Some thoughts arrive faster than you can type. Some arrive while your hands are busy. The built-in chat has a microphone button so you can hand them in by voice.
Tap, talk, send
Tap the mic in the chat composer, say your thing, tap again to stop. The audio uploads in the background and Whisper transcribes it server-side. The transcript appears in the input field — read it, edit it, send it. Voice is the input medium; the wiki still sees clean text.
99 languages, code-switching aware
Whisper-large-v3-turbo handles 99 languages without you picking one. Switch mid-sentence, drop in a German technical term while speaking English, mumble a brand name — the transcript handles it. Daneshbase doesn't add a language picker because Whisper doesn't need one.
Re-record without consequence
Said the wrong thing? Tap the mic again and the new recording silently replaces the previous transcript. There's no broken thread to apologise for, no half-finished sentence in the chat history.
Hands-busy capture
The mic isn't a novelty button — it's how you capture thoughts when you can't type. Walking, cooking, on the phone, driving (responsibly). Speak the note, the agent files it in the right doc. Read the result later when you have a screen.
Audio isn't kept
The recording uploads, gets transcribed, the transcript becomes part of the conversation. The audio itself isn't stored — only the text ends up in your wiki.