Configure voice input

Voice input (speech-to-text) is instance-wide: one key enables dictation for every workspace and every user. It uses Groq’s Whisper transcription (note: Groq, not Grok).

Provide the key

1. Obtain a Groq API key from your Groq account.
2. Provide it to the server as GROQ_API_KEY:
   • Environment / .env — set GROQ_API_KEY=<key> (the standalone Docker Compose deployment reads it from .env).
   • OpenBao — when SECRET_BACKEND=openbao, store GROQ_API_KEY in the supacloud/ai KV map; the server hydrates it at boot.
3. Restart the server so it picks up the key.

The server resolves the key env-first, then from the supacloud/ai vault map. A direct environment value always overrides the vault copy.

Note

The key is instance-wide and lives only on the server — it is never sent to a browser or an agent container. With no key set, the microphone button is inert and dictation falls back to typing. See Environment variables and Secret provisioning.