+```
+
+This covers 90% of use cases. For complex DAGs, use the verbose node DSL.
+
+## Migration Guide
+
+### From Presets
+The hybrid format is very similar to presets:
+
+**Preset:**
+```toml
+[presets.custom]
+source = "headlines"
+effects = ["noise", "fade"]
+display = "terminal"
+```
+
+**Hybrid:**
+```toml
+[pipeline]
+source = "headlines"
+effects = ["noise", "fade"]
+display = "terminal"
+```
+
+The main difference is using `[pipeline]` instead of `[presets.custom]`.
+
+### From Verbose Node DSL
+**Old (39 lines):**
+```toml
+[nodes.source] type = "source" source = "headlines"
+[nodes.camera] type = "camera" mode = "scroll"
+[nodes.noise] type = "effect" effect = "noise" intensity = 0.3
+[nodes.display] type = "display" backend = "terminal"
+[connections] list = ["source -> camera -> noise -> display"]
+```
+
+**New (14 lines):**
+```toml
+[pipeline]
+source = "headlines"
+camera = { mode = "scroll" }
+effects = [{ name = "noise", intensity = 0.3 }]
+display = { backend = "terminal" }
+```
+
+## When to Use Each Format
+
+| Format | Use When | Lines (example) |
+|--------|----------|-----------------|
+| **Preset** | Simple configurations, no effect intensity tuning | 10 |
+| **Hybrid** | Most common use cases, need intensity tuning | 20 |
+| **Verbose Node DSL** | Complex DAGs, branching, custom connections | 39 |
+| **Python API** | Dynamic configuration, programmatic generation | N/A |
+
+## Examples
+
+See `examples/hybrid_config.toml` for a complete working example.
+
+Run the demo:
+```bash
+python examples/hybrid_visualization.py
+```

docs/positioning-analysis.md

@@ -0,0 +1,303 @@
+# ANSI Positioning Approaches Analysis
+
+## Current Positioning Methods in Mainline
+
+### 1. Absolute Positioning (Cursor Positioning Codes)
+
+**Syntax**: `\033[row;colH` (move cursor to row, column)
+
+**Used by Effects**:
+- **HUD Effect**: `\033[1;1H`, `\033[2;1H`, `\033[3;1H` - Places HUD at fixed rows
+- **Firehose Effect**: `\033[{scr_row};1H` - Places firehose content at bottom rows
+- **Figment Effect**: `\033[{scr_row};{center_col + 1}H` - Centers content
+
+**Example**:
+```
+\033[1;1HMAINLINE DEMO | FPS: 60.0 | 16.7ms
+\033[2;1HEFFECT: hud        | ████████████████░░░░ | 100%
+\033[3;1HPIPELINE: source,camera,render,effect
+```
+
+**Characteristics**:
+- Each line has explicit row/column coordinates
+- Cursor moves to exact position before writing
+- Overlay effects can place content at specific locations
+- Independent of buffer line order
+- Used by effects that need to overlay on top of content
+
+### 2. Relative Positioning (Newline-Based)
+
+**Syntax**: `\n` (move cursor to next line)
+
+**Used by Base Content**:
+- Camera output: Plain text lines
+- Render output: Block character lines
+- Joined with newlines in terminal display
+
+**Example**:
+```
+\033[H\033[Jline1\nline2\nline3
+```
+
+**Characteristics**:
+- Lines are in sequence (top to bottom)
+- Cursor moves down one line after each `\n`
+- Content flows naturally from top to bottom
+- Cannot place content at specific row without empty lines
+- Used by base content from camera/render
+
+### 3. Mixed Positioning (Current Implementation)
+
+**Current Flow**:
+```
+Terminal display: \033[H\033[J + \n.join(buffer)
+Buffer structure: [line1, line2, \033[1;1HHUD line, ...]
+```
+
+**Behavior**:
+1. `\033[H\033[J` - Move to (1,1), clear screen
+2. `line1\n` - Write line1, move to line2
+3. `line2\n` - Write line2, move to line3
+4. `\033[1;1H` - Move back to (1,1)
+5. Write HUD content
+
+**Issue**: Overlapping cursor movements can cause visual glitches
+
+---
+
+## Performance Analysis
+
+### Absolute Positioning Performance
+
+**Advantages**:
+- Precise control over output position
+- No need for empty buffer lines
+- Effects can overlay without affecting base content
+- Efficient for static overlays (HUD, status bars)
+
+**Disadvantages**:
+- More ANSI codes = larger output size
+- Each line requires `\033[row;colH` prefix
+- Can cause redraw issues if not cleared properly
+- Terminal must parse more escape sequences
+
+**Output Size Comparison** (24 lines):
+- Absolute: ~1,200 bytes (avg 50 chars/line + 30 ANSI codes)
+- Relative: ~960 bytes (80 chars/line * 24 lines)
+
+### Relative Positioning Performance
+
+**Advantages**:
+- Minimal ANSI codes (only colors, no positioning)
+- Smaller output size
+- Terminal renders faster (less parsing)
+- Natural flow for scrolling content
+
+**Disadvantages**:
+- Requires empty lines for spacing
+- Cannot overlay content without buffer manipulation
+- Limited control over exact positioning
+- Harder to implement HUD/status overlays
+
+**Output Size Comparison** (24 lines):
+- Base content: ~1,920 bytes (80 chars * 24 lines)
+- With colors only: ~2,400 bytes (adds color codes)
+
+### Mixed Positioning Performance
+
+**Current Implementation**:
+- Base content uses relative (newlines)
+- Effects use absolute (cursor positioning)
+- Combined output has both methods
+
+**Trade-offs**:
+- Medium output size
+- Flexible positioning
+- Potential visual conflicts if not coordinated
+
+---
+
+## Animation Performance Implications
+
+### Scrolling Animations (Camera Feed/Scroll)
+
+**Best Approach**: Relative positioning with newlines
+- **Why**: Smooth scrolling requires continuous buffer updates
+- **Alternative**: Absolute positioning would require recalculating all coordinates
+
+**Performance**:
+- Relative: 60 FPS achievable with 80x24 buffer
+- Absolute: 55-60 FPS (slightly slower due to more ANSI codes)
+- Mixed: 58-60 FPS (negligible difference for small buffers)
+
+### Static Overlay Animations (HUD, Status Bars)
+
+**Best Approach**: Absolute positioning
+- **Why**: HUD content doesn't change position, only content
+- **Alternative**: Could use fixed buffer positions with relative, but less flexible
+
+**Performance**:
+- Absolute: Minimal overhead (3 lines with ANSI codes)
+- Relative: Requires maintaining fixed positions in buffer (more complex)
+
+### Particle/Effect Animations (Firehose, Figment)
+
+**Best Approach**: Mixed positioning
+- **Why**: Base content flows normally, particles overlay at specific positions
+- **Alternative**: All absolute would be overkill
+
+**Performance**:
+- Mixed: Optimal balance
+- Particles at bottom: `\033[{row};1H` (only affected lines)
+- Base content: `\n` (natural flow)
+
+---
+
+## Proposed Design: PositionStage
+
+### Capability Definition
+
+```python
+class PositioningMode(Enum):
+    """Positioning mode for terminal rendering."""
+    ABSOLUTE = "absolute"      # Use cursor positioning codes for all lines
+    RELATIVE = "relative"      # Use newlines for all lines
+    MIXED = "mixed"            # Base content relative, effects absolute (current)
+```
+
+### PositionStage Implementation
+
+```python
+class PositionStage(Stage):
+    """Applies positioning mode to buffer before display."""
+    
+    def __init__(self, mode: PositioningMode = PositioningMode.RELATIVE):
+        self.mode = mode
+        self.name = f"position-{mode.value}"
+        self.category = "position"
+    
+    @property
+    def capabilities(self) -> set[str]:
+        return {"position.output"}
+    
+    @property
+    def dependencies(self) -> set[str]:
+        return {"render.output"}  # Needs content before positioning
+    
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        if self.mode == PositioningMode.ABSOLUTE:
+            return self._to_absolute(data, ctx)
+        elif self.mode == PositioningMode.RELATIVE:
+            return self._to_relative(data, ctx)
+        else:  # MIXED
+            return data  # No transformation needed
+    
+    def _to_absolute(self, data: list[str], ctx: PipelineContext) -> list[str]:
+        """Convert buffer to absolute positioning (all lines have cursor codes)."""
+        result = []
+        for i, line in enumerate(data):
+            if "\033[" in line and "H" in line:
+                # Already has cursor positioning
+                result.append(line)
+            else:
+                # Add cursor positioning for this line
+                result.append(f"\033[{i + 1};1H{line}")
+        return result
+    
+    def _to_relative(self, data: list[str], ctx: PipelineContext) -> list[str]:
+        """Convert buffer to relative positioning (use newlines)."""
+        # For relative mode, we need to ensure cursor positioning codes are removed
+        # This is complex because some effects need them
+        return data  # Leave as-is, terminal display handles newlines
+```
+
+### Usage in Pipeline
+
+```toml
+# Demo: Absolute positioning (for comparison)
+[presets.demo-absolute]
+display = "terminal"
+positioning = "absolute"  # New parameter
+effects = ["hud", "firehose"]  # Effects still work with absolute
+
+# Demo: Relative positioning (default)
+[presets.demo-relative]
+display = "terminal"
+positioning = "relative"  # New parameter
+effects = ["hud", "firehose"]  # Effects must adapt
+```
+
+### Terminal Display Integration
+
+```python
+def show(self, buffer: list[str], border: bool = False, mode: PositioningMode = None) -> None:
+    # Apply border if requested
+    if border and border != BorderMode.OFF:
+        buffer = render_border(buffer, self.width, self.height, fps, frame_time)
+    
+    # Apply positioning based on mode
+    if mode == PositioningMode.ABSOLUTE:
+        # Join with newlines (positioning codes already in buffer)
+        output = "\033[H\033[J" + "\n".join(buffer)
+    elif mode == PositioningMode.RELATIVE:
+        # Join with newlines
+        output = "\033[H\033,J" + "\n".join(buffer)
+    else:  # MIXED
+        # Current implementation
+        output = "\033[H\033[J" + "\n".join(buffer)
+    
+    sys.stdout.buffer.write(output.encode())
+    sys.stdout.flush()
+```
+
+---
+
+## Recommendations
+
+### For Different Animation Types
+
+1. **Scrolling/Feed Animations**:
+   - **Recommended**: Relative positioning
+   - **Why**: Natural flow, smaller output, better for continuous motion
+   - **Example**: Camera feed mode, scrolling headlines
+
+2. **Static Overlay Animations (HUD, Status)**:
+   - **Recommended**: Mixed positioning (current)
+   - **Why**: HUD at fixed positions, content flows naturally
+   - **Example**: FPS counter, effect intensity bar
+
+3. **Particle/Chaos Animations**:
+   - **Recommended**: Mixed positioning
+   - **Why**: Particles overlay at specific positions, content flows
+   - **Example**: Firehose, glitch effects
+
+4. **Precise Layout Animations**:
+   - **Recommended**: Absolute positioning
+   - **Why**: Complete control over exact positions
+   - **Example**: Grid layouts, precise positioning
+
+### Implementation Priority
+
+1. **Phase 1**: Document current behavior (done)
+2. **Phase 2**: Create PositionStage with configurable mode
+3. **Phase 3**: Update terminal display to respect positioning mode
+4. **Phase 4**: Create presets for different positioning modes
+5. **Phase 5**: Performance testing and optimization
+
+### Key Considerations
+
+- **Backward Compatibility**: Keep mixed positioning as default
+- **Performance**: Relative is ~20% faster for large buffers
+- **Flexibility**: Absolute allows precise control but increases output size
+- **Simplicity**: Mixed provides best balance for typical use cases
+
+---
+
+## Next Steps
+
+1. Implement `PositioningMode` enum
+2. Create `PositionStage` class with mode configuration
+3. Update terminal display to accept positioning mode parameter
+4. Create test presets for each positioning mode
+5. Performance benchmark each approach
+6. Document best practices for choosing positioning mode

docs/presets-usage.md

@@ -0,0 +1,219 @@
+# Presets Usage Guide
+
+## Overview
+
+The sideline branch introduces a new preset system that allows you to easily configure different pipeline behaviors. This guide explains the available presets and how to use them.
+
+## Available Presets
+
+### 1. upstream-default
+
+**Purpose:** Matches the default upstream Mainline operation for comparison.
+
+**Configuration:**
+- **Display:** Terminal (not pygame)
+- **Camera:** Scroll mode
+- **Effects:** noise, fade, glitch, firehose (classic four effects)
+- **Positioning:** Mixed mode
+- **Message Overlay:** Disabled (matches upstream)
+
+**Usage:**
+```bash
+python -m mainline --preset upstream-default --display terminal
+```
+
+**Best for:**
+- Comparing sideline vs upstream behavior
+- Legacy terminal-based operation
+- Baseline performance testing
+
+### 2. demo
+
+**Purpose:** Showcases sideline features including hotswappable effects and sensors.
+
+**Configuration:**
+- **Display:** Pygame (graphical display)
+- **Camera:** Scroll mode
+- **Effects:** noise, fade, glitch, firehose, hud (with visual feedback)
+- **Positioning:** Mixed mode
+- **Message Overlay:** Enabled (with ntfy integration)
+
+**Features:**
+- **Hotswappable Effects:** Effects can be toggled and modified at runtime
+- **LFO Sensor Modulation:** Oscillator sensor provides smooth intensity modulation
+- **Visual Feedback:** HUD effect shows current effect state and pipeline info
+- **Mixed Positioning:** Optimal balance of performance and control
+
+**Usage:**
+```bash
+python -m mainline --preset demo --display pygame
+```
+
+**Best for:**
+- Exploring sideline capabilities
+- Testing effect hotswapping
+- Demonstrating sensor integration
+
+### 3. demo-base / demo-pygame
+
+**Purpose:** Base presets for custom effect hotswapping experiments.
+
+**Configuration:**
+- **Display:** Terminal (base) or Pygame (pygame variant)
+- **Camera:** Feed mode
+- **Effects:** Empty (add your own)
+- **Positioning:** Mixed mode
+
+**Usage:**
+```bash
+python -m mainline --preset demo-pygame --display pygame
+```
+
+### 4. Other Presets
+
+- `poetry`: Poetry feed with subtle effects
+- `firehose`: High-speed firehose mode
+- `ui`: Interactive UI mode with control panel
+- `fixture`: Uses cached headline fixtures
+- `websocket`: WebSocket display mode
+
+## Positioning Modes
+
+The `--positioning` flag controls how text is positioned in the terminal:
+
+```bash
+# Relative positioning (newlines, good for scrolling)
+python -m mainline --positioning relative --preset demo
+
+# Absolute positioning (cursor codes, good for overlays)
+python -m mainline --positioning absolute --preset demo
+
+# Mixed positioning (default, optimal balance)
+python -m mainline --positioning mixed --preset demo
+```
+
+## Pipeline Stages
+
+### Upstream-Default Pipeline
+
+1. **Source Stage:** Headlines data source
+2. **Viewport Filter:** Filters items to viewport height
+3. **Font Stage:** Renders headlines as block characters
+4. **Camera Stages:** Scrolling animation
+5. **Effect Stages:** noise, fade, glitch, firehose
+6. **Display Stage:** Terminal output
+
+### Demo Pipeline
+
+1. **Source Stage:** Headlines data source
+2. **Viewport Filter:** Filters items to viewport height
+3. **Font Stage:** Renders headlines as block characters
+4. **Camera Stages:** Scrolling animation
+5. **Effect Stages:** noise, fade, glitch, firehose, hud
+6. **Message Overlay:** Ntfy message integration
+7. **Display Stage:** Pygame output
+
+## Command-Line Examples
+
+### Basic Usage
+
+```bash
+# Run upstream-default preset
+python -m mainline --preset upstream-default --display terminal
+
+# Run demo preset
+python -m mainline --preset demo --display pygame
+
+# Run with custom positioning
+python -m mainline --preset demo --display pygame --positioning absolute
+```
+
+### Comparison Testing
+
+```bash
+# Capture upstream output
+python -m mainline --preset upstream-default --display null --viewport 80x24
+
+# Capture sideline output
+python -m mainline --preset demo --display null --viewport 80x24
+```
+
+### Hotswapping Effects
+
+The demo preset supports hotswapping effects at runtime:
+- Use the WebSocket display to send commands
+- Toggle effects on/off
+- Adjust intensity values in real-time
+
+## Configuration Files
+
+### Built-in Presets
+
+Location: `engine/pipeline/presets.py` (Python code)
+
+### User Presets
+
+Location: `~/.config/mainline/presets.toml` or `./presets.toml`
+
+Example user preset:
+```toml
+[presets.my-custom-preset]
+description = "My custom configuration"
+source = "headlines"
+display = "terminal"
+camera = "scroll"
+effects = ["noise", "fade"]
+positioning = "mixed"
+viewport_width = 100
+viewport_height = 30
+```
+
+## Sensor Configuration
+
+### Oscillator Sensor (LFO)
+
+The oscillator sensor provides Low Frequency Oscillator modulation:
+
+```toml
+[sensors.oscillator]
+enabled = true
+waveform = "sine"  # sine, square, triangle, sawtooth
+frequency = 0.05   # 20 second cycle (gentle)
+amplitude = 0.5    # 50% modulation
+```
+
+### Effect Configuration
+
+Effect intensities can be configured with initial values:
+
+```toml
+[effect_configs.noise]
+enabled = true
+intensity = 1.0
+
+[effect_configs.fade]
+enabled = true
+intensity = 1.0
+
+[effect_configs.glitch]
+enabled = true
+intensity = 0.5
+```
+
+## Troubleshooting
+
+### No Display Output
+
+- Check if display backend is available (pygame, terminal, etc.)
+- Use `--display null` for headless testing
+
+### Effects Not Modulating
+
+- Ensure sensor is enabled in presets.toml
+- Check effect intensity values in configuration
+
+### Performance Issues
+
+- Use `--positioning relative` for large buffers
+- Reduce viewport height for better performance
+- Use null display for testing without rendering

docs/proposals/adr-preset-scripting-language.md

@@ -0,0 +1,217 @@
+# ADR: Preset Scripting Language for Mainline
+
+## Status: Draft
+
+## Context
+
+We need to evaluate whether to add a scripting language for authoring presets in Mainline, replacing or augmenting the current TOML-based preset system. The goals are:
+
+1. **Expressiveness**: More powerful than TOML for describing dynamic, procedural, or dataflow-based presets
+2. **Live coding**: Support hot-reloading of presets during runtime (like TidalCycles or Sonic Pi)
+3. **Testing**: Include assertion language to package tests alongside presets
+4. **Toolchain**: Consider packaging and build processes
+
+### Current State
+
+The current preset system uses TOML files (`presets.toml`) with a simple structure:
+
+```toml
+[presets.demo-base]
+description = "Demo: Base preset for effect hot-swapping"
+source = "headlines"
+display = "terminal"
+camera = "feed"
+effects = []  # Demo script will add/remove effects dynamically
+camera_speed = 0.1
+viewport_width = 80
+viewport_height = 24
+```
+
+This is declarative and static. It cannot express:
+- Conditional logic based on runtime state
+- Dataflow between pipeline stages
+- Procedural generation of stage configurations
+- Assertions or validation of preset behavior
+
+### Problems with TOML
+
+- No way to express dependencies between effects or stages
+- Cannot describe temporal/animated behavior
+- No support for sensor bindings or parametric animations
+- Static configuration cannot adapt to runtime conditions
+- No built-in testing/assertion mechanism
+
+## Approaches
+
+### 1. Visual Dataflow Language (PureData-style)
+
+Inspired by Pure Data (Pd), Max/MSP, and TouchDesigner:
+
+**Pros:**
+- Intuitive for creative coding and live performance
+- Strong model for real-time parameter modulation
+- Matches the "patcher" paradigm already seen in pipeline architecture
+- Rich ecosystem of visual programming tools
+
+**Cons:**
+- Complex to implement from scratch
+- Requires dedicated GUI editor
+- Harder to version control (binary/graph formats)
+- Mermaid diagrams alone aren't sufficient for this
+
+**Tools to explore:**
+- libpd (Pure Data bindings for other languages)
+- Node-based frameworks (node-red, various DSP tools)
+- TouchDesigner-like approaches
+
+### 2. Textual DSL (TidalCycles-style)
+
+Domain-specific language focused on pattern transformation:
+
+**Pros:**
+- Lightweight, fast iteration
+- Easy to version control (text files)
+- Can express complex patterns with minimal syntax
+- Proven in livecoding community
+
+**Cons:**
+- Learning curve for non-programmers
+- Less visual than PureData approach
+
+**Example (hypothetical):**
+```
+preset my-show {
+  source: headlines
+  
+  every 8s {
+    effect noise: intensity = (0.5 <-> 1.0)
+  }
+  
+  on mic.level > 0.7 {
+    effect glitch: intensity += 0.2
+  }
+}
+```
+
+### 3. Embed Existing Language
+
+Embed Lua, Python, or JavaScript:
+
+**Pros:**
+- Full power of general-purpose language
+- Existing tooling, testing frameworks
+- Easy to integrate (many embeddable interpreters)
+
+**Cons:**
+- Security concerns with running user code
+- May be overkill for simple presets
+- Testing/assertion system must be built on top
+
+**Tools:**
+- Lua (lightweight, fast)
+- Python (rich ecosystem, but heavier)
+- QuickJS (small, embeddable JS)
+
+### 4. Hybrid Approach
+
+Visual editor generates textual DSL that compiles to Python:
+
+**Pros:**
+- Best of both worlds
+- Can start with simple DSL and add editor later
+
+**Cons:**
+- More complex initial implementation
+
+## Requirements Analysis
+
+### Must Have
+- [ ] Express pipeline stage configurations (source, effects, camera, display)
+- [ ] Support parameter bindings to sensors
+- [ ] Hot-reloading during runtime
+- [ ] Integration with existing Pipeline architecture
+
+### Should Have
+- [ ] Basic assertion language for testing
+- [ ] Ability to define custom abstractions/modules
+- [ ] Version control friendly (text-based)
+
+### Could Have
+- [ ] Visual node-based editor
+- [ ] Real-time visualization of dataflow
+- [ ] MIDI/OSC support for external controllers
+
+## User Stories (Proposed)
+
+### Spike Stories (Investigation)
+
+**Story 1: Evaluate DSL Parsing Tools**
+> As a developer, I want to understand the available Python DSL parsing libraries (Lark, parsy, pyparsing) so that I can choose the right tool for implementing a preset DSL.
+> 
+> **Acceptance**: Document pros/cons of 3+ parsing libraries with small proof-of-concept experiments
+
+**Story 2: Research Livecoding Languages**
+> As a developer, I want to understand how TidalCycles, Sonic Pi, and PureData handle hot-reloading and pattern generation so that I can apply similar techniques to Mainline.
+> 
+> **Acceptance**: Document key architectural patterns from 2+ livecoding systems
+
+**Story 3: Prototype Textual DSL**
+> As a preset author, I want to write presets in a simple textual DSL that supports basic conditionals and sensor bindings.
+> 
+> **Acceptance**: Create a prototype DSL that can parse a sample preset and convert to PipelineConfig
+
+**Story 4: Investigate Assertion/Testing Approaches**
+> As a quality engineer, I want to include assertions with presets so that preset behavior can be validated automatically.
+> 
+> **Acceptance**: Survey testing patterns in livecoding and propose assertion syntax
+
+### Implementation Stories (Future)
+
+**Story 5: Implement Core DSL Parser**
+> As a preset author, I want to write presets in a textual DSL that supports sensors, conditionals, and parameter bindings.
+> 
+> **Acceptance**: DSL parser handles the core syntax, produces valid PipelineConfig
+
+**Story 6: Hot-Reload System**
+> As a performer, I want to edit preset files and see changes reflected in real-time without restarting.
+> 
+> **Acceptance**: File watcher + pipeline mutation API integration works
+
+**Story 7: Assertion Language**
+> As a preset author, I want to include assertions that validate sensor values or pipeline state.
+> 
+> **Acceptance**: Assertions can run as part of preset execution and report pass/fail
+
+**Story 8: Toolchain/Packaging**
+> As a preset distributor, I want to package presets with dependencies for easy sharing.
+> 
+> **Acceptance**: Can create, build, and install a preset package
+
+## Decision
+
+**Recommend: Start with textual DSL approach (Option 2/4)**
+
+Rationale:
+- Lowest barrier to entry (text files, version control)
+- Can evolve to hybrid later if visual editor is needed
+- Strong precedents in livecoding community (TidalCycles, Sonic Pi)
+- Enables hot-reloading naturally
+- Assertion language can be part of the DSL syntax
+
+**Not recommending Mermaid**: Mermaid is excellent for documentation and visualization, but it's a diagramming tool, not a programming language. It cannot express the logic, conditionals, and sensor bindings we need.
+
+## Next Steps
+
+1. Execute Spike Stories 1-4 to reduce uncertainty
+2. Create minimal viable DSL syntax
+3. Prototype hot-reloading with existing preset system
+4. Evaluate whether visual editor adds sufficient value to warrant complexity
+
+## References
+
+- Pure Data: https://puredata.info/
+- TidalCycles: https://tidalcycles.org/
+- Sonic Pi: https://sonic-pi.net/
+- Lark parser: https://lark-parser.readthedocs.io/
+- Mainline Pipeline Architecture: `engine/pipeline/`
+- Current Presets: `presets.toml`

engine/__init__.py

@@ -0,0 +1,10 @@
+# engine — modular internals for mainline
+
+# Import submodules to make them accessible via engine.<name>
+# This is required for unittest.mock.patch to work with "engine.<module>.<function>"
+# strings and for direct attribute access on the engine package.
+import engine.config  # noqa: F401
+import engine.fetch  # noqa: F401
+import engine.filter  # noqa: F401
+import engine.sources  # noqa: F401
+import engine.terminal  # noqa: F401

engine/app.py

@@ -0,0 +1,14 @@
+"""
+Application orchestrator — pipeline mode entry point.
+
+This module provides the main entry point for the application.
+The implementation has been refactored into the engine.app package.
+"""
+
+# Re-export from the new package structure
+from engine.app import main, run_pipeline_mode, run_pipeline_mode_direct
+
+__all__ = ["main", "run_pipeline_mode", "run_pipeline_mode_direct"]
+
+if __name__ == "__main__":
+    main()

engine/app/__init__.py

@@ -0,0 +1,34 @@
+"""
+Application orchestrator — pipeline mode entry point.
+
+This package contains the main application logic for the pipeline mode,
+including pipeline construction, UI controller setup, and the main render loop.
+"""
+
+# Re-export from engine for backward compatibility with tests
+# Re-export effects plugins for backward compatibility with tests
+import engine.effects.plugins as effects_plugins
+from engine import config
+
+# Re-export display registry for backward compatibility with tests
+from engine.display import DisplayRegistry
+
+# Re-export fetch functions for backward compatibility with tests
+from engine.fetch import fetch_all, fetch_poetry, load_cache
+from engine.pipeline import list_presets
+
+from .main import main, run_pipeline_mode_direct
+from .pipeline_runner import run_pipeline_mode
+
+__all__ = [
+    "config",
+    "list_presets",
+    "main",
+    "run_pipeline_mode",
+    "run_pipeline_mode_direct",
+    "fetch_all",
+    "fetch_poetry",
+    "load_cache",
+    "DisplayRegistry",
+    "effects_plugins",
+]

engine/app/main.py

@@ -0,0 +1,595 @@
+"""
+Main entry point and CLI argument parsing for the application.
+"""
+
+import sys
+import time
+
+from engine import config
+from engine.display import BorderMode, DisplayRegistry
+from engine.effects import get_registry
+from engine.fetch import fetch_all, fetch_all_fast, fetch_poetry, load_cache, save_cache
+from engine.pipeline import (
+    Pipeline,
+    PipelineConfig,
+    PipelineContext,
+    list_presets,
+)
+from engine.pipeline.adapters import (
+    CameraStage,
+    DataSourceStage,
+    EffectPluginStage,
+    create_stage_from_display,
+    create_stage_from_effect,
+)
+from engine.pipeline.params import PipelineParams
+from engine.pipeline.ui import UIConfig, UIPanel
+from engine.pipeline.validation import validate_pipeline_config
+
+try:
+    from engine.display.backends.websocket import WebSocketDisplay
+except ImportError:
+    WebSocketDisplay = None
+
+from .pipeline_runner import run_pipeline_mode
+
+
+def _handle_pipeline_mutation(pipeline: Pipeline, command: dict) -> bool:
+    """Handle pipeline mutation commands from REPL or other external control.
+
+    Args:
+        pipeline: The pipeline to mutate
+        command: Command dictionary with 'action' and other parameters
+
+    Returns:
+        True if command was successfully handled, False otherwise
+    """
+    action = command.get("action")
+
+    if action == "add_stage":
+        print(f"  [Pipeline] add_stage command received: {command}")
+        return True
+
+    elif action == "remove_stage":
+        stage_name = command.get("stage")
+        if stage_name:
+            result = pipeline.remove_stage(stage_name)
+            print(f"  [Pipeline] Removed stage '{stage_name}': {result is not None}")
+            return result is not None
+
+    elif action == "replace_stage":
+        stage_name = command.get("stage")
+        print(f"  [Pipeline] replace_stage command received: {command}")
+        return True
+
+    elif action == "swap_stages":
+        stage1 = command.get("stage1")
+        stage2 = command.get("stage2")
+        if stage1 and stage2:
+            result = pipeline.swap_stages(stage1, stage2)
+            print(f"  [Pipeline] Swapped stages '{stage1}' and '{stage2}': {result}")
+            return result
+
+    elif action == "move_stage":
+        stage_name = command.get("stage")
+        after = command.get("after")
+        before = command.get("before")
+        if stage_name:
+            result = pipeline.move_stage(stage_name, after, before)
+            print(f"  [Pipeline] Moved stage '{stage_name}': {result}")
+            return result
+
+    elif action == "enable_stage":
+        stage_name = command.get("stage")
+        if stage_name:
+            result = pipeline.enable_stage(stage_name)
+            print(f"  [Pipeline] Enabled stage '{stage_name}': {result}")
+            return result
+
+    elif action == "disable_stage":
+        stage_name = command.get("stage")
+        if stage_name:
+            result = pipeline.disable_stage(stage_name)
+            print(f"  [Pipeline] Disabled stage '{stage_name}': {result}")
+            return result
+
+    elif action == "cleanup_stage":
+        stage_name = command.get("stage")
+        if stage_name:
+            pipeline.cleanup_stage(stage_name)
+            print(f"  [Pipeline] Cleaned up stage '{stage_name}'")
+            return True
+
+    elif action == "can_hot_swap":
+        stage_name = command.get("stage")
+        if stage_name:
+            can_swap = pipeline.can_hot_swap(stage_name)
+            print(f"  [Pipeline] Can hot-swap '{stage_name}': {can_swap}")
+            return True
+
+    return False
+
+
+def main():
+    """Main entry point - all modes now use presets or CLI construction."""
+    if config.PIPELINE_DIAGRAM:
+        try:
+            from engine.pipeline import generate_pipeline_diagram
+        except ImportError:
+            print("Error: pipeline diagram not available")
+            return
+        print(generate_pipeline_diagram())
+        return
+
+    # Check for direct pipeline construction flags
+    if "--pipeline-source" in sys.argv:
+        # Construct pipeline directly from CLI args
+        run_pipeline_mode_direct()
+        return
+
+    preset_name = None
+
+    if config.PRESET:
+        preset_name = config.PRESET
+    elif config.PIPELINE_MODE:
+        preset_name = config.PIPELINE_PRESET
+    else:
+        preset_name = "demo"
+
+    available = list_presets()
+    if preset_name not in available:
+        print(f"Error: Unknown preset '{preset_name}'")
+        print(f"Available presets: {', '.join(available)}")
+        sys.exit(1)
+
+    run_pipeline_mode(preset_name)
+
+
+def run_pipeline_mode_direct():
+    """Construct and run a pipeline directly from CLI arguments.
+
+    Usage:
+        python -m engine.app --pipeline-source headlines --pipeline-effects noise,fade --display null
+        python -m engine.app --pipeline-source fixture --pipeline-effects glitch --pipeline-ui --display null
+
+    Flags:
+        --pipeline-source <source>: Headlines, fixture, poetry, empty, pipeline-inspect
+        --pipeline-effects <effects>: Comma-separated list (noise, fade, glitch, firehose, hud, tint, border, crop)
+        --pipeline-camera <type>: scroll, feed, horizontal, omni, floating, bounce
+        --pipeline-display <display>: terminal, pygame, websocket, null, multi:term,pygame
+        --pipeline-ui: Enable UI panel (BorderMode.UI)
+        --pipeline-border <mode>: off, simple, ui
+    """
+    import engine.effects.plugins as effects_plugins
+    from engine.camera import Camera
+    from engine.data_sources.pipeline_introspection import PipelineIntrospectionSource
+    from engine.data_sources.sources import EmptyDataSource, ListDataSource
+    from engine.pipeline.adapters import (
+        FontStage,
+        ViewportFilterStage,
+    )
+
+    # Discover and register all effect plugins
+    effects_plugins.discover_plugins()
+
+    # Parse CLI arguments
+    source_name = None
+    effect_names = []
+    camera_type = None
+    display_name = None
+    ui_enabled = False
+    border_mode = BorderMode.OFF
+    source_items = None
+    allow_unsafe = False
+    viewport_width = None
+    viewport_height = None
+
+    i = 1
+    argv = sys.argv
+    while i < len(argv):
+        arg = argv[i]
+        if arg == "--pipeline-source" and i + 1 < len(argv):
+            source_name = argv[i + 1]
+            i += 2
+        elif arg == "--pipeline-effects" and i + 1 < len(argv):
+            effect_names = [e.strip() for e in argv[i + 1].split(",") if e.strip()]
+            i += 2
+        elif arg == "--pipeline-camera" and i + 1 < len(argv):
+            camera_type = argv[i + 1]
+            i += 2
+        elif arg == "--viewport" and i + 1 < len(argv):
+            vp = argv[i + 1]
+            try:
+                viewport_width, viewport_height = map(int, vp.split("x"))
+            except ValueError:
+                print("Error: Invalid viewport format. Use WxH (e.g., 40x15)")
+                sys.exit(1)
+            i += 2
+        elif arg == "--pipeline-display" and i + 1 < len(argv):
+            display_name = argv[i + 1]
+            i += 2
+        elif arg == "--pipeline-ui":
+            ui_enabled = True
+            i += 1
+        elif arg == "--pipeline-border" and i + 1 < len(argv):
+            mode = argv[i + 1]
+            if mode == "simple":
+                border_mode = True
+            elif mode == "ui":
+                border_mode = BorderMode.UI
+            else:
+                border_mode = False
+            i += 2
+        elif arg == "--allow-unsafe":
+            allow_unsafe = True
+            i += 1
+        else:
+            i += 1
+
+    if not source_name:
+        print("Error: --pipeline-source is required")
+        print(
+            "Usage: python -m engine.app --pipeline-source <source> [--pipeline-effects <effects>] ..."
+        )
+        sys.exit(1)
+
+    print("  \033[38;5;245mDirect pipeline construction\033[0m")
+    print(f"  Source: {source_name}")
+    print(f"  Effects: {effect_names}")
+    print(f"  Camera: {camera_type}")
+    print(f"  Display: {display_name}")
+    print(f"  UI Enabled: {ui_enabled}")
+
+    # Create initial config and params
+    params = PipelineParams()
+    params.source = source_name
+    params.camera_mode = camera_type if camera_type is not None else ""
+    params.effect_order = effect_names
+    params.border = border_mode
+
+    # Create minimal config for validation
+    config_obj = PipelineConfig(
+        source=source_name,
+        display=display_name or "",  # Will be filled by validation
+        camera=camera_type if camera_type is not None else "",
+        effects=effect_names,
+    )
+
+    # Run MVP validation
+    result = validate_pipeline_config(config_obj, params, allow_unsafe=allow_unsafe)
+
+    if result.warnings and not allow_unsafe:
+        print("  \033[38;5;226mWarning: MVP validation found issues:\033[0m")
+        for warning in result.warnings:
+            print(f"    - {warning}")
+
+    if result.changes:
+        print("  \033[38;5;226mApplied MVP defaults:\033[0m")
+        for change in result.changes:
+            print(f"    {change}")
+
+    if not result.valid:
+        print(
+            "  \033[38;5;196mPipeline configuration invalid and could not be fixed\033[0m"
+        )
+        sys.exit(1)
+
+    # Show MVP summary
+    print("  \033[38;5;245mMVP Configuration:\033[0m")
+    print(f"    Source: {result.config.source}")
+    print(f"    Display: {result.config.display}")
+    print(f"    Camera: {result.config.camera or 'static (none)'}")
+    print(f"    Effects: {result.config.effects if result.config.effects else 'none'}")
+    print(f"    Border: {result.params.border}")
+
+    # Load source items
+    if source_name == "headlines":
+        cached = load_cache()
+        if cached:
+            source_items = cached
+        else:
+            source_items = fetch_all_fast()
+            if source_items:
+                import threading
+
+                def background_fetch():
+                    full_items, _, _ = fetch_all()
+                    save_cache(full_items)
+
+                background_thread = threading.Thread(
+                    target=background_fetch, daemon=True
+                )
+                background_thread.start()
+    elif source_name == "fixture":
+        source_items = load_cache()
+        if not source_items:
+            print("  \033[38;5;196mNo fixture cache available\033[0m")
+            sys.exit(1)
+    elif source_name == "poetry":
+        source_items, _, _ = fetch_poetry()
+    elif source_name == "empty" or source_name == "pipeline-inspect":
+        source_items = []
+    else:
+        print(f"  \033[38;5;196mUnknown source: {source_name}\033[0m")
+        sys.exit(1)
+
+    if source_items is not None:
+        print(f"  \033[38;5;82mLoaded {len(source_items)} items\033[0m")
+
+    # Set border mode
+    if ui_enabled:
+        border_mode = BorderMode.UI
+
+    # Build pipeline using validated config and params
+    params = result.params
+    params.viewport_width = viewport_width if viewport_width is not None else 80
+    params.viewport_height = viewport_height if viewport_height is not None else 24
+
+    ctx = PipelineContext()
+    ctx.params = params
+
+    # Create display using validated display name
+    display_name = result.config.display or "terminal"  # Default to terminal if empty
+
+    # Warn if display was auto-selected (not explicitly specified)
+    if not display_name:
+        print(
+            "  \033[38;5;226mWarning: No --pipeline-display specified, using default: terminal\033[0m"
+        )
+        print(
+            "  \033[38;5;245mTip: Use --pipeline-display null for headless mode (useful for testing)\033[0m"
+        )
+
+    display = DisplayRegistry.create(display_name)
+
+    # Set positioning mode
+    if "--positioning" in sys.argv:
+        idx = sys.argv.index("--positioning")
+        if idx + 1 < len(sys.argv):
+            params.positioning = sys.argv[idx + 1]
+    if not display:
+        print(f"  \033[38;5;196mFailed to create display: {display_name}\033[0m")
+        sys.exit(1)
+    display.init(0, 0)
+
+    # Create pipeline using validated config
+    pipeline = Pipeline(config=result.config, context=ctx)
+
+    # Add stages
+    # Source stage
+    if source_name == "pipeline-inspect":
+        introspection_source = PipelineIntrospectionSource(
+            pipeline=None,
+            viewport_width=params.viewport_width,
+            viewport_height=params.viewport_height,
+        )
+        pipeline.add_stage(
+            "source", DataSourceStage(introspection_source, name="pipeline-inspect")
+        )
+    elif source_name == "empty":
+        empty_source = EmptyDataSource(
+            width=params.viewport_width, height=params.viewport_height
+        )
+        pipeline.add_stage("source", DataSourceStage(empty_source, name="empty"))
+    else:
+        list_source = ListDataSource(source_items, name=source_name)
+        pipeline.add_stage("source", DataSourceStage(list_source, name=source_name))
+
+    # Add viewport filter and font for headline sources
+    if source_name in ["headlines", "poetry", "fixture"]:
+        pipeline.add_stage(
+            "viewport_filter", ViewportFilterStage(name="viewport-filter")
+        )
+        pipeline.add_stage("font", FontStage(name="font"))
+    else:
+        # Fallback to simple conversion for other sources
+        from engine.pipeline.adapters import SourceItemsToBufferStage
+
+        pipeline.add_stage("render", SourceItemsToBufferStage(name="items-to-buffer"))
+
+    # Add camera
+    speed = getattr(params, "camera_speed", 1.0)
+    camera = None
+    if camera_type == "feed":
+        camera = Camera.feed(speed=speed)
+    elif camera_type == "scroll":
+        camera = Camera.scroll(speed=speed)
+    elif camera_type == "horizontal":
+        camera = Camera.horizontal(speed=speed)
+    elif camera_type == "omni":
+        camera = Camera.omni(speed=speed)
+    elif camera_type == "floating":
+        camera = Camera.floating(speed=speed)
+    elif camera_type == "bounce":
+        camera = Camera.bounce(speed=speed)
+
+    if camera:
+        pipeline.add_stage("camera", CameraStage(camera, name=camera_type))
+
+    # Add effects
+    effect_registry = get_registry()
+    for effect_name in effect_names:
+        effect = effect_registry.get(effect_name)
+        if effect:
+            pipeline.add_stage(
+                f"effect_{effect_name}", create_stage_from_effect(effect, effect_name)
+            )
+
+    # Add display
+    pipeline.add_stage("display", create_stage_from_display(display, display_name))
+
+    pipeline.build()
+
+    if not pipeline.initialize():
+        print("  \033[38;5;196mFailed to initialize pipeline\033[0m")
+        sys.exit(1)
+
+    # Create UI panel if border mode is UI
+    ui_panel = None
+    if params.border == BorderMode.UI:
+        ui_panel = UIPanel(UIConfig(panel_width=24, start_with_preset_picker=True))
+        # Enable raw mode for terminal input if supported
+        if hasattr(display, "set_raw_mode"):
+            display.set_raw_mode(True)
+        for stage in pipeline.stages.values():
+            if isinstance(stage, EffectPluginStage):
+                effect = stage._effect
+                enabled = effect.config.enabled if hasattr(effect, "config") else True
+                stage_control = ui_panel.register_stage(stage, enabled=enabled)
+                stage_control.effect = effect  # type: ignore[attr-defined]
+
+        if ui_panel.stages:
+            first_stage = next(iter(ui_panel.stages))
+            ui_panel.select_stage(first_stage)
+            ctrl = ui_panel.stages[first_stage]
+            if hasattr(ctrl, "effect"):
+                effect = ctrl.effect
+                if hasattr(effect, "config"):
+                    config = effect.config
+                    try:
+                        import dataclasses
+
+                        if dataclasses.is_dataclass(config):
+                            for field_name, field_obj in dataclasses.fields(config):
+                                if field_name == "enabled":
+                                    continue
+                                value = getattr(config, field_name, None)
+                                if value is not None:
+                                    ctrl.params[field_name] = value
+                                    ctrl.param_schema[field_name] = {
+                                        "type": type(value).__name__,
+                                        "min": 0
+                                        if isinstance(value, (int, float))
+                                        else None,
+                                        "max": 1 if isinstance(value, float) else None,
+                                        "step": 0.1 if isinstance(value, float) else 1,
+                                    }
+                    except Exception:
+                        pass
+
+    # Check for REPL effect in pipeline
+    repl_effect = None
+    for stage in pipeline.stages.values():
+        if isinstance(stage, EffectPluginStage) and stage._effect.name == "repl":
+            repl_effect = stage._effect
+            print(
+                "  \033[38;5;46mREPL effect detected - Interactive mode enabled\033[0m"
+            )
+            break
+
+    # Enable raw mode for REPL if present and not already enabled
+    # Also enable for UI border mode (already handled above)
+    if repl_effect and ui_panel is None and hasattr(display, "set_raw_mode"):
+        display.set_raw_mode(True)
+
+    # Run pipeline loop
+    from engine.display import render_ui_panel
+
+    ctx.set("display", display)
+    ctx.set("items", source_items)
+    ctx.set("pipeline", pipeline)
+    ctx.set("pipeline_order", pipeline.execution_order)
+
+    current_width = params.viewport_width
+    current_height = params.viewport_height
+
+    # Only get dimensions from display if viewport wasn't explicitly set
+    if "--viewport" not in sys.argv and hasattr(display, "get_dimensions"):
+        current_width, current_height = display.get_dimensions()
+        params.viewport_width = current_width
+        params.viewport_height = current_height
+
+    print("  \033[38;5;82mStarting pipeline...\033[0m")
+    print("  \033[38;5;245mPress Ctrl+C to exit\033[0m\n")
+
+    try:
+        frame = 0
+        while True:
+            params.frame_number = frame
+            ctx.params = params
+
+            result = pipeline.execute(source_items)
+            if not result.success:
+                error_msg = f" ({result.error})" if result.error else ""
+                print(f"  \033[38;5;196mPipeline execution failed{error_msg}\033[0m")
+                break
+
+            # Render with UI panel
+            if ui_panel is not None:
+                buf = render_ui_panel(
+                    result.data, current_width, current_height, ui_panel
+                )
+                display.show(buf, border=False)
+            else:
+                display.show(result.data, border=border_mode)
+
+            # Handle keyboard events if UI is enabled
+            if ui_panel is not None:
+                # Try pygame first
+                if hasattr(display, "_pygame"):
+                    try:
+                        import pygame
+
+                        for event in pygame.event.get():
+                            if event.type == pygame.KEYDOWN:
+                                ui_panel.process_key_event(event.key, event.mod)
+                    except (ImportError, Exception):
+                        pass
+                # Try terminal input
+                elif hasattr(display, "get_input_keys"):
+                    try:
+                        keys = display.get_input_keys()
+                        for key in keys:
+                            ui_panel.process_key_event(key, 0)
+                    except Exception:
+                        pass
+
+            # --- REPL Input Handling ---
+            if repl_effect and hasattr(display, "get_input_keys"):
+                # Get keyboard input (non-blocking)
+                keys = display.get_input_keys(timeout=0.0)
+
+                for key in keys:
+                    if key == "ctrl_c":
+                        # Request quit when Ctrl+C is pressed
+                        if hasattr(display, "request_quit"):
+                            display.request_quit()
+                        else:
+                            raise KeyboardInterrupt()
+                    elif key == "return":
+                        # Get command string before processing
+                        cmd_str = repl_effect.state.current_command
+                        if cmd_str:
+                            repl_effect.process_command(cmd_str, ctx)
+                            # Check for pending pipeline mutations
+                            pending = repl_effect.get_pending_command()
+                            if pending:
+                                _handle_pipeline_mutation(pipeline, pending)
+                    elif key == "up":
+                        repl_effect.navigate_history(-1)
+                    elif key == "down":
+                        repl_effect.navigate_history(1)
+                    elif key == "backspace":
+                        repl_effect.backspace()
+                    elif len(key) == 1:
+                        repl_effect.append_to_command(key)
+            # --- End REPL Input Handling ---
+
+            # Check for quit request
+            if hasattr(display, "is_quit_requested") and display.is_quit_requested():
+                if hasattr(display, "clear_quit_request"):
+                    display.clear_quit_request()
+                raise KeyboardInterrupt()
+
+            time.sleep(1 / 60)
+            frame += 1
+
+    except KeyboardInterrupt:
+        pipeline.cleanup()
+        display.cleanup()
+        print("\n  \033[38;5;245mPipeline stopped\033[0m")
+        return
+
+    pipeline.cleanup()
+    display.cleanup()
+    print("\n  \033[38;5;245mPipeline stopped\033[0m")

