Is it possible to provide built_in_code_execution ability to read from a file?

[bugthedebugger]

The tile basically. I am looking for a way to allow built_in_code_execution to read a csv file to run analysis rather than me providing the data in the context as the file is bigger than the context window😆. Any way of doing this atm? Or will i have to write custom code executor?

Any thoughts/helps are appreciated. Thanks!

[bugthedebugger]

To anyone who ends up here with the same issue as above, i figured it out. Its not documented anywhere. You NEED to use VertexAiCodeExecutor in order for the code executor to read artifacts/files and pass the artifact as part to be able to read the generated artifacts within the code executor.

Agent(
    name='agent_name',
    model='gemini-2.0-flash',
    description=<description goes here>,
    instruction="""<instruction goes here>""",
    # tools=[built_in_code_execution]
    code_executor=VertexAiCodeExecutor(
        optimize_data_file=True,
        stateful=True,
    )
)

csv_artifact = types.Part(
        inline_data=types.Blob(
            display_name="test_data.csv",
            mime_type="text/csv",
            data=csv_bytes
        )
    )

content = types.Content(role='user', parts=[types.Part(text=query), csv_artifact])

# pass the content into runner 
runner.run_async(user_id=USER_ID, session_id=session.id, new_message=content)

[grishakuznetsov]

thank you so much

[pranjalagg]

Hey, I have the same issue but am not able to solve it.

My setup:

• I have a root agent that talks to the user and transfers the tasks to appropriate agents
• One of the sub agent is an analytics agent (similar to this example).

What I am trying to achieve:

• I want the sub agent to load appropriate csv files (using the tools I have) and pass them to the VertexAiCodeExecutor to do analysis/plotting on it.

I am using adk web for testing my application and am confused on where and how exactly would I use runner as in your example above. Would appreciate any help. TIA

[Mattylam]

hey pranjalagg, Ive got the same issue, were you able to solve this, the closest I got is to save the artifact, using the save_artifact. But it is still not working.