Interface MockModelOptions

Model metadata (e.g. supports, versions).

supports.constrained defaults to 'all' (native constrained generation), so a generate with an output.schema reaches respond with that schema on request.output.schema. Pass supports: { constrained: 'none' } to instead exercise the framework's simulated path, where the schema is injected into the prompt (and thus visible in lastRequestText) and stripped from what respond sees.

Registered model name. Defaults to 'mockModel'.

What to respond with on each generate call. Defaults to empty text.

• A single response (string / object) is returned on every call.
• A callback (request, { sendChunk }) => response is invoked once per call, so you can branch on request history (for tool loops) and stream via sendChunk.
• An array is a queue consumed one item per call, with the last item repeating once exhausted — handy for scripted multi-turn tests. A queued Error is thrown when reached, to inject a failure on a given turn. (Queued items are static, so streaming needs the callback form.)

mockModel(ai, { respond: 'always this' });
mockModel(ai, { respond: ['first', 'second'] });
mockModel(ai, { respond: ['ok', new Error('rate limited')] });
Copy