How nbllt666/cx-o works
CX-O is a desktop AI-companion system: users interact through a browser/Electron pet, which connects to a FastAPI backend for conversational and memory services; the frontend presents responses through voice and Live2D/VRM avatars. Major branches add streaming, meetings, screen perception, autonomy, and distributed failover. Separate workstations provide music/voice generation and model training, while CXO-Tuner supports optional self-evolution. Relationships for unsampled subsystems are included only where the README documents them; paths identify their enclosing source directories.
Main entry points: src Pet and management UI · api HTTP and WebSocket API · chat Conversation and tools · services Voice workstation
Components 23
Desktop experience
- Pet and management UI
APP-Frontend/src - Electron host
APP-Frontend/electron/main.ts - Live2D and VRM
APP-Frontend/src/avatar - Screen perception
APP-Frontend/src/hooks/useVisionPipeline.ts - Meeting experience
APP-Frontend/src/pages/pet/MeetingRoomView.tsx
Companion backend
- HTTP and WebSocket API
CX-O-SERVER/server/api - Conversation and tools
CX-O-SERVER/server/core/chat - Memory and retrieval
CX-O-SERVER/server/core/memory - Speech recognition
CX-O-SERVER/asr_container/streaming_engine.py - Vision processing
CX-O-SERVER/server/core/vision - Autonomous life
CX-O-SERVER/server/autonomy/core/loop/autonomy_engine.py - Dream engine
CX-O-SERVER/server/autonomy/dream/engine.py
Voice and music
- Voice workstation
CX-O-VoiceWorkStation/workstation/services - Song generation
CX-O-VoiceWorkStation/workstation/services/song_pipeline.py - Duet conversion
CX-O-VoiceWorkStation/workstation/services/duet_pipeline.py - Model training
CXO-ModelStation/modelstation/main.py
Management and evolution
- Evolution training
CXO-Tuner/tuner/api/routes.py - CX-A management
CX-O-SERVER/server/core/admin
Integrations and resilience
- Sentinel cluster
CX-O-SERVER/server/core/cluster
Other components
- User
- OBS
- Ollama and LLMs
- ASR and TTS services
Connections 23
- User → Pet and management UI: interacts
README.md - Pet and management UI → HTTP and WebSocket API: sends chat
README.md - HTTP and WebSocket API → Conversation and tools: routes requests
- Conversation and tools → Memory and retrieval: retrieves memories
README.md - Conversation and tools → Ollama and LLMs: requests generation
README.md - HTTP and WebSocket API → Speech recognition: uses speech recognition
README.md - Pet and management UI → ASR and TTS services: uses speech services
README.md - Pet and management UI → Live2D and VRM: renders character
README.md - Pet and management UI → Screen perception: uploads captured clips
APP-Frontend/src/hooks/useVisionPipeline.ts - Screen perception → HTTP and WebSocket API: uploads perception data
- HTTP and WebSocket API → Vision processing: processes screen input
README.md - Pet and management UI → Meeting experience: starts meetings
README.md - Meeting experience → HTTP and WebSocket API: connects to meeting service
README.md - HTTP and WebSocket API → Autonomous life: runs autonomous cycles
README.md - Autonomous life → Dream engine: starts dream sessions
- Pet and management UI → Voice workstation: requests voice work
README.md - Voice workstation → Song generation: generates songs
CX-O-VoiceWorkStation/workstation/services/song_pipeline.py - Voice workstation → Duet conversion: converts duet vocals
CX-O-VoiceWorkStation/workstation/services/duet_pipeline.py - Pet and management UI → Model training: manages voice models
README.md - Pet and management UI → Evolution training: submits evolution feedback
README.md - CX-A management → Sentinel cluster: orchestrates nodes
README.md - HTTP and WebSocket API → Sentinel cluster: replicates state
README.md - Pet and management UI → OBS: provides streaming sources
README.md