How jrsmdev/violet works
Yuki is an AI VTuber companion for live streaming: it accepts chat, voice, and game context, chooses persona-driven reactions, generates dialogue and speech, and presents the result through an avatar and OBS widgets. The graph highlights chat/voice and gameplay as distinct inputs, with mind, memory, and generation control as internal decision layers. The README documents vision and the broader overlay/game capabilities; source evidence is selective, so unsampled implementation links are omitted rather than inferred.
Main entry points: stream.py Stream API · main.py FastAPI Routes · llm_service.py LLM Service · mind_service.py Mind & Reactions
Components 21
User Interfaces
- Web UI
frontend/src/App.tsx - Director Panel
frontend/src/Director.tsx - OBS Widgets
frontend/src/MasterOverlay.tsx
Conversation Runtime
- FastAPI Routes
backend/main.py - Voice Endpoint
backend/api/voice.py - Dialogue Pipeline
backend/pipeline/orchestrator.py - Mind & Reactions
backend/services/mind_service.py - LLM Service
backend/services/llm_service.py - Generation Control
backend/services/generation_controller.py
Streaming & Games
- Stream API
backend/api/stream.py - Game Bridge
backend/agent/game_bridge.py - osu! Reactions
backend/services/osu_service.py - Studio Services
backend/services/studio_service.py - Vision
backend/services/vision_service.py
State & Speech
- Speech Synthesis
backend/services/tts_service.py - Conversation Memory
backend/services/memory_service.py - Event Bus
backend/core/event_bus.py
Other components
- Streamer
- Stream Audience
- Connected Game
- Ollama LLM
Connections 21
- Streamer → Web UI: uses
README.md - Stream Audience → Stream API: sends chat
README.md - Connected Game → Game Bridge: streams context
README.md - Web UI → Voice Endpoint: uploads audio
frontend/src/App.tsx - FastAPI Routes → Voice Endpoint: registers routes
backend/main.py - FastAPI Routes → Stream API: registers routes
backend/main.py - FastAPI Routes → Game Bridge: registers socket
- Dialogue Pipeline → LLM Service: calls
backend/pipeline/orchestrator.py - Dialogue Pipeline → Speech Synthesis: calls
backend/pipeline/orchestrator.py - Dialogue Pipeline → Event Bus: publishes events
backend/pipeline/orchestrator.py - LLM Service → Ollama LLM: requests generation
- LLM Service → Conversation Memory: reads and writes
backend/services/llm_service.py - LLM Service → Event Bus: publishes thoughts
backend/services/llm_service.py - Mind & Reactions → Conversation Memory: uses memory
backend/services/mind_service.py - Mind & Reactions → Generation Control: uses controller
backend/services/mind_service.py - Mind & Reactions → Stream API: uses stream service
backend/services/mind_service.py - osu! Reactions → Stream API: broadcasts reactions
backend/services/osu_service.py - FastAPI Routes → Director Panel: serves panel APIs
- Director Panel → Studio Services: controls studio
- Stream API → OBS Widgets: feeds overlays
README.md - Vision → Mind & Reactions: informs awareness