afourney

If you change the system prompt, I highly recommend also considering this issue: #430 It may also be worthwhile considering: #738. Especially this part: https://github.com/microsoft/autogen/blob/aa051007659959cccbf193a18caea35155555ddc/autogen/agentchat/assistant_agent.py#L21

This is a good opportunity to rethink default. I suggest: allow_speaker_repeat = False last_n_messages = "auto" Make sure all agents have good _descriptions_, distinct from system prompts.

The most common scenario outside of AutoGenStudio where where this comes up is with the UserProxyAgent. GPT-4 GroupChat has an extremely bad habit of spamming the UserProxy, constantly asking for...

Not yet that I am aware of. Let's not develop anything here until we hear from her tomorrow

Hi all. I'm resurrecting this PR, since instrumentation is perhaps the next big missing piece of AutoGenBench. Stay tuned.

I looked at OpenTelemetry earlier, and it seemed a little heavyweight (e.g., we aren't distributed and don't have micro services)-- but maybe our needs have evolved to warrant that. We...

> I agree with having a lightweight local implementation such as this. For any kind of server deployment, we need a standard solution like OpenTelemetry. I might make a follow...

Yes, I need to write a proper readme. There are some instructions here: https://github.com/microsoft/autogen/issues/1943#issuecomment-1987278247 That should hopefully be sufficient to get you started, but let me know if you run...

Looks good, but have you tested these numbers with the OpenAI token calculator they publish? I'm imagining there are some subtleties to confirm (e.g., any tile overlap?) 

Update: I think this issue it related to the ADAL bug here: [https://github.com/AzureAD/azure-activedirectory-library-for-dotnet/issues/222](https://github.com/AzureAD/azure-activedirectory-library-for-dotnet/issues/222)