[Issue]: Can Autogen configure a network proxy

[gmorkgg]

Describe the issue

Can Autogen configure a network proxy to make calls to the official interfaces of AOAI or OpenAI through the network proxy.
Due to poor local network quality, I would like to set up a network proxy. Can it be done in Autogen? Where can I set it up?

Steps to reproduce

The system proxy's https proxy and https proxy have been set, but they are not effective in the Autogen program. Do I need to directly add a proxy network proxy in the program

Screenshots and logs

No response

Additional Information

autogenstudio version: 0.0.47
pyautogen version: 0.2.16

[ekzhu]

Have you checked out LLM proxies like Ollama and LiteLLM?

[sonichi]

Thanks. I remember this question was asked before. The solution was to add some field in the llm_config, if I remember it correctly.

[gmorkgg]

Ok, I will check the configuration of llm_config

[stnbSim]

hi @wiyisun, would you please share the solution on the llm_config's configuration?

[gmorkgg]

llm_config does not do special Settings, only need to set their own application key, I used GPT4 API KEY. You can configure this in the web interface provided by autogenstudio-ui, or in the autogen code root directory, where the configuration file name is required: OAI_CONFIG_LIST. Then launch autogenstudio-ui.

[ekzhu]

Documentation for how to do this: https://microsoft.github.io/autogen/docs/topics/llm_configuration#adding-http-client-in-llm_config-for-proxy

[ekzhu]

Yes. Please see the documentation here: https://microsoft.github.io/autogen/docs/topics/llm_configuration#adding-http-client-in-llm_config-for-proxy

[lc82111]

I have tested the solution in doc for Gemini, it's not working.

[prd-tu-tran]

Hey, I initiated an AgentBuilder class. It requires config_file_or_env. How can I save it to a file or set it up with a proxy? Do you guys know how to fix it?