Configure Ollama inference

Run Ollama on a host reachable from the Payne containers and pull the chosen model before configuring Payne. Local inference reduces external data transit; it does not remove metadata, model, host, or application security risks.

Confirm the selected model against Ollama’s official tool calling and structured output guides. Payne still requires its own live contract check because Ollama feature support does not imply compatibility for every model.

CommandRun from the Ollama host

ollama list

Add the provider, that model identifier, and the URL reachable from Compose:

Environment variablesRun from the `.env.local-live` file

PAYNE_INFERENCE_PROVIDER=ollama
PAYNE_INFERENCE_MODEL=REPLACE_WITH_AN_ALREADY_PULLED_MODEL
PAYNE_OLLAMA_BASE_URL=http://host.docker.internal:11434

Restart the stack and choose Check inference on Connection. Payne does not claim support for arbitrary local models; the configured model must pass the live tool and structured-output check.