Configure Gemini inference

Gemini is optional and is configured only in the Payne web and worker server environment. Never commit the API key or place it in a NEXT_PUBLIC_ variable.

Add these values to .env.local-live:

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

PAYNE_INFERENCE_PROVIDER=gemini
PAYNE_INFERENCE_MODEL=REPLACE_WITH_A_MODEL_ID
GEMINI_API_KEY=REPLACE_WITH_A_SERVER_SIDE_KEY

The model must support the tool-calling and structured-output behavior used by Payne. Check the official Gemini function calling and structured output references when choosing a model. This project does not claim compatibility with every Gemini model.

Restart the stack, open Connection, and choose Check inference.

If the check fails, use Inference troubleshooting.