Stable upgrades #2738
Replies: 3 comments 13 replies
-
There are lot of other which contributed to this projects but thx you for your feedback. |
Beta Was this translation helpful? Give feedback.
-
Hey @rezzie-rich , thanks for rooting for us and the feedback! But actually the most recent agent has the highest quality according to benchmark scores of any of the agents yet, so if you try the most recent version (esp. with clause sonnet 3.5) it may be greatly improved! |
Beta Was this translation helpful? Give feedback.
-
@rezzie-rich Thank you for the feedback! I was initially a bit confused why you say 1.3 is still better, and I think the with-hint vs no-hint issue explains it. It really matters. As best as I can tell, I think it's very likely that currently 1.7 is better. The stuck issue is definitely a pain point, and we're working on it. If you can, it would help immensely if you use the 'thumbs up/down' button to send feedback to developers, when that happens for you. Because it sends us the history of that chat (don't use it if you don't want that sent), and the more real-life histories we have, the more it helps us detect the patterns of behavior that make CodeAct prone to it. Thus it helps fixing them. |
Beta Was this translation helpful? Give feedback.
-
I want to thank you guys for all the efforts and contributions. The project has come a long way since it started, and it's making history as we speak off. However, I can't help but notice the desire for adding new features over perfecting the existing ones.
The original codeact v1 and v1.3 is surely at its peak, but since then, there have been 3 other versions, and they not only perform less but also seem to get stuck in loops and generate false outputs.
Not being able to solve an issue is still better than getting stuck in loops or, worse, giving off wrong outputs as they not only cost ineffective computation but also dilute the code base, making the debugging a nightmare.
The project already has more than enough features to make it somewhat complete. I know open source projects dont pay the bills like a regular job, so the efforts that can be given are very limited. So instead of constantly adding more features ( maybe just one more -memgpt), it would be more effective to improve and perfect the existing ones. There's no point in having newer versions of codeact one after another when they all perform less than the original version. Efforts would be more effective if all the available time is spent identifying the issue and upgrading the versions steadily where each version is out performing the previous one otherwise it will make it even harder to pin point whats causing the newer versions to get stuck.
Taking the apple's developing approach rather than the window will not only improve the overall quality but also make it more reliable. That's the key difference between a production level product and an experimental project.
You guys are heroes for your contributions to the community and dying to see the day when knowing how to use opendevin like excel/powerpoint becomes a job skill.
@rbren @xingyaoww @xcodebuild @enyst @neubig @SmartManoj @iFurySt @PierrunoYT @tobitege
Beta Was this translation helpful? Give feedback.
All reactions