FFFF Voice dictation cannot work on a production start without undocumented env vars · Issue #27 · simon-inkie/inkie-deck · GitHub
[go: up one dir, main page]

Skip to content

Voice dictation cannot work on a production start without undocumented env vars #27

Closed

Description

@simon-inkie

Problem

The voice proxy's default VOICE_ALLOWED_ORIGINS is http://localhost:5174 (packages/deck/server/voice-proxy.mjs:57), an old dev port, so it rejects the page Deck serves on 4242. .env.example has the Deepgram lines but none of the VOICE_* ones. The mic just fails with a WebSocket error in the console.

Evidence

Seen in test: reproduced on a production start; also flagged in an earlier run.

Fix

Default allowed origins to Deck's own http://localhost:$PORT (and the dev port). Add commented VOICE_ALLOWED_ORIGINS, VOICE_PROXY_URL, VOICE_PROXY_HOST to .env.example. Add a documented walkthrough to INSTALL.md, including the case of reaching Deck over Tailscale (wss://). Surface "voice disabled: no DEEPGRAM_API_KEY" in the UI instead of a silent failure.

Acceptance criteria

  • With only DEEPGRAM_API_KEY set, the mic works on a local production start at 4242.
  • Without it, the mic shows a clear disabled state.
  • Test on the default-origin computation.

From the 2026-09-27 Windows/WSL install test.

Metadata

Metadata

Assignees

No one assigned

    Labels

    priority:P2Needs a manual workaroundvoiceVoice dictation

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions

      0