Allow the user to scroll up during LLM generation

[jeremyfowers]

Currently the generation auto-scrolls no matter what. Scrolling up should stop auto-scroll, and scrolling to the bottom should re-enable auto-scroll.

[Muhammad-aidev]

Hi! I would love to take a look at this. Could you please assign this issue to me?

[jeremyfowers]

Sure thing, thank you!

[Muhammad-aidev]

Thank you again for assigning this issue to me. I've spent some time investigating the codebase and wanted to share some feedback from the perspective of a new contributor.

While the scrolling issue itself is in the UI, I found that the main challenge is in creating a suitable testing environment. Specifically:
UI is Tightly Coupled to Setup: The main application interface is not accessible until the entire backend setup process (server checks, model checks, etc.) is complete. For a purely front-end task like this, it makes it difficult to quickly get to the relevant UI components. I was able to work around this by temporarily modifying the code to bypass the setup checks.

No Mock Data for UI Testing: Once the main UI is loaded, the llmOutput box doesn't receive any data without a fully functional LLM backend. To test the scrolling behavior, it's necessary to simulate a long text stream. I achieved this by replacing the createGame function with a local simulation that repeatedly calls setLLMOutput.

For future UI-related "good first issues," it might be very helpful for new contributors if there were a simple developer mode or instructions on how to run the front end with a mocked data stream.
Given the steps required to create a testing environment before the actual fix can be implemented, I feel this is a bit more involved than I'm ready for as a first contribution. I am going to un-assign myself to allow someone with more familiarity with the project to take over.

Thanks again for the opportunity, and I wish you the best of luck with this great project!

[lixiname]

I try to solve this problem. Solving the problem is not difficult, but the previous senior clearly thought more deeply, that is, the current front end heavily relies on backend data, which makes standalone front-end testing difficult.

[jeremyfowers]

@Imran-Ashiq thank you very much for your feedback! The feedback is a great contribution unto itself. In the past I have only worked on backends, so a lot of this is new to me.

It makes complete sense to me to have a “UI development mode” that loads instantly and mocks all the data. I’ll work on that!

[Muhammad-aidev]

@jeremyfowers Thank you so much for your kind words:)