Gemini Model Client in autogen-ext

[ekzhu]

What feature would you like to be added?

Model client for Gemini mirroring the one in v0.2

Why is this needed?

Gemini is a notable provider

[rohanthacker]

If possible, I can i work on this task.
As am in the process of implementing of the Azure Inference API as a ChatCompletionClient, but I believe this client requires some discussion and a little more though into how it is implement due to the large overlap with the OpenAIChatCompletionClient (Check #3902 for more info)

[ekzhu]

@rohanthacker thanks I just assigned! Would be great if you can follow up with us about your plan.

[rohanthacker]

@ekzhu My apologies for the delayed response. The plan I've thought of is as follows:

1. First Implement the Gemini Client with google-generativeai package along with tests
   • This is so that I can get more familiar with the ChatCompletionsClient protocol and other aspects of the 0.4 APIs
2. Introduce Vertex-AI support to the Gemini Client with the google-cloud-aiplatform package
3. Once the above two are discussed and merged, I'll move on to Add the ability to use models deployed on Azure AI Studio #3902

At the moment I'm almost done with the Gemini client and will be opening a draft PR soon.