Expanding upon moatless tools: discussion thread

[zytegalaxy]

First of all hats off to Albert for doing such an amazing job building moatless tools. I wanted to ask if it would be possible to open discussion section under this repo so that larger researcher community can discuss further ideas to build upon moatless tools.

I am exploring and experimenting several angles such as feeding back execution logs (runtime information) as well as other helpful feedback into the context and get the tool retry given that feedback.

I believe it would be also possible to do some RL to fine-tune coding behavior based on run-time feedback for a given codebase.

[aorwall]

Thanks! Sounds like a good idea. I enabled it 🚀 https://github.com/aorwall/moatless-tools/discussions

[aorwall]

I’m thinking about figuring out a way to let others share trajectories and info about experiments done with moatless tools on experiments.moatless.ai. It could maybe work as a nice complement to the discussions there.

I was also planning to share the code for visualizing trajectories in this repo as soon as I’ve got a decent structure for it.