engine/benchmark.py

@@ -0,0 +1,73 @@
+"""
+Benchmark module for performance testing.
+
+Usage:
+    python -m engine.benchmark              # Run all benchmarks
+    python -m engine.benchmark --hook        # Run benchmarks in hook mode (for CI)
+    python -m engine.benchmark --displays null --iterations 20
+"""
+
+import argparse
+import sys
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Run performance benchmarks")
+    parser.add_argument(
+        "--hook",
+        action="store_true",
+        help="Run in hook mode (fail on regression)",
+    )
+    parser.add_argument(
+        "--displays",
+        default="null",
+        help="Comma-separated list of displays to benchmark",
+    )
+    parser.add_argument(
+        "--iterations",
+        type=int,
+        default=100,
+        help="Number of iterations per benchmark",
+    )
+    args = parser.parse_args()
+
+    # Run pytest with benchmark markers
+    pytest_args = [
+        "-v",
+        "-m",
+        "benchmark",
+    ]
+
+    if args.hook:
+        # Hook mode: stricter settings
+        pytest_args.extend(
+            [
+                "--benchmark-only",
+                "--benchmark-compare",
+                "--benchmark-compare-fail=min:5%",  # Fail if >5% slower
+            ]
+        )
+
+    # Add display filter if specified
+    if args.displays:
+        pytest_args.extend(["-k", args.displays])
+
+    # Add iterations
+    if args.iterations:
+        # Set environment variable for benchmark tests
+        import os
+
+        os.environ["BENCHMARK_ITERATIONS"] = str(args.iterations)
+
+    # Run pytest
+    import subprocess
+
+    result = subprocess.run(
+        [sys.executable, "-m", "pytest", "tests/test_benchmark.py"] + pytest_args,
+        cwd=None,  # Current directory
+    )
+    sys.exit(result.returncode)
+
+
+if __name__ == "__main__":
+    main()

engine/camera.py

@@ -0,0 +1,473 @@
+"""
+Camera system for viewport scrolling.
+
+Provides abstraction for camera motion in different modes:
+- Vertical: traditional upward scroll
+- Horizontal: left/right movement
+- Omni: combination of both
+- Floating: sinusoidal/bobbing motion
+
+The camera defines a visible viewport into a larger Canvas.
+"""
+
+import math
+from collections.abc import Callable
+from dataclasses import dataclass, field
+from enum import Enum, auto
+
+
+class CameraMode(Enum):
+    FEED = auto()  # Single item view (static or rapid cycling)
+    SCROLL = auto()  # Smooth vertical scrolling (movie credits style)
+    HORIZONTAL = auto()
+    OMNI = auto()
+    FLOATING = auto()
+    BOUNCE = auto()
+    RADIAL = auto()  # Polar coordinates (r, theta) for radial scanning
+
+
+@dataclass
+class CameraViewport:
+    """Represents the visible viewport."""
+
+    x: int
+    y: int
+    width: int
+    height: int
+
+
+@dataclass
+class Camera:
+    """Camera for viewport scrolling.
+
+    The camera defines a visible viewport into a Canvas.
+    It can be smaller than the canvas to allow scrolling,
+    and supports zoom to scale the view.
+
+    Attributes:
+        x: Current horizontal offset (positive = scroll left)
+        y: Current vertical offset (positive = scroll up)
+        mode: Current camera mode
+        speed: Base scroll speed
+        zoom: Zoom factor (1.0 = 100%, 2.0 = 200% zoom out)
+        canvas_width: Width of the canvas being viewed
+        canvas_height: Height of the canvas being viewed
+        custom_update: Optional custom update function
+    """
+
+    x: int = 0
+    y: int = 0
+    mode: CameraMode = CameraMode.FEED
+    speed: float = 1.0
+    zoom: float = 1.0
+    canvas_width: int = 200  # Larger than viewport for scrolling
+    canvas_height: int = 200
+    custom_update: Callable[["Camera", float], None] | None = None
+    _x_float: float = field(default=0.0, repr=False)
+    _y_float: float = field(default=0.0, repr=False)
+    _time: float = field(default=0.0, repr=False)
+
+    @property
+    def w(self) -> int:
+        """Shorthand for viewport_width."""
+        return self.viewport_width
+
+    def set_speed(self, speed: float) -> None:
+        """Set the camera scroll speed dynamically.
+
+        This allows camera speed to be modulated during runtime
+        via PipelineParams or directly.
+
+        Args:
+            speed: New speed value (0.0 = stopped, >0 = movement)
+        """
+        self.speed = max(0.0, speed)
+
+    @property
+    def h(self) -> int:
+        """Shorthand for viewport_height."""
+        return self.viewport_height
+
+    @property
+    def viewport_width(self) -> int:
+        """Get the visible viewport width.
+
+        This is the canvas width divided by zoom.
+        """
+        return max(1, int(self.canvas_width / self.zoom))
+
+    @property
+    def viewport_height(self) -> int:
+        """Get the visible viewport height.
+
+        This is the canvas height divided by zoom.
+        """
+        return max(1, int(self.canvas_height / self.zoom))
+
+    def get_viewport(self, viewport_height: int | None = None) -> CameraViewport:
+        """Get the current viewport bounds.
+
+        Args:
+            viewport_height: Optional viewport height to use instead of camera's viewport_height
+
+        Returns:
+            CameraViewport with position and size (clamped to canvas bounds)
+        """
+        vw = self.viewport_width
+        vh = viewport_height if viewport_height is not None else self.viewport_height
+
+        clamped_x = max(0, min(self.x, self.canvas_width - vw))
+        clamped_y = max(0, min(self.y, self.canvas_height - vh))
+
+        return CameraViewport(
+            x=clamped_x,
+            y=clamped_y,
+            width=vw,
+            height=vh,
+        )
+
+        return CameraViewport(
+            x=clamped_x,
+            y=clamped_y,
+            width=vw,
+            height=vh,
+        )
+
+    def set_zoom(self, zoom: float) -> None:
+        """Set the zoom factor.
+
+        Args:
+            zoom: Zoom factor (1.0 = 100%, 2.0 = zoomed out 2x, 0.5 = zoomed in 2x)
+        """
+        self.zoom = max(0.1, min(10.0, zoom))
+
+    def update(self, dt: float) -> None:
+        """Update camera position based on mode.
+
+        Args:
+            dt: Delta time in seconds
+        """
+        self._time += dt
+
+        if self.custom_update:
+            self.custom_update(self, dt)
+            return
+
+        if self.mode == CameraMode.FEED:
+            self._update_feed(dt)
+        elif self.mode == CameraMode.SCROLL:
+            self._update_scroll(dt)
+        elif self.mode == CameraMode.HORIZONTAL:
+            self._update_horizontal(dt)
+        elif self.mode == CameraMode.OMNI:
+            self._update_omni(dt)
+        elif self.mode == CameraMode.FLOATING:
+            self._update_floating(dt)
+        elif self.mode == CameraMode.BOUNCE:
+            self._update_bounce(dt)
+        elif self.mode == CameraMode.RADIAL:
+            self._update_radial(dt)
+
+        # Bounce mode handles its own bounds checking
+        if self.mode != CameraMode.BOUNCE:
+            self._clamp_to_bounds()
+
+    def _clamp_to_bounds(self) -> None:
+        """Clamp camera position to stay within canvas bounds.
+
+        Only clamps if the viewport is smaller than the canvas.
+        If viewport equals canvas (no scrolling needed), allows any position
+        for backwards compatibility with original behavior.
+        """
+        vw = self.viewport_width
+        vh = self.viewport_height
+
+        # Only clamp if there's room to scroll
+        if vw < self.canvas_width:
+            self.x = max(0, min(self.x, self.canvas_width - vw))
+        if vh < self.canvas_height:
+            self.y = max(0, min(self.y, self.canvas_height - vh))
+
+    def _update_feed(self, dt: float) -> None:
+        """Feed mode: rapid scrolling (1 row per frame at speed=1.0)."""
+        self.y += int(self.speed * dt * 60)
+
+    def _update_scroll(self, dt: float) -> None:
+        """Scroll mode: smooth vertical scrolling with float accumulation."""
+        self._y_float += self.speed * dt * 60
+        self.y = int(self._y_float)
+
+    def _update_horizontal(self, dt: float) -> None:
+        self.x += int(self.speed * dt * 60)
+
+    def _update_omni(self, dt: float) -> None:
+        speed = self.speed * dt * 60
+        self.y += int(speed)
+        self.x += int(speed * 0.5)
+
+    def _update_floating(self, dt: float) -> None:
+        base = self.speed * 30
+        self.y = int(math.sin(self._time * 2) * base)
+        self.x = int(math.cos(self._time * 1.5) * base * 0.5)
+
+    def _update_bounce(self, dt: float) -> None:
+        """Bouncing DVD-style camera that bounces off canvas edges."""
+        vw = self.viewport_width
+        vh = self.viewport_height
+
+        # Initialize direction if not set
+        if not hasattr(self, "_bounce_dx"):
+            self._bounce_dx = 1
+            self._bounce_dy = 1
+
+        # Calculate max positions
+        max_x = max(0, self.canvas_width - vw)
+        max_y = max(0, self.canvas_height - vh)
+
+        # Move
+        move_speed = self.speed * dt * 60
+
+        # Bounce off edges - reverse direction when hitting bounds
+        self.x += int(move_speed * self._bounce_dx)
+        self.y += int(move_speed * self._bounce_dy)
+
+        # Bounce horizontally
+        if self.x <= 0:
+            self.x = 0
+            self._bounce_dx = 1
+        elif self.x >= max_x:
+            self.x = max_x
+            self._bounce_dx = -1
+
+        # Bounce vertically
+        if self.y <= 0:
+            self.y = 0
+            self._bounce_dy = 1
+        elif self.y >= max_y:
+            self.y = max_y
+            self._bounce_dy = -1
+
+    def _update_radial(self, dt: float) -> None:
+        """Radial camera mode: polar coordinate scrolling (r, theta).
+
+        The camera rotates around the center of the canvas while optionally
+        moving outward/inward along rays. This enables:
+        - Radar sweep animations
+        - Pendulum view oscillation
+        - Spiral scanning motion
+
+        Uses polar coordinates internally:
+        - _r_float: radial distance from center (accumulates smoothly)
+        - _theta_float: angle in radians (accumulates smoothly)
+        - Updates x, y based on conversion from polar to Cartesian
+        """
+        # Initialize radial state if needed
+        if not hasattr(self, "_r_float"):
+            self._r_float = 0.0
+            self._theta_float = 0.0
+
+        # Update angular position (rotation around center)
+        # Speed controls rotation rate
+        theta_speed = self.speed * dt * 1.0  # radians per second
+        self._theta_float += theta_speed
+
+        # Update radial position (inward/outward from center)
+        # Can be modulated by external sensor
+        if hasattr(self, "_radial_input"):
+            r_input = self._radial_input
+        else:
+            # Default: slow outward drift
+            r_input = 0.0
+
+        r_speed = self.speed * dt * 20.0  # pixels per second
+        self._r_float += r_input + r_speed * 0.01
+
+        # Clamp radial position to canvas bounds
+        max_r = min(self.canvas_width, self.canvas_height) / 2
+        self._r_float = max(0.0, min(self._r_float, max_r))
+
+        # Convert polar to Cartesian, centered at canvas center
+        center_x = self.canvas_width / 2
+        center_y = self.canvas_height / 2
+
+        self.x = int(center_x + self._r_float * math.cos(self._theta_float))
+        self.y = int(center_y + self._r_float * math.sin(self._theta_float))
+
+        # Clamp to canvas bounds
+        self._clamp_to_bounds()
+
+    def set_radial_input(self, value: float) -> None:
+        """Set radial input for sensor-driven radius modulation.
+
+        Args:
+            value: Sensor value (0-1) that modulates radial distance
+        """
+        self._radial_input = value * 10.0  # Scale to reasonable pixel range
+
+    def set_radial_angle(self, angle: float) -> None:
+        """Set radial angle directly (for OSC integration).
+
+        Args:
+            angle: Angle in radians (0 to 2π)
+        """
+        self._theta_float = angle
+
+    def reset(self) -> None:
+        """Reset camera position and state."""
+        self.x = 0
+        self.y = 0
+        self._time = 0.0
+        self.zoom = 1.0
+        # Reset bounce direction state
+        if hasattr(self, "_bounce_dx"):
+            self._bounce_dx = 1
+            self._bounce_dy = 1
+        # Reset radial state
+        if hasattr(self, "_r_float"):
+            self._r_float = 0.0
+            self._theta_float = 0.0
+
+    def set_canvas_size(self, width: int, height: int) -> None:
+        """Set the canvas size and clamp position if needed.
+
+        Args:
+            width: New canvas width
+            height: New canvas height
+        """
+        self.canvas_width = width
+        self.canvas_height = height
+        self._clamp_to_bounds()
+
+    def apply(
+        self, buffer: list[str], viewport_width: int, viewport_height: int | None = None
+    ) -> list[str]:
+        """Apply camera viewport to a text buffer.
+
+        Slices the buffer based on camera position (x, y) and viewport dimensions.
+        Handles ANSI escape codes correctly for colored/styled text.
+
+        Args:
+            buffer: List of strings representing lines of text
+            viewport_width: Width of the visible viewport in characters
+            viewport_height: Height of the visible viewport (overrides camera's viewport_height if provided)
+
+        Returns:
+            Sliced buffer containing only the visible lines and columns
+        """
+        from engine.effects.legacy import vis_offset, vis_trunc
+
+        if not buffer:
+            return buffer
+
+        # Get current viewport bounds (clamped to canvas size)
+        viewport = self.get_viewport(viewport_height)
+
+        # Use provided viewport_height if given, otherwise use camera's viewport
+        vh = viewport_height if viewport_height is not None else viewport.height
+
+        # Vertical slice: extract lines that fit in viewport height
+        start_y = viewport.y
+        end_y = min(viewport.y + vh, len(buffer))
+
+        if start_y >= len(buffer):
+            # Scrolled past end of buffer, return empty viewport
+            return [""] * vh
+
+        vertical_slice = buffer[start_y:end_y]
+
+        # Horizontal slice: apply horizontal offset and truncate to width
+        horizontal_slice = []
+        for line in vertical_slice:
+            # Apply horizontal offset (skip first x characters, handling ANSI)
+            offset_line = vis_offset(line, viewport.x)
+            # Truncate to viewport width (handling ANSI)
+            truncated_line = vis_trunc(offset_line, viewport_width)
+
+            # Pad line to full viewport width to prevent ghosting when panning
+            # Skip padding for empty lines to preserve intentional blank lines
+            import re
+
+            visible_len = len(re.sub(r"\x1b\[[0-9;]*m", "", truncated_line))
+            if visible_len < viewport_width and visible_len > 0:
+                truncated_line += " " * (viewport_width - visible_len)
+
+            horizontal_slice.append(truncated_line)
+
+        # Pad with empty lines if needed to fill viewport height
+        while len(horizontal_slice) < vh:
+            horizontal_slice.append("")
+
+        return horizontal_slice
+
+    @classmethod
+    def feed(cls, speed: float = 1.0) -> "Camera":
+        """Create a feed camera (rapid single-item scrolling, 1 row/frame at speed=1.0)."""
+        return cls(mode=CameraMode.FEED, speed=speed, canvas_height=200)
+
+    @classmethod
+    def scroll(cls, speed: float = 0.5) -> "Camera":
+        """Create a smooth scrolling camera (movie credits style).
+
+        Uses float accumulation for sub-integer speeds.
+        Sets canvas_width=0 so it matches viewport_width for proper text wrapping.
+        """
+        return cls(
+            mode=CameraMode.SCROLL, speed=speed, canvas_width=0, canvas_height=200
+        )
+
+    @classmethod
+    def vertical(cls, speed: float = 1.0) -> "Camera":
+        """Deprecated: Use feed() or scroll() instead."""
+        return cls(mode=CameraMode.FEED, speed=speed, canvas_height=200)
+
+    @classmethod
+    def horizontal(cls, speed: float = 1.0) -> "Camera":
+        """Create a horizontal scrolling camera."""
+        return cls(mode=CameraMode.HORIZONTAL, speed=speed, canvas_width=200)
+
+    @classmethod
+    def omni(cls, speed: float = 1.0) -> "Camera":
+        """Create an omnidirectional scrolling camera."""
+        return cls(
+            mode=CameraMode.OMNI, speed=speed, canvas_width=200, canvas_height=200
+        )
+
+    @classmethod
+    def floating(cls, speed: float = 1.0) -> "Camera":
+        """Create a floating/bobbing camera."""
+        return cls(
+            mode=CameraMode.FLOATING, speed=speed, canvas_width=200, canvas_height=200
+        )
+
+    @classmethod
+    def bounce(cls, speed: float = 1.0) -> "Camera":
+        """Create a bouncing DVD-style camera that bounces off canvas edges."""
+        return cls(
+            mode=CameraMode.BOUNCE, speed=speed, canvas_width=200, canvas_height=200
+        )
+
+    @classmethod
+    def radial(cls, speed: float = 1.0) -> "Camera":
+        """Create a radial camera (polar coordinate scanning).
+
+        The camera rotates around the center of the canvas with smooth angular motion.
+        Enables radar sweep, pendulum view, and spiral scanning animations.
+
+        Args:
+            speed: Rotation speed (higher = faster rotation)
+
+        Returns:
+            Camera configured for radial polar coordinate scanning
+        """
+        cam = cls(
+            mode=CameraMode.RADIAL, speed=speed, canvas_width=200, canvas_height=200
+        )
+        # Initialize radial state
+        cam._r_float = 0.0
+        cam._theta_float = 0.0
+        return cam
+
+    @classmethod
+    def custom(cls, update_fn: Callable[["Camera", float], None]) -> "Camera":
+        """Create a camera with custom update function."""
+        return cls(custom_update=update_fn)

engine/canvas.py

@@ -0,0 +1,186 @@
+"""
+Canvas - 2D surface for rendering.
+
+The Canvas represents a full rendered surface that can be larger than the display.
+The Camera then defines the visible viewport into this canvas.
+"""
+
+from dataclasses import dataclass
+
+
+@dataclass
+class CanvasRegion:
+    """A rectangular region on the canvas."""
+
+    x: int
+    y: int
+    width: int
+    height: int
+
+    def is_valid(self) -> bool:
+        """Check if region has positive dimensions."""
+        return self.width > 0 and self.height > 0
+
+    def rows(self) -> set[int]:
+        """Return set of row indices in this region."""
+        return set(range(self.y, self.y + self.height))
+
+
+class Canvas:
+    """2D canvas for rendering content.
+
+    The canvas is a 2D grid of cells that can hold text content.
+    It can be larger than the visible viewport (display).
+
+    Attributes:
+        width: Total width in characters
+        height: Total height in characters
+    """
+
+    def __init__(self, width: int = 80, height: int = 24):
+        self.width = width
+        self.height = height
+        self._grid: list[list[str]] = [
+            [" " for _ in range(width)] for _ in range(height)
+        ]
+        self._dirty_regions: list[CanvasRegion] = []  # Track dirty regions
+
+    def clear(self) -> None:
+        """Clear the entire canvas."""
+        self._grid = [[" " for _ in range(self.width)] for _ in range(self.height)]
+        self._dirty_regions = [CanvasRegion(0, 0, self.width, self.height)]
+
+    def mark_dirty(self, x: int, y: int, width: int, height: int) -> None:
+        """Mark a region as dirty (caller declares what they changed)."""
+        self._dirty_regions.append(CanvasRegion(x, y, width, height))
+
+    def get_dirty_regions(self) -> list[CanvasRegion]:
+        """Get all dirty regions and clear the set."""
+        regions = self._dirty_regions
+        self._dirty_regions = []
+        return regions
+
+    def get_dirty_rows(self) -> set[int]:
+        """Get union of all dirty rows."""
+        rows: set[int] = set()
+        for region in self._dirty_regions:
+            rows.update(region.rows())
+        return rows
+
+    def is_dirty(self) -> bool:
+        """Check if any region is dirty."""
+        return len(self._dirty_regions) > 0
+
+    def get_region(self, x: int, y: int, width: int, height: int) -> list[list[str]]:
+        """Get a rectangular region from the canvas.
+
+        Args:
+            x: Left position
+            y: Top position
+            width: Region width
+            height: Region height
+
+        Returns:
+            2D list of characters (height rows, width columns)
+        """
+        region: list[list[str]] = []
+        for py in range(y, y + height):
+            row: list[str] = []
+            for px in range(x, x + width):
+                if 0 <= py < self.height and 0 <= px < self.width:
+                    row.append(self._grid[py][px])
+                else:
+                    row.append(" ")
+            region.append(row)
+        return region
+
+    def get_region_flat(self, x: int, y: int, width: int, height: int) -> list[str]:
+        """Get a rectangular region as flat list of lines.
+
+        Args:
+            x: Left position
+            y: Top position
+            width: Region width
+            height: Region height
+
+        Returns:
+            List of strings (one per row)
+        """
+        region = self.get_region(x, y, width, height)
+        return ["".join(row) for row in region]
+
+    def put_region(self, x: int, y: int, content: list[list[str]]) -> None:
+        """Put content into a rectangular region on the canvas.
+
+        Args:
+            x: Left position
+            y: Top position
+            content: 2D list of characters to place
+        """
+        height = len(content) if content else 0
+        width = len(content[0]) if height > 0 else 0
+
+        for py, row in enumerate(content):
+            for px, char in enumerate(row):
+                canvas_x = x + px
+                canvas_y = y + py
+                if 0 <= canvas_y < self.height and 0 <= canvas_x < self.width:
+                    self._grid[canvas_y][canvas_x] = char
+
+        if width > 0 and height > 0:
+            self.mark_dirty(x, y, width, height)
+
+    def put_text(self, x: int, y: int, text: str) -> None:
+        """Put a single line of text at position.
+
+        Args:
+            x: Left position
+            y: Row position
+            text: Text to place
+        """
+        text_len = len(text)
+        for i, char in enumerate(text):
+            canvas_x = x + i
+            if 0 <= canvas_x < self.width and 0 <= y < self.height:
+                self._grid[y][canvas_x] = char
+
+        if text_len > 0:
+            self.mark_dirty(x, y, text_len, 1)
+
+    def fill(self, x: int, y: int, width: int, height: int, char: str = " ") -> None:
+        """Fill a rectangular region with a character.
+
+        Args:
+            x: Left position
+            y: Top position
+            width: Region width
+            height: Region height
+            char: Character to fill with
+        """
+        for py in range(y, y + height):
+            for px in range(x, x + width):
+                if 0 <= py < self.height and 0 <= px < self.width:
+                    self._grid[py][px] = char
+
+        if width > 0 and height > 0:
+            self.mark_dirty(x, y, width, height)
+
+    def resize(self, width: int, height: int) -> None:
+        """Resize the canvas.
+
+        Args:
+            width: New width
+            height: New height
+        """
+        if width == self.width and height == self.height:
+            return
+
+        new_grid: list[list[str]] = [[" " for _ in range(width)] for _ in range(height)]
+
+        for py in range(min(self.height, height)):
+            for px in range(min(self.width, width)):
+                new_grid[py][px] = self._grid[py][px]
+
+        self.width = width
+        self.height = height
+        self._grid = new_grid

engine/config.py

@@ -0,0 +1,307 @@
+"""
+Configuration constants, CLI flags, and glyph tables.
+Supports both global constants (backward compatible) and injected config for testing.
+"""
+
+import sys
+from dataclasses import dataclass, field
+from pathlib import Path
+
+_REPO_ROOT = Path(__file__).resolve().parent.parent
+_FONT_EXTENSIONS = {".otf", ".ttf", ".ttc"}
+
+
+def _arg_value(flag, argv: list[str] | None = None):
+    """Get value following a CLI flag, if present."""
+    argv = argv or sys.argv
+    if flag not in argv:
+        return None
+    i = argv.index(flag)
+    return argv[i + 1] if i + 1 < len(argv) else None
+
+
+def _arg_int(flag, default, argv: list[str] | None = None):
+    """Get int CLI argument with safe fallback."""
+    raw = _arg_value(flag, argv)
+    if raw is None:
+        return default
+    try:
+        return int(raw)
+    except ValueError:
+        return default
+
+
+def _resolve_font_path(raw_path):
+    """Resolve font path; relative paths are anchored to repo root."""
+    p = Path(raw_path).expanduser()
+    if p.is_absolute():
+        return str(p)
+    return str((_REPO_ROOT / p).resolve())
+
+
+def _list_font_files(font_dir):
+    """List supported font files within a font directory."""
+    font_root = Path(font_dir)
+    if not font_root.exists() or not font_root.is_dir():
+        return []
+    return [
+        str(path.resolve())
+        for path in sorted(font_root.iterdir())
+        if path.is_file() and path.suffix.lower() in _FONT_EXTENSIONS
+    ]
+
+
+def list_repo_font_files():
+    """Public helper for discovering repository font files."""
+    return _list_font_files(FONT_DIR)
+
+
+def _get_platform_font_paths() -> dict[str, str]:
+    """Get platform-appropriate font paths for non-Latin scripts."""
+    import platform
+
+    system = platform.system()
+
+    if system == "Darwin":
+        return {
+            "zh-cn": "/System/Library/Fonts/STHeiti Medium.ttc",
+            "ja": "/System/Library/Fonts/ヒラギノ角ゴシック W9.ttc",
+            "ko": "/System/Library/Fonts/AppleSDGothicNeo.ttc",
+            "ru": "/System/Library/Fonts/Supplemental/Arial.ttf",
+            "uk": "/System/Library/Fonts/Supplemental/Arial.ttf",
+            "el": "/System/Library/Fonts/Supplemental/Arial.ttf",
+            "he": "/System/Library/Fonts/Supplemental/Arial.ttf",
+            "ar": "/System/Library/Fonts/GeezaPro.ttc",
+            "fa": "/System/Library/Fonts/GeezaPro.ttc",
+            "hi": "/System/Library/Fonts/Kohinoor.ttc",
+            "th": "/System/Library/Fonts/ThonburiUI.ttc",
+        }
+    elif system == "Linux":
+        return {
+            "zh-cn": "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
+            "ja": "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
+            "ko": "/usr/share/fonts/truetype/noto/NotoSansCJK-Regular.ttc",
+            "ru": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "uk": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "el": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "he": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "ar": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "fa": "/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf",
+            "hi": "/usr/share/fonts/truetype/noto/NotoSansDevanagari-Regular.ttf",
+            "th": "/usr/share/fonts/truetype/noto/NotoSansThai-Regular.ttf",
+        }
+    else:
+        return {}
+
+
+@dataclass(frozen=True)
+class Config:
+    """Immutable configuration container for injected config."""
+
+    headline_limit: int = 1000
+    feed_timeout: int = 10
+    mic_threshold_db: int = 50
+    mode: str = "news"
+    firehose: bool = False
+
+    ntfy_topic: str = "https://ntfy.sh/klubhaus_terminal_mainline/json"
+    ntfy_cc_cmd_topic: str = "https://ntfy.sh/klubhaus_terminal_mainline_cc_cmd/json"
+    ntfy_cc_resp_topic: str = "https://ntfy.sh/klubhaus_terminal_mainline_cc_resp/json"
+    ntfy_reconnect_delay: int = 5
+    message_display_secs: int = 30
+
+    font_dir: str = "fonts"
+    font_path: str = ""
+    font_index: int = 0
+    font_picker: bool = True
+    font_sz: int = 60
+    render_h: int = 8
+
+    ssaa: int = 4
+
+    scroll_dur: float = 5.625
+    frame_dt: float = 0.05
+    firehose_h: int = 12
+    grad_speed: float = 0.08
+
+    glitch_glyphs: str = "░▒▓█▌▐╌╍╎╏┃┆┇┊┋"
+    kata_glyphs: str = "ﾊﾐﾋｰｳｼﾅﾓﾆｻﾜﾂｵﾘｱﾎﾃﾏｹﾒｴｶｷﾑﾕﾗｾﾈｽﾀﾇﾍ"
+
+    script_fonts: dict[str, str] = field(default_factory=_get_platform_font_paths)
+
+    display: str = "pygame"
+    positioning: str = "mixed"
+    websocket: bool = False
+    websocket_port: int = 8765
+    theme: str = "green"
+
+    @classmethod
+    def from_args(cls, argv: list[str] | None = None) -> "Config":
+        """Create Config from CLI arguments (or custom argv for testing)."""
+        argv = argv or sys.argv
+
+        font_dir = _resolve_font_path(_arg_value("--font-dir", argv) or "fonts")
+        font_file_arg = _arg_value("--font-file", argv)
+        font_files = _list_font_files(font_dir)
+        font_path = (
+            _resolve_font_path(font_file_arg)
+            if font_file_arg
+            else (font_files[0] if font_files else "")
+        )
+
+        return cls(
+            headline_limit=1000,
+            feed_timeout=10,
+            mic_threshold_db=50,
+            mode="poetry" if "--poetry" in argv or "-p" in argv else "news",
+            firehose="--firehose" in argv,
+            ntfy_topic="https://ntfy.sh/klubhaus_terminal_mainline/json",
+            ntfy_cc_cmd_topic="https://ntfy.sh/klubhaus_terminal_mainline_cc_cmd/json",
+            ntfy_cc_resp_topic="https://ntfy.sh/klubhaus_terminal_mainline_cc_resp/json",
+            ntfy_reconnect_delay=5,
+            message_display_secs=30,
+            font_dir=font_dir,
+            font_path=font_path,
+            font_index=max(0, _arg_int("--font-index", 0, argv)),
+            font_picker="--no-font-picker" not in argv,
+            font_sz=60,
+            render_h=8,
+            ssaa=4,
+            scroll_dur=5.625,
+            frame_dt=0.05,
+            firehose_h=12,
+            grad_speed=0.08,
+            glitch_glyphs="░▒▓█▌▐╌╍╎╏┃┆┇┊┋",
+            kata_glyphs="ﾊﾐﾋｰｳｼﾅﾓﾆｻﾜﾂｵﾘｱﾎﾃﾏｹﾒｴｶｷﾑﾕﾗｾﾈｽﾀﾇﾍ",
+            script_fonts=_get_platform_font_paths(),
+            display=_arg_value("--display", argv) or "terminal",
+            positioning=_arg_value("--positioning", argv) or "mixed",
+            websocket="--websocket" in argv,
+            websocket_port=_arg_int("--websocket-port", 8765, argv),
+            theme=_arg_value("--theme", argv) or "green",
+        )
+
+
+_config: Config | None = None
+
+
+def get_config() -> Config:
+    """Get the global config instance (lazy-loaded)."""
+    global _config
+    if _config is None:
+        _config = Config.from_args()
+    return _config
+
+
+def set_config(config: Config) -> None:
+    """Set the global config instance (for testing)."""
+    global _config
+    _config = config
+
+
+# ─── RUNTIME ──────────────────────────────────────────────
+HEADLINE_LIMIT = 1000
+FEED_TIMEOUT = 10
+MIC_THRESHOLD_DB = 50  # dB above which glitches intensify
+MODE = "poetry" if "--poetry" in sys.argv or "-p" in sys.argv else "news"
+FIREHOSE = "--firehose" in sys.argv
+
+# ─── NTFY MESSAGE QUEUE ──────────────────────────────────
+NTFY_TOPIC = "https://ntfy.sh/klubhaus_terminal_mainline/json"
+NTFY_CC_CMD_TOPIC = "https://ntfy.sh/klubhaus_terminal_mainline_cc_cmd/json"
+NTFY_CC_RESP_TOPIC = "https://ntfy.sh/klubhaus_terminal_mainline_cc_resp/json"
+NTFY_RECONNECT_DELAY = 5  # seconds before reconnecting after a dropped stream
+MESSAGE_DISPLAY_SECS = 30  # how long a message holds the screen
+
+# ─── FONT RENDERING ──────────────────────────────────────
+FONT_DIR = _resolve_font_path(_arg_value("--font-dir") or "fonts")
+_FONT_FILE_ARG = _arg_value("--font-file")
+_FONT_FILES = _list_font_files(FONT_DIR)
+FONT_PATH = (
+    _resolve_font_path(_FONT_FILE_ARG)
+    if _FONT_FILE_ARG
+    else (_FONT_FILES[0] if _FONT_FILES else "")
+)
+FONT_INDEX = max(0, _arg_int("--font-index", 0))
+FONT_PICKER = "--no-font-picker" not in sys.argv
+FONT_SZ = 60
+RENDER_H = 8  # terminal rows per rendered text line
+
+# ─── FONT RENDERING (ADVANCED) ────────────────────────────
+SSAA = 4  # super-sampling factor: render at SSAA× then downsample
+
+# ─── SCROLL / FRAME ──────────────────────────────────────
+SCROLL_DUR = 5.625  # seconds per headline (2/3 original speed)
+FRAME_DT = 0.05  # 50ms base frame rate (20 FPS)
+FIREHOSE_H = 12  # firehose zone height (terminal rows)
+GRAD_SPEED = 0.08  # gradient traversal speed (cycles/sec, ~12s full sweep)
+
+# ─── GLYPHS ───────────────────────────────────────────────
+GLITCH = "░▒▓█▌▐╌╍╎╏┃┆┇┊┋"
+KATA = "ﾊﾐﾋｰｳｼﾅﾓﾆｻﾜﾂｵﾘｱﾎﾃﾏｹﾒｴｶｷﾑﾕﾗｾﾈｽﾀﾇﾍ"
+
+# ─── WEBSOCKET ─────────────────────────────────────────────
+DISPLAY = _arg_value("--display", sys.argv) or "pygame"
+WEBSOCKET = "--websocket" in sys.argv
+WEBSOCKET_PORT = _arg_int("--websocket-port", 8765)
+
+# ─── DEMO MODE ────────────────────────────────────────────
+DEMO = "--demo" in sys.argv
+DEMO_EFFECT_DURATION = 5.0  # seconds per effect
+PIPELINE_DEMO = "--pipeline-demo" in sys.argv
+
+# ─── THEME MANAGEMENT ─────────────────────────────────────────
+ACTIVE_THEME = None
+
+
+def set_active_theme(theme_id: str = "green"):
+    """Set the active theme by ID.
+
+    Args:
+        theme_id: Theme identifier from theme registry (e.g., "green", "orange", "purple")
+
+    Raises:
+        KeyError: If theme_id is not in the theme registry
+
+    Side Effects:
+        Sets the ACTIVE_THEME global variable
+    """
+    global ACTIVE_THEME
+    from engine import themes
+
+    ACTIVE_THEME = themes.get_theme(theme_id)
+
+
+# Initialize theme on module load (lazy to avoid circular dependency)
+def _init_theme():
+    theme_id = _arg_value("--theme", sys.argv) or "green"
+    try:
+        set_active_theme(theme_id)
+    except KeyError:
+        pass  # Theme not found, keep None
+
+
+_init_theme()
+
+
+# ─── PIPELINE MODE (new unified architecture) ─────────────
+PIPELINE_MODE = "--pipeline" in sys.argv
+PIPELINE_PRESET = _arg_value("--pipeline-preset", sys.argv) or "demo"
+
+# ─── PRESET MODE ────────────────────────────────────────────
+PRESET = _arg_value("--preset", sys.argv)
+
+# ─── PIPELINE DIAGRAM ────────────────────────────────────
+PIPELINE_DIAGRAM = "--pipeline-diagram" in sys.argv
+
+# ─── THEME ──────────────────────────────────────────────────
+THEME = _arg_value("--theme", sys.argv) or "green"
+
+
+def set_font_selection(font_path=None, font_index=None):
+    """Set runtime primary font selection."""
+    global FONT_PATH, FONT_INDEX
+    if font_path is not None:
+        FONT_PATH = _resolve_font_path(font_path)
+    if font_index is not None:
+        FONT_INDEX = max(0, int(font_index))

engine/data_sources/__init__.py

@@ -0,0 +1,12 @@
+"""
+Data source implementations for the pipeline architecture.
+
+Import directly from submodules:
+    from engine.data_sources.sources import DataSource, SourceItem, HeadlinesDataSource
+    from engine.data_sources.pipeline_introspection import PipelineIntrospectionSource
+"""
+
+# Re-export for convenience
+from engine.data_sources.sources import ImageItem, SourceItem
+
+__all__ = ["ImageItem", "SourceItem"]

engine/data_sources/checkerboard.py

@@ -0,0 +1,60 @@
+"""Checkerboard data source for visual pattern generation."""
+
+from engine.data_sources.sources import DataSource, SourceItem
+
+
+class CheckerboardDataSource(DataSource):
+    """Data source that generates a checkerboard pattern.
+
+    Creates a grid of alternating characters, useful for testing motion effects
+    and camera movement. The pattern is static; movement comes from camera panning.
+    """
+
+    def __init__(
+        self,
+        width: int = 200,
+        height: int = 200,
+        square_size: int = 10,
+        char_a: str = "#",
+        char_b: str = " ",
+    ):
+        """Initialize checkerboard data source.
+
+        Args:
+            width: Total pattern width in characters
+            height: Total pattern height in lines
+            square_size: Size of each checker square in characters
+            char_a: Character for "filled" squares (default: '#')
+            char_b: Character for "empty" squares (default: ' ')
+        """
+        self.width = width
+        self.height = height
+        self.square_size = square_size
+        self.char_a = char_a
+        self.char_b = char_b
+
+    @property
+    def name(self) -> str:
+        return "checkerboard"
+
+    @property
+    def is_dynamic(self) -> bool:
+        return False
+
+    def fetch(self) -> list[SourceItem]:
+        """Generate the checkerboard pattern as a single SourceItem."""
+        lines = []
+        for y in range(self.height):
+            line_chars = []
+            for x in range(self.width):
+                # Determine which square this position belongs to
+                square_x = x // self.square_size
+                square_y = y // self.square_size
+                # Alternate pattern based on parity of square coordinates
+                if (square_x + square_y) % 2 == 0:
+                    line_chars.append(self.char_a)
+                else:
+                    line_chars.append(self.char_b)
+            lines.append("".join(line_chars))
+        content = "\n".join(lines)
+        return [SourceItem(content=content, source="checkerboard", timestamp="0")]

