Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

How to extend functions when users define agents in the new version #114

Open
wugren opened this issue Jan 15, 2024 · 6 comments
Open

How to extend functions when users define agents in the new version #114

wugren opened this issue Jan 15, 2024 · 6 comments

Comments

@wugren
Copy link
Contributor

wugren commented Jan 15, 2024

In the old version, Agent can define environment.py to extend functions, but it seems that the relevant logic is not seen in the new version.

@waterflier
Copy link
Collaborator

In the current version, the logic for extending functions has undergone a fundamental change. The functions expanded by the Agent are usually composed of underlying functions, while the expansion of native functions typically implies an increase in system's native capabilities, which will be more cautious (involving permission and privacy control). In the future, extensions will be made through the Plugin system.

@waterflier
Copy link
Collaborator

You can read this : https://github.com/fiatrete/OpenDAN-Personal-AI-OS/blob/MVP/doc/function_list.md for more about "System AI Functions"

@wugren
Copy link
Contributor Author

wugren commented Jan 15, 2024

In the current version, the logic for extending functions has undergone a fundamental change. The functions expanded by the Agent are usually composed of underlying functions, while the expansion of native functions typically implies an increase in system's native capabilities, which will be more cautious (involving permission and privacy control). In the future, extensions will be made through the Plugin system.

Does composed of underlying functions mean letting AI generate the code that calls the underlying function? I'm still a little worried about the current AI's ability to generate code. It may not be able to generate it correctly.

@waterflier
Copy link
Collaborator

It is definitely dangerous to rely entirely on AI's ability to write code, so we can divide Agents into two categories: one is an Agent that only uses existing functions in combination, and the other is an Agent that allows it to write code when necessary. I am very much looking forward to seeing the comparison of the capabilities of these two agents in real scenarios.

In fact, in the past 8 weeks of development, I have felt great uncertainty in high-dimensional LLM calls, and I am thinking about how to share these experiences with everyone.

@wugren
Copy link
Contributor Author

wugren commented Jan 17, 2024

I don't expect to use AI to generate a lot of code, I just followed your reply above. My question is, as an agent developer, what should you do when you need to expand functions?
ps:the Plugin system you mentioned has not been implemented yet.

@waterflier
Copy link
Collaborator

We've also upgraded the Function and Action capabilities in version 0.5.2. I will be drafting a detailed document later to introduce these enhancements. I believe this will address most of your concerns and provide deeper insights into the improvements made in this release.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants