If I understand that correctly, I think that can be accomplished in [name-redacted] by viewing the combined diff of your branch by selecting its commits, or selecting one branch tip then another to see the “interdiff”. If this is something else I’d be interested to add support though.
Yeah the TUI ecosystem is much more built out than GUIs for JJ so far. TUIs are great too - over time I've personally found myself ~50% in the CLI and ~50% in some sort of GUI, but a TUI can fill both gaps pretty nicely. I have always hated viewing diffs in the terminal though.
Odd okay, I got a separate report on Linux that the app crashes on startup. So for you, it launches fine, but the Select a repo button does nothing? This may sound weird, but have you waited ~1min to see if a file picker pops up?
It was really great! The thing is, kind of like the Flutter ecosystem, it's as good as its third party libraries. This app is mostly a very advanced UI on top of the CLI with not a whole lot else - I think maybe the only 3rd party library I'm using is the file picker. But for other apps that might have more complex use cases there may be more of a struggle there.
Totally understandable! I'm a solo dev out of LA who's mostly consulted on various apps and sites over the years, and out of my own personal need I began working on the project earlier this year.
The project is indeed closed-source. Personally I've had experiences where I found the project moved forward towards the roadmap much faster - specifically when considering smaller-scale projects that are UI heavy and therefore take in many UI changes in parallel - when I instead implemented feature requests and focused most of my energy on listening to user feedback rather than reviewing code. UI polish can be quite tricky when it comes from many sources :)
That said, nothing is set in stone! If at some point there are enough feature requests to where I am the bottleneck, this is something I would revisit.
Really solid write-up — it’s rare to see someone break down the real tradeoffs of scaling RAG beyond the toy examples. The bit about reranking and chunking actually saving more than fancy LLM tricks hits home to me.