search

posgnu / rci-agent

A codebase for "Language Models can Solve Computer Tasks"
https://posgnu.github.io/rci-web/
MIT License
223 stars 31 forks source link

Executing custom tasks #5

Open RajNS02 opened 1 year ago

RajNS02 commented 1 year ago

Hello! Thanks for this fantastic repo! The paper is also very amazing and insightful.

I was wondering whether it's possible to define custom HTML pages and tasks to be executed. I was thinking of adding the custom HTML file in computergym/miniwob/miniwob_interface/html/miniwob directory and also including it in available_tasks.txt Would this approach work? Please, let me know your thoughts about this.

Regards

posgnu commented 1 year ago

Hello, thank you for your interest in this repo, and sorry for the delay.

Currently, the RCI agent is using the MiniWoB++ benchmark, which is wrapped by the OpenAI Gym interface. Therefore, it will not be easy to add a custom HTML environment to the MiniWoB++ benchmark. Instead, I recommend running the agent directly on your custom website using Selenium. I have already implemented this code for the facebook.com webpage in the main.py. You can play around with the code and let me know if you have more questions.

You can lunch the code with the below argument.

--env facebook