RepoLens

Turn your GitHub repository into a AI-powered mock interview.

Comment

posted an update —

Regarding the issue: "gemini provider error 404, continued with local hash embeddings"

  1. The Issue: The backend logged a 404 error because it requested an outdated embedding model (text-embedding-004).
  2. The Fix: We updated the .env configuration to use the correct and available model (gemini-embedding-2).
  3. The Quota Limit: The fixed API calls then hit a 429 rate limit because processing an entire repository at once exceeded the free-tier quota.
  4. The Validation: Our ProviderRouter successfully caught these failures, seamlessly failing over from Gemini to OpenAI, and finally to local hashes.
  5. The Result: The application degraded gracefully, allowing the user to continue their interview without any crashes despite the API limitations.

Log in or sign up for Devpost to join the conversation.