# Agent Guide for src/

## Context

Contains the main Python package `reachy_mini_conversation_app`. This is the core application code for Mini Minder.

## Structure & Navigation

```
src/reachy_mini_conversation_app/
├── console.py              # FastAPI server (API endpoints + audio pipeline)
├── main.py                 # ReachyMiniApp entry point (robot mode)
├── stream_api.py           # WebSocket streaming + REST endpoints → frontend bridge
├── openai_realtime.py      # OpenAI Realtime API handler
├── realtime_handler.py     # Base handler class
├── handler_factory.py      # Creates OpenAI handler
├── database.py             # SQLite CRUD operations
├── config.py               # Application configuration
├── schemas.py              # Pydantic models / data schemas
├── utils.py                # Shared utilities
├── entry_state.py          # Health entry state machine
├── prompts.py              # Prompt builder/combiner
├── session_enrichment.py   # Neo4j graph enrichment at session start
├── session_summary.py      # Session summary generation
├── appointment_export.py   # Clinical report export
├── email_service.py        # Email delivery service
├── memory_graph.py         # Neo4j knowledge graph operations
├── memory_notes.py         # Session memory notes
├── graph_query_engine.py   # Neo4j query engine
├── pattern_detector.py     # Health pattern detection
├── missed_dose_detection.py # Missed dose tracking
├── pii_guard.py            # PII redaction
├── checkin_scheduler.py    # Proactive check-in scheduling
├── computer_audio.py       # Host audio routing (music playback)
├── wakeword_detector.py    # Wakeword detection
├── camera_worker.py        # Camera frame capture
├── moves.py                # Robot movement definitions
├── dance_emotion_moves.py  # Dance & emotion sequences
├── headless_personality.py # Headless mode personality
├── headless_personality_ui.py # Headless personality UI
├── robot_control_api.py    # Robot control REST API router
├── tools/                  # LLM tools (see tools/AGENT.md)
├── prompts/                # System prompts (see prompts/AGENT.md)
├── profiles/               # Conversation profiles (see profiles/AGENT.md)
│   └── _reachy_mini_minder_locked_profile/  # Clinical tools + GenUI emitters
├── langgraph_agent/        # LangGraph agent integration
├── audio/                  # Audio utilities
├── vision/                 # Vision processing
├── models/                 # Data models
├── static/                 # Static assets
├── templates/              # HTML templates
└── images/                 # Image assets
```

### Key Entry Points

- **API server**: `console.py` → FastAPI on port 7860 (REST + WebSocket)
- **Robot app**: `main.py` → `MiniMinderApp(ReachyMiniApp)`
- **Console**: `console.py` → standalone testing (no robot required)
- **Handler factory**: `handler_factory.py` → creates OpenAI handler

## Development Workflow

```bash
# Run tests for this package
pytest tests/ -v

# Run console mode
python src/reachy_mini_conversation_app/console.py
```

## Dependencies & Connections

- **Internal**: `tools/`, `prompts/`, `profiles/`, `stream_api.py`
- **External packages**: `openai`, `reachy-mini`, `fastapi`, `neo4j`
- **Consumers**:
  - `frontend/` (Next.js) — consumes REST endpoints + WebSocket from `stream_api.py`
  - Reachy Mini daemon — consumes robot app via `main.py`
  - CLI — `reachy-mini-minder` entry point
- **Bridge to frontend**: `stream_api.py` defines all `emit_*()` functions and REST endpoints.
  See `frontend/AGENT.md` for the full endpoint↔hook mapping.

## Configuration

Database stored at project root: `mini_minder.db`.