anyone have any ideas?

Is it a hard stop? Could be a token cutoff? 🤔

I figured it out, there was a default max_tokens settings in the llm provider

making that larger solved the issue