Error: "message" needs to be provided if the generate_init_message method is not overridden.

[hellocaio]

Hello,

Could someone please enligten me about this error? I've been trying to learn AutoGen and create my own application that can answer simple questions, fetch data from API or from local ChromaDB. It is an amazing tool! I had it working but recently I've been getting this error and have no idea how to solve it.

I suspect it is something to do with the API endpoints? Because when I run the notebooks provided in the source code on Github Codespaces I get this same error at times. The following error came from notebook/agentchat_chess.ipynb. All I did was changed OAI_CONFIG_LIST with my keys and ran the code.

Thanks!

[afourney]

This looks like an Azure error, and it has to do with your OAI_CONFIG_LIST. Probably the model name doesn't match the deployment name. I'm not sure why you are seeing this intermittently -- the only thing I can think of is that you have several models in your OAI_CONFIG_LIST file, and it's falling back to a misconfigured Azure endpoint when the earlier models timeout or fail for other intermittent reasons.

[hellocaio]

Thanks for pointing me in the right direction. The error appears related to OpenAI endpoint, I ran out of credit, oops. Can I use only Azure OpenAI or OpenAI is required? I ask because I have a lot of credit in Azure but when I try run it with only Azure OpenAI it fails. Thanks!

[afourney]

You can just use Azure, yes. Just remove any OpenAI entries in the config_list

[hellocaio]

Awesome, thank you! It was lack of credit after all, I ended up running some complex requests on gpt-4 and it drained my credits but it was worth it. :)