Build on top of MetaGPT

[djaym7]

trafficstars

To save significant effort, we should consider building on top of metaGPT:

https://github.com/geekan/MetaGPT/tree/main

[geekan]

Hello, I am the author of MetaGPT, and I will participate in this work, especially the performance optimization of SWE-bench. Yesterday, I wrote a technical route for this dataset, covering the technical points of their technical report today, and also covering the technical points that they did not report.

The truth is, it's not difficult to get there, but it does take some work. Anyone who wants to participate in the performance optimization part is more than welcome.

[z-zeechung]

@geekan i'm currently working on llm apis, what interface i should prepare?

[mach-12]

@geekan Would be definitely up for that. Where is this conversation from, and how can I contribute?

[geekan]

@z-zeechung @mach-12 If you are interested in the current work, you are very welcome to join. We hope to surpass davin's SWE-bench score in the coming weeks. At present, we have a team starting to tackle the problem.

Anyone wants to join can send some basic information to my email, including: name, background, relevant experience, part you want to contribute, and time commitment (in the next 2 weeks). We will review it. For members who successfully join, we will open detailed ideas and documents related to this issue.

As for llm configuration and standardized interaction, you may refer to example and related config

If anyone is interested, you can find me at [email protected]

[JustinLin610]

I'll close this as we are marching fast with community efforts. Reopen this for discussion of metagpt for this project.