Hi there!
I would like to use Gemini 2.0 in my application, in particular I would like to swap it with my current OpenAI Python client (which involves async multiple-function calling).
The problem I encounter is the existence of two different Python packages offered by Google.
First, there is the generative-ai-python
(github: /google-gemini/generative-ai-python), installed as pip install -U google-generativeai
and imported as import google.generativeai as genai
. The README points to these URLs: https://ai.google.dev/ & github: /google-gemini/gemini-api-cookbook/.
Then, there is the python-genai
repo (github: /googleapis/python-genai). It is installed using pip install google-genai
and imported as from google import genai
The README of the Cookbook (another repo!) states that " [Google GenAI SDKs](github: /googleapis/python-genai) will eventually replace the older [Developer SDK](github: /google/generative-ai-python)". When will that happen?
As a developer, I don’t want to waste time building onto a lib that’s going to be deprecated soon. In the other hand, most of the documentation is about the “Developer SDK” and not “Google GenAI SDK”. Although I want to use your models (which I know perform well) this situation is quite a pain.
PS: Is the project of a compatibility with OpenAI’s python client still on tracks? This page has not been updated since 2024-11-26, and for instance, my function calling schema (which works with the OpenAI endpoint) does not work with Gemini’s one.