engine/data_sources/pipeline_introspection.py

@@ -0,0 +1,312 @@
+"""
+Pipeline introspection source - Renders live visualization of pipeline DAG and metrics.
+
+This DataSource introspects one or more Pipeline instances and renders
+an ASCII visualization showing:
+- Stage DAG with signal flow connections
+- Per-stage execution times
+- Sparkline of frame times
+- Stage breakdown bars
+
+Example:
+    source = PipelineIntrospectionSource(pipelines=[my_pipeline])
+    items = source.fetch()  # Returns ASCII visualization
+"""
+
+from typing import TYPE_CHECKING
+
+from engine.data_sources.sources import DataSource, SourceItem
+
+if TYPE_CHECKING:
+    from engine.pipeline.controller import Pipeline
+
+
+SPARKLINE_CHARS = " ▁▂▃▄▅▆▇█"
+BAR_CHARS = " ▁▂▃▄▅▆▇█"
+
+
+class PipelineIntrospectionSource(DataSource):
+    """Data source that renders live pipeline introspection visualization.
+
+    Renders:
+    - DAG of stages with signal flow
+    - Per-stage execution times
+    - Sparkline of frame history
+    - Stage breakdown bars
+    """
+
+    def __init__(
+        self,
+        pipeline: "Pipeline | None" = None,
+        viewport_width: int = 100,
+        viewport_height: int = 35,
+    ):
+        self._pipeline = pipeline  # May be None initially, set later via set_pipeline()
+        self.viewport_width = viewport_width
+        self.viewport_height = viewport_height
+        self.frame = 0
+        self._ready = False
+
+    def set_pipeline(self, pipeline: "Pipeline") -> None:
+        """Set the pipeline to introspect (call after pipeline is built)."""
+        self._pipeline = [pipeline]  # Wrap in list for iteration
+        self._ready = True
+
+    @property
+    def ready(self) -> bool:
+        """Check if source is ready to fetch."""
+        return self._ready
+
+    @property
+    def name(self) -> str:
+        return "pipeline-inspect"
+
+    @property
+    def is_dynamic(self) -> bool:
+        return True
+
+    @property
+    def inlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.NONE}
+
+    @property
+    def outlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.SOURCE_ITEMS}
+
+    def add_pipeline(self, pipeline: "Pipeline") -> None:
+        """Add a pipeline to visualize."""
+        if self._pipeline is None:
+            self._pipeline = [pipeline]
+        elif isinstance(self._pipeline, list):
+            self._pipeline.append(pipeline)
+        else:
+            self._pipeline = [self._pipeline, pipeline]
+        self._ready = True
+
+    def remove_pipeline(self, pipeline: "Pipeline") -> None:
+        """Remove a pipeline from visualization."""
+        if self._pipeline is None:
+            return
+        elif isinstance(self._pipeline, list):
+            self._pipeline = [p for p in self._pipeline if p is not pipeline]
+            if not self._pipeline:
+                self._pipeline = None
+                self._ready = False
+        elif self._pipeline is pipeline:
+            self._pipeline = None
+            self._ready = False
+
+    def fetch(self) -> list[SourceItem]:
+        """Fetch the introspection visualization."""
+        if not self._ready:
+            # Return a placeholder until ready
+            return [
+                SourceItem(
+                    content="Initializing...",
+                    source="pipeline-inspect",
+                    timestamp="init",
+                )
+            ]
+
+        lines = self._render()
+        self.frame += 1
+        content = "\n".join(lines)
+        return [
+            SourceItem(
+                content=content, source="pipeline-inspect", timestamp=f"f{self.frame}"
+            )
+        ]
+
+    def get_items(self) -> list[SourceItem]:
+        return self.fetch()
+
+    def _render(self) -> list[str]:
+        """Render the full visualization."""
+        lines: list[str] = []
+
+        # Header
+        lines.extend(self._render_header())
+
+        # Render pipeline(s) if ready
+        if self._ready and self._pipeline:
+            pipelines = (
+                self._pipeline if isinstance(self._pipeline, list) else [self._pipeline]
+            )
+            for pipeline in pipelines:
+                lines.extend(self._render_pipeline(pipeline))
+
+        # Footer with sparkline
+        lines.extend(self._render_footer())
+
+        return lines
+
+    @property
+    def _pipelines(self) -> list:
+        """Return pipelines as a list for iteration."""
+        if self._pipeline is None:
+            return []
+        elif isinstance(self._pipeline, list):
+            return self._pipeline
+        else:
+            return [self._pipeline]
+
+    def _render_header(self) -> list[str]:
+        """Render the header with frame info and metrics summary."""
+        lines: list[str] = []
+
+        if not self._pipeline:
+            return ["PIPELINE INTROSPECTION"]
+
+        # Get aggregate metrics
+        total_ms = 0.0
+        fps = 0.0
+        frame_count = 0
+
+        for pipeline in self._pipelines:
+            try:
+                metrics = pipeline.get_metrics_summary()
+                if metrics and "error" not in metrics:
+                    # Get avg_ms from pipeline metrics
+                    pipeline_avg = metrics.get("pipeline", {}).get("avg_ms", 0)
+                    total_ms = max(total_ms, pipeline_avg)
+                    # Calculate FPS from avg_ms
+                    if pipeline_avg > 0:
+                        fps = max(fps, 1000.0 / pipeline_avg)
+                    frame_count = max(frame_count, metrics.get("frame_count", 0))
+            except Exception:
+                pass
+
+        header = f"PIPELINE INTROSPECTION -- frame: {self.frame} -- avg: {total_ms:.1f}ms -- fps: {fps:.1f}"
+        lines.append(header)
+
+        return lines
+
+    def _render_pipeline(self, pipeline: "Pipeline") -> list[str]:
+        """Render a single pipeline's DAG."""
+        lines: list[str] = []
+
+        stages = pipeline.stages
+        execution_order = pipeline.execution_order
+
+        if not stages:
+            lines.append("    (no stages)")
+            return lines
+
+        # Build stage info
+        stage_infos: list[dict] = []
+        for name in execution_order:
+            stage = stages.get(name)
+            if not stage:
+                continue
+
+            try:
+                metrics = pipeline.get_metrics_summary()
+                stage_ms = metrics.get("stages", {}).get(name, {}).get("avg_ms", 0.0)
+            except Exception:
+                stage_ms = 0.0
+
+            stage_infos.append(
+                {
+                    "name": name,
+                    "category": stage.category,
+                    "ms": stage_ms,
+                }
+            )
+
+        # Calculate total time for percentages
+        total_time = sum(s["ms"] for s in stage_infos) or 1.0
+
+        # Render DAG - group by category
+        lines.append("")
+        lines.append("  Signal Flow:")
+
+        # Group stages by category for display
+        categories: dict[str, list[dict]] = {}
+        for info in stage_infos:
+            cat = info["category"]
+            if cat not in categories:
+                categories[cat] = []
+            categories[cat].append(info)
+
+        # Render categories in order
+        cat_order = ["source", "render", "effect", "overlay", "display", "system"]
+
+        for cat in cat_order:
+            if cat not in categories:
+                continue
+
+            cat_stages = categories[cat]
+            cat_names = [s["name"] for s in cat_stages]
+            lines.append(f"    {cat}: {' → '.join(cat_names)}")
+
+        # Render timing breakdown
+        lines.append("")
+        lines.append("  Stage Timings:")
+
+        for info in stage_infos:
+            name = info["name"]
+            ms = info["ms"]
+            pct = (ms / total_time) * 100
+            bar = self._render_bar(pct, 20)
+            lines.append(f"    {name:12s} {ms:6.2f}ms {bar} {pct:5.1f}%")
+
+        lines.append("")
+
+        return lines
+
+    def _render_footer(self) -> list[str]:
+        """Render the footer with sparkline."""
+        lines: list[str] = []
+
+        # Get frame history from first pipeline
+        pipelines = self._pipelines
+        if pipelines:
+            try:
+                frame_times = pipelines[0].get_frame_times()
+            except Exception:
+                frame_times = []
+        else:
+            frame_times = []
+
+        if frame_times:
+            sparkline = self._render_sparkline(frame_times[-60:], 50)
+            lines.append(f" Frame Time History (last {len(frame_times[-60:])} frames)")
+            lines.append(f" {sparkline}")
+        else:
+            lines.append(" Frame Time History")
+            lines.append(" (collecting data...)")
+
+        lines.append("")
+
+        return lines
+
+    def _render_bar(self, percentage: float, width: int) -> str:
+        """Render a horizontal bar for percentage."""
+        filled = int((percentage / 100.0) * width)
+        bar = "█" * filled + "░" * (width - filled)
+        return bar
+
+    def _render_sparkline(self, values: list[float], width: int) -> str:
+        """Render a sparkline from values."""
+        if not values:
+            return " " * width
+
+        min_val = min(values)
+        max_val = max(values)
+        range_val = max_val - min_val or 1.0
+
+        result = []
+        for v in values[-width:]:
+            normalized = (v - min_val) / range_val
+            idx = int(normalized * (len(SPARKLINE_CHARS) - 1))
+            idx = max(0, min(idx, len(SPARKLINE_CHARS) - 1))
+            result.append(SPARKLINE_CHARS[idx])
+
+        # Pad to width
+        while len(result) < width:
+            result.insert(0, " ")
+        return "".join(result[:width])

engine/data_sources/sources.py

@@ -0,0 +1,490 @@
+"""
+Data sources for the pipeline architecture.
+
+This module contains all DataSource implementations:
+- DataSource: Abstract base class
+- SourceItem, ImageItem: Data containers
+- HeadlinesDataSource, PoetryDataSource, ImageDataSource: Concrete sources
+- SourceRegistry: Registry for source discovery
+"""
+
+from abc import ABC, abstractmethod
+from collections.abc import Callable
+from dataclasses import dataclass
+from typing import Any
+
+
+@dataclass
+class SourceItem:
+    """A single item from a data source."""
+
+    content: str
+    source: str
+    timestamp: str
+    metadata: dict[str, Any] | None = None
+
+
+@dataclass
+class ImageItem:
+    """An image item from a data source - wraps a PIL Image."""
+
+    image: Any  # PIL Image
+    source: str
+    timestamp: str
+    path: str | None = None  # File path or URL if applicable
+    metadata: dict[str, Any] | None = None
+
+
+class DataSource(ABC):
+    """Abstract base class for data sources.
+
+    Static sources: Data fetched once and cached. Safe to call fetch() multiple times.
+    Dynamic sources: Data changes over time. fetch() should be idempotent.
+    """
+
+    @property
+    @abstractmethod
+    def name(self) -> str:
+        """Display name for this source."""
+        ...
+
+    @property
+    def is_dynamic(self) -> bool:
+        """Whether this source updates dynamically while the app runs. Default False."""
+        return False
+
+    @abstractmethod
+    def fetch(self) -> list[SourceItem]:
+        """Fetch fresh data from the source. Must be idempotent."""
+        ...
+
+    def get_items(self) -> list[SourceItem]:
+        """Get current items. Default implementation returns cached fetch results."""
+        if not hasattr(self, "_items") or self._items is None:
+            self._items = self.fetch()
+        return self._items
+
+    def refresh(self) -> list[SourceItem]:
+        """Force refresh - clear cache and fetch fresh data."""
+        self._items = self.fetch()
+        return self._items
+
+    def stream(self):
+        """Optional: Yield items continuously. Override for streaming sources."""
+        raise NotImplementedError
+
+    def __post_init__(self):
+        self._items: list[SourceItem] | None = None
+
+
+class HeadlinesDataSource(DataSource):
+    """Data source for RSS feed headlines."""
+
+    @property
+    def name(self) -> str:
+        return "headlines"
+
+    def fetch(self) -> list[SourceItem]:
+        from engine.fetch import fetch_all
+
+        items, _, _ = fetch_all()
+        return [SourceItem(content=t, source=s, timestamp=ts) for t, s, ts in items]
+
+
+class EmptyDataSource(DataSource):
+    """Empty data source that produces blank lines for testing.
+
+    Useful for testing display borders, effects, and other pipeline
+    components without needing actual content.
+    """
+
+    def __init__(self, width: int = 80, height: int = 24):
+        self.width = width
+        self.height = height
+
+    @property
+    def name(self) -> str:
+        return "empty"
+
+    @property
+    def is_dynamic(self) -> bool:
+        return False
+
+    def fetch(self) -> list[SourceItem]:
+        # Return empty lines as content
+        content = "\n".join([" " * self.width for _ in range(self.height)])
+        return [SourceItem(content=content, source="empty", timestamp="0")]
+
+
+class ListDataSource(DataSource):
+    """Data source that wraps a pre-fetched list of items.
+
+    Used for bootstrap loading when items are already available in memory.
+    This is a simple wrapper for already-fetched data.
+    """
+
+    def __init__(self, items, name: str = "list"):
+        self._raw_items = items  # Store raw items separately
+        self._items = None  # Cache for converted SourceItem objects
+        self._name = name
+
+    @property
+    def name(self) -> str:
+        return self._name
+
+    @property
+    def is_dynamic(self) -> bool:
+        return False
+
+    def fetch(self) -> list[SourceItem]:
+        # Convert tuple items to SourceItem if needed
+        result = []
+        for item in self._raw_items:
+            if isinstance(item, SourceItem):
+                result.append(item)
+            elif isinstance(item, tuple) and len(item) >= 3:
+                # Assume (content, source, timestamp) tuple format
+                result.append(
+                    SourceItem(content=item[0], source=item[1], timestamp=str(item[2]))
+                )
+            else:
+                # Fallback: treat as string content
+                result.append(
+                    SourceItem(content=str(item), source="list", timestamp="0")
+                )
+        return result
+
+
+class PoetryDataSource(DataSource):
+    """Data source for Poetry DB."""
+
+    @property
+    def name(self) -> str:
+        return "poetry"
+
+    def fetch(self) -> list[SourceItem]:
+        from engine.fetch import fetch_poetry
+
+        items, _, _ = fetch_poetry()
+        return [SourceItem(content=t, source=s, timestamp=ts) for t, s, ts in items]
+
+
+class ImageDataSource(DataSource):
+    """Data source that loads PNG images from file paths or URLs.
+
+    Supports:
+    - Local file paths (e.g., /path/to/image.png)
+    - URLs (e.g., https://example.com/image.png)
+
+    Yields ImageItem objects containing PIL Image objects that can be
+    converted to text buffers by an ImageToTextTransform stage.
+    """
+
+    def __init__(
+        self,
+        path: str | list[str] | None = None,
+        urls: str | list[str] | None = None,
+    ):
+        """
+        Args:
+            path: Single path or list of paths to PNG files
+            urls: Single URL or list of URLs to PNG images
+        """
+        self._paths = [path] if isinstance(path, str) else (path or [])
+        self._urls = [urls] if isinstance(urls, str) else (urls or [])
+        self._images: list[ImageItem] = []
+        self._load_images()
+
+    def _load_images(self) -> None:
+        """Load all images from paths and URLs."""
+        from datetime import datetime
+        from io import BytesIO
+        from urllib.request import urlopen
+
+        timestamp = datetime.now().isoformat()
+
+        for path in self._paths:
+            try:
+                from PIL import Image
+
+                img = Image.open(path)
+                if img.mode != "RGBA":
+                    img = img.convert("RGBA")
+                self._images.append(
+                    ImageItem(
+                        image=img,
+                        source=f"file:{path}",
+                        timestamp=timestamp,
+                        path=path,
+                    )
+                )
+            except Exception:
+                pass
+
+        for url in self._urls:
+            try:
+                from PIL import Image
+
+                with urlopen(url) as response:
+                    img = Image.open(BytesIO(response.read()))
+                if img.mode != "RGBA":
+                    img = img.convert("RGBA")
+                self._images.append(
+                    ImageItem(
+                        image=img,
+                        source=f"url:{url}",
+                        timestamp=timestamp,
+                        path=url,
+                    )
+                )
+            except Exception:
+                pass
+
+    @property
+    def name(self) -> str:
+        return "image"
+
+    @property
+    def is_dynamic(self) -> bool:
+        return False  # Static images, not updating
+
+    def fetch(self) -> list[ImageItem]:
+        """Return loaded images as ImageItem list."""
+        return self._images
+
+    def get_items(self) -> list[ImageItem]:
+        """Return current image items."""
+        return self._images
+
+
+class MetricsDataSource(DataSource):
+    """Data source that renders live pipeline metrics as ASCII art.
+
+    Wraps a Pipeline and displays active stages with their average execution
+    time and approximate FPS impact. Updates lazily when camera is about to
+    focus on a new node (frame % 15 == 12).
+    """
+
+    def __init__(
+        self,
+        pipeline: Any,
+        viewport_width: int = 80,
+        viewport_height: int = 24,
+    ):
+        self.pipeline = pipeline
+        self.viewport_width = viewport_width
+        self.viewport_height = viewport_height
+        self.frame = 0
+        self._cached_metrics: dict | None = None
+
+    @property
+    def name(self) -> str:
+        return "metrics"
+
+    @property
+    def is_dynamic(self) -> bool:
+        return True
+
+    def fetch(self) -> list[SourceItem]:
+        if self.frame % 15 == 12:
+            self._cached_metrics = None
+
+        if self._cached_metrics is None:
+            self._cached_metrics = self._fetch_metrics()
+
+        buffer = self._render_metrics(self._cached_metrics)
+        self.frame += 1
+        content = "\n".join(buffer)
+        return [
+            SourceItem(content=content, source="metrics", timestamp=f"f{self.frame}")
+        ]
+
+    def _fetch_metrics(self) -> dict:
+        if hasattr(self.pipeline, "get_metrics_summary"):
+            metrics = self.pipeline.get_metrics_summary()
+            if "error" not in metrics:
+                return metrics
+        return {"stages": {}, "pipeline": {"avg_ms": 0}}
+
+    def _render_metrics(self, metrics: dict) -> list[str]:
+        stages = metrics.get("stages", {})
+
+        if not stages:
+            return self._render_empty()
+
+        active_stages = {
+            name: stats for name, stats in stages.items() if stats.get("avg_ms", 0) > 0
+        }
+
+        if not active_stages:
+            return self._render_empty()
+
+        total_avg = sum(s["avg_ms"] for s in active_stages.values())
+        if total_avg == 0:
+            total_avg = 1
+
+        lines: list[str] = []
+        lines.append("═" * self.viewport_width)
+        lines.append(" PIPELINE METRICS ".center(self.viewport_width, "─"))
+        lines.append("─" * self.viewport_width)
+
+        header = f"{'STAGE':<20} {'AVG_MS':>8} {'FPS %':>8}"
+        lines.append(header)
+        lines.append("─" * self.viewport_width)
+
+        for name, stats in sorted(active_stages.items()):
+            avg_ms = stats.get("avg_ms", 0)
+            fps_impact = (avg_ms / 16.67) * 100 if avg_ms > 0 else 0
+
+            row = f"{name:<20} {avg_ms:>7.2f} {fps_impact:>7.1f}%"
+            lines.append(row[: self.viewport_width])
+
+        lines.append("─" * self.viewport_width)
+        total_row = (
+            f"{'TOTAL':<20} {total_avg:>7.2f} {(total_avg / 16.67) * 100:>7.1f}%"
+        )
+        lines.append(total_row[: self.viewport_width])
+        lines.append("─" * self.viewport_width)
+        lines.append(
+            f" Frame:{self.frame:04d}  Cache:{'HIT' if self._cached_metrics else 'MISS'}"
+        )
+
+        while len(lines) < self.viewport_height:
+            lines.append(" " * self.viewport_width)
+
+        return lines[: self.viewport_height]
+
+    def _render_empty(self) -> list[str]:
+        lines = [" " * self.viewport_width for _ in range(self.viewport_height)]
+        msg = "No metrics available"
+        y = self.viewport_height // 2
+        x = (self.viewport_width - len(msg)) // 2
+        lines[y] = " " * x + msg + " " * (self.viewport_width - x - len(msg))
+        return lines
+
+    def get_items(self) -> list[SourceItem]:
+        return self.fetch()
+
+
+class CachedDataSource(DataSource):
+    """Data source that wraps another source with caching."""
+
+    def __init__(self, source: DataSource, max_items: int = 100):
+        self.source = source
+        self.max_items = max_items
+
+    @property
+    def name(self) -> str:
+        return f"cached:{self.source.name}"
+
+    def fetch(self) -> list[SourceItem]:
+        items = self.source.fetch()
+        return items[: self.max_items]
+
+    def get_items(self) -> list[SourceItem]:
+        if not hasattr(self, "_items") or self._items is None:
+            self._items = self.fetch()
+        return self._items
+
+
+class TransformDataSource(DataSource):
+    """Data source that transforms items from another source.
+
+    Applies optional filter and map functions to each item.
+    This enables chaining: source → transform → transformed output.
+
+    Args:
+        source: The source to fetch items from
+        filter_fn: Optional function(item: SourceItem) -> bool
+        map_fn: Optional function(item: SourceItem) -> SourceItem
+    """
+
+    def __init__(
+        self,
+        source: DataSource,
+        filter_fn: Callable[[SourceItem], bool] | None = None,
+        map_fn: Callable[[SourceItem], SourceItem] | None = None,
+    ):
+        self.source = source
+        self.filter_fn = filter_fn
+        self.map_fn = map_fn
+
+    @property
+    def name(self) -> str:
+        return f"transform:{self.source.name}"
+
+    def fetch(self) -> list[SourceItem]:
+        items = self.source.fetch()
+
+        if self.filter_fn:
+            items = [item for item in items if self.filter_fn(item)]
+
+        if self.map_fn:
+            items = [self.map_fn(item) for item in items]
+
+        return items
+
+
+class CompositeDataSource(DataSource):
+    """Data source that combines multiple sources."""
+
+    def __init__(self, sources: list[DataSource]):
+        self.sources = sources
+
+    @property
+    def name(self) -> str:
+        return "composite"
+
+    def fetch(self) -> list[SourceItem]:
+        items = []
+        for source in self.sources:
+            items.extend(source.fetch())
+        return items
+
+
+class SourceRegistry:
+    """Registry for data sources."""
+
+    def __init__(self):
+        self._sources: dict[str, DataSource] = {}
+        self._default: str | None = None
+
+    def register(self, source: DataSource, default: bool = False) -> None:
+        self._sources[source.name] = source
+        if default or self._default is None:
+            self._default = source.name
+
+    def get(self, name: str) -> DataSource | None:
+        return self._sources.get(name)
+
+    def list_all(self) -> dict[str, DataSource]:
+        return dict(self._sources)
+
+    def default(self) -> DataSource | None:
+        if self._default:
+            return self._sources.get(self._default)
+        return None
+
+    def create_headlines(self) -> HeadlinesDataSource:
+        return HeadlinesDataSource()
+
+    def create_poetry(self) -> PoetryDataSource:
+        return PoetryDataSource()
+
+
+_global_registry: SourceRegistry | None = None
+
+
+def get_source_registry() -> SourceRegistry:
+    global _global_registry
+    if _global_registry is None:
+        _global_registry = SourceRegistry()
+    return _global_registry
+
+
+def init_default_sources() -> SourceRegistry:
+    """Initialize the default source registry with standard sources."""
+    registry = get_source_registry()
+    registry.register(HeadlinesDataSource(), default=True)
+    registry.register(PoetryDataSource())
+    return registry

engine/display/__init__.py

@@ -0,0 +1,290 @@
+"""
+Display backend system with registry pattern.
+
+Allows swapping output backends via the Display protocol.
+Supports auto-discovery of display backends.
+"""
+
+from enum import Enum, auto
+from typing import Protocol
+
+# Optional backend - requires moderngl package
+try:
+    from engine.display.backends.moderngl import ModernGLDisplay
+
+    _MODERNGL_AVAILABLE = True
+except ImportError:
+    ModernGLDisplay = None
+    _MODERNGL_AVAILABLE = False
+
+from engine.display.backends.multi import MultiDisplay
+from engine.display.backends.null import NullDisplay
+from engine.display.backends.pygame import PygameDisplay
+from engine.display.backends.replay import ReplayDisplay
+from engine.display.backends.terminal import TerminalDisplay
+from engine.display.backends.websocket import WebSocketDisplay
+
+
+class BorderMode(Enum):
+    """Border rendering modes for displays."""
+
+    OFF = auto()  # No border
+    SIMPLE = auto()  # Traditional border with FPS/frame time
+    UI = auto()  # Right-side UI panel with interactive controls
+
+
+class Display(Protocol):
+    """Protocol for display backends.
+
+    Required attributes:
+    - width: int
+    - height: int
+
+    Required methods (duck typing - actual signatures may vary):
+    - init(width, height, reuse=False)
+    - show(buffer, border=False)
+    - clear()
+    - cleanup()
+    - get_dimensions() -> (width, height)
+
+    Optional attributes (for UI mode):
+    - ui_panel: UIPanel instance (set by app when border=UI)
+
+    Optional methods:
+    - is_quit_requested() -> bool
+    - clear_quit_request() -> None
+    """
+
+    width: int
+    height: int
+
+
+class DisplayRegistry:
+    """Registry for display backends with auto-discovery."""
+
+    _backends: dict[str, type[Display]] = {}
+    _initialized = False
+
+    @classmethod
+    def register(cls, name: str, backend_class: type[Display]) -> None:
+        cls._backends[name.lower()] = backend_class
+
+    @classmethod
+    def get(cls, name: str) -> type[Display] | None:
+        return cls._backends.get(name.lower())
+
+    @classmethod
+    def list_backends(cls) -> list[str]:
+        return list(cls._backends.keys())
+
+    @classmethod
+    def create(cls, name: str, **kwargs) -> Display | None:
+        cls.initialize()
+        backend_class = cls.get(name)
+        if backend_class:
+            return backend_class(**kwargs)
+        return None
+
+    @classmethod
+    def initialize(cls) -> None:
+        if cls._initialized:
+            return
+        cls.register("terminal", TerminalDisplay)
+        cls.register("null", NullDisplay)
+        cls.register("replay", ReplayDisplay)
+        cls.register("websocket", WebSocketDisplay)
+        cls.register("pygame", PygameDisplay)
+        if _MODERNGL_AVAILABLE:
+            cls.register("moderngl", ModernGLDisplay)  # type: ignore[arg-type]
+        cls._initialized = True
+
+    @classmethod
+    def create_multi(cls, names: list[str]) -> MultiDisplay | None:
+        displays = []
+        for name in names:
+            backend = cls.create(name)
+            if backend:
+                displays.append(backend)
+            else:
+                return None
+        if not displays:
+            return None
+        return MultiDisplay(displays)
+
+
+def get_monitor():
+    """Get the performance monitor."""
+    try:
+        from engine.effects.performance import get_monitor as _get_monitor
+
+        return _get_monitor()
+    except Exception:
+        return None
+
+
+def _strip_ansi(s: str) -> str:
+    """Strip ANSI escape sequences from string for length calculation."""
+    import re
+
+    return re.sub(r"\x1b\[[0-9;]*[a-zA-Z]", "", s)
+
+
+def _render_simple_border(
+    buf: list[str], width: int, height: int, fps: float = 0.0, frame_time: float = 0.0
+) -> list[str]:
+    """Render a traditional border around the buffer."""
+    if not buf or width < 3 or height < 3:
+        return buf
+
+    inner_w = width - 2
+    inner_h = height - 2
+
+    cropped = []
+    for i in range(min(inner_h, len(buf))):
+        line = buf[i]
+        visible_len = len(_strip_ansi(line))
+        if visible_len > inner_w:
+            cropped.append(line[:inner_w])
+        else:
+            cropped.append(line + " " * (inner_w - visible_len))
+
+    while len(cropped) < inner_h:
+        cropped.append(" " * inner_w)
+
+    if fps > 0:
+        fps_str = f" FPS:{fps:.0f}"
+        if len(fps_str) < inner_w:
+            right_len = inner_w - len(fps_str)
+            top_border = "┌" + "─" * right_len + fps_str + "┐"
+        else:
+            top_border = "┌" + "─" * inner_w + "┐"
+    else:
+        top_border = "┌" + "─" * inner_w + "┐"
+
+    if frame_time > 0:
+        ft_str = f" {frame_time:.1f}ms"
+        if len(ft_str) < inner_w:
+            right_len = inner_w - len(ft_str)
+            bottom_border = "└" + "─" * right_len + ft_str + "┘"
+        else:
+            bottom_border = "└" + "─" * inner_w + "┘"
+    else:
+        bottom_border = "└" + "─" * inner_w + "┘"
+
+    result = [top_border]
+    for line in cropped:
+        if len(line) < inner_w:
+            line = line + " " * (inner_w - len(line))
+        elif len(line) > inner_w:
+            line = line[:inner_w]
+        result.append("│" + line + "│")
+    result.append(bottom_border)
+
+    return result
+
+
+def render_ui_panel(
+    buf: list[str],
+    width: int,
+    height: int,
+    ui_panel,
+    fps: float = 0.0,
+    frame_time: float = 0.0,
+) -> list[str]:
+    """Render buffer with a right-side UI panel."""
+    from engine.pipeline.ui import UIPanel
+
+    if not isinstance(ui_panel, UIPanel):
+        return _render_simple_border(buf, width, height, fps, frame_time)
+
+    panel_width = min(ui_panel.config.panel_width, width - 4)
+    main_width = width - panel_width - 1
+
+    panel_lines = ui_panel.render(panel_width, height)
+
+    main_buf = buf[: height - 2]
+    main_result = _render_simple_border(
+        main_buf, main_width + 2, height, fps, frame_time
+    )
+
+    combined = []
+    for i in range(height):
+        if i < len(main_result):
+            main_line = main_result[i]
+            if len(main_line) >= 2:
+                main_content = (
+                    main_line[1:-1] if main_line[-1] in "│┌┐└┘" else main_line[1:]
+                )
+                main_content = main_content.ljust(main_width)[:main_width]
+            else:
+                main_content = " " * main_width
+        else:
+            main_content = " " * main_width
+
+        panel_idx = i
+        panel_line = (
+            panel_lines[panel_idx][:panel_width].ljust(panel_width)
+            if panel_idx < len(panel_lines)
+            else " " * panel_width
+        )
+
+        separator = "│" if 0 < i < height - 1 else "┼" if i == 0 else "┴"
+        combined.append(main_content + separator + panel_line)
+
+    return combined
+
+
+def render_border(
+    buf: list[str],
+    width: int,
+    height: int,
+    fps: float = 0.0,
+    frame_time: float = 0.0,
+    border_mode: BorderMode | bool = BorderMode.SIMPLE,
+) -> list[str]:
+    """Render a border or UI panel around the buffer.
+
+    Args:
+        buf: Input buffer
+        width: Display width
+        height: Display height
+        fps: FPS for top border
+        frame_time: Frame time for bottom border
+        border_mode: Border rendering mode
+
+    Returns:
+        Buffer with border/panel applied
+    """
+    # Normalize border_mode to BorderMode enum
+    if isinstance(border_mode, bool):
+        border_mode = BorderMode.SIMPLE if border_mode else BorderMode.OFF
+
+    if border_mode == BorderMode.UI:
+        # UI panel requires a UIPanel instance (injected separately)
+        # For now, this will be called by displays that have a ui_panel attribute
+        # This function signature doesn't include ui_panel, so we'll handle it in render_ui_panel
+        # Fall back to simple border if no panel available
+        return _render_simple_border(buf, width, height, fps, frame_time)
+    elif border_mode == BorderMode.SIMPLE:
+        return _render_simple_border(buf, width, height, fps, frame_time)
+    else:
+        return buf
+
+
+__all__ = [
+    "Display",
+    "DisplayRegistry",
+    "get_monitor",
+    "render_border",
+    "render_ui_panel",
+    "BorderMode",
+    "TerminalDisplay",
+    "NullDisplay",
+    "ReplayDisplay",
+    "WebSocketDisplay",
+    "MultiDisplay",
+    "PygameDisplay",
+]
+
+if _MODERNGL_AVAILABLE:
+    __all__.append("ModernGLDisplay")

engine/display/backends/animation_report.py

@@ -0,0 +1,656 @@
+"""
+Animation Report Display Backend
+
+Captures frames from pipeline stages and generates an interactive HTML report
+showing before/after states for each transformative stage.
+"""
+
+import time
+from dataclasses import dataclass, field
+from datetime import datetime
+from pathlib import Path
+from typing import Any
+
+from engine.display.streaming import compute_diff
+
+
+@dataclass
+class CapturedFrame:
+    """A captured frame with metadata."""
+
+    stage: str
+    buffer: list[str]
+    timestamp: float
+    frame_number: int
+    diff_from_previous: dict[str, Any] | None = None
+
+
+@dataclass
+class StageCapture:
+    """Captures frames for a single pipeline stage."""
+
+    name: str
+    frames: list[CapturedFrame] = field(default_factory=list)
+    start_time: float = field(default_factory=time.time)
+    end_time: float = 0.0
+
+    def add_frame(
+        self,
+        buffer: list[str],
+        frame_number: int,
+        previous_buffer: list[str] | None = None,
+    ) -> None:
+        """Add a captured frame."""
+        timestamp = time.time()
+        diff = None
+        if previous_buffer is not None:
+            diff_data = compute_diff(previous_buffer, buffer)
+            diff = {
+                "changed_lines": len(diff_data.changed_lines),
+                "total_lines": len(buffer),
+                "width": diff_data.width,
+                "height": diff_data.height,
+            }
+
+        frame = CapturedFrame(
+            stage=self.name,
+            buffer=list(buffer),
+            timestamp=timestamp,
+            frame_number=frame_number,
+            diff_from_previous=diff,
+        )
+        self.frames.append(frame)
+
+    def finish(self) -> None:
+        """Mark capture as finished."""
+        self.end_time = time.time()
+
+
+class AnimationReportDisplay:
+    """
+    Display backend that captures frames for animation report generation.
+
+    Instead of rendering to terminal, this display captures the buffer at each
+    stage and stores it for later HTML report generation.
+    """
+
+    width: int = 80
+    height: int = 24
+
+    def __init__(self, output_dir: str = "./reports"):
+        """
+        Initialize the animation report display.
+
+        Args:
+            output_dir: Directory where reports will be saved
+        """
+        self.output_dir = Path(output_dir)
+        self.output_dir.mkdir(parents=True, exist_ok=True)
+
+        self._stages: dict[str, StageCapture] = {}
+        self._current_stage: str = ""
+        self._previous_buffer: list[str] | None = None
+        self._frame_number: int = 0
+        self._total_frames: int = 0
+        self._start_time: float = 0.0
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions."""
+        self.width = width
+        self.height = height
+        self._start_time = time.time()
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        """
+        Capture a frame for the current stage.
+
+        Args:
+            buffer: The frame buffer to capture
+            border: Border flag (ignored)
+        """
+        if not self._current_stage:
+            # If no stage is set, use a default name
+            self._current_stage = "final"
+
+        if self._current_stage not in self._stages:
+            self._stages[self._current_stage] = StageCapture(self._current_stage)
+
+        stage = self._stages[self._current_stage]
+        stage.add_frame(buffer, self._frame_number, self._previous_buffer)
+
+        self._previous_buffer = list(buffer)
+        self._frame_number += 1
+        self._total_frames += 1
+
+    def start_stage(self, stage_name: str) -> None:
+        """
+        Start capturing frames for a new stage.
+
+        Args:
+            stage_name: Name of the stage (e.g., "noise", "fade", "firehose")
+        """
+        if self._current_stage and self._current_stage in self._stages:
+            # Finish previous stage
+            self._stages[self._current_stage].finish()
+
+        self._current_stage = stage_name
+        self._previous_buffer = None  # Reset for new stage
+
+    def clear(self) -> None:
+        """Clear the display (no-op for report display)."""
+        pass
+
+    def cleanup(self) -> None:
+        """Cleanup resources."""
+        # Finish current stage
+        if self._current_stage and self._current_stage in self._stages:
+            self._stages[self._current_stage].finish()
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current dimensions."""
+        return (self.width, self.height)
+
+    def get_stages(self) -> dict[str, StageCapture]:
+        """Get all captured stages."""
+        return self._stages
+
+    def generate_report(self, title: str = "Animation Report") -> Path:
+        """
+        Generate an HTML report with captured frames and animations.
+
+        Args:
+            title: Title of the report
+
+        Returns:
+            Path to the generated HTML file
+        """
+        report_path = self.output_dir / f"animation_report_{int(time.time())}.html"
+        html_content = self._build_html(title)
+        report_path.write_text(html_content)
+        return report_path
+
+    def _build_html(self, title: str) -> str:
+        """Build the HTML content for the report."""
+        # Collect all frames across stages
+        all_frames = []
+        for stage_name, stage in self._stages.items():
+            for frame in stage.frames:
+                all_frames.append(frame)
+
+        # Sort frames by timestamp
+        all_frames.sort(key=lambda f: f.timestamp)
+
+        # Build stage sections
+        stages_html = ""
+        for stage_name, stage in self._stages.items():
+            stages_html += self._build_stage_section(stage_name, stage)
+
+        # Build full HTML
+        html = f"""
+<!DOCTYPE html>
+<html lang="en">
+<head>
+    <meta charset="UTF-8">
+    <meta name="viewport" content="width=device-width, initial-scale=1.0">
+    <title>{title}</title>
+    <style>
+        * {{
+            box-sizing: border-box;
+            margin: 0;
+            padding: 0;
+        }}
+        body {{
+            font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', Roboto, sans-serif;
+            background: #1a1a2e;
+            color: #eee;
+            padding: 20px;
+            line-height: 1.6;
+        }}
+        .container {{
+            max-width: 1400px;
+            margin: 0 auto;
+        }}
+        .header {{
+            background: linear-gradient(135deg, #16213e 0%, #1a1a2e 100%);
+            padding: 30px;
+            border-radius: 12px;
+            margin-bottom: 30px;
+            text-align: center;
+            box-shadow: 0 4px 20px rgba(0,0,0,0.3);
+        }}
+        .header h1 {{
+            font-size: 2.5em;
+            margin-bottom: 10px;
+            background: linear-gradient(90deg, #00d4ff, #00ff88);
+            -webkit-background-clip: text;
+            -webkit-text-fill-color: transparent;
+            background-clip: text;
+        }}
+        .header .meta {{
+            color: #888;
+            font-size: 0.9em;
+        }}
+        .stats-grid {{
+            display: grid;
+            grid-template-columns: repeat(auto-fit, minmax(150px, 1fr));
+            gap: 15px;
+            margin: 20px 0;
+        }}
+        .stat-card {{
+            background: #16213e;
+            padding: 15px;
+            border-radius: 8px;
+            text-align: center;
+        }}
+        .stat-value {{
+            font-size: 1.8em;
+            font-weight: bold;
+            color: #00ff88;
+        }}
+        .stat-label {{
+            color: #888;
+            font-size: 0.85em;
+            margin-top: 5px;
+        }}
+        .stage-section {{
+            background: #16213e;
+            border-radius: 12px;
+            margin-bottom: 25px;
+            overflow: hidden;
+            box-shadow: 0 2px 10px rgba(0,0,0,0.2);
+        }}
+        .stage-header {{
+            background: #1f2a48;
+            padding: 15px 20px;
+            display: flex;
+            justify-content: space-between;
+            align-items: center;
+            cursor: pointer;
+            user-select: none;
+        }}
+        .stage-header:hover {{
+            background: #253252;
+        }}
+        .stage-name {{
+            font-weight: bold;
+            font-size: 1.1em;
+            color: #00d4ff;
+        }}
+        .stage-info {{
+            color: #888;
+            font-size: 0.9em;
+        }}
+        .stage-content {{
+            padding: 20px;
+        }}
+        .frames-container {{
+            display: grid;
+            grid-template-columns: repeat(auto-fill, minmax(300px, 1fr));
+            gap: 15px;
+        }}
+        .frame-card {{
+            background: #0f0f1a;
+            border-radius: 8px;
+            overflow: hidden;
+            border: 1px solid #333;
+            transition: transform 0.2s, box-shadow 0.2s;
+        }}
+        .frame-card:hover {{
+            transform: translateY(-2px);
+            box-shadow: 0 4px 15px rgba(0,212,255,0.2);
+        }}
+        .frame-header {{
+            background: #1a1a2e;
+            padding: 10px 15px;
+            font-size: 0.85em;
+            color: #888;
+            border-bottom: 1px solid #333;
+            display: flex;
+            justify-content: space-between;
+        }}
+        .frame-number {{
+            color: #00ff88;
+        }}
+        .frame-diff {{
+            color: #ff6b6b;
+        }}
+        .frame-content {{
+            padding: 10px;
+            font-family: 'Fira Code', 'Consolas', 'Monaco', monospace;
+            font-size: 11px;
+            line-height: 1.3;
+            white-space: pre;
+            overflow-x: auto;
+            max-height: 200px;
+            overflow-y: auto;
+        }}
+        .timeline-section {{
+            background: #16213e;
+            border-radius: 12px;
+            padding: 20px;
+            margin-bottom: 25px;
+        }}
+        .timeline-header {{
+            display: flex;
+            justify-content: space-between;
+            align-items: center;
+            margin-bottom: 15px;
+        }}
+        .timeline-title {{
+            font-weight: bold;
+            color: #00d4ff;
+        }}
+        .timeline-controls {{
+            display: flex;
+            gap: 10px;
+        }}
+        .timeline-controls button {{
+            background: #1f2a48;
+            border: 1px solid #333;
+            color: #eee;
+            padding: 8px 15px;
+            border-radius: 6px;
+            cursor: pointer;
+            transition: all 0.2s;
+        }}
+        .timeline-controls button:hover {{
+            background: #253252;
+            border-color: #00d4ff;
+        }}
+        .timeline-controls button.active {{
+            background: #00d4ff;
+            color: #000;
+        }}
+        .timeline-canvas {{
+            width: 100%;
+            height: 100px;
+            background: #0f0f1a;
+            border-radius: 8px;
+            position: relative;
+            overflow: hidden;
+        }}
+        .timeline-track {{
+            position: absolute;
+            top: 50%;
+            left: 0;
+            right: 0;
+            height: 4px;
+            background: #333;
+            transform: translateY(-50%);
+        }}
+        .timeline-marker {{
+            position: absolute;
+            top: 50%;
+            transform: translate(-50%, -50%);
+            width: 12px;
+            height: 12px;
+            background: #00d4ff;
+            border-radius: 50%;
+            cursor: pointer;
+            transition: all 0.2s;
+        }}
+        .timeline-marker:hover {{
+            transform: translate(-50%, -50%) scale(1.3);
+            box-shadow: 0 0 10px #00d4ff;
+        }}
+        .timeline-marker.stage-{{stage_name}} {{
+            background: var(--stage-color, #00d4ff);
+        }}
+        .comparison-view {{
+            display: grid;
+            grid-template-columns: 1fr 1fr;
+            gap: 20px;
+            margin-top: 20px;
+        }}
+        .comparison-panel {{
+            background: #0f0f1a;
+            border-radius: 8px;
+            padding: 15px;
+            border: 1px solid #333;
+        }}
+        .comparison-panel h4 {{
+            color: #888;
+            margin-bottom: 10px;
+            font-size: 0.9em;
+        }}
+        .comparison-content {{
+            font-family: 'Fira Code', 'Consolas', 'Monaco', monospace;
+            font-size: 11px;
+            line-height: 1.3;
+            white-space: pre;
+        }}
+        .diff-added {{
+            background: rgba(0, 255, 136, 0.2);
+        }}
+        .diff-removed {{
+            background: rgba(255, 107, 107, 0.2);
+        }}
+        @keyframes pulse {{
+            0%, 100% {{ opacity: 1; }}
+            50% {{ opacity: 0.7; }}
+        }}
+        .animating {{
+            animation: pulse 1s infinite;
+        }}
+        .footer {{
+            text-align: center;
+            color: #666;
+            padding: 20px;
+            font-size: 0.9em;
+        }}
+    </style>
+</head>
+<body>
+    <div class="container">
+        <div class="header">
+            <h1>🎬 {title}</h1>
+            <div class="meta">
+                Generated: {datetime.now().strftime("%Y-%m-%d %H:%M:%S")} |
+                Total Frames: {self._total_frames} |
+                Duration: {time.time() - self._start_time:.2f}s
+            </div>
+            <div class="stats-grid">
+                <div class="stat-card">
+                    <div class="stat-value">{len(self._stages)}</div>
+                    <div class="stat-label">Pipeline Stages</div>
+                </div>
+                <div class="stat-card">
+                    <div class="stat-value">{self._total_frames}</div>
+                    <div class="stat-label">Total Frames</div>
+                </div>
+                <div class="stat-card">
+                    <div class="stat-value">{time.time() - self._start_time:.2f}s</div>
+                    <div class="stat-label">Capture Duration</div>
+                </div>
+                <div class="stat-card">
+                    <div class="stat-value">{self.width}x{self.height}</div>
+                    <div class="stat-label">Resolution</div>
+                </div>
+            </div>
+        </div>
+
+        <div class="timeline-section">
+            <div class="timeline-header">
+                <div class="timeline-title">Timeline</div>
+                <div class="timeline-controls">
+                    <button onclick="playAnimation()">▶ Play</button>
+                    <button onclick="pauseAnimation()">⏸ Pause</button>
+                    <button onclick="stepForward()">⏭ Step</button>
+                </div>
+            </div>
+            <div class="timeline-canvas" id="timeline">
+                <div class="timeline-track"></div>
+                <!-- Timeline markers will be added by JavaScript -->
+            </div>
+        </div>
+
+        {stages_html}
+
+        <div class="footer">
+            <p>Animation Report generated by Mainline</p>
+            <p>Use the timeline controls above to play/pause the animation</p>
+        </div>
+    </div>
+
+    <script>
+        // Animation state
+        let currentFrame = 0;
+        let isPlaying = false;
+        let animationInterval = null;
+        const totalFrames = {len(all_frames)};
+
+        // Stage colors for timeline markers
+        const stageColors = {{
+            {self._build_stage_colors()}
+        }};
+
+        // Initialize timeline
+        function initTimeline() {{
+            const timeline = document.getElementById('timeline');
+            const track = timeline.querySelector('.timeline-track');
+
+            {self._build_timeline_markers(all_frames)}
+        }}
+
+        function playAnimation() {{
+            if (isPlaying) return;
+            isPlaying = true;
+            animationInterval = setInterval(() => {{
+                currentFrame = (currentFrame + 1) % totalFrames;
+                updateFrameDisplay();
+            }}, 100);
+        }}
+
+        function pauseAnimation() {{
+            isPlaying = false;
+            if (animationInterval) {{
+                clearInterval(animationInterval);
+                animationInterval = null;
+            }}
+        }}
+
+        function stepForward() {{
+            currentFrame = (currentFrame + 1) % totalFrames;
+            updateFrameDisplay();
+        }}
+
+        function updateFrameDisplay() {{
+            // Highlight current frame in timeline
+            const markers = document.querySelectorAll('.timeline-marker');
+            markers.forEach((marker, index) => {{
+                if (index === currentFrame) {{
+                    marker.style.transform = 'translate(-50%, -50%) scale(1.5)';
+                    marker.style.boxShadow = '0 0 15px #00ff88';
+                }} else {{
+                    marker.style.transform = 'translate(-50%, -50%) scale(1)';
+                    marker.style.boxShadow = 'none';
+                }}
+            }});
+        }}
+
+        // Initialize on page load
+        document.addEventListener('DOMContentLoaded', initTimeline);
+    </script>
+</body>
+</html>
+"""
+        return html
+
+    def _build_stage_section(self, stage_name: str, stage: StageCapture) -> str:
+        """Build HTML for a single stage section."""
+        frames_html = ""
+        for i, frame in enumerate(stage.frames):
+            diff_info = ""
+            if frame.diff_from_previous:
+                changed = frame.diff_from_previous.get("changed_lines", 0)
+                total = frame.diff_from_previous.get("total_lines", 0)
+                diff_info = f'<span class="frame-diff">Δ {changed}/{total}</span>'
+
+            frames_html += f"""
+            <div class="frame-card">
+                <div class="frame-header">
+                    <span>Frame <span class="frame-number">{frame.frame_number}</span></span>
+                    {diff_info}
+                </div>
+                <div class="frame-content">{self._escape_html("".join(frame.buffer))}</div>
+            </div>
+            """
+
+        return f"""
+        <div class="stage-section">
+            <div class="stage-header" onclick="this.nextElementSibling.classList.toggle('hidden')">
+                <span class="stage-name">{stage_name}</span>
+                <span class="stage-info">{len(stage.frames)} frames</span>
+            </div>
+            <div class="stage-content">
+                <div class="frames-container">
+                    {frames_html}
+                </div>
+            </div>
+        </div>
+        """
+
+    def _build_timeline(self, all_frames: list[CapturedFrame]) -> str:
+        """Build timeline HTML."""
+        if not all_frames:
+            return ""
+
+        markers_html = ""
+        for i, frame in enumerate(all_frames):
+            left_percent = (i / len(all_frames)) * 100
+            markers_html += f'<div class="timeline-marker" style="left: {left_percent}%" data-frame="{i}"></div>'
+
+        return markers_html
+
+    def _build_stage_colors(self) -> str:
+        """Build stage color mapping for JavaScript."""
+        colors = [
+            "#00d4ff",
+            "#00ff88",
+            "#ff6b6b",
+            "#ffd93d",
+            "#a855f7",
+            "#ec4899",
+            "#14b8a6",
+            "#f97316",
+            "#8b5cf6",
+            "#06b6d4",
+        ]
+        color_map = ""
+        for i, stage_name in enumerate(self._stages.keys()):
+            color = colors[i % len(colors)]
+            color_map += f'            "{stage_name}": "{color}",\n'
+        return color_map.rstrip(",\n")
+
+    def _build_timeline_markers(self, all_frames: list[CapturedFrame]) -> str:
+        """Build timeline markers in JavaScript."""
+        if not all_frames:
+            return ""
+
+        markers_js = ""
+        for i, frame in enumerate(all_frames):
+            left_percent = (i / len(all_frames)) * 100
+            stage_color = f"stageColors['{frame.stage}']"
+            markers_js += f"""
+            const marker{i} = document.createElement('div');
+            marker{i}.className = 'timeline-marker stage-{{frame.stage}}';
+            marker{i}.style.left = '{left_percent}%';
+            marker{i}.style.setProperty('--stage-color', {stage_color});
+            marker{i}.onclick = () => {{
+                currentFrame = {i};
+                updateFrameDisplay();
+            }};
+            timeline.appendChild(marker{i});
+            """
+
+        return markers_js
+
+    def _escape_html(self, text: str) -> str:
+        """Escape HTML special characters."""
+        return (
+            text.replace("&", "&amp;")
+            .replace("<", "&lt;")
+            .replace(">", "&gt;")
+            .replace('"', "&quot;")
+            .replace("'", "&#39;")
+        )

