Feature Request: Add FunASR/SenseVoice as STT backend

[LauraGPT]

Feature Request

Dograh as an open-source voice AI platform could benefit from FunASR/SenseVoice as an STT backend.

Why

• OpenAI-compatible API — funasr-server provides /v1/audio/transcriptions endpoint
• Real-time WebSocket streaming — built-in streaming server at ws://localhost:10095
• Self-hosted — fully offline, no external APIs
• 50+ languages with automatic language detection
• 170x real-time on GPU — non-autoregressive, very fast
• Speaker diarization — built-in cam++ model for multi-speaker scenarios
• Emotion detection — SenseVoice classifies speech emotion

Quick Setup

pip install funasr
funasr-server --device cuda

Streaming Mode

python -m funasr.bin.ws_server --device cuda
# WebSocket endpoint: ws://localhost:10095

Happy to help with integration!