Problems with the AI service that writes answers — Gemini, Groq, OpenRouter, OpenAI or Claude.
“Add an AI key to get answers”
There’s no key for the AI service selected in Settings → AI for answers.
- Click “Set up keys…” in the message and add a key.
“Choose a model: Settings → Model”
The key is there, but no model is selected yet.
- Settings → Model → pick one (Refresh loads the list).
“Your Gemini key doesn’t work”
The AI service refused the key (the service name in the message is yours).
- Click “Set up keys…” and paste the key again — Realtimer checks it right away.
Technical: HTTP 401 / 403, or 400 “API key not valid” (Gemini)
“Your OpenAI account is out of credit — top up, or switch AI in Settings”
The paid account behind the key has no credit left.
- Top up the account, or pick a free service in Settings → AI for answers.
Technical: HTTP 402
“Gemini’s free daily limit is used up — add another key or switch AI”
Free AI tiers allow a set number of requests per day.
- Add another key, or switch to another service in Settings → AI for answers.
- The limit resets the next day.
Technical: HTTP 429 (per day)
“Too many requests — wait a few seconds”
The per-minute limit of the free tier was hit.
- Wait a few seconds and ask again. Fast uses a lighter model and hits limits less often.
Technical: HTTP 429 (per minute)
“Gemini’s limit is reached — wait a minute, add another key or switch AI”
A usage limit was reached, without details on which one.
- Wait a minute, add another key or switch the AI service.
Technical: HTTP 429
“This model isn’t available anymore — pick another in Settings → Model”
The service retired or renamed the selected model.
- Settings → Model → Refresh → pick another model.
Technical: HTTP 404
“Gemini is overloaded — try again or press Fast”
The AI service is overloaded right now.
- Ask again, or press Fast — it uses a lighter model.
Technical: HTTP 500 / 502 / 503 / 504
“Gemini took too long to answer — try again or press Fast”
No answer within the time limit.
- Ask again, press Fast, or pick a faster model in Settings → Model.
Technical: Request timeout
“Gemini couldn’t answer — try again”
An unexpected error from the AI service.
- Ask again. If it repeats, report a bug with the code.
“Press Start first — answers need the conversation · Nothing heard yet”
Answers are based on what was said, and nothing has been transcribed yet.
- Press Start and let the conversation begin, then press Fast, AI or KB.