Contributions, development, participation and community

[AdamBajger]

I guess the amount of issues is rising, there is a decent amount of forks, and I feel it is a good time to start actively discussing what is going on. I feel like the first step is to move some of the topics discussed in the issues to here. Namely those issues which do not carry any bug report or feature request. I would love to see the issues a little bit more organised and I believe here is a more appropriate place for general discussion.
@Kuberwastaken hope you dont mind I have started this thread. I believe it might help clear up the issue list.

[Kuberwastaken]

@AdamBajger it was needed haha, appreciate it

Will start organizing tomorrow!

[Kuberwastaken]

Also, couldn't find a direct way to contact you but it would be great if you could reach out to me via X or Discord, would love to talk ! It's the same handle -> @ kuberwastaken

[AdamBajger]

cool. I dont really spend too much time on Discrod nor X, but I do have both. I will reach out on discord.
My Discord alias is unfortunatelyyourmom/3XT3NS3 not sure which is visible publicly.

[danihab1b]

Hi guys, probably a dumb/overnoob question:
is it feasible or desireable to bring this modelling type:
https://github.com/lucas-maes/le-wm
as an option? It seems that a self-assisted rust transpiller would benefit largelly from rhis type of architecture. What do you guys think?

[AdamBajger]

not sure how would you want to implement world modelling into a coding agent. Interesting read, tho.

[jhult]

It looks like @thamwangjun is doing some work in https://github.com/thamwangjun/clearest and https://github.com/thamwangjun/clearest-rust. It would be nice for that to get upstreamed back here (if possible and if it makes sense).

[thamwangjun]

@jhult Hi, thanks for looping me in.

I plan to work on my fork to scratch my itches first (bugs that I face - in the middle of fixing one bug), and also implement upstream claude code features (latest stable is 2.1.131 i think - I plan to do analysis of the binary to extract what features are added in new versions, spec them and integrate into claurst).

I am prioritizing these 2 for now, will be happy to prepare patches for upstream contribution.

[AdamBajger]

nice. I am also addressing some missing features and bugs, could not finish quickly enough, though. Just open PR when you are ready.

[watrworld]

Is this project dead? Am I the only one using it LOL, Im surprised not to see further progress

[Kuberwastaken]

@watrworld @danihab1b haha not really dead, I've admittedly been a bit slower, just busy with practicals and exams coming up for college :P

Would love to hear what you want - I wanna make 1.0.0 a bit grand with feature and release it through multiple channels, including npm (with the map file included, of course)

[danihab1b]

@Kuberwastaken thanks for coming back and replying! sorry if I said it in a bad way :(
I´m not a dev so I can only imagine feasibility of things from side studies, including architectures using rust. For example, I had a thought that for the 1.0.0 would be amazing to have, as an opt in, the possibility of using htmx+askama for a quick relatable and manageable IDE, besides of course the already working TUI. Thought that because most of the semi-self-coding environments that are coming to market in github comes as black boxes where people cannot mess with the way agents interact.
Don´t dare propose it as an Issue so I brought it here. What do you think?

[AdamBajger]

this would make for a challenging milestone. I believe we need to address more basic usability issues and some still missing features that are commonly present in other tools.

We can start planning more complex directions after these are addressed.

I would begin with absorbing knowledge from open-code and roo code, while keeping everything rust-native, polish open model support and cross-provider compatibility, while gathering proposals such as yours in the backlog.

[thamwangjun]

I agree with @AdamBajger, and personally I would want to keep the feature set as lean as possible and minimally feature parity with claude code (and claude code itself is a moving target).
It's best for a tool to do only 1 thing, and 1 thing well.

What attracted me to claurst was transparency, I was frustrated with Anthropic engineering introducing bugs repeatedly that harm their users (most harmful were cache breaking bugs that consumed usage limits in mere seconds for Max users).

[Kuberwastaken]

yep, too big a jump imo - we need to keep things manageable and directionally really good for what it is (pun unintended)

[AdamBajger]

Do we want GrowthBook integrated? I am running into some hard decision questions regarding how to wire it in and don't really understand anything about it. I have no real experience using it, I am led to believe that it requires additional infrastructure to be utilized properly and generally provides no value at this stage.

Would you mind if we removed GrowthBook completely?

Otherwise, I would love to get some feedback on how we want to keep and use it.

[Kuberwastaken]

I personally don't have much use for it - I'm not against removing it if it's not used