Feature: Add FunASR/SenseVoice as local voice AI engine

[!NOTE]
License and capability clarification (2026-07-14): FunASR is a toolkit, not a single checkpoint. The FunASR and SenseVoice repository source code is MIT; model weights follow each model card. SenseVoiceSmall supports Chinese, Cantonese, English, Japanese, and Korean, and its weights use the linked FunASR Model Open Source License Agreement. Fun-ASR-Nano-2512 is Apache-2.0. Language coverage, punctuation, and performance depend on the selected model and runtime configuration.

Hi! Great on-device voice AI + RAG for Mac!

Would you consider FunASR or SenseVoice as a local STT engine?

Why?

• Ultra-fast — SenseVoice-Small: ~70ms for 10s audio
• Built-in punctuation — Clean text output for RAG input
• Streaming — Real-time voice processing
• Fully local — No cloud required, perfect for on-device
• Apple Silicon optimized — Runs well on M-series Macs
• 50+ languages, Apache 2.0

References

• FunASR: https://github.com/modelscope/FunASR (16K+ stars)
• SenseVoice: https://github.com/FunAudioLLM/SenseVoice (8K+ stars)

No files, tests, or entry points are named. Read the linked FunASR and SenseVoice references first, then define which engine, model, languages, and runtime are in scope. Done should be an agreed local STT integration with documented model licensing, language coverage, and performance expectations.