Ollama (Local)

Ollama runs open-source models entirely on your machine. No API keys, no network calls, no data leaving your laptop. GROOVE integrates it for offline work and privacy-sensitive projects.

Installation

curl -fsSL https://ollama.ai/install.sh | sh

Pull a model:

ollama pull llama3
ollama pull codellama

Verify it's running:

ollama list

Authentication

None. Ollama runs locally and requires no API keys or accounts.

Models

Model	Notes
Any pulled model	Use ollama list to see available models

Common choices for coding tasks: llama3, codellama, deepseek-coder, mistral. Performance depends on your hardware -- GPU acceleration is recommended for larger models.

GROOVE Integration

GROOVE spawns Ollama agents as managed processes. They participate in the full coordination layer -- introductions, file locks, approval routing, and token tracking. The main difference is that all inference runs locally, so there is no API billing.

Limitations

• No hot-swap -- model changes require a kill-and-respawn cycle
• Hardware dependent -- model quality and speed depend on your CPU/GPU
• Smaller context windows -- most local models have significantly smaller context windows than cloud providers, which means more frequent context rotations

When to Use Ollama

Ollama is the right choice when:

• You're working offline or on a restricted network
• Your project has strict data privacy requirements -- nothing leaves the machine
• You want to experiment with open-source models at zero cost
• You're running a side task that doesn't need the full power of a cloud model

Pair Ollama agents with cloud providers on the same project. Run Claude Code for the core architecture work, and spin up an Ollama agent for local-only tasks like formatting, linting, or generating boilerplate.