engine/display/backends/multi.py

@@ -0,0 +1,50 @@
+"""
+Multi display backend - forwards to multiple displays.
+"""
+
+
+class MultiDisplay:
+    """Display that forwards to multiple displays.
+
+    Supports reuse - passes reuse flag to all child displays.
+    """
+
+    width: int = 80
+    height: int = 24
+
+    def __init__(self, displays: list):
+        self.displays = displays
+        self.width = 80
+        self.height = 24
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize all child displays with dimensions.
+
+        Args:
+            width: Terminal width in characters
+            height: Terminal height in rows
+            reuse: If True, use reuse mode for child displays
+        """
+        self.width = width
+        self.height = height
+        for d in self.displays:
+            d.init(width, height, reuse=reuse)
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        for d in self.displays:
+            d.show(buffer, border=border)
+
+    def clear(self) -> None:
+        for d in self.displays:
+            d.clear()
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get dimensions from the first child display that supports it."""
+        for d in self.displays:
+            if hasattr(d, "get_dimensions"):
+                return d.get_dimensions()
+        return (self.width, self.height)
+
+    def cleanup(self) -> None:
+        for d in self.displays:
+            d.cleanup()

engine/display/backends/null.py

@@ -0,0 +1,183 @@
+"""
+Null/headless display backend.
+"""
+
+import json
+import time
+from pathlib import Path
+from typing import Any
+
+
+class NullDisplay:
+    """Headless/null display - discards all output.
+
+    This display does nothing - useful for headless benchmarking
+    or when no display output is needed. Captures last buffer
+    for testing purposes. Supports frame recording for replay
+    and file export/import.
+    """
+
+    width: int = 80
+    height: int = 24
+    _last_buffer: list[str] | None = None
+
+    def __init__(self):
+        self._last_buffer = None
+        self._is_recording = False
+        self._recorded_frames: list[dict[str, Any]] = []
+        self._frame_count = 0
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions.
+
+        Args:
+            width: Terminal width in characters
+            height: Terminal height in rows
+            reuse: Ignored for NullDisplay (no resources to reuse)
+        """
+        self.width = width
+        self.height = height
+        self._last_buffer = None
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        import sys
+
+        from engine.display import get_monitor, render_border
+
+        fps = 0.0
+        frame_time = 0.0
+        monitor = get_monitor()
+        if monitor:
+            stats = monitor.get_stats()
+            avg_ms = stats.get("pipeline", {}).get("avg_ms", 0) if stats else 0
+            frame_count = stats.get("frame_count", 0) if stats else 0
+            if avg_ms and frame_count > 0:
+                fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+        if border:
+            buffer = render_border(buffer, self.width, self.height, fps, frame_time)
+
+        self._last_buffer = buffer
+
+        if self._is_recording:
+            self._recorded_frames.append(
+                {
+                    "frame_number": self._frame_count,
+                    "buffer": buffer,
+                    "width": self.width,
+                    "height": self.height,
+                }
+            )
+
+        if self._frame_count <= 5 or self._frame_count % 10 == 0:
+            sys.stdout.write("\n" + "=" * 80 + "\n")
+            sys.stdout.write(
+                f"Frame {self._frame_count} (buffer height: {len(buffer)})\n"
+            )
+            sys.stdout.write("=" * 80 + "\n")
+            for i, line in enumerate(buffer[:30]):
+                sys.stdout.write(f"{i:2}: {line}\n")
+            if len(buffer) > 30:
+                sys.stdout.write(f"... ({len(buffer) - 30} more lines)\n")
+            sys.stdout.flush()
+
+        if monitor:
+            t0 = time.perf_counter()
+            chars_in = sum(len(line) for line in buffer)
+            elapsed_ms = (time.perf_counter() - t0) * 1000
+            monitor.record_effect("null_display", elapsed_ms, chars_in, chars_in)
+
+        self._frame_count += 1
+
+    def start_recording(self) -> None:
+        """Begin recording frames."""
+        self._is_recording = True
+        self._recorded_frames = []
+
+    def stop_recording(self) -> None:
+        """Stop recording frames."""
+        self._is_recording = False
+
+    def get_frames(self) -> list[list[str]]:
+        """Get recorded frames as list of buffers.
+
+        Returns:
+            List of buffers, each buffer is a list of strings (lines)
+        """
+        return [frame["buffer"] for frame in self._recorded_frames]
+
+    def get_recorded_data(self) -> list[dict[str, Any]]:
+        """Get full recorded data including metadata.
+
+        Returns:
+            List of frame dicts with 'frame_number', 'buffer', 'width', 'height'
+        """
+        return self._recorded_frames
+
+    def clear_recording(self) -> None:
+        """Clear recorded frames."""
+        self._recorded_frames = []
+
+    def save_recording(self, filepath: str | Path) -> None:
+        """Save recorded frames to a JSON file.
+
+        Args:
+            filepath: Path to save the recording
+        """
+        path = Path(filepath)
+        data = {
+            "version": 1,
+            "display": "null",
+            "width": self.width,
+            "height": self.height,
+            "frame_count": len(self._recorded_frames),
+            "frames": self._recorded_frames,
+        }
+        path.write_text(json.dumps(data, indent=2))
+
+    def load_recording(self, filepath: str | Path) -> list[dict[str, Any]]:
+        """Load recorded frames from a JSON file.
+
+        Args:
+            filepath: Path to load the recording from
+
+        Returns:
+            List of frame dicts
+        """
+        path = Path(filepath)
+        data = json.loads(path.read_text())
+        self._recorded_frames = data.get("frames", [])
+        self.width = data.get("width", 80)
+        self.height = data.get("height", 24)
+        return self._recorded_frames
+
+    def replay_frames(self) -> list[list[str]]:
+        """Get frames for replay.
+
+        Returns:
+            List of buffers for replay
+        """
+        return self.get_frames()
+
+    def clear(self) -> None:
+        pass
+
+    def cleanup(self) -> None:
+        pass
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current dimensions.
+
+        Returns:
+            (width, height) in character cells
+        """
+        return (self.width, self.height)
+
+    def is_quit_requested(self) -> bool:
+        """Check if quit was requested (optional protocol method)."""
+        return False
+
+    def clear_quit_request(self) -> None:
+        """Clear quit request (optional protocol method)."""
+        pass

engine/display/backends/pygame.py

@@ -0,0 +1,369 @@
+"""
+Pygame display backend - renders to a native application window.
+"""
+
+import time
+
+from engine.display.renderer import parse_ansi
+
+
+class PygameDisplay:
+    """Pygame display backend - renders to native window.
+
+    Supports reuse mode - when reuse=True, skips SDL initialization
+    and reuses the existing pygame window from a previous instance.
+    """
+
+    width: int = 80
+    window_width: int = 800
+    window_height: int = 600
+
+    def __init__(
+        self,
+        cell_width: int = 10,
+        cell_height: int = 18,
+        window_width: int = 800,
+        window_height: int = 600,
+        target_fps: float = 30.0,
+    ):
+        self.width = 80
+        self.height = 24
+        self.cell_width = cell_width
+        self.cell_height = cell_height
+        self.window_width = window_width
+        self.window_height = window_height
+        self.target_fps = target_fps
+        self._initialized = False
+        self._pygame = None
+        self._screen = None
+        self._font = None
+        self._resized = False
+        self._quit_requested = False
+        self._last_frame_time = 0.0
+        self._frame_period = 1.0 / target_fps if target_fps > 0 else 0
+        self._glyph_cache = {}
+
+    def _get_font_path(self) -> str | None:
+        """Get font path for rendering."""
+        import os
+        import sys
+        from pathlib import Path
+
+        env_font = os.environ.get("MAINLINE_PYGAME_FONT")
+        if env_font and os.path.exists(env_font):
+            return env_font
+
+        def search_dir(base_path: str) -> str | None:
+            if not os.path.exists(base_path):
+                return None
+            if os.path.isfile(base_path):
+                return base_path
+            for font_file in Path(base_path).rglob("*"):
+                if font_file.suffix.lower() in (".ttf", ".otf", ".ttc"):
+                    name = font_file.stem.lower()
+                    if "geist" in name and ("nerd" in name or "mono" in name):
+                        return str(font_file)
+            return None
+
+        search_dirs = []
+        if sys.platform == "darwin":
+            search_dirs.append(os.path.expanduser("~/Library/Fonts/"))
+        elif sys.platform == "win32":
+            search_dirs.append(
+                os.path.expanduser("~\\AppData\\Local\\Microsoft\\Windows\\Fonts\\")
+            )
+        else:
+            search_dirs.extend(
+                [
+                    os.path.expanduser("~/.local/share/fonts/"),
+                    os.path.expanduser("~/.fonts/"),
+                    "/usr/share/fonts/",
+                ]
+            )
+
+        for search_dir_path in search_dirs:
+            found = search_dir(search_dir_path)
+            if found:
+                return found
+
+        return None
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions.
+
+        Args:
+            width: Terminal width in characters
+            height: Terminal height in rows
+            reuse: If True, attach to existing pygame window instead of creating new
+        """
+        self.width = width
+        self.height = height
+
+        try:
+            import pygame
+        except ImportError:
+            return
+
+        if reuse and PygameDisplay._pygame_initialized:
+            self._pygame = pygame
+            self._initialized = True
+            return
+
+        pygame.init()
+        pygame.display.set_caption("Mainline")
+
+        self._screen = pygame.display.set_mode(
+            (self.window_width, self.window_height),
+            pygame.RESIZABLE,
+        )
+        self._pygame = pygame
+        PygameDisplay._pygame_initialized = True
+
+        # Calculate character dimensions from actual window size
+        self.width = max(1, self.window_width // self.cell_width)
+        self.height = max(1, self.window_height // self.cell_height)
+
+        font_path = self._get_font_path()
+        if font_path:
+            try:
+                self._font = pygame.font.Font(font_path, self.cell_height - 2)
+            except Exception:
+                self._font = pygame.font.SysFont("monospace", self.cell_height - 2)
+        else:
+            self._font = pygame.font.SysFont("monospace", self.cell_height - 2)
+
+        # Check if font supports box-drawing characters; if not, try to find one
+        self._use_fallback_border = False
+        if self._font:
+            try:
+                # Test rendering some key box-drawing characters
+                test_chars = ["┌", "─", "┐", "│", "└", "┘"]
+                for ch in test_chars:
+                    surf = self._font.render(ch, True, (255, 255, 255))
+                    # If surface is empty (width=0 or all black), font lacks glyph
+                    if surf.get_width() == 0:
+                        raise ValueError("Missing glyph")
+            except Exception:
+                # Font doesn't support box-drawing, will use line drawing fallback
+                self._use_fallback_border = True
+
+        self._initialized = True
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        if not self._initialized or not self._pygame:
+            return
+
+        t0 = time.perf_counter()
+
+        for event in self._pygame.event.get():
+            if event.type == self._pygame.QUIT:
+                self._quit_requested = True
+            elif event.type == self._pygame.KEYDOWN:
+                if event.key in (self._pygame.K_ESCAPE, self._pygame.K_c):
+                    if event.key == self._pygame.K_c and not (
+                        event.mod & self._pygame.KMOD_LCTRL
+                        or event.mod & self._pygame.KMOD_RCTRL
+                    ):
+                        continue
+                    self._quit_requested = True
+            elif event.type == self._pygame.VIDEORESIZE:
+                self.window_width = event.w
+                self.window_height = event.h
+                self.width = max(1, self.window_width // self.cell_width)
+                self.height = max(1, self.window_height // self.cell_height)
+                self._resized = True
+
+        # FPS limiting - skip frame if we're going too fast
+        if self._frame_period > 0:
+            now = time.perf_counter()
+            elapsed = now - self._last_frame_time
+            if elapsed < self._frame_period:
+                return  # Skip this frame
+            self._last_frame_time = now
+
+        # Get metrics for border display
+        fps = 0.0
+        frame_time = 0.0
+        from engine.display import get_monitor
+
+        monitor = get_monitor()
+        if monitor:
+            stats = monitor.get_stats()
+            avg_ms = stats.get("pipeline", {}).get("avg_ms", 0) if stats else 0
+            frame_count = stats.get("frame_count", 0) if stats else 0
+            if avg_ms and frame_count > 0:
+                fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+        self._screen.fill((0, 0, 0))
+
+        # If border requested but font lacks box-drawing glyphs, use graphical fallback
+        if border and self._use_fallback_border:
+            self._draw_fallback_border(fps, frame_time)
+            # Adjust content area to fit inside border
+            content_offset_x = self.cell_width
+            content_offset_y = self.cell_height
+            self.window_width - 2 * self.cell_width
+            self.window_height - 2 * self.cell_height
+        else:
+            # Normal rendering (with or without text border)
+            content_offset_x = 0
+            content_offset_y = 0
+
+            if border:
+                from engine.display import render_border
+
+                buffer = render_border(buffer, self.width, self.height, fps, frame_time)
+
+        blit_list = []
+
+        for row_idx, line in enumerate(buffer[: self.height]):
+            if row_idx >= self.height:
+                break
+
+            tokens = parse_ansi(line)
+            x_pos = content_offset_x
+
+            for text, fg, bg, _bold in tokens:
+                if not text:
+                    continue
+
+                # Use None as key for no background
+                bg_key = bg if bg != (0, 0, 0) else None
+                cache_key = (text, fg, bg_key)
+
+                if cache_key not in self._glyph_cache:
+                    # Render and cache
+                    if bg_key is not None:
+                        self._glyph_cache[cache_key] = self._font.render(
+                            text, True, fg, bg_key
+                        )
+                    else:
+                        self._glyph_cache[cache_key] = self._font.render(text, True, fg)
+
+                surface = self._glyph_cache[cache_key]
+                blit_list.append(
+                    (surface, (x_pos, content_offset_y + row_idx * self.cell_height))
+                )
+                x_pos += self._font.size(text)[0]
+
+        self._screen.blits(blit_list)
+
+        # Draw fallback border using graphics if needed
+        if border and self._use_fallback_border:
+            self._draw_fallback_border(fps, frame_time)
+
+        self._pygame.display.flip()
+
+        elapsed_ms = (time.perf_counter() - t0) * 1000
+
+        if monitor:
+            chars_in = sum(len(line) for line in buffer)
+            monitor.record_effect("pygame_display", elapsed_ms, chars_in, chars_in)
+
+    def _draw_fallback_border(self, fps: float, frame_time: float) -> None:
+        """Draw border using pygame graphics primitives instead of text."""
+        if not self._screen or not self._pygame:
+            return
+
+        # Colors
+        border_color = (0, 255, 0)  # Green (like terminal border)
+        text_color = (255, 255, 255)
+
+        # Calculate dimensions
+        x1 = 0
+        y1 = 0
+        x2 = self.window_width - 1
+        y2 = self.window_height - 1
+
+        # Draw outer rectangle
+        self._pygame.draw.rect(
+            self._screen, border_color, (x1, y1, x2 - x1 + 1, y2 - y1 + 1), 1
+        )
+
+        # Draw top border with FPS
+        if fps > 0:
+            fps_text = f" FPS:{fps:.0f}"
+        else:
+            fps_text = ""
+        # We need to render this text with a fallback font that has basic ASCII
+        # Use system font which should have these characters
+        try:
+            font = self._font  # May not have box chars but should have alphanumeric
+            text_surf = font.render(fps_text, True, text_color, (0, 0, 0))
+            text_rect = text_surf.get_rect()
+            # Position on top border, right-aligned
+            text_x = x2 - text_rect.width - 5
+            text_y = y1 + 2
+            self._screen.blit(text_surf, (text_x, text_y))
+        except Exception:
+            pass
+
+        # Draw bottom border with frame time
+        if frame_time > 0:
+            ft_text = f" {frame_time:.1f}ms"
+            try:
+                ft_surf = self._font.render(ft_text, True, text_color, (0, 0, 0))
+                ft_rect = ft_surf.get_rect()
+                ft_x = x2 - ft_rect.width - 5
+                ft_y = y2 - ft_rect.height - 2
+                self._screen.blit(ft_surf, (ft_x, ft_y))
+            except Exception:
+                pass
+
+    def clear(self) -> None:
+        if self._screen and self._pygame:
+            self._screen.fill((0, 0, 0))
+            self._pygame.display.flip()
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current terminal dimensions based on window size.
+
+        Returns:
+            (width, height) in character cells
+        """
+        # Query actual window size and recalculate character cells
+        if self._screen and self._pygame:
+            try:
+                w, h = self._screen.get_size()
+                if w != self.window_width or h != self.window_height:
+                    self.window_width = w
+                    self.window_height = h
+                    self.width = max(1, w // self.cell_width)
+                    self.height = max(1, h // self.cell_height)
+            except Exception:
+                pass
+        return self.width, self.height
+
+    def cleanup(self, quit_pygame: bool = True) -> None:
+        """Cleanup display resources.
+
+        Args:
+            quit_pygame: If True, quit pygame entirely. Set to False when
+                        reusing the display to avoid closing shared window.
+        """
+        if quit_pygame and self._pygame:
+            self._pygame.quit()
+            PygameDisplay._pygame_initialized = False
+
+    @classmethod
+    def reset_state(cls) -> None:
+        """Reset pygame state - useful for testing."""
+        cls._pygame_initialized = False
+
+    def is_quit_requested(self) -> bool:
+        """Check if user requested quit (Ctrl+C, Ctrl+Q, or Escape).
+
+        Returns True if the user pressed Ctrl+C, Ctrl+Q, or Escape.
+        The main loop should check this and raise KeyboardInterrupt.
+        """
+        return self._quit_requested
+
+    def clear_quit_request(self) -> bool:
+        """Clear the quit request flag after handling.
+
+        Returns the previous quit request state.
+        """
+        was_requested = self._quit_requested
+        self._quit_requested = False
+        return was_requested

engine/display/backends/replay.py

@@ -0,0 +1,122 @@
+"""
+Replay display backend - plays back recorded frames.
+"""
+
+from typing import Any
+
+
+class ReplayDisplay:
+    """Replay display - plays back recorded frames.
+
+    This display reads frames from a recording (list of frame data)
+    and yields them sequentially, useful for testing and demo purposes.
+    """
+
+    width: int = 80
+    height: int = 24
+
+    def __init__(self):
+        self._frames: list[dict[str, Any]] = []
+        self._current_frame = 0
+        self._playback_index = 0
+        self._loop = False
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions.
+
+        Args:
+            width: Terminal width in characters
+            height: Terminal height in rows
+            reuse: Ignored for ReplayDisplay
+        """
+        self.width = width
+        self.height = height
+
+    def set_frames(self, frames: list[dict[str, Any]]) -> None:
+        """Set frames to replay.
+
+        Args:
+            frames: List of frame dicts with 'buffer', 'width', 'height'
+        """
+        self._frames = frames
+        self._current_frame = 0
+        self._playback_index = 0
+
+    def set_loop(self, loop: bool) -> None:
+        """Set loop playback mode.
+
+        Args:
+            loop: True to loop, False to stop at end
+        """
+        self._loop = loop
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        """Display a frame (ignored in replay mode).
+
+        Args:
+            buffer: Buffer to display (ignored)
+            border: Border flag (ignored)
+        """
+        pass
+
+    def get_next_frame(self) -> list[str] | None:
+        """Get the next frame in the recording.
+
+        Returns:
+            Buffer list of strings, or None if playback is done
+        """
+        if not self._frames:
+            return None
+
+        if self._playback_index >= len(self._frames):
+            if self._loop:
+                self._playback_index = 0
+            else:
+                return None
+
+        frame = self._frames[self._playback_index]
+        self._playback_index += 1
+        return frame.get("buffer")
+
+    def reset(self) -> None:
+        """Reset playback to the beginning."""
+        self._playback_index = 0
+
+    def seek(self, index: int) -> None:
+        """Seek to a specific frame.
+
+        Args:
+            index: Frame index to seek to
+        """
+        if 0 <= index < len(self._frames):
+            self._playback_index = index
+
+    def is_finished(self) -> bool:
+        """Check if playback is finished.
+
+        Returns:
+            True if at end of frames and not looping
+        """
+        return not self._loop and self._playback_index >= len(self._frames)
+
+    def clear(self) -> None:
+        pass
+
+    def cleanup(self) -> None:
+        pass
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current dimensions.
+
+        Returns:
+            (width, height) in character cells
+        """
+        return (self.width, self.height)
+
+    def is_quit_requested(self) -> bool:
+        """Check if quit was requested (optional protocol method)."""
+        return False
+
+    def clear_quit_request(self) -> None:
+        """Clear quit request (optional protocol method)."""
+        pass

engine/display/backends/terminal.py

@@ -0,0 +1,262 @@
+"""
+ANSI terminal display backend.
+"""
+
+import os
+import select
+import sys
+import termios
+import tty
+
+
+class TerminalDisplay:
+    """ANSI terminal display backend.
+
+    Renders buffer to stdout using ANSI escape codes.
+    Supports reuse - when reuse=True, skips re-initializing terminal state.
+    Auto-detects terminal dimensions on init.
+    """
+
+    width: int = 80
+    height: int = 24
+    _initialized: bool = False
+
+    def __init__(self, target_fps: float = 30.0):
+        self.target_fps = target_fps
+        self._frame_period = 1.0 / target_fps if target_fps > 0 else 0
+        self._last_frame_time = 0.0
+        self._cached_dimensions: tuple[int, int] | None = None
+        self._raw_mode_enabled: bool = False
+        self._original_termios: list = []
+        self._quit_requested: bool = False
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions.
+
+        If width/height are not provided (0/None), auto-detects terminal size.
+        Otherwise uses provided dimensions or falls back to terminal size
+        if the provided dimensions exceed terminal capacity.
+
+        Args:
+            width: Desired terminal width (0 = auto-detect)
+            height: Desired terminal height (0 = auto-detect)
+            reuse: If True, skip terminal re-initialization
+        """
+        from engine.terminal import CURSOR_OFF
+
+        # Auto-detect terminal size (handle case where no terminal)
+        try:
+            term_size = os.get_terminal_size()
+            term_width = term_size.columns
+            term_height = term_size.lines
+        except OSError:
+            # No terminal available (e.g., in tests)
+            term_width = width if width > 0 else 80
+            term_height = height if height > 0 else 24
+
+        # Use provided dimensions if valid, otherwise use terminal size
+        if width > 0 and height > 0:
+            self.width = min(width, term_width)
+            self.height = min(height, term_height)
+        else:
+            self.width = term_width
+            self.height = term_height
+
+        if not reuse or not self._initialized:
+            print(CURSOR_OFF, end="", flush=True)
+            self._initialized = True
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current terminal dimensions.
+
+        Returns cached dimensions to avoid querying terminal every frame,
+        which can cause inconsistent results. Dimensions are only refreshed
+        when they actually change.
+
+        Returns:
+            (width, height) in character cells
+        """
+        try:
+            term_size = os.get_terminal_size()
+            new_dims = (term_size.columns, term_size.lines)
+        except OSError:
+            new_dims = (self.width, self.height)
+
+        # Only update cached dimensions if they actually changed
+        if self._cached_dimensions is None or self._cached_dimensions != new_dims:
+            self._cached_dimensions = new_dims
+            self.width = new_dims[0]
+            self.height = new_dims[1]
+
+        return self._cached_dimensions
+
+    def show(
+        self, buffer: list[str], border: bool = False, positioning: str = "mixed"
+    ) -> None:
+        """Display buffer with optional border and positioning mode.
+
+        Args:
+            buffer: List of lines to display
+            border: Whether to apply border
+            positioning: Positioning mode - "mixed" (default), "absolute", or "relative"
+        """
+        import sys
+
+        from engine.display import get_monitor, render_border
+
+        # Note: Frame rate limiting is handled by the caller (e.g., FrameTimer).
+        # This display renders every frame it receives.
+
+        # Get metrics for border display
+        fps = 0.0
+        frame_time = 0.0
+        monitor = get_monitor()
+        if monitor:
+            stats = monitor.get_stats()
+            avg_ms = stats.get("pipeline", {}).get("avg_ms", 0) if stats else 0
+            frame_count = stats.get("frame_count", 0) if stats else 0
+            if avg_ms and frame_count > 0:
+                fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+        # Apply border if requested
+        from engine.display import BorderMode
+
+        if border and border != BorderMode.OFF:
+            buffer = render_border(buffer, self.width, self.height, fps, frame_time)
+
+        # Apply positioning based on mode
+        if positioning == "absolute":
+            # All lines should have cursor positioning codes
+            # Join with newlines (cursor codes already in buffer)
+            output = "\033[H\033[J" + "\n".join(buffer)
+        elif positioning == "relative":
+            # Remove cursor positioning codes (except colors) and join with newlines
+            import re
+
+            cleaned_buffer = []
+            for line in buffer:
+                # Remove cursor positioning codes but keep color codes
+                # Pattern: \033[row;colH or \033[row;col;...H
+                cleaned = re.sub(r"\033\[[0-9;]*H", "", line)
+                cleaned_buffer.append(cleaned)
+            output = "\033[H\033[J" + "\n".join(cleaned_buffer)
+        else:  # mixed (default)
+            # Current behavior: join with newlines
+            # Effects that need absolute positioning have their own cursor codes
+            output = "\033[H\033[J" + "\n".join(buffer)
+
+        sys.stdout.buffer.write(output.encode())
+        sys.stdout.flush()
+
+    def clear(self) -> None:
+        from engine.terminal import CLR
+
+        print(CLR, end="", flush=True)
+
+    def cleanup(self) -> None:
+        from engine.terminal import CURSOR_ON
+
+        # Restore normal terminal mode if raw mode was enabled
+        self.set_raw_mode(False)
+
+        print(CURSOR_ON, end="", flush=True)
+
+    def is_quit_requested(self) -> bool:
+        """Check if quit was requested (optional protocol method)."""
+        return self._quit_requested
+
+    def clear_quit_request(self) -> None:
+        """Clear quit request (optional protocol method)."""
+        self._quit_requested = False
+
+    def request_quit(self) -> None:
+        """Request quit (e.g., when Ctrl+C is pressed)."""
+        self._quit_requested = True
+
+    def set_raw_mode(self, enable: bool = True) -> None:
+        """Enable/disable raw terminal mode for input capture.
+
+        When raw mode is enabled:
+        - Keystrokes are read immediately without echo
+        - Special keys (arrows, Ctrl+C, etc.) are captured
+        - Terminal is not in cooked/canonical mode
+
+        Args:
+            enable: True to enable raw mode, False to restore normal mode
+        """
+        try:
+            if enable and not self._raw_mode_enabled:
+                # Save original terminal settings
+                self._original_termios = termios.tcgetattr(sys.stdin)
+                # Set raw mode
+                tty.setraw(sys.stdin.fileno())
+                self._raw_mode_enabled = True
+            elif not enable and self._raw_mode_enabled:
+                # Restore original terminal settings
+                if self._original_termios:
+                    termios.tcsetattr(
+                        sys.stdin, termios.TCSADRAIN, self._original_termios
+                    )
+                self._raw_mode_enabled = False
+        except (termios.error, OSError):
+            # Terminal might not support raw mode (e.g., in tests)
+            pass
+
+    def get_input_keys(self, timeout: float = 0.0) -> list[str]:
+        """Get available keyboard input.
+
+        Reads available keystrokes from stdin. Should be called
+        with raw mode enabled for best results.
+
+        Args:
+            timeout: Maximum time to wait for input (seconds)
+
+        Returns:
+            List of key symbols as strings
+        """
+        keys = []
+
+        try:
+            # Check if input is available
+            if select.select([sys.stdin], [], [], timeout)[0]:
+                char = sys.stdin.read(1)
+
+                if char == "\x1b":  # Escape sequence
+                    # Read next character to determine key
+                    seq = sys.stdin.read(2)
+                    if seq == "[A":
+                        keys.append("up")
+                    elif seq == "[B":
+                        keys.append("down")
+                    elif seq == "[C":
+                        keys.append("right")
+                    elif seq == "[D":
+                        keys.append("left")
+                    else:
+                        # Unknown escape sequence
+                        keys.append("escape")
+                elif char == "\n" or char == "\r":
+                    keys.append("return")
+                elif char == "\t":
+                    keys.append("tab")
+                elif char == " ":
+                    keys.append(" ")
+                elif char == "\x7f" or char == "\x08":  # Backspace or Ctrl+H
+                    keys.append("backspace")
+                elif char == "\x03":  # Ctrl+C
+                    keys.append("ctrl_c")
+                elif char == "\x04":  # Ctrl+D
+                    keys.append("ctrl_d")
+                elif char == "\x1b":  # Escape
+                    keys.append("escape")
+                elif char.isprintable():
+                    keys.append(char)
+        except OSError:
+            pass
+
+        return keys
+
+    def is_raw_mode_enabled(self) -> bool:
+        """Check if raw mode is currently enabled."""
+        return self._raw_mode_enabled

engine/display/backends/websocket.py

@@ -0,0 +1,464 @@
+"""
+WebSocket display backend - broadcasts frame buffer to connected web clients.
+
+Supports streaming protocols:
+- Full frame (JSON) - default for compatibility
+- Binary streaming - efficient binary protocol
+- Diff streaming - only sends changed lines
+
+TODO: Transform to a true streaming backend with:
+- Proper WebSocket message streaming (currently sends full buffer each frame)
+- Connection pooling and backpressure handling
+- Binary protocol for efficiency (instead of JSON)
+- Client management with proper async handling
+- Mark for deprecation if replaced by a new streaming implementation
+
+Current implementation: Simple broadcast of text frames to all connected clients.
+"""
+
+import asyncio
+import base64
+import json
+import threading
+import time
+from enum import IntFlag
+
+from engine.display.streaming import (
+    MessageType,
+    compress_frame,
+    compute_diff,
+    encode_binary_message,
+    encode_diff_message,
+)
+
+
+class StreamingMode(IntFlag):
+    """Streaming modes for WebSocket display."""
+
+    JSON = 0x01  # Full JSON frames (default, compatible)
+    BINARY = 0x02  # Binary compression
+    DIFF = 0x04  # Differential updates
+
+
+try:
+    import websockets
+except ImportError:
+    websockets = None
+
+
+def get_monitor():
+    """Get the performance monitor."""
+    try:
+        from engine.effects.performance import get_monitor as _get_monitor
+
+        return _get_monitor()
+    except Exception:
+        return None
+
+
+class WebSocketDisplay:
+    """WebSocket display backend - broadcasts to HTML Canvas clients."""
+
+    width: int = 80
+    height: int = 24
+
+    def __init__(
+        self,
+        host: str = "0.0.0.0",
+        port: int = 8765,
+        http_port: int = 8766,
+        streaming_mode: StreamingMode = StreamingMode.JSON,
+    ):
+        self.host = host
+        self.port = port
+        self.http_port = http_port
+        self.width = 80
+        self.height = 24
+        self._clients: set = set()
+        self._server_running = False
+        self._http_running = False
+        self._server_thread: threading.Thread | None = None
+        self._http_thread: threading.Thread | None = None
+        self._available = True
+        self._max_clients = 10
+        self._client_connected_callback = None
+        self._client_disconnected_callback = None
+        self._command_callback = None
+        self._controller = None  # Reference to UI panel or pipeline controller
+        self._frame_delay = 0.0
+        self._httpd = None  # HTTP server instance
+
+        # Streaming configuration
+        self._streaming_mode = streaming_mode
+        self._last_buffer: list[str] = []
+        self._client_capabilities: dict = {}  # Track client capabilities
+
+        try:
+            import websockets as _ws
+
+            self._available = _ws is not None
+        except ImportError:
+            self._available = False
+
+    def is_available(self) -> bool:
+        """Check if WebSocket support is available."""
+        return self._available
+
+    def init(self, width: int, height: int, reuse: bool = False) -> None:
+        """Initialize display with dimensions and start server.
+
+        Args:
+            width: Terminal width in characters
+            height: Terminal height in rows
+            reuse: If True, skip starting servers (assume already running)
+        """
+        self.width = width
+        self.height = height
+
+        if not reuse or not self._server_running:
+            self.start_server()
+            self.start_http_server()
+
+    def show(self, buffer: list[str], border: bool = False) -> None:
+        """Broadcast buffer to all connected clients using streaming protocol."""
+        t0 = time.perf_counter()
+
+        # Get metrics for border display
+        fps = 0.0
+        frame_time = 0.0
+        monitor = get_monitor()
+        if monitor:
+            stats = monitor.get_stats()
+            avg_ms = stats.get("pipeline", {}).get("avg_ms", 0) if stats else 0
+            frame_count = stats.get("frame_count", 0) if stats else 0
+            if avg_ms and frame_count > 0:
+                fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+        # Apply border if requested
+        if border:
+            from engine.display import render_border
+
+            buffer = render_border(buffer, self.width, self.height, fps, frame_time)
+
+        if not self._clients:
+            self._last_buffer = buffer
+            return
+
+        # Send to each client based on their capabilities
+        disconnected = set()
+        for client in list(self._clients):
+            try:
+                client_id = id(client)
+                client_mode = self._client_capabilities.get(
+                    client_id, StreamingMode.JSON
+                )
+
+                if client_mode & StreamingMode.DIFF:
+                    self._send_diff_frame(client, buffer)
+                elif client_mode & StreamingMode.BINARY:
+                    self._send_binary_frame(client, buffer)
+                else:
+                    self._send_json_frame(client, buffer)
+            except Exception:
+                disconnected.add(client)
+
+        for client in disconnected:
+            self._clients.discard(client)
+            if self._client_disconnected_callback:
+                self._client_disconnected_callback(client)
+
+        self._last_buffer = buffer
+
+        elapsed_ms = (time.perf_counter() - t0) * 1000
+        if monitor:
+            chars_in = sum(len(line) for line in buffer)
+            monitor.record_effect("websocket_display", elapsed_ms, chars_in, chars_in)
+
+    def _send_json_frame(self, client, buffer: list[str]) -> None:
+        """Send frame as JSON."""
+        frame_data = {
+            "type": "frame",
+            "width": self.width,
+            "height": self.height,
+            "lines": buffer,
+        }
+        message = json.dumps(frame_data)
+        asyncio.run(client.send(message))
+
+    def _send_binary_frame(self, client, buffer: list[str]) -> None:
+        """Send frame as compressed binary."""
+        compressed = compress_frame(buffer)
+        message = encode_binary_message(
+            MessageType.FULL_FRAME, self.width, self.height, compressed
+        )
+        encoded = base64.b64encode(message).decode("utf-8")
+        asyncio.run(client.send(encoded))
+
+    def _send_diff_frame(self, client, buffer: list[str]) -> None:
+        """Send frame as diff."""
+        diff = compute_diff(self._last_buffer, buffer)
+
+        if not diff.changed_lines:
+            return
+
+        diff_payload = encode_diff_message(diff)
+        message = encode_binary_message(
+            MessageType.DIFF_FRAME, self.width, self.height, diff_payload
+        )
+        encoded = base64.b64encode(message).decode("utf-8")
+        asyncio.run(client.send(encoded))
+
+    def set_streaming_mode(self, mode: StreamingMode) -> None:
+        """Set the default streaming mode for new clients."""
+        self._streaming_mode = mode
+
+    def get_streaming_mode(self) -> StreamingMode:
+        """Get the current streaming mode."""
+        return self._streaming_mode
+
+    def clear(self) -> None:
+        """Broadcast clear command to all clients."""
+        if self._clients:
+            clear_data = {"type": "clear"}
+            message = json.dumps(clear_data)
+            for client in list(self._clients):
+                try:
+                    asyncio.run(client.send(message))
+                except Exception:
+                    pass
+
+    def cleanup(self) -> None:
+        """Stop the servers."""
+        self.stop_server()
+        self.stop_http_server()
+
+    async def _websocket_handler(self, websocket):
+        """Handle WebSocket connections."""
+        if len(self._clients) >= self._max_clients:
+            await websocket.close()
+            return
+
+        self._clients.add(websocket)
+        if self._client_connected_callback:
+            self._client_connected_callback(websocket)
+
+        try:
+            async for message in websocket:
+                try:
+                    data = json.loads(message)
+                    msg_type = data.get("type")
+
+                    if msg_type == "resize":
+                        self.width = data.get("width", 80)
+                        self.height = data.get("height", 24)
+                    elif msg_type == "command" and self._command_callback:
+                        # Forward commands to the pipeline controller
+                        command = data.get("command", {})
+                        self._command_callback(command)
+                    elif msg_type == "state_request":
+                        # Send current state snapshot
+                        state = self._get_state_snapshot()
+                        if state:
+                            response = {"type": "state", "state": state}
+                            await websocket.send(json.dumps(response))
+                except json.JSONDecodeError:
+                    pass
+        except Exception:
+            pass
+        finally:
+            self._clients.discard(websocket)
+            if self._client_disconnected_callback:
+                self._client_disconnected_callback(websocket)
+
+    async def _run_websocket_server(self):
+        """Run the WebSocket server."""
+        if not websockets:
+            return
+        async with websockets.serve(self._websocket_handler, self.host, self.port):
+            while self._server_running:
+                await asyncio.sleep(0.1)
+
+    async def _run_http_server(self):
+        """Run simple HTTP server for the client."""
+        import os
+        from http.server import HTTPServer, SimpleHTTPRequestHandler
+
+        # Find the project root by locating 'engine' directory in the path
+        websocket_file = os.path.abspath(__file__)
+        parts = websocket_file.split(os.sep)
+        if "engine" in parts:
+            engine_idx = parts.index("engine")
+            project_root = os.sep.join(parts[:engine_idx])
+            client_dir = os.path.join(project_root, "client")
+        else:
+            # Fallback: go up 4 levels from websocket.py
+            # websocket.py: .../engine/display/backends/websocket.py
+            # We need: .../client
+            client_dir = os.path.join(
+                os.path.dirname(
+                    os.path.dirname(os.path.dirname(os.path.dirname(__file__)))
+                ),
+                "client",
+            )
+
+        class Handler(SimpleHTTPRequestHandler):
+            def __init__(self, *args, **kwargs):
+                super().__init__(*args, directory=client_dir, **kwargs)
+
+            def log_message(self, format, *args):
+                pass
+
+        httpd = HTTPServer((self.host, self.http_port), Handler)
+        # Store reference for shutdown
+        self._httpd = httpd
+        # Serve requests continuously
+        httpd.serve_forever()
+
+    def _run_async(self, coro):
+        """Run coroutine in background."""
+        try:
+            asyncio.run(coro)
+        except Exception as e:
+            print(f"WebSocket async error: {e}")
+
+    def start_server(self):
+        """Start the WebSocket server in a background thread."""
+        if not self._available:
+            return
+        if self._server_thread is not None:
+            return
+
+        self._server_running = True
+        self._server_thread = threading.Thread(
+            target=self._run_async, args=(self._run_websocket_server(),), daemon=True
+        )
+        self._server_thread.start()
+
+    def stop_server(self):
+        """Stop the WebSocket server."""
+        self._server_running = False
+        self._server_thread = None
+
+    def start_http_server(self):
+        """Start the HTTP server in a background thread."""
+        if not self._available:
+            return
+        if self._http_thread is not None:
+            return
+
+        self._http_running = True
+
+        self._http_running = True
+        self._http_thread = threading.Thread(
+            target=self._run_async, args=(self._run_http_server(),), daemon=True
+        )
+        self._http_thread.start()
+
+    def stop_http_server(self):
+        """Stop the HTTP server."""
+        self._http_running = False
+        if hasattr(self, "_httpd") and self._httpd:
+            self._httpd.shutdown()
+        self._http_thread = None
+
+    def client_count(self) -> int:
+        """Return number of connected clients."""
+        return len(self._clients)
+
+    def get_ws_port(self) -> int:
+        """Return WebSocket port."""
+        return self.port
+
+    def get_http_port(self) -> int:
+        """Return HTTP port."""
+        return self.http_port
+
+    def set_frame_delay(self, delay: float) -> None:
+        """Set delay between frames in seconds."""
+        self._frame_delay = delay
+
+    def get_frame_delay(self) -> float:
+        """Get delay between frames."""
+        return self._frame_delay
+
+    def set_client_connected_callback(self, callback) -> None:
+        """Set callback for client connections."""
+        self._client_connected_callback = callback
+
+    def set_client_disconnected_callback(self, callback) -> None:
+        """Set callback for client disconnections."""
+        self._client_disconnected_callback = callback
+
+    def set_command_callback(self, callback) -> None:
+        """Set callback for incoming command messages from clients."""
+        self._command_callback = callback
+
+    def set_controller(self, controller) -> None:
+        """Set controller (UI panel or pipeline) for state queries and command execution."""
+        self._controller = controller
+
+    def broadcast_state(self, state: dict) -> None:
+        """Broadcast state update to all connected clients.
+
+        Args:
+            state: Dictionary containing state data to send to clients
+        """
+        if not self._clients:
+            return
+
+        message = json.dumps({"type": "state", "state": state})
+
+        disconnected = set()
+        for client in list(self._clients):
+            try:
+                asyncio.run(client.send(message))
+            except Exception:
+                disconnected.add(client)
+
+        for client in disconnected:
+            self._clients.discard(client)
+            if self._client_disconnected_callback:
+                self._client_disconnected_callback(client)
+
+    def _get_state_snapshot(self) -> dict | None:
+        """Get current state snapshot from controller."""
+        if not self._controller:
+            return None
+
+        try:
+            # Expect controller to have methods we need
+            state = {}
+
+            # Get stages info if UIPanel
+            if hasattr(self._controller, "stages"):
+                state["stages"] = {
+                    name: {
+                        "enabled": ctrl.enabled,
+                        "params": ctrl.params,
+                        "selected": ctrl.selected,
+                    }
+                    for name, ctrl in self._controller.stages.items()
+                }
+
+            # Get current preset
+            if hasattr(self._controller, "_current_preset"):
+                state["preset"] = self._controller._current_preset
+            if hasattr(self._controller, "_presets"):
+                state["presets"] = self._controller._presets
+
+            # Get selected stage
+            if hasattr(self._controller, "selected_stage"):
+                state["selected_stage"] = self._controller.selected_stage
+
+            return state
+        except Exception:
+            return None
+
+    def get_dimensions(self) -> tuple[int, int]:
+        """Get current dimensions.
+
+        Returns:
+            (width, height) in character cells
+        """
+        return (self.width, self.height)

