Hello everyone and thank you for your

iianez

Hello everyone and thank you for your valuable contribution to this community. I am relatively new to the world of LlamaIndex (about 1 month), but I am already developing a chatbot with good results. At the moment, I am using the VectorIndexAutoRetriever with CondensePlusContextChatEngine, and it works great with GPT-4-turbo, GPT-3.5-turbo, and Claude Opus on AWS Bedrock. Unfortunately, it always returns the same exception when I try to use it with GPT-4o. Can anyone tell me if this depends on the libraries, perhaps not yet updated for use with this model?

7 comments

WWhiteFang_Jr

Can you share the error ? That will help to understand

iianez

yes of course, this is the output

LLogan M

This just means the LLM failed to predict the proper outputs, this can happen sometimes

iianez

Thank you Logan. It happens every time with gpt4o. Never with other models. I can't find an explanation to this situation.

RRoland Tannous

@ianez what happens if you use chat() instead of stream_chat() in your app? (ie turn off streaming temporarily) . do you still get the same error?

iianez

hei Roland, thank you for reply. With chat() instead of stream_chat() same problem.. VectorIndexAutoretriever() seems to be the problem here... If I use classic VectorIndex with base retriever all works fine

RRoland Tannous

oh not sure what the issue is here then? maybe it's a bug or something that could be resolved with the latest version. try to update your llama-index pakcages.

Add a reply

Find answers from the community

Hello everyone and thank you for your