grepai-ollama-setup

Install Ollama for your platform, start it with ollama serve, then pull an embedding model with ollama pull nomic-embed-text. GrepAI's default configuration already points to the Ollama endpoint at localhost:11434, so no config changes are needed.

nomic-embed-text is the recommended default with 768 dimensions and a 274 MB download, offering a good balance for code search. Use bge-m3 or mxbai-embed-large for larger codebases or higher accuracy, and nomic-embed-text-v2-moe for multilingual code.

Yes, once Ollama and the embedding model are downloaded, embedding generation runs entirely on your machine. Your code never leaves your device, and no internet connection or API key is required during search indexing.

Connection refused on localhost:11434 means the Ollama server is not running. Start it with ollama serve in the foreground, or run it as a background service with nohup or systemd, then verify with curl http://localhost:11434/api/tags.

Memory usage depends on the model: nomic-embed-text uses about 500 MB of RAM, mxbai-embed-large about 1 GB, and bge-m3 about 1.5 GB. If you hit out-of-memory errors, switch to a smaller model or free up system memory.