Troubleshooting

  1. Server won’t start: “Postgres connection failed”
  2. “AI rate limit reached — wait a minute and try again”
  3. The interviewer’s voice sounds robotic
  4. No voice at all
  5. The mic doesn’t work / “Microphone unavailable”
  6. “That answer is already being evaluated”
  7. Google sign-in: “The given origin is not allowed for the given client ID”
  8. Vite fails with “Cannot find native binding”

Server won’t start: “Postgres connection failed”

  • DATABASE_URL is not set: add it to server/.env. The server reads that file even when started from the repo root.
  • password authentication failed: the password inside the URI is wrong. For Supabase, that’s the database password, not your account password.
  • tenant/user … not found: the project ref in the Supabase pooler username (postgres.<project-ref>) or the region in the host is wrong. Copy the URI again from Connect → Session pooler.
  • self-signed certificate in certificate chain: you’re using a non-Supabase host with a private CA. Set DATABASE_SSL_CA=/path/to/ca.pem, or as a last resort DATABASE_SSL=no-verify.
  • ENETUNREACH / IPv6 address: you used Supabase’s direct connection. Use the Session pooler URI.

“AI rate limit reached — wait a minute and try again”

You hit Groq’s free-tier limit for the chat model (30 requests a minute, 1,000 a day). Wait a minute, or upgrade your Groq plan.

The interviewer’s voice sounds robotic

The browser voice has taken over from Groq Orpheus. Check the browser console for Groq TTS unavailable, using the browser voice:. The reason follows:

  • …terms accepted in the Groq console: accept them once at the Orpheus playground.
  • AI rate limit reached: the Orpheus free tier is about 100 requests and ~3.6K characters a day. The Groq voice is retried 2 minutes after a rate limit, and resets daily.
  • Server-side speech is disabled: TTS_PROVIDER=browser is set on the server.

No voice at all

  • Make sure Voice On is shown in the interview header. It’s a mute toggle.
  • Browsers block audio until you interact with the page; the Start Interview click counts. If you reloaded mid-interview, start a new one.
  • On Linux without speech-dispatcher, the browser has no voices, so the fallback is silent. The interview still continues; read the question on screen.

The mic doesn’t work / “Microphone unavailable”

  • The page must be served over https:// or from localhost. On http://<lan-ip> browsers disable the microphone, and you’ll get the text box instead.
  • Allow microphone access in the browser’s site settings, then reload.
  • “No audio was captured” means the recording was empty. Check that the right input device is selected in your OS.

“That answer is already being evaluated”

The same answer was submitted twice, for example a double click. The first submission is still being scored, so wait for it. If the server crashed mid-evaluation, the claim expires after 2 minutes.

Google sign-in: “The given origin is not allowed for the given client ID”

Add the exact origin you’re using (scheme, host and port, e.g. http://localhost:5173) under Authorized JavaScript origins for your OAuth client in Google Cloud Console. Google accepts only localhost or https origins, so a LAN IP over plain HTTP can’t be used. Email and password sign-in always works.

Vite fails with “Cannot find native binding”

node_modules was installed on a different OS (for example Windows) and lacks the Linux native packages. Reinstall on this machine:

cd client && rm -rf node_modules && npm ci