Claude response cut short

​@hohoho 

Use the built in “AI by Zapier” step, you can select Claude as your LLM inside that and you won’t get the response cutoff issue.

Thanks mate, I have tried doing that but now it’s saying timed out.

All I wanted it to do is to process scraped data and feed that to Claude for analysis, that’s it, but it’s not allowing me to.

Is the response cut short also a known issue?

​@hohoho 

Can you screenshot what you’ve setup and the cutoff output you’re seeing.

Output was cut short and it’s been quite challenging.

Separately, thanks for the advice on AI by Zapier, it helped for sure by reducing the data field and now returning better insights and output!
 

Hi there ​@hohoho 👋

Sorry to hear the response from Claude is being cut-off. I did some checking and found a similar report of this which was due to the current limitation mentioned in our Test Zap steps guide here:

This isn’t specific to the Claude app, it applies to other app actions as well. If you switch on the Zap and run a live test then it should pass the full response from Claude over to subsequent actions. 

Can you give try running a live test and let us know whether that works as expected?

This is great feedback, thanks ​@SamB! I did exactly this and the values did expand further than testing phase however, it did eventually truncated. I am suspecting because the output was too long.

Thanks for giving that a try ​@hohoho. I think you’re right, if it’s still being truncating on the live Zap runs it could be due to the Apify output being too long. It may be using up most of Claude’s available context tokens just to read the input, leaving too few tokens to generate the full response 🤔

In which case, I’d suggest splitting the output from it into smaller chunks before passing to Claude. That should help to stop Claude’s responses getting cut off. We have a guide on how to do that here: Modify large data for your AI prompts

Keep us posted on how it goes!