What region does Google ADK use with Gemini when running a model locally?

[mattmcc-attest]

If I run an agent locally using run_async with a root_agent that points to gemini-2.5-flash, how do I know which region the local agent runner will send requests to Gemini?

[adk-bot]

Response from ADK Answering Agent (experimental, answer may be inaccurate)

TLDR: When running an agent locally with run_async, the Google Cloud region for Gemini requests is determined by the GOOGLE_CLOUD_LOCATION environment variable. You should set this in a .env file in your project's root directory.

---

When you run an agent locally using run_async and it calls a Gemini model, the ADK (Agent Development Kit) uses the underlying Google Cloud client libraries. The region for these API requests is configured through an environment variable.

To specify the region, you should create a .env file in the root of your project and add the following line:

GOOGLE_CLOUD_LOCATION="your-google-cloud-region"

For example:

GOOGLE_CLOUD_LOCATION="us-central1"

You will also need to ensure that you have configured ADK to use the Vertex AI backend by adding this line to your .env file:

GOOGLE_GENAI_USE_VERTEXAI=TRUE

And of course, your project ID:

GOOGLE_CLOUD_PROJECT="your-google-cloud-project-id"

You can see examples and more details in the quickstart and streaming documentation. [1][2]

If the GOOGLE_CLOUD_LOCATION is not set, the system might default to us-central1 [3], or fall back to the region configured in your local gcloud settings (gcloud config get-value location). However, it is best practice to explicitly define the region in your project's .env file to ensure your requests are sent to the correct region.

---

[1] https://google.github.io/adk-docs/get-started/quickstart/
[2] https://google.github.io/adk-docs/streaming/custom-streaming/
[3] https://github.com/google/adk-python/blob/main/src/google/adk/cli/cli_create.py

[mattmcc-attest]

I am not using vertex AI