Gpt4

Did you update your openai package? Did you use the ChatOpenAI class from langchain for gpt4?

(There are notebook examples in the llama index repo with gpt4, so it should work 😅)

Yes for both 🙂

@Logan M I'm actually trying to use it directly from Langchain, without an underlying index

Hmm I'm not sure what the problem is then 🤔 I would double check that you ran pip install --upgrade langchain openai and double check the token.

Otherwise, it seems like an issue on openai's end 👀

Thanks, will try that again 🙂

@Logan M It didn't work directly through Langchain but I thought I should maybe try it using the LLMPredictor wrapper. However, now I get an error even for gpt3.5 model. Any idea what am I doing wrong ?

llm_predictor = LLMPredictor(llm=ChatOpenAI(temperature=0, model_name="gpt-3.5-turbo")) query = "What's Daniel's address?" response = llm_predictor.predict(Prompt(query))

And I get the error:

AttributeError: 'Prompt' object has no attribute 'input_variables'

Ohhh, i think the input to the llm_predictor needs a few special things, it's not intended to be used directly 😅

I read in the docs it has a "predict" function, so I thought I can use it directly

The error I'm getting is for the last line

Yea, you need to use a properly initialized prompt object, like in the repo in gpt_index/prompts/prompts.py

I think the docs just show all our object apis, instead of the actual user-facing ones 🙃