engine/display/renderer.py

@@ -0,0 +1,280 @@
+"""
+Shared display rendering utilities.
+
+Provides common functionality for displays that render text to images
+(Pygame, Sixel, Kitty displays).
+"""
+
+from typing import Any
+
+ANSI_COLORS = {
+    0: (0, 0, 0),
+    1: (205, 49, 49),
+    2: (13, 188, 121),
+    3: (229, 229, 16),
+    4: (36, 114, 200),
+    5: (188, 63, 188),
+    6: (17, 168, 205),
+    7: (229, 229, 229),
+    8: (102, 102, 102),
+    9: (241, 76, 76),
+    10: (35, 209, 139),
+    11: (245, 245, 67),
+    12: (59, 142, 234),
+    13: (214, 112, 214),
+    14: (41, 184, 219),
+    15: (255, 255, 255),
+}
+
+
+def parse_ansi(
+    text: str,
+) -> list[tuple[str, tuple[int, int, int], tuple[int, int, int], bool]]:
+    """Parse ANSI escape sequences into text tokens with colors.
+
+    Args:
+        text: Text containing ANSI escape sequences
+
+    Returns:
+        List of (text, fg_rgb, bg_rgb, bold) tuples
+    """
+    tokens = []
+    current_text = ""
+    fg = (204, 204, 204)
+    bg = (0, 0, 0)
+    bold = False
+    i = 0
+
+    ANSI_COLORS_4BIT = {
+        0: (0, 0, 0),
+        1: (205, 49, 49),
+        2: (13, 188, 121),
+        3: (229, 229, 16),
+        4: (36, 114, 200),
+        5: (188, 63, 188),
+        6: (17, 168, 205),
+        7: (229, 229, 229),
+        8: (102, 102, 102),
+        9: (241, 76, 76),
+        10: (35, 209, 139),
+        11: (245, 245, 67),
+        12: (59, 142, 234),
+        13: (214, 112, 214),
+        14: (41, 184, 219),
+        15: (255, 255, 255),
+    }
+
+    while i < len(text):
+        char = text[i]
+
+        if char == "\x1b" and i + 1 < len(text) and text[i + 1] == "[":
+            if current_text:
+                tokens.append((current_text, fg, bg, bold))
+                current_text = ""
+
+            i += 2
+            code = ""
+            while i < len(text):
+                c = text[i]
+                if c.isalpha():
+                    break
+                code += c
+                i += 1
+
+            if code:
+                codes = code.split(";")
+                for c in codes:
+                    if c == "0":
+                        fg = (204, 204, 204)
+                        bg = (0, 0, 0)
+                        bold = False
+                    elif c == "1":
+                        bold = True
+                    elif c == "22":
+                        bold = False
+                    elif c == "39":
+                        fg = (204, 204, 204)
+                    elif c == "49":
+                        bg = (0, 0, 0)
+                    elif c.isdigit():
+                        color_idx = int(c)
+                        if color_idx in ANSI_COLORS_4BIT:
+                            fg = ANSI_COLORS_4BIT[color_idx]
+                        elif 30 <= color_idx <= 37:
+                            fg = ANSI_COLORS_4BIT.get(color_idx - 30, fg)
+                        elif 40 <= color_idx <= 47:
+                            bg = ANSI_COLORS_4BIT.get(color_idx - 40, bg)
+                        elif 90 <= color_idx <= 97:
+                            fg = ANSI_COLORS_4BIT.get(color_idx - 90 + 8, fg)
+                        elif 100 <= color_idx <= 107:
+                            bg = ANSI_COLORS_4BIT.get(color_idx - 100 + 8, bg)
+                    elif c.startswith("38;5;"):
+                        idx = int(c.split(";")[-1])
+                        if idx < 256:
+                            if idx < 16:
+                                fg = ANSI_COLORS_4BIT.get(idx, fg)
+                            elif idx < 232:
+                                c_idx = idx - 16
+                                fg = (
+                                    (c_idx >> 4) * 51,
+                                    ((c_idx >> 2) & 7) * 51,
+                                    (c_idx & 3) * 85,
+                                )
+                            else:
+                                gray = (idx - 232) * 10 + 8
+                                fg = (gray, gray, gray)
+                    elif c.startswith("48;5;"):
+                        idx = int(c.split(";")[-1])
+                        if idx < 256:
+                            if idx < 16:
+                                bg = ANSI_COLORS_4BIT.get(idx, bg)
+                            elif idx < 232:
+                                c_idx = idx - 16
+                                bg = (
+                                    (c_idx >> 4) * 51,
+                                    ((c_idx >> 2) & 7) * 51,
+                                    (c_idx & 3) * 85,
+                                )
+                            else:
+                                gray = (idx - 232) * 10 + 8
+                                bg = (gray, gray, gray)
+            i += 1
+        else:
+            current_text += char
+            i += 1
+
+    if current_text:
+        tokens.append((current_text, fg, bg, bold))
+
+    return tokens if tokens else [("", fg, bg, bold)]
+
+
+def get_default_font_path() -> str | None:
+    """Get the path to a default monospace font."""
+    import os
+    import sys
+    from pathlib import Path
+
+    def search_dir(base_path: str) -> str | None:
+        if not os.path.exists(base_path):
+            return None
+        if os.path.isfile(base_path):
+            return base_path
+        for font_file in Path(base_path).rglob("*"):
+            if font_file.suffix.lower() in (".ttf", ".otf", ".ttc"):
+                name = font_file.stem.lower()
+                if "geist" in name and ("nerd" in name or "mono" in name):
+                    return str(font_file)
+                if "mono" in name or "courier" in name or "terminal" in name:
+                    return str(font_file)
+        return None
+
+    search_dirs = []
+    if sys.platform == "darwin":
+        search_dirs.extend(
+            [
+                os.path.expanduser("~/Library/Fonts/"),
+                "/System/Library/Fonts/",
+            ]
+        )
+    elif sys.platform == "win32":
+        search_dirs.extend(
+            [
+                os.path.expanduser("~\\AppData\\Local\\Microsoft\\Windows\\Fonts\\"),
+                "C:\\Windows\\Fonts\\",
+            ]
+        )
+    else:
+        search_dirs.extend(
+            [
+                os.path.expanduser("~/.local/share/fonts/"),
+                os.path.expanduser("~/.fonts/"),
+                "/usr/share/fonts/",
+            ]
+        )
+
+    for search_dir_path in search_dirs:
+        found = search_dir(search_dir_path)
+        if found:
+            return found
+
+    if sys.platform != "win32":
+        try:
+            import subprocess
+
+            for pattern in ["monospace", "DejaVuSansMono", "LiberationMono"]:
+                result = subprocess.run(
+                    ["fc-match", "-f", "%{file}", pattern],
+                    capture_output=True,
+                    text=True,
+                    timeout=5,
+                )
+                if result.returncode == 0 and result.stdout.strip():
+                    font_file = result.stdout.strip()
+                    if os.path.exists(font_file):
+                        return font_file
+        except Exception:
+            pass
+
+    return None
+
+
+def render_to_pil(
+    buffer: list[str],
+    width: int,
+    height: int,
+    cell_width: int = 10,
+    cell_height: int = 18,
+    font_path: str | None = None,
+) -> Any:
+    """Render buffer to a PIL Image.
+
+    Args:
+        buffer: List of text lines to render
+        width: Terminal width in characters
+        height: Terminal height in rows
+        cell_width: Width of each character cell in pixels
+        cell_height: Height of each character cell in pixels
+        font_path: Path to TTF/OTF font file (optional)
+
+    Returns:
+        PIL Image object
+    """
+    from PIL import Image, ImageDraw, ImageFont
+
+    img_width = width * cell_width
+    img_height = height * cell_height
+
+    img = Image.new("RGBA", (img_width, img_height), (0, 0, 0, 255))
+    draw = ImageDraw.Draw(img)
+
+    if font_path:
+        try:
+            font = ImageFont.truetype(font_path, cell_height - 2)
+        except Exception:
+            font = ImageFont.load_default()
+    else:
+        font = ImageFont.load_default()
+
+    for row_idx, line in enumerate(buffer[:height]):
+        if row_idx >= height:
+            break
+
+        tokens = parse_ansi(line)
+        x_pos = 0
+        y_pos = row_idx * cell_height
+
+        for text, fg, bg, _bold in tokens:
+            if not text:
+                continue
+
+            if bg != (0, 0, 0):
+                bbox = draw.textbbox((x_pos, y_pos), text, font=font)
+                draw.rectangle(bbox, fill=(*bg, 255))
+
+            draw.text((x_pos, y_pos), text, fill=(*fg, 255), font=font)
+
+            if font:
+                x_pos += draw.textlength(text, font=font)
+
+    return img

engine/display/streaming.py

@@ -0,0 +1,268 @@
+"""
+Streaming protocol utilities for efficient frame transmission.
+
+Provides:
+- Frame differencing: Only send changed lines
+- Run-length encoding: Compress repeated lines
+- Binary encoding: Compact message format
+"""
+
+import json
+import zlib
+from dataclasses import dataclass
+from enum import IntEnum
+
+
+class MessageType(IntEnum):
+    """Message types for streaming protocol."""
+
+    FULL_FRAME = 1
+    DIFF_FRAME = 2
+    STATE = 3
+    CLEAR = 4
+    PING = 5
+    PONG = 6
+
+
+@dataclass
+class FrameDiff:
+    """Represents a diff between two frames."""
+
+    width: int
+    height: int
+    changed_lines: list[tuple[int, str]]  # (line_index, content)
+
+
+def compute_diff(old_buffer: list[str], new_buffer: list[str]) -> FrameDiff:
+    """Compute differences between old and new buffer.
+
+    Args:
+        old_buffer: Previous frame buffer
+        new_buffer: Current frame buffer
+
+    Returns:
+        FrameDiff with only changed lines
+    """
+    height = len(new_buffer)
+    changed_lines = []
+
+    for i, line in enumerate(new_buffer):
+        if i >= len(old_buffer) or line != old_buffer[i]:
+            changed_lines.append((i, line))
+
+    return FrameDiff(
+        width=len(new_buffer[0]) if new_buffer else 0,
+        height=height,
+        changed_lines=changed_lines,
+    )
+
+
+def encode_rle(lines: list[tuple[int, str]]) -> list[tuple[int, str, int]]:
+    """Run-length encode consecutive identical lines.
+
+    Args:
+        lines: List of (index, content) tuples (must be sorted by index)
+
+    Returns:
+        List of (start_index, content, run_length) tuples
+    """
+    if not lines:
+        return []
+
+    encoded = []
+    start_idx = lines[0][0]
+    current_line = lines[0][1]
+    current_rle = 1
+
+    for idx, line in lines[1:]:
+        if line == current_line:
+            current_rle += 1
+        else:
+            encoded.append((start_idx, current_line, current_rle))
+            start_idx = idx
+            current_line = line
+            current_rle = 1
+
+    encoded.append((start_idx, current_line, current_rle))
+    return encoded
+
+
+def decode_rle(encoded: list[tuple[int, str, int]]) -> list[tuple[int, str]]:
+    """Decode run-length encoded lines.
+
+    Args:
+        encoded: List of (start_index, content, run_length) tuples
+
+    Returns:
+        List of (index, content) tuples
+    """
+    result = []
+    for start_idx, line, rle in encoded:
+        for i in range(rle):
+            result.append((start_idx + i, line))
+    return result
+
+
+def compress_frame(buffer: list[str], level: int = 6) -> bytes:
+    """Compress a frame buffer using zlib.
+
+    Args:
+        buffer: Frame buffer (list of lines)
+        level: Compression level (0-9)
+
+    Returns:
+        Compressed bytes
+    """
+    content = "\n".join(buffer)
+    return zlib.compress(content.encode("utf-8"), level)
+
+
+def decompress_frame(data: bytes, height: int) -> list[str]:
+    """Decompress a frame buffer.
+
+    Args:
+        data: Compressed bytes
+        height: Number of lines in original buffer
+
+    Returns:
+        Frame buffer (list of lines)
+    """
+    content = zlib.decompress(data).decode("utf-8")
+    lines = content.split("\n")
+    if len(lines) > height:
+        lines = lines[:height]
+    while len(lines) < height:
+        lines.append("")
+    return lines
+
+
+def encode_binary_message(
+    msg_type: MessageType, width: int, height: int, payload: bytes
+) -> bytes:
+    """Encode a binary message.
+
+    Message format:
+    - 1 byte: message type
+    - 2 bytes: width (uint16)
+    - 2 bytes: height (uint16)
+    - 4 bytes: payload length (uint32)
+    - N bytes: payload
+
+    Args:
+        msg_type: Message type
+        width: Frame width
+        height: Frame height
+        payload: Message payload
+
+    Returns:
+        Encoded binary message
+    """
+    import struct
+
+    header = struct.pack("!BHHI", msg_type.value, width, height, len(payload))
+    return header + payload
+
+
+def decode_binary_message(data: bytes) -> tuple[MessageType, int, int, bytes]:
+    """Decode a binary message.
+
+    Args:
+        data: Binary message data
+
+    Returns:
+        Tuple of (msg_type, width, height, payload)
+    """
+    import struct
+
+    msg_type_val, width, height, payload_len = struct.unpack("!BHHI", data[:9])
+    payload = data[9 : 9 + payload_len]
+    return MessageType(msg_type_val), width, height, payload
+
+
+def encode_diff_message(diff: FrameDiff, use_rle: bool = True) -> bytes:
+    """Encode a diff message for transmission.
+
+    Args:
+        diff: Frame diff
+        use_rle: Whether to use run-length encoding
+
+    Returns:
+        Encoded diff payload
+    """
+
+    if use_rle:
+        encoded_lines = encode_rle(diff.changed_lines)
+        data = [[idx, line, rle] for idx, line, rle in encoded_lines]
+    else:
+        data = [[idx, line] for idx, line in diff.changed_lines]
+
+    payload = json.dumps(data).encode("utf-8")
+    return payload
+
+
+def decode_diff_message(payload: bytes, use_rle: bool = True) -> list[tuple[int, str]]:
+    """Decode a diff message.
+
+    Args:
+        payload: Encoded diff payload
+        use_rle: Whether run-length encoding was used
+
+    Returns:
+        List of (line_index, content) tuples
+    """
+
+    data = json.loads(payload.decode("utf-8"))
+
+    if use_rle:
+        return decode_rle([(idx, line, rle) for idx, line, rle in data])
+    else:
+        return [(idx, line) for idx, line in data]
+
+
+def should_use_diff(
+    old_buffer: list[str], new_buffer: list[str], threshold: float = 0.3
+) -> bool:
+    """Determine if diff or full frame is more efficient.
+
+    Args:
+        old_buffer: Previous frame
+        new_buffer: Current frame
+        threshold: Max changed ratio to use diff (0.0-1.0)
+
+    Returns:
+        True if diff is more efficient
+    """
+    if not old_buffer or not new_buffer:
+        return False
+
+    diff = compute_diff(old_buffer, new_buffer)
+    total_lines = len(new_buffer)
+    changed_ratio = len(diff.changed_lines) / total_lines if total_lines > 0 else 1.0
+
+    return changed_ratio <= threshold
+
+
+def apply_diff(old_buffer: list[str], diff: FrameDiff) -> list[str]:
+    """Apply a diff to an old buffer to get the new buffer.
+
+    Args:
+        old_buffer: Previous frame buffer
+        diff: Frame diff to apply
+
+    Returns:
+        New frame buffer
+    """
+    new_buffer = list(old_buffer)
+
+    for line_idx, content in diff.changed_lines:
+        if line_idx < len(new_buffer):
+            new_buffer[line_idx] = content
+        else:
+            while len(new_buffer) < line_idx:
+                new_buffer.append("")
+            new_buffer.append(content)
+
+    while len(new_buffer) < diff.height:
+        new_buffer.append("")
+
+    return new_buffer[: diff.height]

engine/effects/__init__.py

@@ -0,0 +1,42 @@
+from engine.effects.chain import EffectChain
+from engine.effects.controller import handle_effects_command, show_effects_menu
+from engine.effects.legacy import (
+    fade_line,
+    firehose_line,
+    glitch_bar,
+    next_headline,
+    noise,
+    vis_offset,
+    vis_trunc,
+)
+from engine.effects.performance import PerformanceMonitor, get_monitor, set_monitor
+from engine.effects.registry import EffectRegistry, get_registry, set_registry
+from engine.effects.types import (
+    EffectConfig,
+    EffectContext,
+    PipelineConfig,
+    create_effect_context,
+)
+
+__all__ = [
+    "EffectChain",
+    "EffectRegistry",
+    "EffectConfig",
+    "EffectContext",
+    "PipelineConfig",
+    "create_effect_context",
+    "get_registry",
+    "set_registry",
+    "get_monitor",
+    "set_monitor",
+    "PerformanceMonitor",
+    "handle_effects_command",
+    "show_effects_menu",
+    "fade_line",
+    "firehose_line",
+    "glitch_bar",
+    "noise",
+    "next_headline",
+    "vis_trunc",
+    "vis_offset",
+]

engine/effects/chain.py

@@ -0,0 +1,87 @@
+import time
+
+from engine.effects.performance import PerformanceMonitor, get_monitor
+from engine.effects.registry import EffectRegistry
+from engine.effects.types import EffectContext, PartialUpdate
+
+
+class EffectChain:
+    def __init__(
+        self, registry: EffectRegistry, monitor: PerformanceMonitor | None = None
+    ):
+        self._registry = registry
+        self._order: list[str] = []
+        self._monitor = monitor
+
+    def _get_monitor(self) -> PerformanceMonitor:
+        if self._monitor is not None:
+            return self._monitor
+        return get_monitor()
+
+    def set_order(self, names: list[str]) -> None:
+        self._order = list(names)
+
+    def get_order(self) -> list[str]:
+        return self._order.copy()
+
+    def add_effect(self, name: str, position: int | None = None) -> bool:
+        if name not in self._registry.list_all():
+            return False
+        if position is None:
+            self._order.append(name)
+        else:
+            self._order.insert(position, name)
+        return True
+
+    def remove_effect(self, name: str) -> bool:
+        if name in self._order:
+            self._order.remove(name)
+            return True
+        return False
+
+    def reorder(self, new_order: list[str]) -> bool:
+        all_plugins = set(self._registry.list_all().keys())
+        if not all(name in all_plugins for name in new_order):
+            return False
+        self._order = list(new_order)
+        return True
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        monitor = self._get_monitor()
+        frame_number = ctx.frame_number
+        monitor.start_frame(frame_number)
+
+        # Get dirty regions from canvas via context (set by CanvasStage)
+        dirty_rows = ctx.get_state("canvas.dirty_rows")
+
+        # Create PartialUpdate for effects that support it
+        full_buffer = dirty_rows is None or len(dirty_rows) == 0
+        partial = PartialUpdate(
+            rows=None,
+            cols=None,
+            dirty=dirty_rows,
+            full_buffer=full_buffer,
+        )
+
+        frame_start = time.perf_counter()
+        result = list(buf)
+        for name in self._order:
+            plugin = self._registry.get(name)
+            if plugin and plugin.config.enabled:
+                chars_in = sum(len(line) for line in result)
+                effect_start = time.perf_counter()
+                try:
+                    # Use process_partial if supported, otherwise fall back to process
+                    if getattr(plugin, "supports_partial_updates", False):
+                        result = plugin.process_partial(result, ctx, partial)
+                    else:
+                        result = plugin.process(result, ctx)
+                except Exception:
+                    plugin.config.enabled = False
+                elapsed = time.perf_counter() - effect_start
+                chars_out = sum(len(line) for line in result)
+                monitor.record_effect(name, elapsed * 1000, chars_in, chars_out)
+
+        total_elapsed = time.perf_counter() - frame_start
+        monitor.end_frame(frame_number, total_elapsed * 1000)
+        return result

engine/effects/controller.py

@@ -0,0 +1,137 @@
+from engine.effects.performance import get_monitor
+from engine.effects.registry import get_registry
+
+_effect_chain_ref = None
+
+
+def _get_effect_chain():
+    global _effect_chain_ref
+    return _effect_chain_ref
+
+
+def set_effect_chain_ref(chain) -> None:
+    global _effect_chain_ref
+    _effect_chain_ref = chain
+
+
+def handle_effects_command(cmd: str) -> str:
+    """Handle /effects command from NTFY message.
+
+    Commands:
+        /effects list - list all effects and their status
+        /effects <name> on - enable an effect
+        /effects <name> off - disable an effect
+        /effects <name> intensity <0.0-1.0> - set intensity
+        /effects reorder <name1>,<name2>,... - reorder pipeline
+        /effects stats - show performance statistics
+    """
+    parts = cmd.strip().split()
+    if not parts or parts[0] != "/effects":
+        return "Unknown command"
+
+    registry = get_registry()
+    chain = _get_effect_chain()
+
+    if len(parts) == 1 or parts[1] == "list":
+        result = ["Effects:"]
+        for name, plugin in registry.list_all().items():
+            status = "ON" if plugin.config.enabled else "OFF"
+            intensity = plugin.config.intensity
+            result.append(f"  {name}: {status} (intensity={intensity})")
+        if chain:
+            result.append(f"Order: {chain.get_order()}")
+        return "\n".join(result)
+
+    if parts[1] == "stats":
+        return _format_stats()
+
+    if parts[1] == "reorder" and len(parts) >= 3:
+        new_order = parts[2].split(",")
+        if chain and chain.reorder(new_order):
+            return f"Reordered pipeline: {new_order}"
+        return "Failed to reorder pipeline"
+
+    if len(parts) < 3:
+        return "Usage: /effects <name> on|off|intensity <value>"
+
+    effect_name = parts[1]
+    action = parts[2]
+
+    if effect_name not in registry.list_all():
+        return f"Unknown effect: {effect_name}"
+
+    if action == "on":
+        registry.enable(effect_name)
+        return f"Enabled: {effect_name}"
+
+    if action == "off":
+        registry.disable(effect_name)
+        return f"Disabled: {effect_name}"
+
+    if action == "intensity" and len(parts) >= 4:
+        try:
+            value = float(parts[3])
+            if not 0.0 <= value <= 1.0:
+                return "Intensity must be between 0.0 and 1.0"
+            plugin = registry.get(effect_name)
+            if plugin:
+                plugin.config.intensity = value
+                return f"Set {effect_name} intensity to {value}"
+        except ValueError:
+            return "Invalid intensity value"
+
+    return f"Unknown action: {action}"
+
+
+def _format_stats() -> str:
+    monitor = get_monitor()
+    stats = monitor.get_stats()
+
+    if "error" in stats:
+        return stats["error"]
+
+    lines = ["Performance Stats:"]
+
+    pipeline = stats["pipeline"]
+    lines.append(
+        f"  Pipeline: avg={pipeline['avg_ms']:.2f}ms min={pipeline['min_ms']:.2f}ms max={pipeline['max_ms']:.2f}ms (over {stats['frame_count']} frames)"
+    )
+
+    if stats["effects"]:
+        lines.append("  Per-effect (avg ms):")
+        for name, effect_stats in stats["effects"].items():
+            lines.append(
+                f"    {name}: avg={effect_stats['avg_ms']:.2f}ms min={effect_stats['min_ms']:.2f}ms max={effect_stats['max_ms']:.2f}ms"
+            )
+
+    return "\n".join(lines)
+
+
+def show_effects_menu() -> str:
+    """Generate effects menu text for display."""
+    registry = get_registry()
+    chain = _get_effect_chain()
+
+    lines = [
+        "\033[1;38;5;231m=== EFFECTS MENU ===\033[0m",
+        "",
+        "Effects:",
+    ]
+
+    for name, plugin in registry.list_all().items():
+        status = "ON" if plugin.config.enabled else "OFF"
+        intensity = plugin.config.intensity
+        lines.append(f"  [{status:3}] {name}: intensity={intensity:.2f}")
+
+    if chain:
+        lines.append("")
+        lines.append(f"Pipeline order: {' -> '.join(chain.get_order())}")
+
+    lines.append("")
+    lines.append("Controls:")
+    lines.append("  /effects <name> on|off")
+    lines.append("  /effects <name> intensity <0.0-1.0>")
+    lines.append("  /effects reorder name1,name2,...")
+    lines.append("")
+
+    return "\n".join(lines)

engine/effects/legacy.py

@@ -0,0 +1,173 @@
+"""
+Visual effects: noise, glitch, fade, ANSI-aware truncation, firehose, headline pool.
+Depends on: config, terminal, sources.
+
+These are low-level functional implementations of visual effects. They are used
+internally by the EffectPlugin system (effects_plugins/*.py) and also directly
+by layers.py and scroll.py for rendering.
+
+The plugin system provides a higher-level OOP interface with configuration
+support, while these legacy functions provide direct functional access.
+Both systems coexist - there are no current plans to deprecate the legacy functions.
+"""
+
+import random
+from datetime import datetime
+
+from engine import config
+from engine.sources import FEEDS, POETRY_SOURCES
+from engine.terminal import C_DIM, DIM, G_DIM, G_LO, RST, W_GHOST
+
+
+def noise(w):
+    d = random.choice([0.15, 0.25, 0.35, 0.12])
+    return "".join(
+        f"{random.choice([G_LO, G_DIM, C_DIM, W_GHOST])}"
+        f"{random.choice(config.GLITCH + config.KATA)}{RST}"
+        if random.random() < d
+        else " "
+        for _ in range(w)
+    )
+
+
+def glitch_bar(w):
+    c = random.choice(["░", "▒", "─", "╌"])
+    n = random.randint(3, w // 2)
+    o = random.randint(0, w - n)
+    return " " * o + f"{G_LO}{DIM}" + c * n + RST
+
+
+def fade_line(s, fade):
+    """Dissolve a rendered line by probabilistically dropping characters."""
+    if fade >= 1.0:
+        return s
+    if fade <= 0.0:
+        return ""
+    result = []
+    i = 0
+    while i < len(s):
+        if s[i] == "\033" and i + 1 < len(s) and s[i + 1] == "[":
+            j = i + 2
+            while j < len(s) and not s[j].isalpha():
+                j += 1
+            result.append(s[i : j + 1])
+            i = j + 1
+        elif s[i] == " ":
+            result.append(" ")
+            i += 1
+        else:
+            result.append(s[i] if random.random() < fade else " ")
+            i += 1
+    return "".join(result)
+
+
+def vis_trunc(s, w):
+    """Truncate string to visual width w, skipping ANSI escape codes."""
+    result = []
+    vw = 0
+    i = 0
+    while i < len(s):
+        if vw >= w:
+            break
+        if s[i] == "\033" and i + 1 < len(s) and s[i + 1] == "[":
+            j = i + 2
+            while j < len(s) and not s[j].isalpha():
+                j += 1
+            result.append(s[i : j + 1])
+            i = j + 1
+        else:
+            result.append(s[i])
+            vw += 1
+            i += 1
+    return "".join(result)
+
+
+def vis_offset(s, offset):
+    """Offset string by skipping first offset visual characters, skipping ANSI escape codes."""
+    if offset <= 0:
+        return s
+    result = []
+    vw = 0
+    i = 0
+    skipping = True
+    while i < len(s):
+        if s[i] == "\033" and i + 1 < len(s) and s[i + 1] == "[":
+            j = i + 2
+            while j < len(s) and not s[j].isalpha():
+                j += 1
+            if skipping:
+                i = j + 1
+                continue
+            result.append(s[i : j + 1])
+            i = j + 1
+        else:
+            if skipping:
+                if vw >= offset:
+                    skipping = False
+                    result.append(s[i])
+                vw += 1
+                i += 1
+            else:
+                result.append(s[i])
+                i += 1
+    return "".join(result)
+
+
+def next_headline(pool, items, seen):
+    """Pull the next unique headline from pool, refilling as needed."""
+    while True:
+        if not pool:
+            pool.extend(items)
+            random.shuffle(pool)
+            seen.clear()
+        title, src, ts = pool.pop()
+        sig = title.lower().strip()
+        if sig not in seen:
+            seen.add(sig)
+            return title, src, ts
+
+
+def firehose_line(items, w):
+    """Generate one line of rapidly cycling firehose content."""
+    r = random.random()
+    if r < 0.35:
+        # Raw headline text
+        title, src, ts = random.choice(items)
+        text = title[: w - 1]
+        color = random.choice([G_LO, G_DIM, W_GHOST, C_DIM])
+        return f"{color}{text}{RST}"
+    elif r < 0.55:
+        # Dense glitch noise
+        d = random.choice([0.45, 0.55, 0.65, 0.75])
+        return "".join(
+            f"{random.choice([G_LO, G_DIM, C_DIM, W_GHOST])}"
+            f"{random.choice(config.GLITCH + config.KATA)}{RST}"
+            if random.random() < d
+            else " "
+            for _ in range(w)
+        )
+    elif r < 0.78:
+        # Status / program output
+        sources = FEEDS if config.MODE == "news" else POETRY_SOURCES
+        src = random.choice(list(sources.keys()))
+        msgs = [
+            f"  SIGNAL :: {src} :: {datetime.now().strftime('%H:%M:%S.%f')[:-3]}",
+            f"  ░░ FEED ACTIVE :: {src}",
+            f"  >> DECODE 0x{random.randint(0x1000, 0xFFFF):04X} :: {src[:24]}",
+            f"  ▒▒ ACQUIRE :: {random.choice(['TCP', 'UDP', 'RSS', 'ATOM', 'XML'])} :: {src}",
+            f"  {''.join(random.choice(config.KATA) for _ in range(3))} STRM "
+            f"{random.randint(0, 255):02X}:{random.randint(0, 255):02X}",
+        ]
+        text = random.choice(msgs)[: w - 1]
+        color = random.choice([G_LO, G_DIM, W_GHOST])
+        return f"{color}{text}{RST}"
+    else:
+        # Headline fragment with glitch prefix
+        title, _, _ = random.choice(items)
+        start = random.randint(0, max(0, len(title) - 20))
+        frag = title[start : start + random.randint(10, 35)]
+        pad = random.randint(0, max(0, w - len(frag) - 8))
+        gp = "".join(random.choice(config.GLITCH) for _ in range(random.randint(1, 3)))
+        text = (" " * pad + gp + " " + frag)[: w - 1]
+        color = random.choice([G_LO, C_DIM, W_GHOST])
+        return f"{color}{text}{RST}"

engine/effects/performance.py

@@ -0,0 +1,103 @@
+from collections import deque
+from dataclasses import dataclass
+
+
+@dataclass
+class EffectTiming:
+    name: str
+    duration_ms: float
+    buffer_chars_in: int
+    buffer_chars_out: int
+
+
+@dataclass
+class FrameTiming:
+    frame_number: int
+    total_ms: float
+    effects: list[EffectTiming]
+
+
+class PerformanceMonitor:
+    """Collects and stores performance metrics for effect pipeline."""
+
+    def __init__(self, max_frames: int = 60):
+        self._max_frames = max_frames
+        self._frames: deque[FrameTiming] = deque(maxlen=max_frames)
+        self._current_frame: list[EffectTiming] = []
+
+    def start_frame(self, frame_number: int) -> None:
+        self._current_frame = []
+
+    def record_effect(
+        self, name: str, duration_ms: float, chars_in: int, chars_out: int
+    ) -> None:
+        self._current_frame.append(
+            EffectTiming(
+                name=name,
+                duration_ms=duration_ms,
+                buffer_chars_in=chars_in,
+                buffer_chars_out=chars_out,
+            )
+        )
+
+    def end_frame(self, frame_number: int, total_ms: float) -> None:
+        self._frames.append(
+            FrameTiming(
+                frame_number=frame_number,
+                total_ms=total_ms,
+                effects=self._current_frame,
+            )
+        )
+
+    def get_stats(self) -> dict:
+        if not self._frames:
+            return {"error": "No timing data available"}
+
+        total_times = [f.total_ms for f in self._frames]
+        avg_total = sum(total_times) / len(total_times)
+        min_total = min(total_times)
+        max_total = max(total_times)
+
+        effect_stats: dict[str, dict] = {}
+        for frame in self._frames:
+            for effect in frame.effects:
+                if effect.name not in effect_stats:
+                    effect_stats[effect.name] = {"times": [], "total_chars": 0}
+                effect_stats[effect.name]["times"].append(effect.duration_ms)
+                effect_stats[effect.name]["total_chars"] += effect.buffer_chars_out
+
+        for name, stats in effect_stats.items():
+            times = stats["times"]
+            stats["avg_ms"] = sum(times) / len(times)
+            stats["min_ms"] = min(times)
+            stats["max_ms"] = max(times)
+            del stats["times"]
+
+        return {
+            "frame_count": len(self._frames),
+            "pipeline": {
+                "avg_ms": avg_total,
+                "min_ms": min_total,
+                "max_ms": max_total,
+            },
+            "effects": effect_stats,
+        }
+
+    def reset(self) -> None:
+        self._frames.clear()
+        self._current_frame = []
+
+
+_monitor: PerformanceMonitor | None = None
+
+
+def get_monitor() -> PerformanceMonitor:
+    global _monitor
+    if _monitor is None:
+        _monitor = PerformanceMonitor()
+    return _monitor
+
+
+def set_monitor(monitor: PerformanceMonitor) -> None:
+    global _monitor
+    _monitor = monitor

engine/effects/plugins/__init__.py

@@ -0,0 +1,38 @@
+from pathlib import Path
+
+PLUGIN_DIR = Path(__file__).parent
+
+
+def discover_plugins():
+    from engine.effects.registry import get_registry
+    from engine.effects.types import EffectPlugin
+
+    registry = get_registry()
+    imported = {}
+
+    for file_path in PLUGIN_DIR.glob("*.py"):
+        if file_path.name.startswith("_"):
+            continue
+        module_name = file_path.stem
+        if module_name in ("base", "types"):
+            continue
+
+        try:
+            module = __import__(f"engine.effects.plugins.{module_name}", fromlist=[""])
+            for attr_name in dir(module):
+                attr = getattr(module, attr_name)
+                if (
+                    isinstance(attr, type)
+                    and issubclass(attr, EffectPlugin)
+                    and attr is not EffectPlugin
+                    and attr_name.endswith("Effect")
+                ):
+                    plugin = attr()
+                    if not isinstance(plugin, EffectPlugin):
+                        continue
+                    registry.register(plugin)
+                    imported[plugin.name] = plugin
+        except Exception:
+            pass
+
+    return imported

engine/effects/plugins/afterimage.py

@@ -0,0 +1,122 @@
+"""Afterimage effect using previous frame."""
+
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class AfterimageEffect(EffectPlugin):
+    """Show a faint ghost of the previous frame.
+
+    This effect requires a FrameBufferStage to be present in the pipeline.
+    It shows a dimmed version of the previous frame super-imposed on the
+    current frame.
+
+    Attributes:
+        name: "afterimage"
+        config: EffectConfig with intensity parameter (0.0-1.0)
+        param_bindings: Optional sensor bindings for intensity modulation
+
+    Example:
+        >>> effect = AfterimageEffect()
+        >>> effect.configure(EffectConfig(intensity=0.3))
+        >>> result = effect.process(buffer, ctx)
+    """
+
+    name = "afterimage"
+    config: EffectConfig = EffectConfig(enabled=True, intensity=0.3)
+    param_bindings: dict[str, dict[str, str | float]] = {}
+    supports_partial_updates = False
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Apply afterimage effect using the previous frame.
+
+        Args:
+            buf: Current text buffer (list of strings)
+            ctx: Effect context with access to framebuffer history
+
+        Returns:
+            Buffer with ghost of previous frame overlaid
+        """
+        if not buf:
+            return buf
+
+        # Get framebuffer history from context
+        history = None
+
+        for key in ctx.state:
+            if key.startswith("framebuffer.") and key.endswith(".history"):
+                history = ctx.state[key]
+                break
+
+        if not history or len(history) < 1:
+            # No previous frame available
+            return buf
+
+        # Get intensity from config
+        intensity = self.config.params.get("intensity", self.config.intensity)
+        intensity = max(0.0, min(1.0, intensity))
+
+        if intensity <= 0.0:
+            return buf
+
+        # Get the previous frame (index 1, since index 0 is current)
+        prev_frame = history[1] if len(history) > 1 else None
+        if not prev_frame:
+            return buf
+
+        # Blend current and previous frames
+        viewport_height = ctx.terminal_height - ctx.ticker_height
+        result = []
+
+        for row in range(len(buf)):
+            if row >= viewport_height:
+                result.append(buf[row])
+                continue
+
+            current_line = buf[row]
+            prev_line = prev_frame[row] if row < len(prev_frame) else ""
+
+            if not prev_line:
+                result.append(current_line)
+                continue
+
+            # Apply dimming effect by reducing ANSI color intensity or adding transparency
+            # For a simple text version, we'll use a blend strategy
+            blended = self._blend_lines(current_line, prev_line, intensity)
+            result.append(blended)
+
+        return result
+
+    def _blend_lines(self, current: str, previous: str, intensity: float) -> str:
+        """Blend current and previous line with given intensity.
+
+        For text with ANSI codes, true blending is complex. This is a simplified
+        version that uses color averaging when possible.
+
+        A more sophisticated implementation would:
+        1. Parse ANSI color codes from both lines
+        2. Blend RGB values based on intensity
+        3. Reconstruct the line with blended colors
+
+        For now, we'll use a heuristic: if lines are similar, return current.
+        If they differ, we alternate or use the previous as a faint overlay.
+        """
+        if current == previous:
+            return current
+
+        # Simple blending: intensity determines mix
+        # intensity=1.0 => fully current
+        # intensity=0.3 => 70% previous ghost, 30% current
+
+        if intensity > 0.7:
+            return current
+        elif intensity < 0.3:
+            # Show previous but dimmed (simulate by adding faint color/gray)
+            return previous  # Would need to dim ANSI colors
+        else:
+            # For medium intensity, alternate based on character pattern
+            # This is a placeholder for proper blending
+            return current
+
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect."""
+        self.config = config

engine/effects/plugins/border.py

@@ -0,0 +1,105 @@
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class BorderEffect(EffectPlugin):
+    """Simple border effect for terminal display.
+
+    Draws a border around the buffer and optionally displays
+    performance metrics in the border corners.
+
+    Internally crops to display dimensions to ensure border fits.
+    """
+
+    name = "border"
+    config = EffectConfig(enabled=True, intensity=1.0)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not buf:
+            return buf
+
+        # Get actual display dimensions from context
+        display_w = ctx.terminal_width
+        display_h = ctx.terminal_height
+
+        # If dimensions are reasonable, crop first - use slightly smaller to ensure fit
+        if display_w >= 10 and display_h >= 3:
+            # Subtract 2 for border characters (left and right)
+            crop_w = display_w - 2
+            crop_h = display_h - 2
+            buf = self._crop_to_size(buf, crop_w, crop_h)
+            w = display_w
+            h = display_h
+        else:
+            # Use buffer dimensions
+            h = len(buf)
+            w = max(len(line) for line in buf) if buf else 0
+
+        if w < 3 or h < 3:
+            return buf
+
+        inner_w = w - 2
+
+        # Get metrics from context
+        fps = 0.0
+        frame_time = 0.0
+        metrics = ctx.get_state("metrics")
+        if metrics:
+            avg_ms = metrics.get("avg_ms")
+            frame_count = metrics.get("frame_count", 0)
+            if avg_ms and frame_count > 0:
+                fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+        # Build borders
+        # Top border: ┌────────────────────┐ or with FPS
+        if fps > 0:
+            fps_str = f" FPS:{fps:.0f}"
+            if len(fps_str) < inner_w:
+                right_len = inner_w - len(fps_str)
+                top_border = "┌" + "─" * right_len + fps_str + "┐"
+            else:
+                top_border = "┌" + "─" * inner_w + "┐"
+        else:
+            top_border = "┌" + "─" * inner_w + "┐"
+
+        # Bottom border: └────────────────────┘ or with frame time
+        if frame_time > 0:
+            ft_str = f" {frame_time:.1f}ms"
+            if len(ft_str) < inner_w:
+                right_len = inner_w - len(ft_str)
+                bottom_border = "└" + "─" * right_len + ft_str + "┘"
+            else:
+                bottom_border = "└" + "─" * inner_w + "┘"
+        else:
+            bottom_border = "└" + "─" * inner_w + "┘"
+
+        # Build result with left/right borders
+        result = [top_border]
+        for line in buf[: h - 2]:
+            if len(line) >= inner_w:
+                result.append("│" + line[:inner_w] + "│")
+            else:
+                result.append("│" + line + " " * (inner_w - len(line)) + "│")
+
+        result.append(bottom_border)
+
+        return result
+
+    def _crop_to_size(self, buf: list[str], w: int, h: int) -> list[str]:
+        """Crop buffer to fit within w x h."""
+        result = []
+        for i in range(min(h, len(buf))):
+            line = buf[i]
+            if len(line) > w:
+                result.append(line[:w])
+            else:
+                result.append(line + " " * (w - len(line)))
+
+        # Pad with empty lines if needed (for border)
+        while len(result) < h:
+            result.append(" " * w)
+
+        return result
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/crop.py

@@ -0,0 +1,42 @@
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class CropEffect(EffectPlugin):
+    """Crop effect that crops the input buffer to fit the display.
+
+    This ensures the output buffer matches the actual display dimensions,
+    useful when the source produces a buffer larger than the viewport.
+    """
+
+    name = "crop"
+    config = EffectConfig(enabled=True, intensity=1.0)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not buf:
+            return buf
+
+        # Get actual display dimensions from context
+        w = (
+            ctx.terminal_width
+            if ctx.terminal_width > 0
+            else max(len(line) for line in buf)
+        )
+        h = ctx.terminal_height if ctx.terminal_height > 0 else len(buf)
+
+        # Crop buffer to fit
+        result = []
+        for i in range(min(h, len(buf))):
+            line = buf[i]
+            if len(line) > w:
+                result.append(line[:w])
+            else:
+                result.append(line + " " * (w - len(line)))
+
+        # Pad with empty lines if needed
+        while len(result) < h:
+            result.append(" " * w)
+
+        return result
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/fade.py

