Proposal for IDE-Based Control and Interactive Code Revision

[alpgul]

I believe it would be much more efficient to implement the control mechanism through an IDE extension. This structure would allow for leaving comments directly on the additions, deletions, or modifications made by the AI, enabling the code to be re-coded or revised based on those comments in the future.

This way, developers can iterate quickly within the editor without needing a browser. Constant context switching between the browser and the IDE would significantly hinder the development process compared to the speed and fluidity of working directly within the IDE.

[backnotprop]

@alpgul thanks

Do you mean for the entire experience.

Or just the code review feature, plannotator-review

[alpgul]

Yes, I mean review feature. I believe performing checks via an IDE extension would be more efficient than a browser-based solution. Most review tools I have experienced operate individually through manual triggers. With the bulk commenting feature used in this tool, we can add multiple comments and create a more consistent code structure.

[alpgul]

Let me provide a step-by-step example scenario for better understanding:

1. Suppose you have a plan that modifies the wrong files and code.
2. In this plan, we open a global or task-specific comment.
3. We then select a code block (from project files or from uncommitted code change files) within the IDE and choose the "create comment" option.
4. We add this code block and the associated comment as a sub-comment within the main task comment.
5. This approach enables more reliable and faster code production.
6. This way, while editing the plan, we would be doing so with the support of code blocks.
7. Of course, before implementing this, a plugin needs to be developed for the IDE to manage these controls.
8. This feature can be added at a later stage.

You can think of this as the popup prompt feature with code transformation capabilities found in IDEs. However, while IDEs execute these manually and individually, we run them in bulk here, which will ensure higher code consistency.