Using a Custom model_client Instead of LLM in AutoGen (v0.4.9)

[galahad42]

Hi everyone,
In my current implementation, I'm using Redis to communicate with a locally hosted LLM server. However, while working with AutoGen, I've encountered a challenge:
AutoGen seems to require a model to be defined via model_client.
Previously, in version 0.2, I could simply set:
llm_config = False
to skip using an LLM, which worked well for my custom setup.

My questions are:
Is there a way to plug in a custom function instead of defining a model in llm_config?
(Since my communication is handled via Redis and I don't need a standard LLM client.)
Is llm_config = False still supported in AutoGen v0.4.9, or is there an alternative approach in the latest version?

Any help or insights would be appreciated!

Thanks!

[ekzhu]

AutoGen seems to require a model to be defined via model_client.

AssistantAgent is just one implementation of agent. It's not meant to be used for all scenarios. Perhaps you need a custom agent?

https://microsoft.github.io/autogen/stable/user-guide/agentchat-user-guide/custom-agents.html

Can you explain how you are using redis with your model?