@@ -0,0 +1,58 @@
+import random
+
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class FadeEffect(EffectPlugin):
+    name = "fade"
+    config = EffectConfig(enabled=True, intensity=1.0, entropy=0.1)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not ctx.ticker_height:
+            return buf
+        result = list(buf)
+        intensity = self.config.intensity
+
+        top_zone = max(1, int(ctx.ticker_height * 0.25))
+        bot_zone = max(1, int(ctx.ticker_height * 0.10))
+
+        for r in range(len(result)):
+            if r >= ctx.ticker_height:
+                continue
+            top_f = min(1.0, r / top_zone) if top_zone > 0 else 1.0
+            bot_f = (
+                min(1.0, (ctx.ticker_height - 1 - r) / bot_zone)
+                if bot_zone > 0
+                else 1.0
+            )
+            row_fade = min(top_f, bot_f) * intensity
+
+            if row_fade < 1.0 and result[r].strip():
+                result[r] = self._fade_line(result[r], row_fade)
+
+        return result
+
+    def _fade_line(self, s: str, fade: float) -> str:
+        if fade >= 1.0:
+            return s
+        if fade <= 0.0:
+            return s  # Preserve original line length - don't return empty
+        result = []
+        i = 0
+        while i < len(s):
+            if s[i] == "\033" and i + 1 < len(s) and s[i + 1] == "[":
+                j = i + 2
+                while j < len(s) and not s[j].isalpha():
+                    j += 1
+                result.append(s[i : j + 1])
+                i = j + 1
+            elif s[i] == " ":
+                result.append(" ")
+                i += 1
+            else:
+                result.append(s[i] if random.random() < fade else " ")
+                i += 1
+        return "".join(result)
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/figment.py

@@ -0,0 +1,332 @@
+"""
+Figment overlay effect for modern pipeline architecture.
+
+Provides periodic SVG glyph overlays with reveal/hold/dissolve animation phases.
+Integrates directly with the pipeline's effect system without legacy dependencies.
+"""
+
+import random
+from dataclasses import dataclass
+from enum import Enum, auto
+from pathlib import Path
+
+from engine import config
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+from engine.figment_render import rasterize_svg
+from engine.figment_trigger import FigmentAction, FigmentCommand, FigmentTrigger
+from engine.terminal import RST
+from engine.themes import THEME_REGISTRY
+
+
+class FigmentPhase(Enum):
+    """Animation phases for figment overlay."""
+
+    REVEAL = auto()
+    HOLD = auto()
+    DISSOLVE = auto()
+
+
+@dataclass
+class FigmentState:
+    """State of a figment overlay at a given frame."""
+
+    phase: FigmentPhase
+    progress: float
+    rows: list[str]
+    gradient: list[int]
+    center_row: int
+    center_col: int
+
+
+def _color_codes_to_ansi(gradient: list[int]) -> list[str]:
+    """Convert gradient list to ANSI color codes.
+
+    Args:
+        gradient: List of 256-color palette codes
+
+    Returns:
+        List of ANSI escape code strings
+    """
+    codes = []
+    for color in gradient:
+        if isinstance(color, int):
+            codes.append(f"\033[38;5;{color}m")
+        else:
+            # Fallback to green
+            codes.append("\033[38;5;46m")
+    return codes if codes else ["\033[38;5;46m"]
+
+
+def render_figment_overlay(figment_state: FigmentState, w: int, h: int) -> list[str]:
+    """Render figment overlay as ANSI cursor-positioning commands.
+
+    Args:
+        figment_state: FigmentState with phase, progress, rows, gradient, centering.
+        w: terminal width
+        h: terminal height
+
+    Returns:
+        List of ANSI strings to append to display buffer.
+    """
+    rows = figment_state.rows
+    if not rows:
+        return []
+
+    phase = figment_state.phase
+    progress = figment_state.progress
+    gradient = figment_state.gradient
+    center_row = figment_state.center_row
+    center_col = figment_state.center_col
+
+    cols = _color_codes_to_ansi(gradient)
+
+    # Build a list of non-space cell positions
+    cell_positions = []
+    for r_idx, row in enumerate(rows):
+        for c_idx, ch in enumerate(row):
+            if ch != " ":
+                cell_positions.append((r_idx, c_idx))
+
+    n_cells = len(cell_positions)
+    if n_cells == 0:
+        return []
+
+    # Use a deterministic seed so the reveal/dissolve pattern is stable per-figment
+    rng = random.Random(hash(tuple(rows[0][:10])) if rows[0] else 42)
+    shuffled = list(cell_positions)
+    rng.shuffle(shuffled)
+
+    # Phase-dependent visibility
+    if phase == FigmentPhase.REVEAL:
+        visible_count = int(n_cells * progress)
+        visible = set(shuffled[:visible_count])
+    elif phase == FigmentPhase.HOLD:
+        visible = set(cell_positions)
+        # Strobe: dim some cells periodically
+        if int(progress * 20) % 3 == 0:
+            dim_count = int(n_cells * 0.3)
+            visible -= set(shuffled[:dim_count])
+    elif phase == FigmentPhase.DISSOLVE:
+        remaining_count = int(n_cells * (1.0 - progress))
+        visible = set(shuffled[:remaining_count])
+    else:
+        visible = set(cell_positions)
+
+    # Build overlay commands
+    overlay: list[str] = []
+    n_cols = len(cols)
+    max_x = max((len(r.rstrip()) for r in rows if r.strip()), default=1)
+
+    for r_idx, row in enumerate(rows):
+        scr_row = center_row + r_idx + 1  # 1-indexed
+        if scr_row < 1 or scr_row > h:
+            continue
+
+        line_buf: list[str] = []
+        has_content = False
+
+        for c_idx, ch in enumerate(row):
+            scr_col = center_col + c_idx + 1
+            if scr_col < 1 or scr_col > w:
+                continue
+
+            if ch != " " and (r_idx, c_idx) in visible:
+                # Apply gradient color
+                shifted = (c_idx / max(max_x - 1, 1)) % 1.0
+                idx = min(round(shifted * (n_cols - 1)), n_cols - 1)
+                line_buf.append(f"{cols[idx]}{ch}{RST}")
+                has_content = True
+            else:
+                line_buf.append(" ")
+
+        if has_content:
+            line_str = "".join(line_buf).rstrip()
+            if line_str.strip():
+                overlay.append(f"\033[{scr_row};{center_col + 1}H{line_str}{RST}")
+
+    return overlay
+
+
+class FigmentEffect(EffectPlugin):
+    """Figment overlay effect for pipeline architecture.
+
+    Provides periodic SVG overlays with reveal/hold/dissolve animation.
+    """
+
+    name = "figment"
+    config = EffectConfig(
+        enabled=True,
+        intensity=1.0,
+        params={
+            "interval_secs": 60,
+            "display_secs": 4.5,
+            "figment_dir": "figments",
+        },
+    )
+    supports_partial_updates = False
+    is_overlay = True  # Figment is an overlay effect that composes on top of the buffer
+
+    def __init__(
+        self,
+        figment_dir: str | None = None,
+        triggers: list[FigmentTrigger] | None = None,
+    ):
+        self.config = EffectConfig(
+            enabled=True,
+            intensity=1.0,
+            params={
+                "interval_secs": 60,
+                "display_secs": 4.5,
+                "figment_dir": figment_dir or "figments",
+            },
+        )
+        self._triggers = triggers or []
+        self._phase: FigmentPhase | None = None
+        self._progress: float = 0.0
+        self._rows: list[str] = []
+        self._gradient: list[int] = []
+        self._center_row: int = 0
+        self._center_col: int = 0
+        self._timer: float = 0.0
+        self._last_svg: str | None = None
+        self._svg_files: list[str] = []
+        self._scan_svgs()
+
+    def _scan_svgs(self) -> None:
+        """Scan figment directory for SVG files."""
+        figment_dir = Path(self.config.params["figment_dir"])
+        if figment_dir.is_dir():
+            self._svg_files = sorted(str(p) for p in figment_dir.glob("*.svg"))
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Add figment overlay to buffer."""
+        if not self.config.enabled:
+            return buf
+
+        # Get figment state using frame number from context
+        figment_state = self.get_figment_state(
+            ctx.frame_number, ctx.terminal_width, ctx.terminal_height
+        )
+
+        if figment_state:
+            # Render overlay and append to buffer
+            overlay = render_figment_overlay(
+                figment_state, ctx.terminal_width, ctx.terminal_height
+            )
+            buf = buf + overlay
+
+        return buf
+
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect."""
+        # Preserve figment_dir if the new config doesn't supply one
+        figment_dir = config.params.get(
+            "figment_dir", self.config.params.get("figment_dir", "figments")
+        )
+        self.config = config
+        if "figment_dir" not in self.config.params:
+            self.config.params["figment_dir"] = figment_dir
+        self._scan_svgs()
+
+    def trigger(self, w: int, h: int) -> None:
+        """Manually trigger a figment display."""
+        if not self._svg_files:
+            return
+
+        # Pick a random SVG, avoid repeating
+        candidates = [s for s in self._svg_files if s != self._last_svg]
+        if not candidates:
+            candidates = self._svg_files
+        svg_path = random.choice(candidates)
+        self._last_svg = svg_path
+
+        # Rasterize
+        try:
+            self._rows = rasterize_svg(svg_path, w, h)
+        except Exception:
+            return
+
+        # Pick random theme gradient
+        theme_key = random.choice(list(THEME_REGISTRY.keys()))
+        self._gradient = THEME_REGISTRY[theme_key].main_gradient
+
+        # Center in viewport
+        figment_h = len(self._rows)
+        figment_w = max((len(r) for r in self._rows), default=0)
+        self._center_row = max(0, (h - figment_h) // 2)
+        self._center_col = max(0, (w - figment_w) // 2)
+
+        # Start reveal phase
+        self._phase = FigmentPhase.REVEAL
+        self._progress = 0.0
+
+    def get_figment_state(
+        self, frame_number: int, w: int, h: int
+    ) -> FigmentState | None:
+        """Tick the state machine and return current state, or None if idle."""
+        if not self.config.enabled:
+            return None
+
+        # Poll triggers
+        for trig in self._triggers:
+            cmd = trig.poll()
+            if cmd is not None:
+                self._handle_command(cmd, w, h)
+
+        # Tick timer when idle
+        if self._phase is None:
+            self._timer += config.FRAME_DT
+            interval = self.config.params.get("interval_secs", 60)
+            if self._timer >= interval:
+                self._timer = 0.0
+                self.trigger(w, h)
+
+        # Tick animation — snapshot current phase/progress, then advance
+        if self._phase is not None:
+            # Capture the state at the start of this frame
+            current_phase = self._phase
+            current_progress = self._progress
+
+            # Advance for next frame
+            display_secs = self.config.params.get("display_secs", 4.5)
+            phase_duration = display_secs / 3.0
+            self._progress += config.FRAME_DT / phase_duration
+
+            if self._progress >= 1.0:
+                self._progress = 0.0
+                if self._phase == FigmentPhase.REVEAL:
+                    self._phase = FigmentPhase.HOLD
+                elif self._phase == FigmentPhase.HOLD:
+                    self._phase = FigmentPhase.DISSOLVE
+                elif self._phase == FigmentPhase.DISSOLVE:
+                    self._phase = None
+
+            return FigmentState(
+                phase=current_phase,
+                progress=current_progress,
+                rows=self._rows,
+                gradient=self._gradient,
+                center_row=self._center_row,
+                center_col=self._center_col,
+            )
+
+        return None
+
+    def _handle_command(self, cmd: FigmentCommand, w: int, h: int) -> None:
+        """Handle a figment command."""
+        if cmd.action == FigmentAction.TRIGGER:
+            self.trigger(w, h)
+        elif cmd.action == FigmentAction.SET_INTENSITY and isinstance(
+            cmd.value, (int, float)
+        ):
+            self.config.intensity = float(cmd.value)
+        elif cmd.action == FigmentAction.SET_INTERVAL and isinstance(
+            cmd.value, (int, float)
+        ):
+            self.config.params["interval_secs"] = float(cmd.value)
+        elif cmd.action == FigmentAction.SET_COLOR and isinstance(cmd.value, str):
+            if cmd.value in THEME_REGISTRY:
+                self._gradient = THEME_REGISTRY[cmd.value].main_gradient
+        elif cmd.action == FigmentAction.STOP:
+            self._phase = None
+            self._progress = 0.0

engine/effects/plugins/firehose.py

@@ -0,0 +1,72 @@
+import random
+from datetime import datetime
+
+from engine import config
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+from engine.sources import FEEDS, POETRY_SOURCES
+from engine.terminal import C_DIM, G_DIM, G_LO, RST, W_GHOST
+
+
+class FirehoseEffect(EffectPlugin):
+    name = "firehose"
+    config = EffectConfig(enabled=True, intensity=1.0, entropy=0.9)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        firehose_h = config.FIREHOSE_H if config.FIREHOSE else 0
+        if firehose_h <= 0 or not ctx.items:
+            return buf
+
+        result = list(buf)
+        intensity = self.config.intensity
+        h = ctx.terminal_height
+
+        for fr in range(firehose_h):
+            scr_row = h - firehose_h + fr + 1
+            fline = self._firehose_line(ctx.items, ctx.terminal_width, intensity)
+            result.append(f"\033[{scr_row};1H{fline}\033[K")
+        return result
+
+    def _firehose_line(self, items: list, w: int, intensity: float) -> str:
+        r = random.random()
+        if r < 0.35 * intensity:
+            title, src, ts = random.choice(items)
+            text = title[: w - 1]
+            color = random.choice([G_LO, G_DIM, W_GHOST, C_DIM])
+            return f"{color}{text}{RST}"
+        elif r < 0.55 * intensity:
+            d = random.choice([0.45, 0.55, 0.65, 0.75])
+            return "".join(
+                f"{random.choice([G_LO, G_DIM, C_DIM, W_GHOST])}"
+                f"{random.choice(config.GLITCH + config.KATA)}{RST}"
+                if random.random() < d
+                else " "
+                for _ in range(w)
+            )
+        elif r < 0.78 * intensity:
+            sources = FEEDS if config.MODE == "news" else POETRY_SOURCES
+            src = random.choice(list(sources.keys()))
+            msgs = [
+                f"  SIGNAL :: {src} :: {datetime.now().strftime('%H:%M:%S.%f')[:-3]}",
+                f"  ░░ FEED ACTIVE :: {src}",
+                f"  >> DECODE 0x{random.randint(0x1000, 0xFFFF):04X} :: {src[:24]}",
+                f"  ▒▒ ACQUIRE :: {random.choice(['TCP', 'UDP', 'RSS', 'ATOM', 'XML'])} :: {src}",
+                f"  {''.join(random.choice(config.KATA) for _ in range(3))} STRM "
+                f"{random.randint(0, 255):02X}:{random.randint(0, 255):02X}",
+            ]
+            text = random.choice(msgs)[: w - 1]
+            color = random.choice([G_LO, G_DIM, W_GHOST])
+            return f"{color}{text}{RST}"
+        else:
+            title, _, _ = random.choice(items)
+            start = random.randint(0, max(0, len(title) - 20))
+            frag = title[start : start + random.randint(10, 35)]
+            pad = random.randint(0, max(0, w - len(frag) - 8))
+            gp = "".join(
+                random.choice(config.GLITCH) for _ in range(random.randint(1, 3))
+            )
+            text = (" " * pad + gp + " " + frag)[: w - 1]
+            color = random.choice([G_LO, C_DIM, W_GHOST])
+            return f"{color}{text}{RST}"
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/glitch.py

@@ -0,0 +1,52 @@
+import random
+
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+from engine.terminal import DIM, G_LO, RST
+
+
+class GlitchEffect(EffectPlugin):
+    name = "glitch"
+    config = EffectConfig(enabled=True, intensity=1.0, entropy=0.8)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not buf:
+            return buf
+        result = list(buf)
+        intensity = self.config.intensity
+
+        glitch_prob = 0.32 + min(0.9, ctx.mic_excess * 0.16)
+        glitch_prob = glitch_prob * intensity
+        n_hits = 4 + int(ctx.mic_excess / 2)
+        n_hits = int(n_hits * intensity)
+
+        if random.random() < glitch_prob:
+            # Store original visible lengths before any modifications
+            # Strip ANSI codes to get visible length
+            import re
+
+            ansi_pattern = re.compile(r"\x1b\[[0-9;]*[a-zA-Z]")
+            original_lengths = [len(ansi_pattern.sub("", line)) for line in result]
+            for _ in range(min(n_hits, len(result))):
+                gi = random.randint(0, len(result) - 1)
+                result[gi]
+                target_len = original_lengths[gi]  # Use stored original length
+                glitch_bar = self._glitch_bar(target_len)
+                result[gi] = glitch_bar
+        return result
+
+    def _glitch_bar(self, target_len: int) -> str:
+        c = random.choice(["░", "▒", "─", "\xc2"])
+        n = random.randint(3, max(3, target_len // 2))
+        o = random.randint(0, max(0, target_len - n))
+
+        glitch_chars = c * n
+        trailing_spaces = target_len - o - n
+        trailing_spaces = max(0, trailing_spaces)
+
+        glitch_part = f"{G_LO}{DIM}" + glitch_chars + RST
+        result = " " * o + glitch_part + " " * trailing_spaces
+
+        return result
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/hud.py

@@ -0,0 +1,102 @@
+from engine.effects.types import (
+    EffectConfig,
+    EffectContext,
+    EffectPlugin,
+    PartialUpdate,
+)
+
+
+class HudEffect(EffectPlugin):
+    name = "hud"
+    config = EffectConfig(enabled=True, intensity=1.0)
+    supports_partial_updates = True  # Enable partial update optimization
+
+    # Cache last HUD content to detect changes
+    _last_hud_content: tuple | None = None
+
+    def process_partial(
+        self, buf: list[str], ctx: EffectContext, partial: PartialUpdate
+    ) -> list[str]:
+        # If full buffer requested, process normally
+        if partial.full_buffer:
+            return self.process(buf, ctx)
+
+        # If HUD rows (0, 1, 2) aren't dirty, skip processing
+        if partial.dirty:
+            hud_rows = {0, 1, 2}
+            dirty_hud_rows = partial.dirty & hud_rows
+            if not dirty_hud_rows:
+                return buf  # Nothing for HUD to do
+
+        # Proceed with full processing
+        return self.process(buf, ctx)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        result = list(buf)
+
+        # Read metrics from pipeline context (first-class citizen)
+        # Falls back to global monitor for backwards compatibility
+        metrics = ctx.get_state("metrics")
+        if not metrics:
+            # Fallback to global monitor for backwards compatibility
+            from engine.effects.performance import get_monitor
+
+            monitor = get_monitor()
+            if monitor:
+                stats = monitor.get_stats()
+                if stats and "pipeline" in stats:
+                    metrics = stats
+
+        fps = 0.0
+        frame_time = 0.0
+        if metrics:
+            if "error" in metrics:
+                pass  # No metrics available yet
+            elif "pipeline" in metrics:
+                frame_time = metrics["pipeline"].get("avg_ms", 0.0)
+                frame_count = metrics.get("frame_count", 0)
+                if frame_count > 0 and frame_time > 0:
+                    fps = 1000.0 / frame_time
+            elif "avg_ms" in metrics:
+                # Direct metrics format
+                frame_time = metrics.get("avg_ms", 0.0)
+                frame_count = metrics.get("frame_count", 0)
+                if frame_count > 0 and frame_time > 0:
+                    fps = 1000.0 / frame_time
+
+        effect_name = self.config.params.get("display_effect", "none")
+        effect_intensity = self.config.params.get("display_intensity", 0.0)
+
+        hud_lines = []
+        hud_lines.append(
+            f"\033[1;1H\033[38;5;46mMAINLINE DEMO\033[0m \033[38;5;245m|\033[0m \033[38;5;39mFPS: {fps:.1f}\033[0m \033[38;5;245m|\033[0m \033[38;5;208m{frame_time:.1f}ms\033[0m"
+        )
+
+        bar_width = 20
+        filled = int(bar_width * effect_intensity)
+        bar = (
+            "\033[38;5;82m"
+            + "█" * filled
+            + "\033[38;5;240m"
+            + "░" * (bar_width - filled)
+            + "\033[0m"
+        )
+        hud_lines.append(
+            f"\033[2;1H\033[38;5;45mEFFECT:\033[0m \033[1;38;5;227m{effect_name:12s}\033[0m \033[38;5;245m|\033[0m {bar} \033[38;5;245m|\033[0m \033[38;5;219m{effect_intensity * 100:.0f}%\033[0m"
+        )
+
+        # Get pipeline order from context
+        pipeline_order = ctx.get_state("pipeline_order")
+        pipeline_str = ",".join(pipeline_order) if pipeline_order else "(none)"
+        hud_lines.append(f"\033[3;1H\033[38;5;44mPIPELINE:\033[0m {pipeline_str}")
+
+        for i, line in enumerate(hud_lines):
+            if i < len(result):
+                result[i] = line
+            else:
+                result.append(line)
+
+        return result
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/motionblur.py

@@ -0,0 +1,119 @@
+"""Motion blur effect using frame history."""
+
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class MotionBlurEffect(EffectPlugin):
+    """Apply motion blur by blending current frame with previous frames.
+
+    This effect requires a FrameBufferStage to be present in the pipeline.
+    The framebuffer provides frame history which is blended with the current
+    frame based on intensity.
+
+    Attributes:
+        name: "motionblur"
+        config: EffectConfig with intensity parameter (0.0-1.0)
+        param_bindings: Optional sensor bindings for intensity modulation
+
+    Example:
+        >>> effect = MotionBlurEffect()
+        >>> effect.configure(EffectConfig(intensity=0.5))
+        >>> result = effect.process(buffer, ctx)
+    """
+
+    name = "motionblur"
+    config: EffectConfig = EffectConfig(enabled=True, intensity=0.5)
+    param_bindings: dict[str, dict[str, str | float]] = {}
+    supports_partial_updates = False
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Apply motion blur by blending with previous frames.
+
+        Args:
+            buf: Current text buffer (list of strings)
+            ctx: Effect context with access to framebuffer history
+
+        Returns:
+            Blended buffer with motion blur effect applied
+        """
+        if not buf:
+            return buf
+
+        # Get framebuffer history from context
+        # We'll look for the first available framebuffer history
+        history = None
+
+        for key in ctx.state:
+            if key.startswith("framebuffer.") and key.endswith(".history"):
+                history = ctx.state[key]
+                break
+
+        if not history:
+            # No framebuffer available, return unchanged
+            return buf
+
+        # Get intensity from config
+        intensity = self.config.params.get("intensity", self.config.intensity)
+        intensity = max(0.0, min(1.0, intensity))
+
+        if intensity <= 0.0:
+            return buf
+
+        # Get decay factor (how quickly older frames fade)
+        decay = self.config.params.get("decay", 0.7)
+
+        # Build output buffer
+        result = []
+        viewport_height = ctx.terminal_height - ctx.ticker_height
+
+        # Determine how many frames to blend (up to history depth)
+        max_frames = min(len(history), 5)  # Cap at 5 frames for performance
+
+        for row in range(len(buf)):
+            if row >= viewport_height:
+                # Beyond viewport, just copy
+                result.append(buf[row])
+                continue
+
+            # Start with current frame
+            blended = buf[row]
+
+            # Blend with historical frames
+            weight_sum = 1.0
+            if max_frames > 0 and intensity > 0:
+                for i in range(max_frames):
+                    frame_weight = intensity * (decay**i)
+                    if frame_weight < 0.01:  # Skip negligible weights
+                        break
+
+                    hist_row = history[i][row] if row < len(history[i]) else ""
+                    # Simple string blending: we'll concatenate with space
+                    # For a proper effect, we'd need to blend ANSI colors
+                    # This is a simplified version that just adds the frames
+                    blended = self._blend_strings(blended, hist_row, frame_weight)
+                    weight_sum += frame_weight
+
+            result.append(blended)
+
+        return result
+
+    def _blend_strings(self, current: str, historical: str, weight: float) -> str:
+        """Blend two strings with given weight.
+
+        This is a simplified blending that works with ANSI codes.
+        For proper blending we'd need to parse colors, but for now
+        we use a heuristic: if strings are identical, return one.
+        If they differ, we alternate or concatenate based on weight.
+        """
+        if current == historical:
+            return current
+
+        # If weight is high, show current; if low, show historical
+        if weight > 0.5:
+            return current
+        else:
+            return historical
+
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect."""
+        self.config = config

engine/effects/plugins/noise.py

@@ -0,0 +1,37 @@
+import random
+
+from engine import config
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+from engine.terminal import C_DIM, G_DIM, G_LO, RST, W_GHOST
+
+
+class NoiseEffect(EffectPlugin):
+    name = "noise"
+    config = EffectConfig(enabled=True, intensity=0.15, entropy=0.4)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not ctx.ticker_height:
+            return buf
+        result = list(buf)
+        intensity = self.config.intensity
+        probability = intensity * 0.15
+
+        for r in range(len(result)):
+            cy = ctx.scroll_cam + r
+            if random.random() < probability:
+                original_line = result[r]
+                result[r] = self._generate_noise(len(original_line), cy)
+        return result
+
+    def _generate_noise(self, w: int, cy: int) -> str:
+        d = random.choice([0.15, 0.25, 0.35, 0.12])
+        return "".join(
+            f"{random.choice([G_LO, G_DIM, C_DIM, W_GHOST])}"
+            f"{random.choice(config.GLITCH + config.KATA)}{RST}"
+            if random.random() < d
+            else " "
+            for _ in range(w)
+        )
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/plugins/repl.py

@@ -0,0 +1,410 @@
+"""REPL Effect Plugin
+
+A HUD-style command-line interface for interactive pipeline control.
+
+This effect provides a Read-Eval-Print Loop (REPL) that allows users to:
+- View pipeline status and metrics
+- Toggle effects on/off
+- Adjust effect parameters in real-time
+- Inspect pipeline configuration
+- Execute commands for pipeline manipulation
+
+Usage:
+    Add 'repl' to the effects list in your configuration.
+
+Commands:
+    help - Show available commands
+    status - Show pipeline status
+    effects - List all effects
+    effect <name> <on|off> - Toggle an effect
+    param <effect> <param> <value> - Set effect parameter
+    pipeline - Show current pipeline order
+    clear - Clear output buffer
+    quit - Exit REPL
+
+Keyboard:
+    Enter - Execute command
+    Up/Down - Navigate command history
+    Tab - Auto-complete (if implemented)
+    Ctrl+C - Clear current input
+"""
+
+from dataclasses import dataclass, field
+
+from engine.effects.types import (
+    EffectConfig,
+    EffectContext,
+    EffectPlugin,
+    PartialUpdate,
+)
+
+
+@dataclass
+class REPLState:
+    """State of the REPL interface."""
+
+    command_history: list[str] = field(default_factory=list)
+    current_command: str = ""
+    history_index: int = -1
+    output_buffer: list[str] = field(default_factory=list)
+    max_history: int = 50
+    max_output_lines: int = 20
+
+
+class ReplEffect(EffectPlugin):
+    """REPL effect with HUD-style overlay for interactive pipeline control."""
+
+    name = "repl"
+    config = EffectConfig(
+        enabled=True,
+        intensity=1.0,
+        params={
+            "display_height": 8,  # Height of REPL area in lines
+            "show_hud": True,  # Show HUD header lines
+        },
+    )
+    supports_partial_updates = True
+
+    def __init__(self):
+        super().__init__()
+        self.state = REPLState()
+        self._last_metrics: dict | None = None
+
+    def process_partial(
+        self, buf: list[str], ctx: EffectContext, partial: PartialUpdate
+    ) -> list[str]:
+        """Handle partial updates efficiently."""
+        if partial.full_buffer:
+            return self.process(buf, ctx)
+        # Always process REPL since it needs to stay visible
+        return self.process(buf, ctx)
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Render buffer with REPL overlay."""
+        # Get display dimensions from context
+        height = ctx.terminal_height if hasattr(ctx, "terminal_height") else len(buf)
+        width = ctx.terminal_width if hasattr(ctx, "terminal_width") else 80
+
+        # Calculate areas
+        repl_height = self.config.params.get("display_height", 8)
+        show_hud = self.config.params.get("show_hud", True)
+
+        # Reserve space for REPL at bottom
+        # HUD uses top 3 lines if enabled
+        content_height = max(1, height - repl_height)
+
+        # Build output
+        output = []
+
+        # Add content (truncated or padded)
+        for i in range(content_height):
+            if i < len(buf):
+                output.append(buf[i][:width])
+            else:
+                output.append(" " * width)
+
+        # Add HUD lines if enabled
+        if show_hud:
+            hud_output = self._render_hud(width, ctx)
+            # Overlay HUD on first lines of content
+            for i, line in enumerate(hud_output):
+                if i < len(output):
+                    output[i] = line[:width]
+
+        # Add separator
+        output.append("─" * width)
+
+        # Add REPL area
+        repl_lines = self._render_repl(width, repl_height - 1)
+        output.extend(repl_lines)
+
+        # Ensure correct height
+        while len(output) < height:
+            output.append(" " * width)
+        output = output[:height]
+
+        return output
+
+    def _render_hud(self, width: int, ctx: EffectContext) -> list[str]:
+        """Render HUD-style header with metrics."""
+        lines = []
+
+        # Get metrics
+        metrics = self._get_metrics(ctx)
+        fps = metrics.get("fps", 0.0)
+        frame_time = metrics.get("frame_time", 0.0)
+
+        # Line 1: Title + FPS + Frame time
+        fps_str = f"FPS: {fps:.1f}" if fps > 0 else "FPS: --"
+        time_str = f"{frame_time:.1f}ms" if frame_time > 0 else "--ms"
+        line1 = (
+            f"\033[38;5;46mMAINLINE REPL\033[0m "
+            f"\033[38;5;245m|\033[0m \033[38;5;39m{fps_str}\033[0m "
+            f"\033[38;5;245m|\033[0m \033[38;5;208m{time_str}\033[0m"
+        )
+        lines.append(line1[:width])
+
+        # Line 2: Command count + History index
+        cmd_count = len(self.state.command_history)
+        hist_idx = (
+            f"[{self.state.history_index + 1}/{cmd_count}]" if cmd_count > 0 else ""
+        )
+        line2 = (
+            f"\033[38;5;45mCOMMANDS:\033[0m "
+            f"\033[1;38;5;227m{cmd_count}\033[0m "
+            f"\033[38;5;245m|\033[0m \033[38;5;219m{hist_idx}\033[0m"
+        )
+        lines.append(line2[:width])
+
+        # Line 3: Output buffer count
+        out_count = len(self.state.output_buffer)
+        line3 = f"\033[38;5;44mOUTPUT:\033[0m \033[1;38;5;227m{out_count}\033[0m lines"
+        lines.append(line3[:width])
+
+        return lines
+
+    def _render_repl(self, width: int, height: int) -> list[str]:
+        """Render REPL interface."""
+        lines = []
+
+        # Calculate how many output lines to show
+        # Reserve 1 line for input prompt
+        output_height = height - 1
+        output_start = max(0, len(self.state.output_buffer) - output_height)
+
+        # Render output buffer
+        for i in range(output_height):
+            idx = output_start + i
+            if idx < len(self.state.output_buffer):
+                line = self.state.output_buffer[idx][:width]
+                lines.append(line)
+            else:
+                lines.append(" " * width)
+
+        # Render input prompt
+        prompt = "> "
+        input_line = f"{prompt}{self.state.current_command}"
+        # Add cursor indicator
+        cursor = "█" if len(self.state.current_command) % 2 == 0 else " "
+        input_line += cursor
+        lines.append(input_line[:width])
+
+        return lines
+
+    def _get_metrics(self, ctx: EffectContext) -> dict:
+        """Get pipeline metrics from context."""
+        metrics = ctx.get_state("metrics")
+        if metrics:
+            self._last_metrics = metrics
+
+        if self._last_metrics:
+            # Extract FPS and frame time
+            fps = 0.0
+            frame_time = 0.0
+
+            if "pipeline" in self._last_metrics:
+                avg_ms = self._last_metrics["pipeline"].get("avg_ms", 0.0)
+                frame_count = self._last_metrics.get("frame_count", 0)
+                if frame_count > 0 and avg_ms > 0:
+                    fps = 1000.0 / avg_ms
+                frame_time = avg_ms
+
+            return {"fps": fps, "frame_time": frame_time}
+
+        return {"fps": 0.0, "frame_time": 0.0}
+
+    def process_command(self, command: str, ctx: EffectContext | None = None) -> None:
+        """Process a REPL command."""
+        cmd = command.strip()
+        if not cmd:
+            return
+
+        # Add to history
+        self.state.command_history.append(cmd)
+        if len(self.state.command_history) > self.state.max_history:
+            self.state.command_history.pop(0)
+
+        self.state.history_index = len(self.state.command_history)
+        self.state.current_command = ""
+
+        # Add to output buffer
+        self.state.output_buffer.append(f"> {cmd}")
+
+        # Parse command
+        parts = cmd.split()
+        cmd_name = parts[0].lower()
+        cmd_args = parts[1:] if len(parts) > 1 else []
+
+        # Execute command
+        try:
+            if cmd_name == "help":
+                self._cmd_help()
+            elif cmd_name == "status":
+                self._cmd_status(ctx)
+            elif cmd_name == "effects":
+                self._cmd_effects(ctx)
+            elif cmd_name == "effect":
+                self._cmd_effect(cmd_args, ctx)
+            elif cmd_name == "param":
+                self._cmd_param(cmd_args, ctx)
+            elif cmd_name == "pipeline":
+                self._cmd_pipeline(ctx)
+            elif cmd_name == "clear":
+                self.state.output_buffer.clear()
+            elif cmd_name == "quit" or cmd_name == "exit":
+                self.state.output_buffer.append("Use Ctrl+C to exit")
+            else:
+                self.state.output_buffer.append(f"Unknown command: {cmd_name}")
+                self.state.output_buffer.append("Type 'help' for available commands")
+
+        except Exception as e:
+            self.state.output_buffer.append(f"Error: {e}")
+
+    def _cmd_help(self):
+        """Show help message."""
+        self.state.output_buffer.append("Available commands:")
+        self.state.output_buffer.append("  help - Show this help")
+        self.state.output_buffer.append("  status - Show pipeline status")
+        self.state.output_buffer.append("  effects - List all effects")
+        self.state.output_buffer.append("  effect <name> <on|off> - Toggle effect")
+        self.state.output_buffer.append(
+            "  param <effect> <param> <value> - Set parameter"
+        )
+        self.state.output_buffer.append("  pipeline - Show current pipeline order")
+        self.state.output_buffer.append("  clear - Clear output buffer")
+        self.state.output_buffer.append("  quit - Show exit message")
+
+    def _cmd_status(self, ctx: EffectContext | None):
+        """Show pipeline status."""
+        if ctx:
+            metrics = self._get_metrics(ctx)
+            self.state.output_buffer.append(f"FPS: {metrics['fps']:.1f}")
+            self.state.output_buffer.append(
+                f"Frame time: {metrics['frame_time']:.1f}ms"
+            )
+
+        self.state.output_buffer.append(
+            f"Output lines: {len(self.state.output_buffer)}"
+        )
+        self.state.output_buffer.append(
+            f"History: {len(self.state.command_history)} commands"
+        )
+
+    def _cmd_effects(self, ctx: EffectContext | None):
+        """List all effects."""
+        if ctx:
+            # Try to get effect list from context
+            effects = ctx.get_state("pipeline_order")
+            if effects:
+                self.state.output_buffer.append("Pipeline effects:")
+                for i, name in enumerate(effects):
+                    self.state.output_buffer.append(f"  {i + 1}. {name}")
+            else:
+                self.state.output_buffer.append("No pipeline information available")
+        else:
+            self.state.output_buffer.append("No context available")
+
+    def _cmd_effect(self, args: list[str], ctx: EffectContext | None):
+        """Toggle effect on/off."""
+        if len(args) < 2:
+            self.state.output_buffer.append("Usage: effect <name> <on|off>")
+            return
+
+        effect_name = args[0]
+        state = args[1].lower()
+
+        if state not in ("on", "off"):
+            self.state.output_buffer.append("State must be 'on' or 'off'")
+            return
+
+        # Emit event to toggle effect
+        enabled = state == "on"
+        self.state.output_buffer.append(f"Effect '{effect_name}' set to {state}")
+
+        # Store command for external handling
+        self._pending_command = {
+            "action": "enable_stage" if enabled else "disable_stage",
+            "stage": effect_name,
+        }
+
+    def _cmd_param(self, args: list[str], ctx: EffectContext | None):
+        """Set effect parameter."""
+        if len(args) < 3:
+            self.state.output_buffer.append("Usage: param <effect> <param> <value>")
+            return
+
+        effect_name = args[0]
+        param_name = args[1]
+        try:
+            param_value = float(args[2])
+        except ValueError:
+            self.state.output_buffer.append("Value must be a number")
+            return
+
+        self.state.output_buffer.append(
+            f"Setting {effect_name}.{param_name} = {param_value}"
+        )
+
+        # Store command for external handling
+        self._pending_command = {
+            "action": "adjust_param",
+            "stage": effect_name,
+            "param": param_name,
+            "delta": param_value,  # Note: This sets absolute value, need adjustment
+        }
+
+    def _cmd_pipeline(self, ctx: EffectContext | None):
+        """Show current pipeline order."""
+        if ctx:
+            pipeline_order = ctx.get_state("pipeline_order")
+            if pipeline_order:
+                self.state.output_buffer.append(
+                    "Pipeline: " + " → ".join(pipeline_order)
+                )
+            else:
+                self.state.output_buffer.append("Pipeline information not available")
+        else:
+            self.state.output_buffer.append("No context available")
+
+    def get_pending_command(self) -> dict | None:
+        """Get and clear pending command for external handling."""
+        cmd = getattr(self, "_pending_command", None)
+        if cmd:
+            self._pending_command = None
+        return cmd
+
+    def navigate_history(self, direction: int) -> None:
+        """Navigate command history (up/down)."""
+        if not self.state.command_history:
+            return
+
+        if direction > 0:  # Down
+            self.state.history_index = min(
+                len(self.state.command_history), self.state.history_index + 1
+            )
+        else:  # Up
+            self.state.history_index = max(0, self.state.history_index - 1)
+
+        if self.state.history_index < len(self.state.command_history):
+            self.state.current_command = self.state.command_history[
+                self.state.history_index
+            ]
+        else:
+            self.state.current_command = ""
+
+    def append_to_command(self, char: str) -> None:
+        """Append character to current command."""
+        if len(char) == 1:  # Single character
+            self.state.current_command += char
+
+    def backspace(self) -> None:
+        """Remove last character from command."""
+        self.state.current_command = self.state.current_command[:-1]
+
+    def clear_command(self) -> None:
+        """Clear current command."""
+        self.state.current_command = ""
+
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect."""
+        self.config = config

engine/effects/plugins/tint.py

@@ -0,0 +1,99 @@
+from engine.effects.types import EffectConfig, EffectContext, EffectPlugin
+
+
+class TintEffect(EffectPlugin):
+    """Tint effect that applies an RGB color overlay to the buffer.
+
+    Uses ANSI escape codes to tint text with the specified RGB values.
+    Supports transparency (0-100%) for blending.
+
+    Inlets:
+    - r: Red component (0-255)
+    - g: Green component (0-255)
+    - b: Blue component (0-255)
+    - a: Alpha/transparency (0.0-1.0, where 0.0 = fully transparent)
+    """
+
+    name = "tint"
+    config = EffectConfig(enabled=True, intensity=1.0)
+
+    # Define inlet types for PureData-style typing
+    @property
+    def inlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.TEXT_BUFFER}
+
+    @property
+    def outlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        if not buf:
+            return buf
+
+        # Get tint values from effect params or sensors
+        r = self.config.params.get("r", 255)
+        g = self.config.params.get("g", 255)
+        b = self.config.params.get("b", 255)
+        a = self.config.params.get("a", 0.3)  # Default 30% tint
+
+        # Clamp values
+        r = max(0, min(255, int(r)))
+        g = max(0, min(255, int(g)))
+        b = max(0, min(255, int(b)))
+        a = max(0.0, min(1.0, float(a)))
+
+        if a <= 0:
+            return buf
+
+        # Convert RGB to ANSI 256 color
+        ansi_color = self._rgb_to_ansi256(r, g, b)
+
+        # Apply tint with transparency effect
+        result = []
+        for line in buf:
+            if not line.strip():
+                result.append(line)
+                continue
+
+            # Check if line already has ANSI codes
+            if "\033[" in line:
+                # For lines with existing colors, wrap the whole line
+                result.append(f"\033[38;5;{ansi_color}m{line}\033[0m")
+            else:
+                # Apply tint to plain text lines
+                result.append(f"\033[38;5;{ansi_color}m{line}\033[0m")
+
+        return result
+
+    def _rgb_to_ansi256(self, r: int, g: int, b: int) -> int:
+        """Convert RGB (0-255 each) to ANSI 256 color code."""
+        if r == g == b == 0:
+            return 16
+        if r == g == b == 255:
+            return 231
+
+        # Calculate grayscale
+        gray = int((0.299 * r + 0.587 * g + 0.114 * b) / 255 * 24) + 232
+
+        # Calculate color cube
+        ri = int(r / 51)
+        gi = int(g / 51)
+        bi = int(b / 51)
+        color = 16 + 36 * ri + 6 * gi + bi
+
+        # Use whichever is closer - gray or color
+        gray_dist = abs(r - gray)
+        color_dist = (
+            (r - ri * 51) ** 2 + (g - gi * 51) ** 2 + (b - bi * 51) ** 2
+        ) ** 0.5
+
+        if gray_dist < color_dist:
+            return gray
+        return color
+
+    def configure(self, config: EffectConfig) -> None:
+        self.config = config

engine/effects/registry.py

@@ -0,0 +1,59 @@
+from engine.effects.types import EffectConfig, EffectPlugin
+
+
+class EffectRegistry:
+    def __init__(self):
+        self._plugins: dict[str, EffectPlugin] = {}
+        self._discovered: bool = False
+
+    def register(self, plugin: EffectPlugin) -> None:
+        self._plugins[plugin.name] = plugin
+
+    def get(self, name: str) -> EffectPlugin | None:
+        return self._plugins.get(name)
+
+    def list_all(self) -> dict[str, EffectPlugin]:
+        return self._plugins.copy()
+
+    def list_enabled(self) -> list[EffectPlugin]:
+        return [p for p in self._plugins.values() if p.config.enabled]
+
+    def enable(self, name: str) -> bool:
+        plugin = self._plugins.get(name)
+        if plugin:
+            plugin.config.enabled = True
+            return True
+        return False
+
+    def disable(self, name: str) -> bool:
+        plugin = self._plugins.get(name)
+        if plugin:
+            plugin.config.enabled = False
+            return True
+        return False
+
+    def configure(self, name: str, config: EffectConfig) -> bool:
+        plugin = self._plugins.get(name)
+        if plugin:
+            plugin.configure(config)
+            return True
+        return False
+
+    def is_enabled(self, name: str) -> bool:
+        plugin = self._plugins.get(name)
+        return plugin.config.enabled if plugin else False
+
+
+_registry: EffectRegistry | None = None
+
+
+def get_registry() -> EffectRegistry:
+    global _registry
+    if _registry is None:
+        _registry = EffectRegistry()
+    return _registry
+
+
+def set_registry(registry: EffectRegistry) -> None:
+    global _registry
+    _registry = registry

