How emmdrole/tars works
Project type and purpose
• This is a Python-based real-time audio processing application (named “realtime_robot”) packaged as a CLI/service.
• Purpose: capture live audio, send it to OpenAI via WebSocket, receive text or commands, then synthesize voice using ElevenLabs TTS and potentially route output back to the user (or downstream systems).
File structure
• Top level:
• pyproject.toml & .venv → Python project with isolated environment and packaging info
• src/main.py → application entry point (orchestrator)
• src/audio/audio_input.py → audio capture module
• src/ws/openai_ws.py → WebSocket client to OpenAI
• src/voice/elevenlabs_ws.py → WebSocket client to ElevenLabs TTS
• debug_audio.raw → sample or test audio data
README for additional insights
• README.md is a placeholder (“tars… coming soon”), so no formal architecture docs. All inference comes from code layout and file names.