[Feature Req] Voice input support (via Amazon Transcribe?)

[athewsey]

I have a demo use-case with a touchscreen but no keyboard, for which it'd be useful to support audio input.

I had a working prototype of this some time ago, but would need to clean up the code before being able to propose here directly. In general, it was along the lines of:

• Use the amazon-transcribe Python SDK
• Set up the transcription client in Chainlit's @cl.on_audio_start hook
• send_audio_event in Chainlit's @cl.on_audio_chunk hook
• handle_events in Chainlit's @cl.on_audio_end hook (I couldn't get this working in on_audio_chunk at the time on CL v2.6.2, because both the Chainlit hook and the Transcribe handler were async generators so didn't like being nested).