grepai-troubleshooting

The index-not-found error means no index exists for the project. Run grepai init to create the configuration, then run grepai watch to build the index before searching.

Connection failures usually mean Ollama is not running, the endpoint in .grepai/config.yaml is wrong, or a firewall blocks the port. Start Ollama with ollama serve and verify connectivity with curl http://localhost:11434/api/tags.

Empty results usually come from an empty index, overly broad ignore patterns, or a too-specific query. Check grepai status for file and chunk counts, review ignore rules in .grepai/config.yaml, and retry with a broader query.

Yes, GrepAI supports OpenAI as an embedding provider by setting provider to openai with a model like text-embedding-3-small in the config. You must export a valid OPENAI_API_KEY environment variable or you will get 401 Unauthorized errors.

Slow indexing comes from large codebases, slow embedding providers, or missing ignore patterns. Add directories like node_modules and dist to the ignore list, use a smaller model such as nomic-embed-text, or switch to OpenAI with higher parallelism.

Delete the entire .grepai directory with rm -rf .grepai, then run grepai init followed by grepai watch to rebuild configuration and index from scratch.