Being into C and C++ my self. I find this very impressive. Though, The only thing is that when I pressed my arrow keys to try get the previous command or prompt it entered the key code for the key press. Then if I entered that it thought I was talking about the particle theory. But yeah. Besides that. It is a genuinly good tool and you should be proud of it.
Thank you so much for trying out my project! Also, thank you for your feedback! That happens because the program uses an alternate terminal buffer in order to make it look styled better, and in the alternate terminal buffer, those types of key presses (and mouse wheel scrolling) will be sent directly to the terminal, unless raw input mode is enabled. Doing that would require a lot of extra boilerplate code or an external TUI library.
You can absolutely contribute to the project! The way these types of hackathons work is a bit different from the normal types of hackathons, because your payout is based on a precise measurement of the hours in the project. And this hackathon ends in 4 days. It'll probably finish reviewing just after the hackathon concludes.
Well because i don't have a compiler currently I will just go through and find you all the librarys and stick in an issue or something with what to include them because I can be asked to find a compiler for windows.
I’m not very experienced with C++, but this was really impressive! 😸🙌
Building a Markov Chain model from scratch with just C++20 STL is already cool, but I especially liked seeing how you kept profiling and optimizing it as the project grew.
It really feels like the kind of project that teaches you much more than just syntax.
Great work! Keep building weird things like this. 🚀
For me, the lowest-level thing I made has to be this. This was the first time I've worked with bare C++, with no dependencies (like Qt). What was the lowest-level thing you've ever worked on before? If you've never done anything low-level, what was the highest-level thing you've done before?
Really impressive project! Building a configurable N-gram Markov Chain from scratch using C++20 is a great way to strengthen your understanding of algorithms, data structures, and performance optimization. I especially liked how you improved corpus loading and integrated static analysis into your development workflow. I'm curious, what was the biggest technical challenge you faced while optimizing the model? Would be great to connect and exchange ideas!
The biggest technical challenge I faced while optimizing the model was absolutely the loading of the tokens. There were just so many things to optimize. Near the start, I was using a vector and manually checking if tokens are already in it to prevent having duplicates, instead of just using an unordered set, which is a much faster implementation. I will probably also add a setting to preserve or not preserve the token frequency in a bit. And I was generating a whole new container of cleaned tokens every prompt, instead of just generating it once during loading. Another thing was that before I refactored (slightly), there was just one code file, which was 1,000 lines long. While technically not an optimization issue, it made it very hard to scroll to wherever I needed to edit in Neovim.
We're a place where coders share, stay up-to-date and grow their careers.
Being into C and C++ my self. I find this very impressive. Though, The only thing is that when I pressed my arrow keys to try get the previous command or prompt it entered the key code for the key press. Then if I entered that it thought I was talking about the particle theory. But yeah. Besides that. It is a genuinly good tool and you should be proud of it.
Thank you so much for trying out my project! Also, thank you for your feedback! That happens because the program uses an alternate terminal buffer in order to make it look styled better, and in the alternate terminal buffer, those types of key presses (and mouse wheel scrolling) will be sent directly to the terminal, unless raw input mode is enabled. Doing that would require a lot of extra boilerplate code or an external TUI library.
Is the project still a hackathon project. because if not. I will work on it.
You can absolutely contribute to the project! The way these types of hackathons work is a bit different from the normal types of hackathons, because your payout is based on a precise measurement of the hours in the project. And this hackathon ends in 4 days. It'll probably finish reviewing just after the hackathon concludes.
Well because i don't have a compiler currently I will just go through and find you all the librarys and stick in an issue or something with what to include them because I can be asked to find a compiler for windows.
A possible fix for that problem @tyleruploads , It might or might not work. github.com/tyleruploads/corpustein...
I’m not very experienced with C++, but this was really impressive! 😸🙌
Building a Markov Chain model from scratch with just C++20 STL is already cool, but I especially liked seeing how you kept profiling and optimizing it as the project grew.
It really feels like the kind of project that teaches you much more than just syntax.
Great work! Keep building weird things like this. 🚀
Thank you so much for the support!
For me, the lowest-level thing I made has to be this. This was the first time I've worked with bare C++, with no dependencies (like Qt). What was the lowest-level thing you've ever worked on before? If you've never done anything low-level, what was the highest-level thing you've done before?
Really impressive project! Building a configurable N-gram Markov Chain from scratch using C++20 is a great way to strengthen your understanding of algorithms, data structures, and performance optimization. I especially liked how you improved corpus loading and integrated static analysis into your development workflow. I'm curious, what was the biggest technical challenge you faced while optimizing the model? Would be great to connect and exchange ideas!
The biggest technical challenge I faced while optimizing the model was absolutely the loading of the tokens. There were just so many things to optimize. Near the start, I was using a vector and manually checking if tokens are already in it to prevent having duplicates, instead of just using an unordered set, which is a much faster implementation. I will probably also add a setting to preserve or not preserve the token frequency in a bit. And I was generating a whole new container of cleaned tokens every prompt, instead of just generating it once during loading. Another thing was that before I refactored (slightly), there was just one code file, which was 1,000 lines long. While technically not an optimization issue, it made it very hard to scroll to wherever I needed to edit in Neovim.