Add pluggable ASR provider interface (Qwen3-ASR / FireRedASR support)

[yxs]

I’m exploring using Handy for Chinese-English mixed input, where models like Qwen3-ASR and FireRedASR significantly outperform Whisper in accuracy and code-switching scenarios.

Currently, integrating these models is difficult because they rely on different runtime stacks.

Any ideas?

[cjpais]

They will be integrated in the next month or two! I've been working on a new backend

I also want better Chinese support so it's coming!

[yxs]

Really excited to see this, would be a big step forward for Handy.
Handy seems to be building and maintaining its own inference/runtime layer, rather than relying on a platform-specific runtime like MLX for Apple.
That feels quite different from projects like Voxt, which lean on MLX/WhisperKit and mainly act as a frontend/model orchestrator.

[Gin-F]

That's great! I really love this software. It’s well-designed, lightweight, and fast. If you could add ASR support to improve Chinese language capabilities, it would be absolutely perfect.😀 Would it be possible to include FunASR and Paraformer?

[cjpais]

FunASR will be supported in 0.9.0 which is coming out later today