engine/effects/types.py

@@ -0,0 +1,281 @@
+"""
+Visual effects type definitions and base classes.
+
+EffectPlugin Architecture:
+- Uses ABC (Abstract Base Class) for interface enforcement
+- Runtime discovery via directory scanning (effects_plugins/)
+- Configuration via EffectConfig dataclass
+- Context passed through EffectContext dataclass
+
+Plugin System Research (see AGENTS.md for references):
+- VST: Standardized audio interfaces, chaining, presets (FXP/FXB)
+- Python Entry Points: Namespace packages, importlib.metadata discovery
+- Shadertoy: Shader-based with uniforms as context
+
+Current gaps vs industry patterns:
+- No preset save/load system
+- No external plugin distribution via entry points
+- No plugin metadata (version, author, description)
+"""
+
+from abc import ABC, abstractmethod
+from dataclasses import dataclass, field
+from typing import Any
+
+
+@dataclass
+class PartialUpdate:
+    """Represents a partial buffer update for optimized rendering.
+
+    Instead of processing the full buffer every frame, effects that support
+    partial updates can process only changed regions.
+
+    Attributes:
+        rows: Row indices that changed (None = all rows)
+        cols: Column range that changed (None = full width)
+        dirty: Set of dirty row indices
+    """
+
+    rows: tuple[int, int] | None = None  # (start, end) inclusive
+    cols: tuple[int, int] | None = None  # (start, end) inclusive
+    dirty: set[int] | None = None  # Set of dirty row indices
+    full_buffer: bool = True  # If True, process entire buffer
+
+
+@dataclass
+class EffectContext:
+    """Context passed to effect plugins during processing.
+
+    Contains terminal dimensions, camera state, frame info, and real-time sensor values.
+    """
+
+    terminal_width: int
+    terminal_height: int
+    scroll_cam: int
+    ticker_height: int
+    camera_x: int = 0
+    mic_excess: float = 0.0
+    grad_offset: float = 0.0
+    frame_number: int = 0
+    has_message: bool = False
+    items: list = field(default_factory=list)
+    _state: dict[str, Any] = field(default_factory=dict, repr=False)
+
+    def compute_entropy(self, effect_name: str, data: Any) -> float:
+        """Compute entropy score for an effect based on its output.
+
+        Args:
+            effect_name: Name of the effect
+            data: Processed buffer or effect-specific data
+
+        Returns:
+            Entropy score 0.0-1.0 representing visual chaos
+        """
+        # Default implementation: use effect name as seed for deterministic randomness
+        # Better implementations can analyze actual buffer content
+        import hashlib
+
+        data_str = str(data)[:100] if data else ""
+        hash_val = hashlib.md5(f"{effect_name}:{data_str}".encode()).hexdigest()
+        # Convert hash to float 0.0-1.0
+        entropy = int(hash_val[:8], 16) / 0xFFFFFFFF
+        return min(max(entropy, 0.0), 1.0)
+
+    def get_sensor_value(self, sensor_name: str) -> float | None:
+        """Get a sensor value from context state.
+
+        Args:
+            sensor_name: Name of the sensor (e.g., "mic", "camera")
+
+        Returns:
+            Sensor value as float, or None if not available.
+        """
+        return self._state.get(f"sensor.{sensor_name}")
+
+    def set_state(self, key: str, value: Any) -> None:
+        """Set a state value in the context."""
+        self._state[key] = value
+
+    def get_state(self, key: str, default: Any = None) -> Any:
+        """Get a state value from the context."""
+        return self._state.get(key, default)
+
+    @property
+    def state(self) -> dict[str, Any]:
+        """Get the state dictionary for direct access by effects."""
+        return self._state
+
+
+@dataclass
+class EffectConfig:
+    enabled: bool = True
+    intensity: float = 1.0
+    entropy: float = 0.0  # Visual chaos metric (0.0 = calm, 1.0 = chaotic)
+    params: dict[str, Any] = field(default_factory=dict)
+
+
+class EffectPlugin(ABC):
+    """Abstract base class for effect plugins.
+
+    Subclasses must define:
+    - name: str - unique identifier for the effect
+    - config: EffectConfig - current configuration
+
+    Optional class attribute:
+    - param_bindings: dict - Declarative sensor-to-param bindings
+        Example:
+            param_bindings = {
+                "intensity": {"sensor": "mic", "transform": "linear"},
+                "rate": {"sensor": "mic", "transform": "exponential"},
+            }
+
+    And implement:
+    - process(buf, ctx) -> list[str]
+    - configure(config) -> None
+
+    Effect Behavior with ticker_height=0:
+    - NoiseEffect: Returns buffer unchanged (no ticker to apply noise to)
+    - FadeEffect: Returns buffer unchanged (no ticker to fade)
+    - GlitchEffect: Processes normally (doesn't depend on ticker_height)
+    - FirehoseEffect: Returns buffer unchanged if no items in context
+
+    Effects should handle missing or zero context values gracefully by
+    returning the input buffer unchanged rather than raising errors.
+
+    The param_bindings system enables PureData-style signal routing:
+    - Sensors emit values that effects can bind to
+    - Transform functions map sensor values to param ranges
+    - Effects read bound values from context.state["sensor.{name}"]
+    """
+
+    name: str
+    config: EffectConfig
+    param_bindings: dict[str, dict[str, str | float]] = {}
+    supports_partial_updates: bool = False  # Override in subclasses for optimization
+
+    @abstractmethod
+    def process(self, buf: list[str], ctx: EffectContext) -> list[str]:
+        """Process the buffer with this effect applied.
+
+        Args:
+            buf: List of lines to process
+            ctx: Effect context with terminal state
+
+        Returns:
+            Processed buffer (may be same object or new list)
+        """
+        ...
+
+    def process_partial(
+        self, buf: list[str], ctx: EffectContext, partial: PartialUpdate
+    ) -> list[str]:
+        """Process a partial buffer for optimized rendering.
+
+        Override this in subclasses that support partial updates for performance.
+        Default implementation falls back to full buffer processing.
+
+        Args:
+            buf: List of lines to process
+            ctx: Effect context with terminal state
+            partial: PartialUpdate indicating which regions changed
+
+        Returns:
+            Processed buffer (may be same object or new list)
+        """
+        # Default: fall back to full processing
+        return self.process(buf, ctx)
+
+    @abstractmethod
+    def configure(self, config: EffectConfig) -> None:
+        """Configure the effect with new settings.
+
+        Args:
+            config: New configuration to apply
+        """
+        ...
+
+
+def create_effect_context(
+    terminal_width: int = 80,
+    terminal_height: int = 24,
+    scroll_cam: int = 0,
+    ticker_height: int = 0,
+    mic_excess: float = 0.0,
+    grad_offset: float = 0.0,
+    frame_number: int = 0,
+    has_message: bool = False,
+    items: list | None = None,
+) -> EffectContext:
+    """Factory function to create EffectContext with sensible defaults."""
+    return EffectContext(
+        terminal_width=terminal_width,
+        terminal_height=terminal_height,
+        scroll_cam=scroll_cam,
+        ticker_height=ticker_height,
+        mic_excess=mic_excess,
+        grad_offset=grad_offset,
+        frame_number=frame_number,
+        has_message=has_message,
+        items=items or [],
+    )
+
+
+@dataclass
+class PipelineConfig:
+    order: list[str] = field(default_factory=list)
+    effects: dict[str, EffectConfig] = field(default_factory=dict)
+
+
+def apply_param_bindings(
+    effect: "EffectPlugin",
+    ctx: EffectContext,
+) -> EffectConfig:
+    """Apply sensor bindings to effect config.
+
+    This resolves param_bindings declarations by reading sensor values
+    from the context and applying transform functions.
+
+    Args:
+        effect: The effect with param_bindings to apply
+        ctx: EffectContext containing sensor values
+
+    Returns:
+        Modified EffectConfig with sensor-driven values applied.
+    """
+    import copy
+
+    if not effect.param_bindings:
+        return effect.config
+
+    config = copy.deepcopy(effect.config)
+
+    for param_name, binding in effect.param_bindings.items():
+        sensor_name: str = binding.get("sensor", "")
+        transform: str = binding.get("transform", "linear")
+
+        if not sensor_name:
+            continue
+
+        sensor_value = ctx.get_sensor_value(sensor_name)
+        if sensor_value is None:
+            continue
+
+        if transform == "linear":
+            applied_value: float = sensor_value
+        elif transform == "exponential":
+            applied_value = sensor_value**2
+        elif transform == "threshold":
+            threshold = float(binding.get("threshold", 0.5))
+            applied_value = 1.0 if sensor_value > threshold else 0.0
+        elif transform == "inverse":
+            applied_value = 1.0 - sensor_value
+        else:
+            applied_value = sensor_value
+
+        config.params[f"{param_name}_sensor"] = applied_value
+
+        if param_name == "intensity":
+            base_intensity = effect.config.intensity
+            config.intensity = base_intensity * (0.5 + applied_value * 0.5)
+
+    return config

engine/eventbus.py

@@ -0,0 +1,72 @@
+"""
+Event bus - pub/sub messaging for decoupled component communication.
+"""
+
+import threading
+from collections import defaultdict
+from collections.abc import Callable
+from typing import Any
+
+from engine.events import EventType
+
+
+class EventBus:
+    """Thread-safe event bus for publish-subscribe messaging."""
+
+    def __init__(self):
+        self._subscribers: dict[EventType, list[Callable[[Any], None]]] = defaultdict(
+            list
+        )
+        self._lock = threading.Lock()
+
+    def subscribe(self, event_type: EventType, callback: Callable[[Any], None]) -> None:
+        """Register a callback for a specific event type."""
+        with self._lock:
+            self._subscribers[event_type].append(callback)
+
+    def unsubscribe(
+        self, event_type: EventType, callback: Callable[[Any], None]
+    ) -> None:
+        """Remove a callback for a specific event type."""
+        with self._lock:
+            if callback in self._subscribers[event_type]:
+                self._subscribers[event_type].remove(callback)
+
+    def publish(self, event_type: EventType, event: Any = None) -> None:
+        """Publish an event to all subscribers."""
+        with self._lock:
+            callbacks = list(self._subscribers.get(event_type, []))
+        for callback in callbacks:
+            try:
+                callback(event)
+            except Exception:
+                pass
+
+    def clear(self) -> None:
+        """Remove all subscribers."""
+        with self._lock:
+            self._subscribers.clear()
+
+    def subscriber_count(self, event_type: EventType | None = None) -> int:
+        """Get subscriber count for an event type, or total if None."""
+        with self._lock:
+            if event_type is None:
+                return sum(len(cb) for cb in self._subscribers.values())
+            return len(self._subscribers.get(event_type, []))
+
+
+_event_bus: EventBus | None = None
+
+
+def get_event_bus() -> EventBus:
+    """Get the global event bus instance."""
+    global _event_bus
+    if _event_bus is None:
+        _event_bus = EventBus()
+    return _event_bus
+
+
+def set_event_bus(bus: EventBus) -> None:
+    """Set the global event bus instance (for testing)."""
+    global _event_bus
+    _event_bus = bus

engine/events.py

@@ -0,0 +1,67 @@
+"""
+Event types for the mainline application.
+Defines the core events that flow through the system.
+These types support a future migration to an event-driven architecture.
+"""
+
+from dataclasses import dataclass
+from datetime import datetime
+from enum import Enum, auto
+
+
+class EventType(Enum):
+    """Core event types in the mainline application."""
+
+    NEW_HEADLINE = auto()
+    FRAME_TICK = auto()
+    MIC_LEVEL = auto()
+    NTFY_MESSAGE = auto()
+    STREAM_START = auto()
+    STREAM_END = auto()
+
+
+@dataclass
+class HeadlineEvent:
+    """Event emitted when a new headline is ready for display."""
+
+    title: str
+    source: str
+    timestamp: str
+    language: str | None = None
+
+
+@dataclass
+class FrameTickEvent:
+    """Event emitted on each render frame."""
+
+    frame_number: int
+    timestamp: datetime
+    delta_seconds: float
+
+
+@dataclass
+class MicLevelEvent:
+    """Event emitted when microphone level changes significantly."""
+
+    db_level: float
+    excess_above_threshold: float
+    timestamp: datetime
+
+
+@dataclass
+class NtfyMessageEvent:
+    """Event emitted when an ntfy message is received."""
+
+    title: str
+    body: str
+    message_id: str | None = None
+    timestamp: datetime | None = None
+
+
+@dataclass
+class StreamEvent:
+    """Event emitted when stream starts or ends."""
+
+    event_type: EventType
+    headline_count: int = 0
+    timestamp: datetime | None = None

engine/fetch.py

@@ -0,0 +1,198 @@
+"""
+RSS feed fetching, Project Gutenberg parsing, and headline caching.
+Depends on: config, sources, filter, terminal.
+"""
+
+import json
+import pathlib
+import re
+import urllib.request
+from concurrent.futures import ThreadPoolExecutor, as_completed
+from datetime import datetime
+from typing import Any
+
+import feedparser
+
+from engine import config
+from engine.filter import skip, strip_tags
+from engine.sources import FEEDS, POETRY_SOURCES
+from engine.terminal import boot_ln
+
+HeadlineTuple = tuple[str, str, str]
+
+DEFAULT_MAX_WORKERS = 10
+FAST_START_SOURCES = 5
+FAST_START_TIMEOUT = 3
+
+
+def fetch_feed(url: str) -> tuple[str, Any] | tuple[None, None]:
+    """Fetch and parse a single RSS feed URL. Returns (url, feed) tuple."""
+    try:
+        req = urllib.request.Request(url, headers={"User-Agent": "mainline/0.1"})
+        timeout = FAST_START_TIMEOUT if url in _fast_start_urls else config.FEED_TIMEOUT
+        resp = urllib.request.urlopen(req, timeout=timeout)
+        return (url, feedparser.parse(resp.read()))
+    except Exception:
+        return (url, None)
+
+
+def _parse_feed(feed: Any, src: str) -> list[HeadlineTuple]:
+    """Parse a feed and return list of headline tuples."""
+    items = []
+    if feed is None or (feed.bozo and not feed.entries):
+        return items
+
+    for e in feed.entries:
+        t = strip_tags(e.get("title", ""))
+        if not t or skip(t):
+            continue
+        pub = e.get("published_parsed") or e.get("updated_parsed")
+        try:
+            ts = datetime(*pub[:6]).strftime("%H:%M") if pub else "——:——"
+        except Exception:
+            ts = "——:——"
+        items.append((t, src, ts))
+    return items
+
+
+def fetch_all_fast() -> list[HeadlineTuple]:
+    """Fetch only the first N sources for fast startup."""
+    global _fast_start_urls
+    _fast_start_urls = set(list(FEEDS.values())[:FAST_START_SOURCES])
+
+    items: list[HeadlineTuple] = []
+    with ThreadPoolExecutor(max_workers=FAST_START_SOURCES) as executor:
+        futures = {
+            executor.submit(fetch_feed, url): src
+            for src, url in list(FEEDS.items())[:FAST_START_SOURCES]
+        }
+        for future in as_completed(futures):
+            src = futures[future]
+            url, feed = future.result()
+            if feed is None or (feed.bozo and not feed.entries):
+                boot_ln(src, "DARK", False)
+                continue
+            parsed = _parse_feed(feed, src)
+            if parsed:
+                items.extend(parsed)
+                boot_ln(src, f"LINKED [{len(parsed)}]", True)
+            else:
+                boot_ln(src, "EMPTY", False)
+    return items
+
+
+def fetch_all() -> tuple[list[HeadlineTuple], int, int]:
+    """Fetch all RSS feeds concurrently and return items, linked count, failed count."""
+    global _fast_start_urls
+    _fast_start_urls = set()
+
+    items: list[HeadlineTuple] = []
+    linked = failed = 0
+
+    with ThreadPoolExecutor(max_workers=DEFAULT_MAX_WORKERS) as executor:
+        futures = {executor.submit(fetch_feed, url): src for src, url in FEEDS.items()}
+        for future in as_completed(futures):
+            src = futures[future]
+            url, feed = future.result()
+            if feed is None or (feed.bozo and not feed.entries):
+                boot_ln(src, "DARK", False)
+                failed += 1
+                continue
+            parsed = _parse_feed(feed, src)
+            if parsed:
+                items.extend(parsed)
+                boot_ln(src, f"LINKED [{len(parsed)}]", True)
+                linked += 1
+            else:
+                boot_ln(src, "EMPTY", False)
+                failed += 1
+
+    return items, linked, failed
+
+
+def _fetch_gutenberg(url: str, label: str) -> list[HeadlineTuple]:
+    """Download and parse stanzas/passages from a Project Gutenberg text."""
+    try:
+        req = urllib.request.Request(url, headers={"User-Agent": "mainline/0.1"})
+        resp = urllib.request.urlopen(req, timeout=15)
+        text = (
+            resp.read()
+            .decode("utf-8", errors="replace")
+            .replace("\r\n", "\n")
+            .replace("\r", "\n")
+        )
+        m = re.search(r"\*\*\*\s*START OF[^\n]*\n", text)
+        if m:
+            text = text[m.end() :]
+        m = re.search(r"\*\*\*\s*END OF", text)
+        if m:
+            text = text[: m.start()]
+        blocks = re.split(r"\n{2,}", text.strip())
+        items = []
+        for blk in blocks:
+            blk = " ".join(blk.split())
+            if len(blk) < 20 or len(blk) > 280:
+                continue
+            if blk.isupper():
+                continue
+            if re.match(r"^[IVXLCDM]+\.?\s*$", blk):
+                continue
+            items.append((blk, label, ""))
+        return items
+    except Exception:
+        return []
+
+
+def fetch_poetry() -> tuple[list[HeadlineTuple], int, int]:
+    """Fetch all poetry/literature sources concurrently."""
+    items = []
+    linked = failed = 0
+
+    with ThreadPoolExecutor(max_workers=DEFAULT_MAX_WORKERS) as executor:
+        futures = {
+            executor.submit(_fetch_gutenberg, url, label): label
+            for label, url in POETRY_SOURCES.items()
+        }
+        for future in as_completed(futures):
+            label = futures[future]
+            stanzas = future.result()
+            if stanzas:
+                boot_ln(label, f"LOADED [{len(stanzas)}]", True)
+                items.extend(stanzas)
+                linked += 1
+            else:
+                boot_ln(label, "DARK", False)
+                failed += 1
+
+    return items, linked, failed
+
+
+_cache_dir = pathlib.Path(__file__).resolve().parent / "fixtures"
+
+
+def _cache_path():
+    return _cache_dir / "headlines.json"
+
+
+def load_cache():
+    """Load cached items from disk if available."""
+    p = _cache_path()
+    if not p.exists():
+        return None
+    try:
+        data = json.loads(p.read_text())
+        items = [tuple(i) for i in data["items"]]
+        return items if items else None
+    except Exception:
+        return None
+
+
+def save_cache(items):
+    """Save fetched items to disk for fast subsequent runs."""
+    try:
+        _cache_path().write_text(json.dumps({"items": items}))
+    except Exception:
+        pass
+
+
+_fast_start_urls: set = set()

engine/figment_render.py

@@ -0,0 +1,90 @@
+"""
+SVG to half-block terminal art rasterization.
+
+Pipeline: SVG -> cairosvg -> PIL -> greyscale threshold -> half-block encode.
+Follows the same pixel-pair approach as engine/render.py for OTF fonts.
+"""
+
+from __future__ import annotations
+
+import os
+import sys
+from io import BytesIO
+
+# cairocffi (used by cairosvg) calls dlopen() to find the Cairo C library.
+# On macOS with Homebrew, Cairo lives in /opt/homebrew/lib (Apple Silicon) or
+# /usr/local/lib (Intel), which are not in dyld's default search path.
+# Setting DYLD_LIBRARY_PATH before the import directs dlopen() to those paths.
+if sys.platform == "darwin" and not os.environ.get("DYLD_LIBRARY_PATH"):
+    for _brew_lib in ("/opt/homebrew/lib", "/usr/local/lib"):
+        if os.path.exists(os.path.join(_brew_lib, "libcairo.2.dylib")):
+            os.environ["DYLD_LIBRARY_PATH"] = _brew_lib
+            break
+
+import cairosvg
+from PIL import Image
+
+_cache: dict[tuple[str, int, int], list[str]] = {}
+
+
+def rasterize_svg(svg_path: str, width: int, height: int) -> list[str]:
+    """Convert SVG file to list of half-block terminal rows (uncolored).
+
+    Args:
+        svg_path: Path to SVG file.
+        width: Target terminal width in columns.
+        height: Target terminal height in rows.
+
+    Returns:
+        List of strings, one per terminal row, containing block characters.
+    """
+    cache_key = (svg_path, width, height)
+    if cache_key in _cache:
+        return _cache[cache_key]
+
+    # SVG -> PNG in memory
+    png_bytes = cairosvg.svg2png(
+        url=svg_path,
+        output_width=width,
+        output_height=height * 2,  # 2 pixel rows per terminal row
+    )
+
+    # PNG -> greyscale PIL image
+    # Composite RGBA onto white background so transparent areas become white (255)
+    # and drawn pixels retain their luminance values.
+    img_rgba = Image.open(BytesIO(png_bytes)).convert("RGBA")
+    img_rgba = img_rgba.resize((width, height * 2), Image.Resampling.LANCZOS)
+    background = Image.new("RGBA", img_rgba.size, (255, 255, 255, 255))
+    background.paste(img_rgba, mask=img_rgba.split()[3])
+    img = background.convert("L")
+
+    data = img.tobytes()
+    pix_w = width
+    pix_h = height * 2
+    # White (255) = empty space, dark (< threshold) = filled pixel
+    threshold = 128
+
+    # Half-block encode: walk pixel pairs
+    rows: list[str] = []
+    for y in range(0, pix_h, 2):
+        row: list[str] = []
+        for x in range(pix_w):
+            top = data[y * pix_w + x] < threshold
+            bot = data[(y + 1) * pix_w + x] < threshold if y + 1 < pix_h else False
+            if top and bot:
+                row.append("█")
+            elif top:
+                row.append("▀")
+            elif bot:
+                row.append("▄")
+            else:
+                row.append(" ")
+        rows.append("".join(row))
+
+    _cache[cache_key] = rows
+    return rows
+
+
+def clear_cache() -> None:
+    """Clear the rasterization cache (e.g., on terminal resize)."""
+    _cache.clear()

engine/figment_trigger.py

@@ -0,0 +1,36 @@
+"""
+Figment trigger protocol and command types.
+
+Defines the extensible input abstraction for triggering figment displays
+from any control surface (ntfy, MQTT, serial, etc.).
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+from enum import Enum
+from typing import Protocol
+
+
+class FigmentAction(Enum):
+    TRIGGER = "trigger"
+    SET_INTENSITY = "set_intensity"
+    SET_INTERVAL = "set_interval"
+    SET_COLOR = "set_color"
+    STOP = "stop"
+
+
+@dataclass
+class FigmentCommand:
+    action: FigmentAction
+    value: float | str | None = None
+
+
+class FigmentTrigger(Protocol):
+    """Protocol for figment trigger sources.
+
+    Any input source (ntfy, MQTT, serial) can implement this
+    to trigger and control figment displays.
+    """
+
+    def poll(self) -> FigmentCommand | None: ...

engine/filter.py

@@ -0,0 +1,60 @@
+"""
+HTML stripping and content filter (sports, vapid, insipid).
+No internal dependencies.
+"""
+
+import re
+from html import unescape
+from html.parser import HTMLParser
+
+
+# ─── HTML STRIPPING ───────────────────────────────────────
+class _Strip(HTMLParser):
+    def __init__(self):
+        super().__init__()
+        self._t = []
+
+    def handle_data(self, d):
+        self._t.append(d)
+
+    def text(self):
+        return "".join(self._t).strip()
+
+
+def strip_tags(html):
+    s = _Strip()
+    s.feed(unescape(html or ""))
+    return s.text()
+
+
+# ─── CONTENT FILTER ───────────────────────────────────────
+_SKIP_RE = re.compile(
+    r"\b(?:"
+    # ── sports ──
+    r"football|soccer|basketball|baseball|softball|tennis|golf|cricket|rugby|"
+    r"hockey|lacrosse|volleyball|badminton|"
+    r"nba|nfl|nhl|mlb|mls|fifa|uefa|"
+    r"premier league|champions league|la liga|serie a|bundesliga|"
+    r"world cup|super bowl|world series|stanley cup|"
+    r"playoff|playoffs|touchdown|goalkeeper|striker|quarterback|"
+    r"slam dunk|home run|grand slam|offside|halftime|"
+    r"batting|wicket|innings|"
+    r"formula 1|nascar|motogp|"
+    r"boxing|ufc|mma|"
+    r"marathon|tour de france|"
+    r"transfer window|draft pick|relegation|"
+    # ── vapid / insipid ──
+    r"kardashian|jenner|reality tv|reality show|"
+    r"influencer|viral video|tiktok|instagram|"
+    r"best dressed|worst dressed|red carpet|"
+    r"horoscope|zodiac|gossip|bikini|selfie|"
+    r"you won.t believe|what happened next|"
+    r"celebrity couple|celebrity feud|baby bump"
+    r")\b",
+    re.IGNORECASE,
+)
+
+
+def skip(title):
+    """Return True if headline is sports, vapid, or insipid."""
+    return bool(_SKIP_RE.search(title))

engine/frame.py

@@ -0,0 +1,57 @@
+"""
+Frame timing utilities — FPS control and precise timing.
+"""
+
+import time
+
+
+class FrameTimer:
+    """Frame timer for consistent render loop timing."""
+
+    def __init__(self, target_frame_dt: float = 0.05):
+        self.target_frame_dt = target_frame_dt
+        self._frame_count = 0
+        self._start_time = time.monotonic()
+        self._last_frame_time = self._start_time
+
+    @property
+    def fps(self) -> float:
+        """Current FPS based on elapsed frames."""
+        elapsed = time.monotonic() - self._start_time
+        if elapsed > 0:
+            return self._frame_count / elapsed
+        return 0.0
+
+    def sleep_until_next_frame(self) -> float:
+        """Sleep to maintain target frame rate. Returns actual elapsed time."""
+        now = time.monotonic()
+        elapsed = now - self._last_frame_time
+        self._last_frame_time = now
+        self._frame_count += 1
+
+        sleep_time = max(0, self.target_frame_dt - elapsed)
+        if sleep_time > 0:
+            time.sleep(sleep_time)
+        return elapsed
+
+    def reset(self) -> None:
+        """Reset frame counter and start time."""
+        self._frame_count = 0
+        self._start_time = time.monotonic()
+        self._last_frame_time = self._start_time
+
+
+def calculate_scroll_step(
+    scroll_dur: float, view_height: int, padding: int = 15
+) -> float:
+    """Calculate scroll step interval for smooth scrolling.
+
+    Args:
+        scroll_dur: Duration in seconds for one headline to scroll through view
+        view_height: Terminal height in rows
+        padding: Extra rows for off-screen content
+
+    Returns:
+        Time in seconds between scroll steps
+    """
+    return scroll_dur / (view_height + padding) * 2

engine/interfaces/__init__.py

@@ -0,0 +1,73 @@
+"""
+Core interfaces for the mainline pipeline architecture.
+
+This module provides all abstract base classes and protocols that define
+the contracts between pipeline components:
+
+- Stage: Base class for pipeline components (imported from pipeline.core)
+- DataSource: Abstract data providers (imported from data_sources.sources)
+- EffectPlugin: Visual effects interface (imported from effects.types)
+- Sensor: Real-time input interface (imported from sensors)
+- Display: Output backend protocol (imported from display)
+
+This module provides a centralized import location for all interfaces.
+"""
+
+from engine.data_sources.sources import (
+    DataSource,
+    ImageItem,
+    SourceItem,
+)
+from engine.display import Display
+from engine.effects.types import (
+    EffectConfig,
+    EffectContext,
+    EffectPlugin,
+    PartialUpdate,
+    PipelineConfig,
+    apply_param_bindings,
+    create_effect_context,
+)
+from engine.pipeline.core import (
+    DataType,
+    Stage,
+    StageConfig,
+    StageError,
+    StageResult,
+    create_stage_error,
+)
+from engine.sensors import (
+    Sensor,
+    SensorStage,
+    SensorValue,
+    create_sensor_stage,
+)
+
+__all__ = [
+    # Stage interfaces
+    "DataType",
+    "Stage",
+    "StageConfig",
+    "StageError",
+    "StageResult",
+    "create_stage_error",
+    # Data source interfaces
+    "DataSource",
+    "ImageItem",
+    "SourceItem",
+    # Effect interfaces
+    "EffectConfig",
+    "EffectContext",
+    "EffectPlugin",
+    "PartialUpdate",
+    "PipelineConfig",
+    "apply_param_bindings",
+    "create_effect_context",
+    # Sensor interfaces
+    "Sensor",
+    "SensorStage",
+    "SensorValue",
+    "create_sensor_stage",
+    # Display protocol
+    "Display",
+]

engine/ntfy.py

@@ -0,0 +1,122 @@
+"""
+ntfy.sh SSE stream listener — standalone, zero internal dependencies.
+Reusable by any visualizer:
+
+    from engine.ntfy import NtfyPoller
+    poller = NtfyPoller("https://ntfy.sh/my_topic/json")
+    poller.start()
+    # in render loop:
+    msg = poller.get_active_message()
+    if msg:
+        title, body, ts = msg
+        render_my_message(title, body)
+"""
+
+import json
+import threading
+import time
+import urllib.request
+from collections.abc import Callable
+from datetime import datetime
+from urllib.parse import parse_qs, urlencode, urlparse, urlunparse
+
+from engine.events import NtfyMessageEvent
+
+
+class NtfyPoller:
+    """SSE stream listener for ntfy.sh topics. Messages arrive in ~1s (network RTT)."""
+
+    def __init__(self, topic_url, reconnect_delay=5, display_secs=30):
+        self.topic_url = topic_url
+        self.reconnect_delay = reconnect_delay
+        self.display_secs = display_secs
+        self._message = None  # (title, body, monotonic_timestamp) or None
+        self._lock = threading.Lock()
+        self._subscribers: list[Callable[[NtfyMessageEvent], None]] = []
+
+    def subscribe(self, callback: Callable[[NtfyMessageEvent], None]) -> None:
+        """Register a callback to be called when a message is received."""
+        self._subscribers.append(callback)
+
+    def unsubscribe(self, callback: Callable[[NtfyMessageEvent], None]) -> None:
+        """Remove a registered callback."""
+        if callback in self._subscribers:
+            self._subscribers.remove(callback)
+
+    def _emit(self, event: NtfyMessageEvent) -> None:
+        """Emit an event to all subscribers."""
+        for cb in self._subscribers:
+            try:
+                cb(event)
+            except Exception:
+                pass
+
+    def start(self):
+        """Start background stream thread. Returns True."""
+        t = threading.Thread(target=self._stream_loop, daemon=True)
+        t.start()
+        return True
+
+    def get_active_message(self):
+        """Return (title, body, timestamp) if a message is active and not expired, else None."""
+        with self._lock:
+            if self._message is None:
+                return None
+            title, body, ts = self._message
+            if time.monotonic() - ts < self.display_secs:
+                return self._message
+            self._message = None
+            return None
+
+    def dismiss(self):
+        """Manually dismiss the current message."""
+        with self._lock:
+            self._message = None
+
+    def _build_url(self, last_id=None):
+        """Build the stream URL, substituting since= to avoid message replays on reconnect."""
+        parsed = urlparse(self.topic_url)
+        params = parse_qs(parsed.query, keep_blank_values=True)
+        params["since"] = [last_id if last_id else "20s"]
+        new_query = urlencode({k: v[0] for k, v in params.items()})
+        return urlunparse(parsed._replace(query=new_query))
+
+    def _stream_loop(self):
+        last_id = None
+        while True:
+            try:
+                url = self._build_url(last_id)
+                req = urllib.request.Request(
+                    url, headers={"User-Agent": "mainline/0.1"}
+                )
+                # timeout=90 keeps the socket alive through ntfy.sh keepalive heartbeats
+                resp = urllib.request.urlopen(req, timeout=90)
+                while True:
+                    line = resp.readline()
+                    if not line:
+                        break  # server closed connection — reconnect
+                    try:
+                        data = json.loads(line.decode("utf-8", errors="replace"))
+                    except json.JSONDecodeError:
+                        continue
+                    # Advance cursor on every event (message + keepalive) to
+                    # avoid replaying already-seen events after a reconnect.
+                    if "id" in data:
+                        last_id = data["id"]
+                    if data.get("event") == "message":
+                        with self._lock:
+                            self._message = (
+                                data.get("title", ""),
+                                data.get("message", ""),
+                                time.monotonic(),
+                            )
+                        event = NtfyMessageEvent(
+                            title=data.get("title", ""),
+                            body=data.get("message", ""),
+                            message_id=data.get("id"),
+                            timestamp=datetime.now(),
+                        )
+                        self._emit(event)
+            except Exception:
+                pass
+            time.sleep(self.reconnect_delay)

engine/pipeline/__init__.py

@@ -0,0 +1,106 @@
+"""
+Unified Pipeline Architecture.
+
+This module provides a clean, dependency-managed pipeline system:
+- Stage: Base class for all pipeline components
+- Pipeline: DAG-based execution orchestrator
+- PipelineParams: Runtime configuration for animation
+- PipelinePreset: Pre-configured pipeline configurations
+- StageRegistry: Unified registration for all stage types
+
+The pipeline architecture supports:
+- Sources: Data providers (headlines, poetry, pipeline viz)
+- Effects: Post-processors (noise, fade, glitch, hud)
+- Displays: Output backends (terminal, pygame, websocket)
+- Cameras: Viewport controllers (vertical, horizontal, omni)
+
+Example:
+    from engine.pipeline import Pipeline, PipelineConfig, StageRegistry
+
+    pipeline = Pipeline(PipelineConfig(source="headlines", display="terminal"))
+    pipeline.add_stage("source", StageRegistry.create("source", "headlines"))
+    pipeline.add_stage("display", StageRegistry.create("display", "terminal"))
+    pipeline.build().initialize()
+
+    result = pipeline.execute(initial_data)
+"""
+
+from engine.pipeline.controller import (
+    Pipeline,
+    PipelineConfig,
+    PipelineRunner,
+    create_default_pipeline,
+    create_pipeline_from_params,
+)
+from engine.pipeline.core import (
+    PipelineContext,
+    Stage,
+    StageConfig,
+    StageError,
+    StageResult,
+)
+from engine.pipeline.params import (
+    DEFAULT_HEADLINE_PARAMS,
+    DEFAULT_PIPELINE_PARAMS,
+    DEFAULT_PYGAME_PARAMS,
+    PipelineParams,
+)
+from engine.pipeline.presets import (
+    DEMO_PRESET,
+    FIREHOSE_PRESET,
+    PIPELINE_VIZ_PRESET,
+    POETRY_PRESET,
+    UI_PRESET,
+    WEBSOCKET_PRESET,
+    PipelinePreset,
+    create_preset_from_params,
+    get_preset,
+    list_presets,
+)
+from engine.pipeline.registry import (
+    StageRegistry,
+    discover_stages,
+    register_camera,
+    register_display,
+    register_effect,
+    register_source,
+)
+
+__all__ = [
+    # Core
+    "Stage",
+    "StageConfig",
+    "StageError",
+    "StageResult",
+    "PipelineContext",
+    # Controller
+    "Pipeline",
+    "PipelineConfig",
+    "PipelineRunner",
+    "create_default_pipeline",
+    "create_pipeline_from_params",
+    # Params
+    "PipelineParams",
+    "DEFAULT_HEADLINE_PARAMS",
+    "DEFAULT_PIPELINE_PARAMS",
+    "DEFAULT_PYGAME_PARAMS",
+    # Presets
+    "PipelinePreset",
+    "PRESETS",
+    "DEMO_PRESET",
+    "POETRY_PRESET",
+    "PIPELINE_VIZ_PRESET",
+    "WEBSOCKET_PRESET",
+    "FIREHOSE_PRESET",
+    "UI_PRESET",
+    "get_preset",
+    "list_presets",
+    "create_preset_from_params",
+    # Registry
+    "StageRegistry",
+    "discover_stages",
+    "register_source",
+    "register_effect",
+    "register_display",
+    "register_camera",
+]

engine/pipeline/adapters.py

@@ -0,0 +1,50 @@
+"""
+Stage adapters - Bridge existing components to the Stage interface.
+
+This module provides adapters that wrap existing components
+(EffectPlugin, Display, DataSource, Camera) as Stage implementations.
+
+DEPRECATED: This file is now a compatibility wrapper.
+Use `engine.pipeline.adapters` package instead.
+"""
+
+# Re-export from the new package structure for backward compatibility
+from engine.pipeline.adapters import (
+    # Adapter classes
+    CameraStage,
+    CanvasStage,
+    DataSourceStage,
+    DisplayStage,
+    EffectPluginStage,
+    FontStage,
+    ImageToTextStage,
+    PassthroughStage,
+    SourceItemsToBufferStage,
+    ViewportFilterStage,
+    # Factory functions
+    create_stage_from_camera,
+    create_stage_from_display,
+    create_stage_from_effect,
+    create_stage_from_font,
+    create_stage_from_source,
+)
+
+__all__ = [
+    # Adapter classes
+    "EffectPluginStage",
+    "DisplayStage",
+    "DataSourceStage",
+    "PassthroughStage",
+    "SourceItemsToBufferStage",
+    "CameraStage",
+    "ViewportFilterStage",
+    "FontStage",
+    "ImageToTextStage",
+    "CanvasStage",
+    # Factory functions
+    "create_stage_from_display",
+    "create_stage_from_effect",
+    "create_stage_from_source",
+    "create_stage_from_camera",
+    "create_stage_from_font",
+]

engine/pipeline/adapters/__init__.py

@@ -0,0 +1,55 @@
+"""Stage adapters - Bridge existing components to the Stage interface.
+
+This module provides adapters that wrap existing components
+(EffectPlugin, Display, DataSource, Camera) as Stage implementations.
+"""
+
+from .camera import CameraClockStage, CameraStage
+from .data_source import DataSourceStage, PassthroughStage, SourceItemsToBufferStage
+from .display import DisplayStage
+from .effect_plugin import EffectPluginStage
+from .factory import (
+    create_stage_from_camera,
+    create_stage_from_display,
+    create_stage_from_effect,
+    create_stage_from_font,
+    create_stage_from_source,
+)
+from .message_overlay import MessageOverlayConfig, MessageOverlayStage
+from .positioning import (
+    PositioningMode,
+    PositionStage,
+    create_position_stage,
+)
+from .transform import (
+    CanvasStage,
+    FontStage,
+    ImageToTextStage,
+    ViewportFilterStage,
+)
+
+__all__ = [
+    # Adapter classes
+    "EffectPluginStage",
+    "DisplayStage",
+    "DataSourceStage",
+    "PassthroughStage",
+    "SourceItemsToBufferStage",
+    "CameraStage",
+    "CameraClockStage",
+    "ViewportFilterStage",
+    "FontStage",
+    "ImageToTextStage",
+    "CanvasStage",
+    "MessageOverlayStage",
+    "MessageOverlayConfig",
+    "PositionStage",
+    "PositioningMode",
+    # Factory functions
+    "create_stage_from_display",
+    "create_stage_from_effect",
+    "create_stage_from_source",
+    "create_stage_from_camera",
+    "create_stage_from_font",
+    "create_position_stage",
+]

engine/pipeline/adapters/camera.py

@@ -0,0 +1,219 @@
+"""Adapter for camera stage."""
+
+import time
+from typing import Any
+
+from engine.pipeline.core import DataType, PipelineContext, Stage
+
+
+class CameraClockStage(Stage):
+    """Per-frame clock stage that updates camera state.
+
+    This stage runs once per frame and updates the camera's internal state
+    (position, time). It makes camera_y/camera_x available to subsequent
+    stages via the pipeline context.
+
+    Unlike other stages, this is a pure clock stage and doesn't process
+    data - it just updates camera state and passes data through unchanged.
+    """
+
+    def __init__(self, camera, name: str = "camera-clock"):
+        self._camera = camera
+        self.name = name
+        self.category = "camera"
+        self.optional = False
+        self._last_frame_time: float | None = None
+
+    @property
+    def stage_type(self) -> str:
+        return "camera"
+
+    @property
+    def capabilities(self) -> set[str]:
+        # Provides camera state info only
+        # NOTE: Do NOT provide "source" as it conflicts with viewport_filter's "source.filtered"
+        return {"camera.state"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        # Clock stage - no dependencies (updates every frame regardless of data flow)
+        return set()
+
+    @property
+    def inlet_types(self) -> set:
+        # Accept any data type - this is a pass-through stage
+        return {DataType.ANY}
+
+    @property
+    def outlet_types(self) -> set:
+        # Pass through whatever was received
+        return {DataType.ANY}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Update camera state and pass data through.
+
+        This stage updates the camera's internal state (position, time) and
+        makes the updated camera_y/camera_x available to subsequent stages
+        via the pipeline context.
+
+        The data is passed through unchanged - this stage only updates
+        camera state, it doesn't transform the data.
+        """
+        if data is None:
+            return data
+
+        # Update camera speed from params if explicitly set (for dynamic modulation)
+        # Only update if camera_speed in params differs from the default (1.0)
+        # This preserves camera speed set during construction
+        if (
+            ctx.params
+            and hasattr(ctx.params, "camera_speed")
+            and ctx.params.camera_speed != 1.0
+        ):
+            self._camera.set_speed(ctx.params.camera_speed)
+
+        current_time = time.perf_counter()
+        dt = 0.0
+        if self._last_frame_time is not None:
+            dt = current_time - self._last_frame_time
+            self._camera.update(dt)
+        self._last_frame_time = current_time
+
+        # Update context with current camera position
+        ctx.set_state("camera_y", self._camera.y)
+        ctx.set_state("camera_x", self._camera.x)
+
+        # Pass data through unchanged
+        return data
+
+
+class CameraStage(Stage):
+    """Adapter wrapping Camera as a Stage.
+
+    This stage applies camera viewport transformation to the rendered buffer.
+    Camera state updates are handled by CameraClockStage.
+    """
+
+    def __init__(self, camera, name: str = "vertical"):
+        self._camera = camera
+        self.name = name
+        self.category = "camera"
+        self.optional = True
+        self._last_frame_time: float | None = None
+
+    def save_state(self) -> dict[str, Any]:
+        """Save camera state for restoration after pipeline rebuild.
+
+        Returns:
+            Dictionary containing camera state that can be restored
+        """
+        state = {
+            "x": self._camera.x,
+            "y": self._camera.y,
+            "mode": self._camera.mode.value
+            if hasattr(self._camera.mode, "value")
+            else self._camera.mode,
+            "speed": self._camera.speed,
+            "zoom": self._camera.zoom,
+            "canvas_width": self._camera.canvas_width,
+            "canvas_height": self._camera.canvas_height,
+            "_x_float": getattr(self._camera, "_x_float", 0.0),
+            "_y_float": getattr(self._camera, "_y_float", 0.0),
+            "_time": getattr(self._camera, "_time", 0.0),
+        }
+        # Save radial camera state if present
+        if hasattr(self._camera, "_r_float"):
+            state["_r_float"] = self._camera._r_float
+        if hasattr(self._camera, "_theta_float"):
+            state["_theta_float"] = self._camera._theta_float
+        if hasattr(self._camera, "_radial_input"):
+            state["_radial_input"] = self._camera._radial_input
+        return state
+
+    def restore_state(self, state: dict[str, Any]) -> None:
+        """Restore camera state from saved state.
+
+        Args:
+            state: Dictionary containing camera state from save_state()
+        """
+        from engine.camera import CameraMode
+
+        self._camera.x = state.get("x", 0)
+        self._camera.y = state.get("y", 0)
+
+        # Restore mode - handle both enum value and direct enum
+        mode_value = state.get("mode", 0)
+        if isinstance(mode_value, int):
+            self._camera.mode = CameraMode(mode_value)
+        else:
+            self._camera.mode = mode_value
+
+        self._camera.speed = state.get("speed", 1.0)
+        self._camera.zoom = state.get("zoom", 1.0)
+        self._camera.canvas_width = state.get("canvas_width", 200)
+        self._camera.canvas_height = state.get("canvas_height", 200)
+
+        # Restore internal state
+        if hasattr(self._camera, "_x_float"):
+            self._camera._x_float = state.get("_x_float", 0.0)
+        if hasattr(self._camera, "_y_float"):
+            self._camera._y_float = state.get("_y_float", 0.0)
+        if hasattr(self._camera, "_time"):
+            self._camera._time = state.get("_time", 0.0)
+
+        # Restore radial camera state if present
+        if hasattr(self._camera, "_r_float"):
+            self._camera._r_float = state.get("_r_float", 0.0)
+        if hasattr(self._camera, "_theta_float"):
+            self._camera._theta_float = state.get("_theta_float", 0.0)
+        if hasattr(self._camera, "_radial_input"):
+            self._camera._radial_input = state.get("_radial_input", 0.0)
+
+    @property
+    def stage_type(self) -> str:
+        return "camera"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"camera"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return {"render.output", "camera.state"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Apply camera transformation to items."""
+        if data is None:
+            return data
+
+        # Camera state is updated by CameraClockStage
+        # We only apply the viewport transformation here
+
+        if hasattr(self._camera, "apply"):
+            viewport_width = ctx.params.viewport_width if ctx.params else 80
+            viewport_height = ctx.params.viewport_height if ctx.params else 24
+
+            # Use filtered camera position if available (from ViewportFilterStage)
+            # This handles the case where the buffer has been filtered and starts at row 0
+            filtered_camera_y = ctx.get("camera_y", self._camera.y)
+
+            # Temporarily adjust camera position for filtering
+            original_y = self._camera.y
+            self._camera.y = filtered_camera_y
+
+            try:
+                result = self._camera.apply(data, viewport_width, viewport_height)
+            finally:
+                # Restore original camera position
+                self._camera.y = original_y
+
+            return result
+        return data

