Is there a way to connect to two different GCP accounts using same ADK running instance

[abhishek-1309]

Hi, I have two cloud GCP accounts, currently I have an agent (google.adk.agents -> Agent )setup for both of them hosted seperately. I want to combine them into a single server, so two different agents for two different cloud credentials? Also so is there a way to pass credentials other than the env variable approach?

[ankursharmas]

Is it possible for you to provide details on why you have such a setup?

There is at least one approach i can think of, but that requires doing some gymnastics (that i would like to avoid, hence my question above):

Assuming that you are using GCP credentials in the context of talking of Gemini model. You could create two instances of Gemini adapter (https://github.com/google/adk-python/blob/main/src/google/adk/models/google_llm.py#L52). This will require you to override the api_client method (https://github.com/google/adk-python/blob/main/src/google/adk/models/google_llm.py#L205) and create new client that takes in GCP project details that you can pass in explicitly (https://googleapis.github.io/python-genai/#api-selection).

Once you get to create your own Gemini adapter using this way, you can pass the instance to the Agent (instead of model id string).

[abhishek-1309]

I'm building a chatbot for two different customers. The chatbot is implemented using a common Agent class from ADK, which includes tools and model callbacks. While the core Agent logic remains the same for both customers, the prompts and data inputs are customized for each one.

Now, both customers are requesting to use their own Gemini models instead of the shared model I initially provided, which previously served both of them.

So essentially, I will receive 2 different credentials to use gemini, so for each of the customer the agent should connect to their own instance of Gemini