Migrate to new higher-level google-genai SDK?

[Saran33]

I see this issue was previously closed in #734 but I'm wondering if there are any further plans to switch from google-ai-generativelanguage to the new higher-level python-genai SDK?

My understanding is that the former is for lower-level API access, while the later is more of a facade, as per this comment.

On a personal note, I see the former uses gRPC, while the latter uses REST transport. Although gRPC is faster and switching to REST could arguably be considered a regression, personally I find gRPC is more of a pain for testing purposes, given that most other LLM providers are using REST. Personally I would prefer to be able to use REST for consistency and maintainability in testing (recording event streams with tools like VCRpy etc). I'm not sure if there are plans to support gRPC in the new SDK though, going by this issue googleapis/python-genai#557

[Smityz]

I've also encountered this problem and I'm looking forward to seeing it resolved.

[sergioshev]

Same here, we relay on langchain_google_vertexai wich seems to be using google-cloud-aiplatform under the hood. And as per this doc google-genai is the new way to go. Also I'm looking forward how this will be resolved, since there is an explict way to use vertexAI instead of generative ai (developer mode).

client = genai.Client(
    vertexai=True, project='your-project-id', location='us-central1'
)

[emhagman]

I've started the work for this here #1077