Code execution (interpreter) support

[operand]

It's easy enough today to implement an "eval" action on an agent and use it to run arbitrary code in python or the shell, but given the power of code execution in general, I want to explore how it can be supported as a first class feature.

Following the React app updates, I plan to explore how this could be more directly supported.

In the meantime, this is all it takes to create a simple code-interpreter-like agent if you'd like to play around with this concept.

@action
def eval(self, code: str):
    """Evaluate python code within this agent's context"""
    eval(code)

[operand]

I've decided that this feature will be addressed in a new independent project dedicated to creating a new UI for experimentation and development. See the discussion here for more details.

If you'd like to sign up to be emailed when a prototype is ready, please click here.