The documentation is good. Can you provide a score?

[Deng-Xian-Sheng]

I can't use your question template, sorry.

The documentation is good. Can you provide a score?

Openhands is something similar to Devin, but it is open source.

Can your project be exported with OpenAI API?

Or, Openhands has used something like a "test set" to test performance.

This represents the degree of completion of LLM calling tools or LLM independently completing programming tasks.

It was 19% before, and then the score reached 30%. I don't know if it has reached 50% now, but it shouldn't.

I want to know if you can use such a test to further prove the performance improvement of your framework for LLM?

In short, does your project improve LLM's ability to complete programming tasks by itself?

I looked through the documentation and didn't see the core things for the time being, such as how various "memories" interact with LLM, is it called by Tools? Or through other methods.

I also didn't see the generation and query methods of various memories.

This, I guess, can be controlled by the user.

It looks very interesting.

Another point I am more concerned about is what is the context of LLM? When using this framework.

I think that when using the framework, it is necessary to see the content of each request sent to LLM, which is crucial to understanding the principles of the framework.

Some offensive requests, this is quite interesting, but as you know, there are a lot of Agent frameworks, yours should be more special and innovative.

There are also some frameworks that are less customized, but completely GUI, such as dify.

There are also some memGPT or similar things. I have a deep understanding of memGPT, which is not innovative.

Some people let LLM automatically formulate agent processes.

There are too many, and it is impossible for all frameworks to become standards.

How effective is your framework in actual use? Have you built a chat to play with as a demo? It's very good.

Since you all compare it with chatgpt o1, having a demo that can be experienced online will greatly expand the influence. ! !