Ollama
Ollama runs open-weight models locally on your machine. Jorv Builder can route drones (and optionally the brain) through Ollama — useful for fully offline work, privacy-sensitive codebases, or just experimenting.
What Jorv Builder supports
- Drone routing — set Ollama as your drone provider in Settings → AI Models → Drones.
- Brain routing — possible but not recommended; Claude Opus is meaningfully better at the planning + tool-use loop than current open-weight models.
- Auto-discovery — Jorv Builder queries
localhost:11434and lists the models you have installed.
Setup
- Install Ollama:
curl -fsSL https://ollama.com/install.sh | sh(macOS / Linux) or download from ollama.com. - Pull a model you want to use:
ollama pull qwen2.5-coder:14b - Start Ollama (
ollama serveif it didn't auto-start). - In Jorv Builder, Settings → AI Models → Ollama. Jorv Builder lists installed models. Pick one for the drone slot.
Performance
Local models are slower than hosted Claude / GPT for the same task. Plan for ~3–10× longer drone calls. The trade-off is zero per-call cost and full offline operation.
Related
- Concept: Orbit (the brain)
- Settings → AI Models (in-app)