Documentation

[MichalBortkiewicz]

Hi!

I really like the idea of defining agent policy as a graph and your implementation seems highly generic. However, it is also quite hard to navigate the method calls without a proper overview/documentation of the code.

Can you maybe provide some additional info about code structure, abstractions and nodes activations/interruptions?

Regards

[nicoguertler]

Hi, thank you for your feedback! I agree that without a documentation it is currently hard to navigate and use the code. While I won't be able to write a full documentation right away I can start adding some material. Here are some things I could work on:

• An introduction outlining the idea of defining the policy via a graph
• A discussion of the control flow in the graph during the execution of the policy
• A tutorial on how to run graph_rl on a new environment
• Writing documentation for the classes and methods

What would you be most interested in?

[MichalBortkiewicz]

Thank you for your response!

From my perspective, documentation for classes and methods is alright.

However, the most confusing part for me was the control flow in the graph during the execution of the policy. I had a difficult time understanding the communication between high-level and low-level policies/nodes. I guess a more detailed diagram (maybe UML) of the communication (with message carriers between the nodes) will be super helpful.

I haven't yet checked running graph_rl on a new environment, but a quick tutorial on this would also be nice :)