gonuLocal Ollama inference
Use a supported model already installed on your machine through a local client.
WHEN IT HELPS
Put it to work.
Keep model inference on a development machine when that model and the selected task are compatible.
Before you begin
- Worker, CLI or VS Code on the machine running Ollama.
- An installed model, a reachable loopback daemon and suitable model capabilities for the task.
WALK THROUGH IT
See the workflow, step by step.
Synthetic sample content, with simplified screens. No model call, upload, account change or paid task runs here.
Install and start Ollama through its official setup, then install a suitable model yourself. Gonu does not silently download models for you.
Use the local model settings in Worker, or the provider/model options in CLI or VS Code. Select the exact installed model and start a new task.
Start with a short Ask task. Read connection/capability errors. Only expand to file tools when the model supports them and you have reviewed context sharing.
Local setup / Checklist
THE USEFUL RESULT
What you should leave with.
Local inference for an eligible task with explicit model selection.
Know the boundaries.
- Local inference is not fully offline orchestration: selected context and session history can still pass through the Gonu API.
- The local bridge does not silently fall back to a cloud model. Hosted GitHub jobs cannot use the Ollama instance on your laptop.
IF SOMETHING DOESN’T WORK
Model unavailable or incompatible.
Check that Ollama is running, the exact model is installed, and its metadata/capabilities support the task. Try a small Ask request before tool execution.
Contact support