The response JSON is always truncated, resulting in an unexpected answer.

[ChambersXDU]

I choose litellm and use a gpt-4o-mini model.

I found that every time I asked a question, it seem to return many similar contents, at the end of each content, the JSON is incomplete.

What happened? how can I solve it?

[tcsenpai]

I will have to find a key for 4o-mini to reproduce this.

I think is (as usual in these cases) prompt related.

I will try to add a specific provider for OpenAI to stop having this issue too.

If anyone wants to try to solve this, you are most welcome.