It is returning empty message from time to time. My settings were reasoning enabled, cache enabled or disabled, reasoning level medium or high.
Hi @Kuane,
Thanks for flagging that, and apologies for the inconvenience.
Could you share the debugger log links to runs where you saw this behavior?
Please see dm
Hi @Kuane,
Thanks for sharing the link. From what I can see, Claude Sonnet 5.5 refuses to generate a response, but the same prompt works fine on its own, so the system prompt may be causing the issue.
Could you share the system prompt so I can take a closer look?
Hi @Kuane,
Thanks for checking that. I can confirm there’s nothing wrong with the API settings. The empty message comes from Claude Sonnet 5.5 itself declining to generate a response for those runs, not from MindStudio.
Different model versions can respond differently to the same input, which is why you’re getting a response from Claude Sonnet 5 but not 5.5.
Hi Alex,
That’s weird. Generally when a model declines, it will say why and not just return an empty message.
Hi @Kuane,
That’s right, when a run errors out, you can see the details in the debugger. However, in your run, the model simply didn’t generate any output
