gemini-2.0 does not recognize google_search_retrieval tool

[Kvit]

Description of the bug:

While switching model from gemini-1.5-pro to gemini-2.0-flash-exp google_search_retrieval tool is no longer recognized

Actual vs expected behavior:

Using "gemini-2.0-flash-exp"

model = genai.GenerativeModel(
    model_name="gemini-2.0-flash-exp",
    tools='google_search_retrieval')

response = model.generate_content("Search latest news")
print(response.text)

InvalidArgument: 400 Unable to submit request because Please use google_search field instead of google_search_retrieval field.. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/gemini

Expected:
Here's a summary of the latest news as of February 1, 2025...

Any other information you'd like to share?

Package Version

Name: google-generativeai
Version: 0.8.4
Summary: Google Generative AI High level API client library and tools.
Home-page: https://github.com/google/generative-ai-python

[Kvit]

Code with search tool generated by Google AI Studio https://aistudio.google.com/ for model gemini-2.0-flash-exp also does not work

[gmKeshari]

Hi @Kvit,

My hunch is search grounding with 2.0 models only work with the new SDK. Have you tried with new SDK?? Here is the gist. You can also follow the quickstart colab notebook for grounding using search tool.

I checked with the AI Studio code, it's not working. I will escalate this to our internal team to update this.

Thanks.

[Kvit]

@gmKeshari I think I figured out the issue: Gemini 2.0 API structure is different from the previous version. I suppose AI Studio code generator should be updated. Otherwise, it is confusing.

https://ai.google.dev/gemini-api/docs/models/gemini-v2

[tqdo]

@Kvit so how did you get it to work? I am having the same error

[gmKeshari]

Hi @tqdo,

Please use this doc which uses new SDK.

Thanks.

[Wazbat]

I've arrived here while struggling to figure out how to enable grounding when using langchain. It seems like the documentation is a little all over the place, including the docs on the GCP page itself for grounding

[meher-gajula]

https://cloud.google.com/vertex-ai/generative-ai/docs/model-reference/grounding#ground-on-public-data

I just tried the example on Google Docs and I get the same error.