Agents are going to be constantly changing. Hopefully, new ones will be added all the time. I could see an agent hub being created in the future. Currently, when you look at examples and see a fix or new config value for an agent, it's tough to find the version which the change was included in - if it was included at all. I found a feature in the examples that isn't a part of a release. That can be very frustrating for someone just learning about the project.
IMO agents should each have their own version and each release of LangStream should include the versions of each agent in the change notes. That creates a clear, documentable way for someone to understand what each agent supports in the given LangStream version.
I think it's better to keep a mono-repo. A multi-repo will quickly lead to a dependency hell.
We should find other ways to solve the problems you describe.
Agents are going to be constantly changing. Hopefully, new ones will be added all the time. I could see an agent hub being created in the future. Currently, when you look at examples and see a fix or new config value for an agent, it's tough to find the version which the change was included in - if it was included at all. I found a feature in the examples that isn't a part of a release. That can be very frustrating for someone just learning about the project.
IMO agents should each have their own version and each release of LangStream should include the versions of each agent in the change notes. That creates a clear, documentable way for someone to understand what each agent supports in the given LangStream version.