Configure model profiles

A model profile is the configuration unit Agents and CodeAct Agents use to call a model. Keeping keys and model selection in a profile rather than copying them into every node makes credential rotation, model replacement, and diagnosis much clearer.

Create a profile

1. Open Settings → Model Profiles.
2. Select New.
3. Enter a name based on its purpose, such as prod-review, local-ollama, or eval-cheap.
4. Choose a provider.
5. Enter the model ID and required credentials. Compatible APIs also need a Base URL.
6. Save, then validate the connection with a minimal workflow run.

Use it in an Agent

Select an Agent node and open Model & instructions → Model configuration in the right inspector, then choose the profile. A model profile defines only which model to call. Role, task objective, output format, and tool boundaries belong in node instructions.

Configuration guidance

• Use separate profiles for local, development, evaluation, and production work.
• Never put a key in a node instruction, workflow input, or Python code.
• After changing a model, temperature, or provider, run the same evaluation suite before publishing.
• Temperature and Top P are node-level generation controls. Leave them empty for model defaults unless you specifically need stability or diversity.

Media placeholder · screenshot model-profiles/01-profile-list.png
Show the Settings profile list with local, evaluation, and production naming examples. All real provider credentials must be redacted.