Question: does tool calling conflict with structured output?

[ktrapeznikov]

It seems I can either have agent use tools or return structured output (by calling .structured) but not both. Is that by design? I tested with the latest anthropic and openai models?

[evalstate]

It's by design at the moment. What model/provider are you using?

[ktrapeznikov]

I tried both anthropic and openai. So I ended up just calling agent once to get the response as text and then second time to convert text response to specified pydantic object using .structured method