engine/pipeline/adapters/data_source.py

@@ -0,0 +1,143 @@
+"""
+Stage adapters - Bridge existing components to the Stage interface.
+
+This module provides adapters that wrap existing components
+(DataSource) as Stage implementations.
+"""
+
+from typing import Any
+
+from engine.data_sources import SourceItem
+from engine.pipeline.core import DataType, PipelineContext, Stage
+
+
+class DataSourceStage(Stage):
+    """Adapter wrapping DataSource as a Stage."""
+
+    def __init__(self, data_source, name: str = "headlines"):
+        self._source = data_source
+        self.name = name
+        self.category = "source"
+        self.optional = False
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {f"source.{self.name}"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return set()
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.NONE}  # Sources don't take input
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Fetch data from source."""
+        if hasattr(self._source, "get_items"):
+            return self._source.get_items()
+        return data
+
+
+class PassthroughStage(Stage):
+    """Simple stage that passes data through unchanged.
+
+    Used for sources that already provide the data in the correct format
+    (e.g., pipeline introspection that outputs text directly).
+    """
+
+    def __init__(self, name: str = "passthrough"):
+        self.name = name
+        self.category = "render"
+        self.optional = True
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"render.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return {"source"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Pass data through unchanged."""
+        return data
+
+
+class SourceItemsToBufferStage(Stage):
+    """Convert SourceItem objects to text buffer.
+
+    Takes a list of SourceItem objects and extracts their content,
+    splitting on newlines to create a proper text buffer for display.
+    """
+
+    def __init__(self, name: str = "items-to-buffer"):
+        self.name = name
+        self.category = "render"
+        self.optional = True
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"render.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return {"source"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Convert SourceItem list to text buffer."""
+        if data is None:
+            return []
+
+        # If already a list of strings, return as-is
+        if isinstance(data, list) and data and isinstance(data[0], str):
+            return data
+
+        # If it's a list of SourceItem, extract content
+        if isinstance(data, list):
+            result = []
+            for item in data:
+                if isinstance(item, SourceItem):
+                    # Split content by newline to get individual lines
+                    lines = item.content.split("\n")
+                    result.extend(lines)
+                elif hasattr(item, "content"):  # Has content attribute
+                    lines = str(item.content).split("\n")
+                    result.extend(lines)
+                else:
+                    result.append(str(item))
+            return result
+
+        # Single item
+        if isinstance(data, SourceItem):
+            return data.content.split("\n")
+
+        return [str(data)]

engine/pipeline/adapters/display.py

@@ -0,0 +1,108 @@
+"""Adapter wrapping Display as a Stage."""
+
+from typing import Any
+
+from engine.pipeline.core import PipelineContext, Stage
+
+
+class DisplayStage(Stage):
+    """Adapter wrapping Display as a Stage."""
+
+    def __init__(self, display, name: str = "terminal", positioning: str = "mixed"):
+        self._display = display
+        self.name = name
+        self.category = "display"
+        self.optional = False
+        self._initialized = False
+        self._init_width = 80
+        self._init_height = 24
+        self._positioning = positioning
+
+    def save_state(self) -> dict[str, Any]:
+        """Save display state for restoration after pipeline rebuild.
+
+        Returns:
+            Dictionary containing display state that can be restored
+        """
+        return {
+            "initialized": self._initialized,
+            "init_width": self._init_width,
+            "init_height": self._init_height,
+            "width": getattr(self._display, "width", 80),
+            "height": getattr(self._display, "height", 24),
+        }
+
+    def restore_state(self, state: dict[str, Any]) -> None:
+        """Restore display state from saved state.
+
+        Args:
+            state: Dictionary containing display state from save_state()
+        """
+        self._initialized = state.get("initialized", False)
+        self._init_width = state.get("init_width", 80)
+        self._init_height = state.get("init_height", 24)
+
+        # Restore display dimensions if the display supports it
+        if hasattr(self._display, "width"):
+            self._display.width = state.get("width", 80)
+        if hasattr(self._display, "height"):
+            self._display.height = state.get("height", 24)
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"display.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        # Display needs rendered content and camera transformation
+        return {"render.output", "camera"}
+
+    @property
+    def inlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.TEXT_BUFFER}  # Display consumes rendered text
+
+    @property
+    def outlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.NONE}  # Display is a terminal stage (no output)
+
+    def init(self, ctx: PipelineContext) -> bool:
+        w = ctx.params.viewport_width if ctx.params else 80
+        h = ctx.params.viewport_height if ctx.params else 24
+
+        # Try to reuse display if already initialized
+        reuse = self._initialized
+        result = self._display.init(w, h, reuse=reuse)
+
+        # Update initialization state
+        if result is not False:
+            self._initialized = True
+            self._init_width = w
+            self._init_height = h
+
+        return result is not False
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Output data to display."""
+        if data is not None:
+            # Check if positioning mode is specified in context params
+            positioning = self._positioning
+            if ctx and ctx.params and hasattr(ctx.params, "positioning"):
+                positioning = ctx.params.positioning
+
+            # Pass positioning to display if supported
+            if (
+                hasattr(self._display, "show")
+                and "positioning" in self._display.show.__code__.co_varnames
+            ):
+                self._display.show(data, positioning=positioning)
+            else:
+                # Fallback for displays that don't support positioning parameter
+                self._display.show(data)
+        return data
+
+    def cleanup(self) -> None:
+        self._display.cleanup()

engine/pipeline/adapters/effect_plugin.py

@@ -0,0 +1,124 @@
+"""Adapter wrapping EffectPlugin as a Stage."""
+
+from typing import Any
+
+from engine.pipeline.core import PipelineContext, Stage
+
+
+class EffectPluginStage(Stage):
+    """Adapter wrapping EffectPlugin as a Stage.
+
+    Supports capability-based dependencies through the dependencies parameter.
+    """
+
+    def __init__(
+        self,
+        effect_plugin,
+        name: str = "effect",
+        dependencies: set[str] | None = None,
+    ):
+        self._effect = effect_plugin
+        self.name = name
+        self.category = "effect"
+        self.optional = False
+        self._dependencies = dependencies or set()
+
+    @property
+    def stage_type(self) -> str:
+        """Return stage_type based on effect name.
+
+        Overlay effects have stage_type "overlay".
+        """
+        if self.is_overlay:
+            return "overlay"
+        return self.category
+
+    @property
+    def render_order(self) -> int:
+        """Return render_order based on effect type.
+
+        Overlay effects have high render_order to appear on top.
+        """
+        if self.is_overlay:
+            return 100  # High order for overlays
+        return 0
+
+    @property
+    def is_overlay(self) -> bool:
+        """Return True for overlay effects.
+
+        Overlay effects compose on top of the buffer
+        rather than transforming it for the next stage.
+        """
+        # Check if the effect has an is_overlay attribute that is explicitly True
+        # (not just any truthy value from a mock object)
+        if hasattr(self._effect, "is_overlay"):
+            effect_overlay = self._effect.is_overlay
+            # Only return True if it's explicitly set to True
+            if effect_overlay is True:
+                return True
+        return self.name == "hud"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {f"effect.{self.name}"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return self._dependencies
+
+    @property
+    def inlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.TEXT_BUFFER}
+
+    @property
+    def outlet_types(self) -> set:
+        from engine.pipeline.core import DataType
+
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Process data through the effect."""
+        if data is None:
+            return None
+        from engine.effects.types import EffectContext, apply_param_bindings
+
+        w = ctx.params.viewport_width if ctx.params else 80
+        h = ctx.params.viewport_height if ctx.params else 24
+        frame = ctx.params.frame_number if ctx.params else 0
+
+        effect_ctx = EffectContext(
+            terminal_width=w,
+            terminal_height=h,
+            scroll_cam=0,
+            ticker_height=h,
+            camera_x=0,
+            mic_excess=0.0,
+            grad_offset=(frame * 0.01) % 1.0,
+            frame_number=frame,
+            has_message=False,
+            items=ctx.get("items", []),
+        )
+
+        # Copy sensor state from PipelineContext to EffectContext
+        for key, value in ctx.state.items():
+            if key.startswith("sensor."):
+                effect_ctx.set_state(key, value)
+
+        # Copy metrics from PipelineContext to EffectContext
+        if "metrics" in ctx.state:
+            effect_ctx.set_state("metrics", ctx.state["metrics"])
+
+        # Copy pipeline_order from PipelineContext services to EffectContext state
+        pipeline_order = ctx.get("pipeline_order")
+        if pipeline_order:
+            effect_ctx.set_state("pipeline_order", pipeline_order)
+
+        # Apply sensor param bindings if effect has them
+        if hasattr(self._effect, "param_bindings") and self._effect.param_bindings:
+            bound_config = apply_param_bindings(self._effect, effect_ctx)
+            self._effect.configure(bound_config)
+
+        return self._effect.process(data, effect_ctx)

engine/pipeline/adapters/factory.py

@@ -0,0 +1,38 @@
+"""Factory functions for creating stage instances."""
+
+from engine.pipeline.adapters.camera import CameraStage
+from engine.pipeline.adapters.data_source import DataSourceStage
+from engine.pipeline.adapters.display import DisplayStage
+from engine.pipeline.adapters.effect_plugin import EffectPluginStage
+from engine.pipeline.adapters.transform import FontStage
+
+
+def create_stage_from_display(display, name: str = "terminal") -> DisplayStage:
+    """Create a DisplayStage from a display instance."""
+    return DisplayStage(display, name=name)
+
+
+def create_stage_from_effect(effect_plugin, name: str) -> EffectPluginStage:
+    """Create an EffectPluginStage from an effect plugin."""
+    return EffectPluginStage(effect_plugin, name=name)
+
+
+def create_stage_from_source(data_source, name: str = "headlines") -> DataSourceStage:
+    """Create a DataSourceStage from a data source."""
+    return DataSourceStage(data_source, name=name)
+
+
+def create_stage_from_camera(camera, name: str = "vertical") -> CameraStage:
+    """Create a CameraStage from a camera instance."""
+    return CameraStage(camera, name=name)
+
+
+def create_stage_from_font(
+    font_path: str | None = None,
+    font_size: int | None = None,
+    font_ref: str | None = "default",
+    name: str = "font",
+) -> FontStage:
+    """Create a FontStage with specified font configuration."""
+    # FontStage currently doesn't use these parameters but keeps them for compatibility
+    return FontStage(name=name)

engine/pipeline/adapters/frame_capture.py

@@ -0,0 +1,165 @@
+"""
+Frame Capture Stage Adapter
+
+Wraps pipeline stages to capture frames for animation report generation.
+"""
+
+from typing import Any
+
+from engine.display.backends.animation_report import AnimationReportDisplay
+from engine.pipeline.core import PipelineContext, Stage
+
+
+class FrameCaptureStage(Stage):
+    """
+    Wrapper stage that captures frames before and after a wrapped stage.
+
+    This allows generating animation reports showing how each stage
+    transforms the data.
+    """
+
+    def __init__(
+        self,
+        wrapped_stage: Stage,
+        display: AnimationReportDisplay,
+        name: str | None = None,
+    ):
+        """
+        Initialize frame capture stage.
+
+        Args:
+            wrapped_stage: The stage to wrap and capture frames from
+            display: The animation report display to send frames to
+            name: Optional name for this capture stage
+        """
+        self._wrapped_stage = wrapped_stage
+        self._display = display
+        self.name = name or f"capture_{wrapped_stage.name}"
+        self.category = wrapped_stage.category
+        self.optional = wrapped_stage.optional
+
+        # Capture state
+        self._captured_input = False
+        self._captured_output = False
+
+    @property
+    def stage_type(self) -> str:
+        return self._wrapped_stage.stage_type
+
+    @property
+    def capabilities(self) -> set[str]:
+        return self._wrapped_stage.capabilities
+
+    @property
+    def dependencies(self) -> set[str]:
+        return self._wrapped_stage.dependencies
+
+    @property
+    def inlet_types(self) -> set:
+        return self._wrapped_stage.inlet_types
+
+    @property
+    def outlet_types(self) -> set:
+        return self._wrapped_stage.outlet_types
+
+    def init(self, ctx: PipelineContext) -> bool:
+        """Initialize the wrapped stage."""
+        return self._wrapped_stage.init(ctx)
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """
+        Process data through wrapped stage and capture frames.
+
+        Args:
+            data: Input data (typically a text buffer)
+            ctx: Pipeline context
+
+        Returns:
+            Output data from wrapped stage
+        """
+        # Capture input frame (before stage processing)
+        if isinstance(data, list) and all(isinstance(line, str) for line in data):
+            self._display.start_stage(f"{self._wrapped_stage.name}_input")
+            self._display.show(data)
+            self._captured_input = True
+
+        # Process through wrapped stage
+        result = self._wrapped_stage.process(data, ctx)
+
+        # Capture output frame (after stage processing)
+        if isinstance(result, list) and all(isinstance(line, str) for line in result):
+            self._display.start_stage(f"{self._wrapped_stage.name}_output")
+            self._display.show(result)
+            self._captured_output = True
+
+        return result
+
+    def cleanup(self) -> None:
+        """Cleanup the wrapped stage."""
+        self._wrapped_stage.cleanup()
+
+
+class FrameCaptureController:
+    """
+    Controller for managing frame capture across the pipeline.
+
+    This class provides an easy way to enable frame capture for
+    specific stages or the entire pipeline.
+    """
+
+    def __init__(self, display: AnimationReportDisplay):
+        """
+        Initialize frame capture controller.
+
+        Args:
+            display: The animation report display to use for capture
+        """
+        self._display = display
+        self._captured_stages: list[FrameCaptureStage] = []
+
+    def wrap_stage(self, stage: Stage, name: str | None = None) -> FrameCaptureStage:
+        """
+        Wrap a stage with frame capture.
+
+        Args:
+            stage: The stage to wrap
+            name: Optional name for the capture stage
+
+        Returns:
+            Wrapped stage that captures frames
+        """
+        capture_stage = FrameCaptureStage(stage, self._display, name)
+        self._captured_stages.append(capture_stage)
+        return capture_stage
+
+    def wrap_stages(self, stages: dict[str, Stage]) -> dict[str, Stage]:
+        """
+        Wrap multiple stages with frame capture.
+
+        Args:
+            stages: Dictionary of stage names to stages
+
+        Returns:
+            Dictionary of stage names to wrapped stages
+        """
+        wrapped = {}
+        for name, stage in stages.items():
+            wrapped[name] = self.wrap_stage(stage, name)
+        return wrapped
+
+    def get_captured_stages(self) -> list[FrameCaptureStage]:
+        """Get list of all captured stages."""
+        return self._captured_stages
+
+    def generate_report(self, title: str = "Pipeline Animation Report") -> str:
+        """
+        Generate the animation report.
+
+        Args:
+            title: Title for the report
+
+        Returns:
+            Path to the generated HTML file
+        """
+        report_path = self._display.generate_report(title)
+        return str(report_path)

engine/pipeline/adapters/message_overlay.py

@@ -0,0 +1,185 @@
+"""
+Message overlay stage - Renders ntfy messages as an overlay on the buffer.
+
+This stage provides message overlay capability for displaying ntfy.sh messages
+as a centered panel with pink/magenta gradient, matching upstream/main aesthetics.
+"""
+
+import re
+import time
+from dataclasses import dataclass
+from datetime import datetime
+
+from engine import config
+from engine.effects.legacy import vis_trunc
+from engine.pipeline.core import DataType, PipelineContext, Stage
+from engine.render.blocks import big_wrap
+from engine.render.gradient import msg_gradient
+
+
+@dataclass
+class MessageOverlayConfig:
+    """Configuration for MessageOverlayStage."""
+
+    enabled: bool = True
+    display_secs: int = 30  # How long to display messages
+    topic_url: str | None = None  # Ntfy topic URL (None = use config default)
+
+
+class MessageOverlayStage(Stage):
+    """Stage that renders ntfy message overlay on the buffer.
+
+    Provides:
+    - message.overlay capability (optional)
+    - Renders centered panel with pink/magenta gradient
+    - Shows title, body, timestamp, and remaining time
+    """
+
+    name = "message_overlay"
+    category = "overlay"
+
+    def __init__(
+        self, config: MessageOverlayConfig | None = None, name: str = "message_overlay"
+    ):
+        self.config = config or MessageOverlayConfig()
+        self._ntfy_poller = None
+        self._msg_cache = (None, None)  # (cache_key, rendered_rows)
+
+    @property
+    def capabilities(self) -> set[str]:
+        """Provides message overlay capability."""
+        return {"message.overlay"} if self.config.enabled else set()
+
+    @property
+    def dependencies(self) -> set[str]:
+        """Needs rendered buffer and camera transformation to overlay onto."""
+        return {"render.output", "camera"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def init(self, ctx: PipelineContext) -> bool:
+        """Initialize ntfy poller if topic URL is configured."""
+        if not self.config.enabled:
+            return True
+
+        # Get or create ntfy poller
+        topic_url = self.config.topic_url or config.NTFY_TOPIC
+        if topic_url:
+            from engine.ntfy import NtfyPoller
+
+            self._ntfy_poller = NtfyPoller(
+                topic_url=topic_url,
+                reconnect_delay=getattr(config, "NTFY_RECONNECT_DELAY", 5),
+                display_secs=self.config.display_secs,
+            )
+            self._ntfy_poller.start()
+            ctx.set("ntfy_poller", self._ntfy_poller)
+
+        return True
+
+    def process(self, data: list[str], ctx: PipelineContext) -> list[str]:
+        """Render message overlay on the buffer."""
+        if not self.config.enabled or not data:
+            return data
+
+        # Get active message from poller
+        msg = None
+        if self._ntfy_poller:
+            msg = self._ntfy_poller.get_active_message()
+
+        if msg is None:
+            return data
+
+        # Render overlay
+        w = ctx.terminal_width if hasattr(ctx, "terminal_width") else 80
+        h = ctx.terminal_height if hasattr(ctx, "terminal_height") else 24
+
+        overlay, self._msg_cache = self._render_message_overlay(
+            msg, w, h, self._msg_cache
+        )
+
+        # Composite overlay onto buffer
+        result = list(data)
+        for line in overlay:
+            # Overlay uses ANSI cursor positioning, just append
+            result.append(line)
+
+        return result
+
+    def _render_message_overlay(
+        self,
+        msg: tuple[str, str, float] | None,
+        w: int,
+        h: int,
+        msg_cache: tuple,
+    ) -> tuple[list[str], tuple]:
+        """Render ntfy message overlay.
+
+        Args:
+            msg: (title, body, timestamp) or None
+            w: terminal width
+            h: terminal height
+            msg_cache: (cache_key, rendered_rows) for caching
+
+        Returns:
+            (list of ANSI strings, updated cache)
+        """
+        overlay = []
+        if msg is None:
+            return overlay, msg_cache
+
+        m_title, m_body, m_ts = msg
+        display_text = m_body or m_title or "(empty)"
+        display_text = re.sub(r"\s+", " ", display_text.upper())
+
+        cache_key = (display_text, w)
+        if msg_cache[0] != cache_key:
+            msg_rows = big_wrap(display_text, w - 4)
+            msg_cache = (cache_key, msg_rows)
+        else:
+            msg_rows = msg_cache[1]
+
+        msg_rows = msg_gradient(msg_rows, (time.monotonic() * config.GRAD_SPEED) % 1.0)
+
+        elapsed_s = int(time.monotonic() - m_ts)
+        remaining = max(0, self.config.display_secs - elapsed_s)
+        ts_str = datetime.now().strftime("%H:%M:%S")
+        panel_h = len(msg_rows) + 2
+        panel_top = max(0, (h - panel_h) // 2)
+
+        row_idx = 0
+        for mr in msg_rows:
+            ln = vis_trunc(mr, w)
+            overlay.append(f"\033[{panel_top + row_idx + 1};1H  {ln}\033[0m\033[K")
+            row_idx += 1
+
+        meta_parts = []
+        if m_title and m_title != m_body:
+            meta_parts.append(m_title)
+        meta_parts.append(f"ntfy \u00b7 {ts_str} \u00b7 {remaining}s")
+        meta = (
+            "  " + " \u00b7 ".join(meta_parts)
+            if len(meta_parts) > 1
+            else "  " + meta_parts[0]
+        )
+        overlay.append(
+            f"\033[{panel_top + row_idx + 1};1H\033[38;5;245m{meta}\033[0m\033[K"
+        )
+        row_idx += 1
+
+        bar = "\u2500" * (w - 4)
+        overlay.append(
+            f"\033[{panel_top + row_idx + 1};1H  \033[2;38;5;37m{bar}\033[0m\033[K"
+        )
+
+        return overlay, msg_cache
+
+    def cleanup(self) -> None:
+        """Cleanup resources."""
+        pass

engine/pipeline/adapters/positioning.py

@@ -0,0 +1,185 @@
+"""PositionStage - Configurable positioning mode for terminal rendering.
+
+This module provides positioning stages that allow choosing between
+different ANSI positioning approaches:
+- ABSOLUTE: Use cursor positioning codes (\\033[row;colH) for all lines
+- RELATIVE: Use newlines for all lines
+- MIXED: Base content uses newlines, effects use cursor positioning (default)
+"""
+
+from enum import Enum
+from typing import Any
+
+from engine.pipeline.core import DataType, PipelineContext, Stage
+
+
+class PositioningMode(Enum):
+    """Positioning mode for terminal rendering."""
+
+    ABSOLUTE = "absolute"  # All lines have cursor positioning codes
+    RELATIVE = "relative"  # Lines use newlines (no cursor codes)
+    MIXED = "mixed"  # Mixed: newlines for base, cursor codes for overlays (default)
+
+
+class PositionStage(Stage):
+    """Applies positioning mode to buffer before display.
+
+    This stage allows configuring how lines are positioned in the terminal:
+    - ABSOLUTE: Each line has \\033[row;colH prefix (precise control)
+    - RELATIVE: Lines are joined with \\n (natural flow)
+    - MIXED: Leaves buffer as-is (effects add their own positioning)
+    """
+
+    def __init__(
+        self, mode: PositioningMode = PositioningMode.RELATIVE, name: str = "position"
+    ):
+        self.mode = mode
+        self.name = name
+        self.category = "position"
+        self._mode_str = mode.value
+
+    def save_state(self) -> dict[str, Any]:
+        """Save positioning mode for restoration."""
+        return {"mode": self.mode.value}
+
+    def restore_state(self, state: dict[str, Any]) -> None:
+        """Restore positioning mode from saved state."""
+        mode_value = state.get("mode", "relative")
+        self.mode = PositioningMode(mode_value)
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"position.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        # Position stage typically runs after render but before effects
+        # Effects may add their own positioning codes
+        return {"render.output"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def init(self, ctx: PipelineContext) -> bool:
+        """Initialize the positioning stage."""
+        return True
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Apply positioning mode to the buffer.
+
+        Args:
+            data: List of strings (buffer lines)
+            ctx: Pipeline context
+
+        Returns:
+            Buffer with applied positioning mode
+        """
+        if data is None:
+            return data
+
+        if not isinstance(data, list):
+            return data
+
+        if self.mode == PositioningMode.ABSOLUTE:
+            return self._to_absolute(data, ctx)
+        elif self.mode == PositioningMode.RELATIVE:
+            return self._to_relative(data, ctx)
+        else:  # MIXED
+            return data  # No transformation
+
+    def _to_absolute(self, data: list[str], ctx: PipelineContext) -> list[str]:
+        """Convert buffer to absolute positioning (all lines have cursor codes).
+
+        This mode prefixes each line with \\033[row;colH to move cursor
+        to the exact position before writing the line.
+
+        Args:
+            data: List of buffer lines
+            ctx: Pipeline context (provides terminal dimensions)
+
+        Returns:
+            Buffer with cursor positioning codes for each line
+        """
+        result = []
+        viewport_height = ctx.params.viewport_height if ctx.params else 24
+
+        for i, line in enumerate(data):
+            if i >= viewport_height:
+                break  # Don't exceed viewport
+
+            # Check if line already has cursor positioning
+            if "\033[" in line and "H" in line:
+                # Already has cursor positioning - leave as-is
+                result.append(line)
+            else:
+                # Add cursor positioning for this line
+                # Row is 1-indexed
+                result.append(f"\033[{i + 1};1H{line}")
+
+        return result
+
+    def _to_relative(self, data: list[str], ctx: PipelineContext) -> list[str]:
+        """Convert buffer to relative positioning (use newlines).
+
+        This mode removes explicit cursor positioning codes from lines
+        (except for effects that specifically add them).
+
+        Note: Effects like HUD add their own cursor positioning codes,
+        so we can't simply remove all of them. We rely on the terminal
+        display to join lines with newlines.
+
+        Args:
+            data: List of buffer lines
+            ctx: Pipeline context (unused)
+
+        Returns:
+            Buffer with minimal cursor positioning (only for overlays)
+        """
+        # For relative mode, we leave the buffer as-is
+        # The terminal display handles joining with newlines
+        # Effects that need absolute positioning will add their own codes
+
+        # Filter out lines that would cause double-positioning
+        result = []
+        for i, line in enumerate(data):
+            # Check if this line looks like base content (no cursor code at start)
+            # vs an effect line (has cursor code at start)
+            if line.startswith("\033[") and "H" in line[:20]:
+                # This is an effect with positioning - keep it
+                result.append(line)
+            else:
+                # Base content - strip any inline cursor codes (rare)
+                # but keep color codes
+                result.append(line)
+
+        return result
+
+    def cleanup(self) -> None:
+        """Clean up positioning stage."""
+        pass
+
+
+# Convenience function to create positioning stage
+def create_position_stage(
+    mode: str = "relative", name: str = "position"
+) -> PositionStage:
+    """Create a positioning stage with the specified mode.
+
+    Args:
+        mode: Positioning mode ("absolute", "relative", or "mixed")
+        name: Name for the stage
+
+    Returns:
+        PositionStage instance
+    """
+    try:
+        positioning_mode = PositioningMode(mode)
+    except ValueError:
+        positioning_mode = PositioningMode.RELATIVE
+
+    return PositionStage(mode=positioning_mode, name=name)

engine/pipeline/adapters/transform.py

@@ -0,0 +1,293 @@
+"""Adapters for transform stages (viewport, font, image, canvas)."""
+
+from typing import Any
+
+import engine.render
+from engine.data_sources import SourceItem
+from engine.pipeline.core import DataType, PipelineContext, Stage
+
+
+def estimate_simple_height(text: str, width: int) -> int:
+    """Estimate height in terminal rows using simple word wrap.
+
+    Uses conservative estimation suitable for headlines.
+    Each wrapped line is approximately 6 terminal rows (big block rendering).
+    """
+    words = text.split()
+    if not words:
+        return 6
+
+    lines = 1
+    current_len = 0
+    for word in words:
+        word_len = len(word)
+        if current_len + word_len + 1 > width - 4:  # -4 for margins
+            lines += 1
+            current_len = word_len
+        else:
+            current_len += word_len + 1
+
+    return lines * 6  # 6 rows per line for big block rendering
+
+
+class ViewportFilterStage(Stage):
+    """Filter items to viewport height based on rendered height."""
+
+    def __init__(self, name: str = "viewport-filter"):
+        self.name = name
+        self.category = "render"
+        self.optional = True
+        self._layout: list[int] = []
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"source.filtered"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        # Always requires camera.state for viewport filtering
+        # CameraUpdateStage provides this (auto-injected if missing)
+        return {"source", "camera.state"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Filter items to viewport height based on rendered height."""
+        if data is None:
+            return data
+
+        if not isinstance(data, list):
+            return data
+
+        if not data:
+            return []
+
+        # Get viewport parameters from context
+        viewport_height = ctx.params.viewport_height if ctx.params else 24
+        viewport_width = ctx.params.viewport_width if ctx.params else 80
+        camera_y = ctx.get("camera_y", 0)
+
+        # Estimate height for each item and cache layout
+        self._layout = []
+        cumulative_heights = []
+        current_height = 0
+
+        for item in data:
+            title = item.content if isinstance(item, SourceItem) else str(item)
+            # Use simple height estimation (not PIL-based)
+            estimated_height = estimate_simple_height(title, viewport_width)
+            self._layout.append(estimated_height)
+            current_height += estimated_height
+            cumulative_heights.append(current_height)
+
+        # Find visible range based on camera_y and viewport_height
+        # camera_y is the scroll offset (how many rows are scrolled up)
+        start_y = camera_y
+        end_y = camera_y + viewport_height
+
+        # Find start index (first item that intersects with visible range)
+        start_idx = 0
+        start_item_y = 0  # Y position where the first visible item starts
+        for i, total_h in enumerate(cumulative_heights):
+            if total_h > start_y:
+                start_idx = i
+                # Calculate the Y position of the start of this item
+                if i > 0:
+                    start_item_y = cumulative_heights[i - 1]
+                break
+
+        # Find end index (first item that extends beyond visible range)
+        end_idx = len(data)
+        for i, total_h in enumerate(cumulative_heights):
+            if total_h >= end_y:
+                end_idx = i + 1
+                break
+
+        # Adjust camera_y for the filtered buffer
+        # The filtered buffer starts at row 0, but the camera position
+        # needs to be relative to where the first visible item starts
+        filtered_camera_y = camera_y - start_item_y
+
+        # Update context with the filtered camera position
+        # This ensures CameraStage can correctly slice the filtered buffer
+        ctx.set_state("camera_y", filtered_camera_y)
+        ctx.set_state("camera_x", ctx.get("camera_x", 0))  # Keep camera_x unchanged
+
+        # Return visible items
+        return data[start_idx:end_idx]
+
+
+class FontStage(Stage):
+    """Render items using font."""
+
+    def __init__(self, name: str = "font"):
+        self.name = name
+        self.category = "render"
+        self.optional = False
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"render.output"}
+
+    @property
+    def stage_dependencies(self) -> set[str]:
+        # Must connect to viewport_filter stage to get filtered source
+        return {"viewport_filter"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        # Depend on source.filtered (provided by viewport_filter)
+        # This ensures we get the filtered/processed source, not raw source
+        return {"source.filtered"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Render items to text buffer using font."""
+        if data is None:
+            return []
+
+        if not isinstance(data, list):
+            return [str(data)]
+
+        import os
+
+        if os.environ.get("DEBUG_CAMERA"):
+            print(f"FontStage: input items={len(data)}")
+
+        viewport_width = ctx.params.viewport_width if ctx.params else 80
+
+        result = []
+        for item in data:
+            if isinstance(item, SourceItem):
+                title = item.content
+                src = item.source
+                ts = item.timestamp
+                content_lines, _, _ = engine.render.make_block(
+                    title, src, ts, viewport_width
+                )
+                result.extend(content_lines)
+            elif hasattr(item, "content"):
+                title = str(item.content)
+                content_lines, _, _ = engine.render.make_block(
+                    title, "", "", viewport_width
+                )
+                result.extend(content_lines)
+            else:
+                result.append(str(item))
+        return result
+
+
+class ImageToTextStage(Stage):
+    """Convert image items to text."""
+
+    def __init__(self, name: str = "image-to-text"):
+        self.name = name
+        self.category = "render"
+        self.optional = True
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"render.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return {"source"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Convert image items to text representation."""
+        if data is None:
+            return []
+
+        if not isinstance(data, list):
+            return [str(data)]
+
+        result = []
+        for item in data:
+            # Check if item is an image
+            if hasattr(item, "image_path") or hasattr(item, "image_data"):
+                # Placeholder: would normally render image to ASCII art
+                result.append(f"[Image: {getattr(item, 'image_path', 'data')}]")
+            elif isinstance(item, SourceItem):
+                result.extend(item.content.split("\n"))
+            else:
+                result.append(str(item))
+        return result
+
+
+class CanvasStage(Stage):
+    """Render items to canvas."""
+
+    def __init__(self, name: str = "canvas"):
+        self.name = name
+        self.category = "render"
+        self.optional = False
+
+    @property
+    def stage_type(self) -> str:
+        return "render"
+
+    @property
+    def capabilities(self) -> set[str]:
+        return {"render.output"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        return {"source"}
+
+    @property
+    def inlet_types(self) -> set:
+        return {DataType.SOURCE_ITEMS}
+
+    @property
+    def outlet_types(self) -> set:
+        return {DataType.TEXT_BUFFER}
+
+    def process(self, data: Any, ctx: PipelineContext) -> Any:
+        """Render items to canvas."""
+        if data is None:
+            return []
+
+        if not isinstance(data, list):
+            return [str(data)]
+
+        # Simple canvas rendering
+        result = []
+        for item in data:
+            if isinstance(item, SourceItem):
+                result.extend(item.content.split("\n"))
+            else:
+                result.append(str(item))
+        return result

engine/pipeline/core.py

@@ -0,0 +1,321 @@
+"""
+Pipeline core - Unified Stage abstraction and PipelineContext.
+
+This module provides the foundation for a clean, dependency-managed pipeline:
+- Stage: Base class for all pipeline components (sources, effects, displays, cameras)
+- PipelineContext: Dependency injection context for runtime data exchange
+- Capability system: Explicit capability declarations with duck-typing support
+- DataType: PureData-style inlet/outlet typing for validation
+"""
+
+from abc import ABC, abstractmethod
+from collections.abc import Callable
+from dataclasses import dataclass, field
+from enum import Enum, auto
+from typing import TYPE_CHECKING, Any
+
+if TYPE_CHECKING:
+    from engine.pipeline.params import PipelineParams
+
+
+class DataType(Enum):
+    """PureData-style data types for inlet/outlet validation.
+
+    Each type represents a specific data format that flows through the pipeline.
+    This enables compile-time-like validation of connections.
+
+    Examples:
+        SOURCE_ITEMS: List[SourceItem] - raw items from sources
+        ITEM_TUPLES: List[tuple] - (title, source, timestamp) tuples
+        TEXT_BUFFER: List[str] - rendered ANSI buffer for display
+        RAW_TEXT: str - raw text strings
+        PIL_IMAGE: PIL Image object
+    """
+
+    SOURCE_ITEMS = auto()  # List[SourceItem] - from DataSource
+    ITEM_TUPLES = auto()  # List[tuple] - (title, source, ts)
+    TEXT_BUFFER = auto()  # List[str] - ANSI buffer
+    RAW_TEXT = auto()  # str - raw text
+    PIL_IMAGE = auto()  # PIL Image object
+    ANY = auto()  # Accepts any type
+    NONE = auto()  # No data (terminator)
+
+
+@dataclass
+class StageConfig:
+    """Configuration for a single stage."""
+
+    name: str
+    category: str
+    enabled: bool = True
+    optional: bool = False
+    params: dict[str, Any] = field(default_factory=dict)
+
+
+class Stage(ABC):
+    """Abstract base class for all pipeline stages.
+
+    A Stage is a single component in the rendering pipeline. Stages can be:
+    - Sources: Data providers (headlines, poetry, pipeline viz)
+    - Effects: Post-processors (noise, fade, glitch, hud)
+    - Displays: Output backends (terminal, pygame, websocket)
+    - Cameras: Viewport controllers (vertical, horizontal, omni)
+    - Overlays: UI elements that compose on top (HUD)
+
+    Stages declare:
+    - capabilities: What they provide to other stages
+    - dependencies: What they need from other stages
+    - stage_type: Category of stage (source, effect, overlay, display)
+    - render_order: Execution order within category
+    - is_overlay: If True, output is composited on top, not passed downstream
+
+    Duck-typing is supported: any class with the required methods can act as a Stage.
+    """
+
+    name: str
+    category: str  # "source", "effect", "overlay", "display", "camera"
+    optional: bool = False  # If True, pipeline continues even if stage fails
+
+    @property
+    def stage_type(self) -> str:
+        """Category of stage for ordering.
+
+        Valid values: "source", "effect", "overlay", "display", "camera"
+        Defaults to category for backwards compatibility.
+        """
+        return self.category
+
+    @property
+    def render_order(self) -> int:
+        """Execution order within stage_type group.
+
+        Higher values execute later. Useful for ordering overlays
+        or effects that need specific execution order.
+        """
+        return 0
+
+    @property
+    def is_overlay(self) -> bool:
+        """If True, this stage's output is composited on top of the buffer.
+
+        Overlay stages don't pass their output to the next stage.
+        Instead, their output is layered on top of the final buffer.
+        Use this for HUD, status displays, and similar UI elements.
+        """
+        return False
+
+    @property
+    def inlet_types(self) -> set[DataType]:
+        """Return set of data types this stage accepts.
+
+        PureData-style inlet typing. If the connected upstream stage's
+        outlet_type is not in this set, the pipeline will raise an error.
+
+        Examples:
+            - Source stages: {DataType.NONE} (no input needed)
+            - Transform stages: {DataType.ITEM_TUPLES, DataType.TEXT_BUFFER}
+            - Display stages: {DataType.TEXT_BUFFER}
+        """
+        return {DataType.ANY}
+
+    @property
+    def outlet_types(self) -> set[DataType]:
+        """Return set of data types this stage produces.
+
+        PureData-style outlet typing. Downstream stages must accept
+        this type in their inlet_types.
+
+        Examples:
+            - Source stages: {DataType.SOURCE_ITEMS}
+            - Transform stages: {DataType.TEXT_BUFFER}
+            - Display stages: {DataType.NONE} (consumes data)
+        """
+        return {DataType.ANY}
+
+    @property
+    def capabilities(self) -> set[str]:
+        """Return set of capabilities this stage provides.
+
+        Examples:
+            - "source.headlines"
+            - "effect.noise"
+            - "display.output"
+            - "camera"
+        """
+        return {f"{self.category}.{self.name}"}
+
+    @property
+    def dependencies(self) -> set[str]:
+        """Return set of capability names this stage needs.
+
+        Examples:
+            - {"display.output"}
+            - {"source.headlines"}
+            - {"camera"}
+        """
+        return set()
+
+    @property
+    def stage_dependencies(self) -> set[str]:
+        """Return set of stage names this stage must connect to directly.
+
+        This allows explicit stage-to-stage dependencies, useful for enforcing
+        pipeline structure when capability matching alone is insufficient.
+
+        Examples:
+            - {"viewport_filter"}  # Must connect to viewport_filter stage
+            - {"camera_update"}    # Must connect to camera_update stage
+
+        NOTE: These are stage names (as added to pipeline), not capabilities.
+        """
+        return set()
+
+    def init(self, ctx: "PipelineContext") -> bool:
+        """Initialize stage with pipeline context.
+
+        Args:
+            ctx: PipelineContext for accessing services
+
+        Returns:
+            True if initialization succeeded, False otherwise
+        """
+        return True
+
+    @abstractmethod
+    def process(self, data: Any, ctx: "PipelineContext") -> Any:
+        """Process input data and return output.
+
+        Args:
+            data: Input data from previous stage (or initial data for first stage)
+            ctx: PipelineContext for accessing services and state
+
+        Returns:
+            Processed data for next stage
+        """
+        ...
+
+    def cleanup(self) -> None:  # noqa: B027
+        """Clean up resources when pipeline shuts down."""
+        pass
+
+    def get_config(self) -> StageConfig:
+        """Return current configuration of this stage."""
+        return StageConfig(
+            name=self.name,
+            category=self.category,
+            optional=self.optional,
+        )
+
+    def set_enabled(self, enabled: bool) -> None:
+        """Enable or disable this stage."""
+        self._enabled = enabled  # type: ignore[attr-defined]
+
+    def is_enabled(self) -> bool:
+        """Check if stage is enabled."""
+        return getattr(self, "_enabled", True)
+
+
+@dataclass
+class StageResult:
+    """Result of stage processing, including success/failure info."""
+
+    success: bool
+    data: Any
+    error: str | None = None
+    stage_name: str = ""
+
+
+class PipelineContext:
+    """Dependency injection context passed through the pipeline.
+
+    Provides:
+    - services: Named services (display, config, event_bus, etc.)
+    - state: Runtime state shared between stages
+    - params: PipelineParams for animation-driven config
+
+    Services can be injected at construction time or lazily resolved.
+    """
+
+    def __init__(
+        self,
+        services: dict[str, Any] | None = None,
+        initial_state: dict[str, Any] | None = None,
+    ):
+        self.services: dict[str, Any] = services or {}
+        self.state: dict[str, Any] = initial_state or {}
+        self._params: PipelineParams | None = None
+
+        # Lazy resolvers for common services
+        self._lazy_resolvers: dict[str, Callable[[], Any]] = {
+            "config": self._resolve_config,
+            "event_bus": self._resolve_event_bus,
+        }
+
+    def _resolve_config(self) -> Any:
+        from engine.config import get_config
+
+        return get_config()
+
+    def _resolve_event_bus(self) -> Any:
+        from engine.eventbus import get_event_bus
+
+        return get_event_bus()
+
+    def get(self, key: str, default: Any = None) -> Any:
+        """Get a service or state value by key.
+
+        First checks services, then state, then lazy resolution.
+        """
+        if key in self.services:
+            return self.services[key]
+        if key in self.state:
+            return self.state[key]
+        if key in self._lazy_resolvers:
+            try:
+                return self._lazy_resolvers[key]()
+            except Exception:
+                return default
+        return default
+
+    def set(self, key: str, value: Any) -> None:
+        """Set a service or state value."""
+        self.services[key] = value
+
+    def set_state(self, key: str, value: Any) -> None:
+        """Set a runtime state value."""
+        self.state[key] = value
+
+    def get_state(self, key: str, default: Any = None) -> Any:
+        """Get a runtime state value."""
+        return self.state.get(key, default)
+
+    @property
+    def params(self) -> "PipelineParams | None":
+        """Get current pipeline params (for animation)."""
+        return self._params
+
+    @params.setter
+    def params(self, value: "PipelineParams") -> None:
+        """Set pipeline params (from animation controller)."""
+        self._params = value
+
+    def has_capability(self, capability: str) -> bool:
+        """Check if a capability is available."""
+        return capability in self.services or capability in self._lazy_resolvers
+
+
+class StageError(Exception):
+    """Raised when a stage fails to process."""
+
+    def __init__(self, stage_name: str, message: str, is_optional: bool = False):
+        self.stage_name = stage_name
+        self.message = message
+        self.is_optional = is_optional
+        super().__init__(f"Stage '{stage_name}' failed: {message}")
+
+
+def create_stage_error(
+    stage_name: str, error: Exception, is_optional: bool = False
+) -> StageError:
+    """Helper to create a StageError from an exception."""
+    return StageError(stage_name, str(error), is_optional)