grepai-ollama-setup_skill

Overview

This skill installs and configures Ollama as the local embedding provider for GrepAI so you can generate private embeddings on your machine. It guides you through installation on macOS, Linux, and Windows, model downloads, verification steps, and GrepAI configuration. The goal is a fast, private setup where code never leaves your device.

How this skill works

The skill walks you through installing the Ollama service, pulling an embedding model (recommended: nomic-embed-text), verifying the local API is responding, and updating GrepAI’s config to point at the local endpoint. It includes commands to test embedding generation and common troubleshooting steps for connection, model availability, and resource limits.

When to use it

• Setting up GrepAI for private, local embeddings
• Installing Ollama for the first time on macOS, Linux, or Windows
• Choosing and downloading an embedding model for code search
• Troubleshooting Ollama connectivity or model issues
• Optimizing local embedding performance and memory usage

Best practices

• Start Ollama before launching GrepAI (run ollama serve or enable the service)
• Use nomic-embed-text for a balance of size and performance unless you need multilingual or higher-accuracy models
• Run Ollama in the background as a service for continuous availability
• Monitor memory usage and pick a smaller model if you hit RAM limits
• Keep models updated with ollama pull to benefit from fixes and improvements

Example use cases

• Private code search in a local repo where code must never leave the machine
• Setting up an offline development environment that generates embeddings without network access
• Switching to a multilingual embedding model for repositories with mixed-language comments
• Testing embedding generation locally during CI or pre-commit checks
• Troubleshooting slow or failing GrepAI searches by validating the Ollama endpoint and model