This PR contains:

• New features
• Changes to dev-tools e.g. CI config / github tooling
• Docs
• Bug fixes
• Code refactor

What is the current behavior? (You can also link to an open issue here)

No aideml agent

What is the new behavior?

Yes aideml agent!

tl;dr: implements a version of https://github.com/WecoAI/aideml using inspect's new agents protocol.

Running the agent currently has a few requirements:

• install igraph
• your dockerfile must contain pandas, genson, and humanize pip packages

This implementation is fairly rough right now, so happy to do a bit of iteration to clean it up / document it better. I would like to make it more of a general framework, take away the need to install extra dependencies, and move the default prompts away from being so ML focused. Providing a default aide docker image would be a nice bonus, similar to how the web browser tool works currently, but I do not know how to do this.

Below is some example code documenting how to use the agent currently:

from pathlib import Path

from rich import print

from inspect_ai import Task, eval, task
from inspect_ai.agent import aide_agent
from inspect_ai.dataset import Sample
from inspect_ai.scorer import exact


@task
def hello_world():
    return Task(
        dataset=[
            Sample(
                input="Just reply with Hello World",
                target="Hello World",
            )
        ],
        solver=[
            aide_agent(data_dir=Path("/home/ubuntu/workspace/aideml-inspect/aide_inspect/assets/test")),
        ],
        scorer=exact(),
        sandbox="docker",
    )


if __name__ == "__main__":
    log = eval(hello_world(), model="anthropic/claude-3-7-sonnet-20250219")
    print(".")

@jjallaire let me know what you'd need here for this to be mergeable!

Does this PR introduce a breaking change? (What changes might users need to make in their application due to this PR?)

no

Other information:

n/a