Yes, runtime debugging helps to solve a lot of bugs. I was so annoyed with buggy codes generated by AI so at the end I actually developed something not sure similar to your or not:both AI coder and AI debugger two-in-one, a fork of Roo-Code. So it (Zentara-Code) can code and debug it code at the same time in runtime. It automatically set breakpoints, do stack tracing, inspect stack frame variables. As it leverages Debug Adapter Protocol (DAP), it works for any language in VSCode. https://github.com/Zentar-Ai/Zentara-Code

Thank you for the review. I see that currently code assistants do not have access to a lot of tools that a normal software developer has during software development life cycle. Once we enable it, code assistants would be much more smarter. Debugger is one of the tools that code assistants need.

Would it double your codebase? Do you think it would work for a large codebase?

Not anywhere close. It's basically just maintaining a simple descriptive index the model can later use to decide what files it needs to read given the task you've given it.

Any new features comparing with existing ones?