Support voice input

Using an in-browser whisper implementation such as https://www.ratchet.sh/whisper-turbo

The issue suggests integrating an in-browser Whisper implementation like ratchet.sh/whisper-turbo for voice input. Start by examining the project's frontend codebase to understand how user input is currently handled. Look for audio capture and processing hooks, and assess how to integrate the Whisper model into the existing TypeScript/JavaScript flow. 'Done' means voice input is captured, transcribed, and fed into the LLM chat interface without server calls.

Vom Indexierungsmodell aus dem Issue-Text verfasst.