[Feature Request ]: AutoGen Studio - Introduce a configurable timeout parameter for the local code executor, adjustable directly from the UI.

[jesusdeljim]

What feature would you like to be added?

When using Autogen directly from code and initializing an agent capable of executing code locally, the LocalCommandLineExecutor class is utilized, which can be initialized with a timeout parameter to control the code execution timeout. However, in Autogen Studio, the default timeout value of 60 seconds is always used, and it is currently not possible to increase this value without modifying the source code.

Specifically, in the file:
../venv/lib/python3.10/site-packages/autogenstudio/utils/utils.py, the LocalCommandLineExecutor is initialized without passing the timeout parameter.

Are there any plans to add support for this feature in a future release?

Why is this needed?

It would be highly beneficial to allow the configuration of a timeout parameter when defining an agent responsible for code execution, especially when handling time-intensive tasks. Since LocalCommandLineExecutor supports a timeout parameter when used in code, enabling the same feature in Autogen Studio would ensure consistency and parity between direct code-based usage and UI-driven workflows.

[rysweet]

thanks @jesusdeljim - we will consider this and update the issue soon.

[rysweet]

also @jesusdeljim since you have pinpointed the issue you could also consider creating a PR:

• fork the repo
• make your proposed fix
• test that it works (write a test if appropriate)
• validate that all the CI is greeen
• submit the PR

[victordibia]

@jesusdeljim

This is good feedback, thanks.

Two important things

• Version: Can you share what version of AutoGen Studio you are using?
  We have released a new version (0.4.0x ..) and currently building that out (see recording below). Can you confirm the error you are referencing is based on this?

• Improving Alignment AG Python Python Library and Studio.
  The issue you mention is related to a broader effort to improve alignment between what you can do in the low level library in python and what is possible in AutoGen Studio. Specifically, we want to support a process where you can build out your team/workflow in code and then run team.dump_component() to get a specification that you can use in AGS and vice versa. This will let us easily do things like also define a LocalCommandLineExecutor as a tool and then call tool.dump_component which will let us use the exact parameters in AGS.

That work is tracked here #4439

ags_update.mp4

[jesusdeljim]

Thanks for the quick response. I was using the previous version working on a project where our tools were dependant on the secrets variables defined in the tools section to work properly. After updating to the latest version (0.4.0.3), I am unable to locate where these secrets are now defined. Additionally, I have not found any mention of them in the documentation. Are these variables being completely removed from Autogen Studio?