models.get() uses credentials' project instead of configured Vertex AI project for API enablement check

[tharun208]

Environment

• Python: 3.13
• google-genai version: latest (via llama-index-llms-google-genai)
• Platform: GKE with Workload Identity
• OS: Linux (container)

Description

When creating a google.genai.Client with vertexai=True and an explicit project, the client.models.get() call checks Vertex AI API enablement against the credentials' GCP project (from Workload Identity / ADC), not the project passed in the config.

This does not happen with the Go SDK (google.golang.org/genai) using the same cross-project setup.

Steps to reproduce

import google.genai

client = google.genai.Client(
    vertexai=True,
    project="project-a",       # Vertex AI API enabled here
    location="us-central1",
    api_key=None,
)
# Credentials come from Workload Identity in "project-b"
# (Vertex AI API NOT enabled in project-b)

client.models.get(model="gemini-2.5-flash-lite")

Expected behavior

The models.get() call should use project-a (the explicitly configured project) for the API request, where Vertex AI API is enabled.

Actual behavior

The SDK sends the request against project-b (the credentials' project), resulting in:

google.genai.errors.ClientError: 403 PERMISSION_DENIED.
Vertex AI API has not been used in project <project-b-number> before or it is disabled.

Stack trace

File "google/genai/models.py", line 4633, in get
    response = self._api_client.request('get', path, request_dict, http_options)
File "google/genai/_api_client.py", line 1375, in request
    response = self._request(http_request, http_options, stream=False)
...
google.genai.errors.ClientError: 403 PERMISSION_DENIED.
{'error': {'code': 403, 'message': 'Vertex AI API has not been used in project <project-b-number>...'}}

Additional context

• The Go SDK (google.golang.org/genai) works correctly with the identical cross-project setup (same service account, same Workload Identity binding, same target project).
• The credentials' service account has the necessary IAM roles in the target project.
• Setting GOOGLE_GENAI_USE_VERTEXAI=true does not help since vertexai_config already sets vertexai=True internally.

[Venkaiahbabuneelam]

Hi @tharun208
Thanks for reaching out us, we tried to reproduce the issue but we didn't observe any issue, we were able to get the response successfully, use the below code and see if you are able to get the response or not.

try:
response = client.models.generate_content(
model="gemini-2.5-flash-lite",
contents="Hello, are you there?"
)
print("Success! Request went through.")
except Exception as e:
print(f"Error caught: {e}")

Thanks