Replies: 1 comment
|
Hello,
Now the important part:
This way, every time I open a new session, I just say for example let's continue working on X project. Everything gets read and picked up and ready to go Here, I've attached my project and skills management skill. I would argue they could be/should be part of odysseus itself for a better management of skills, memories and projects for users. |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
(yes, I had AI write it all. But I did some review on it.)
Hey Felix — I have a somewhat abstract architecture idea that I thought might be relevant to the kind of local/self-hosted agent system you're building with Odysseus.
I want to be clear up front: I don't have a working implementation yet, and I'm not presenting this as a finished solution or asking Odysseus to integrate it.
The idea is a Persistent Context Manager, with a task-oriented component I call Task Maxer.
The basic premise is:
Models/workers are temporary; task state should be persistent.
Instead of making an agent reconstruct a long project from its entire conversation history, a persistent task record would maintain things such as:
Different workers could then read only the state relevant to their assignment. A worker could disappear, be replaced by another model, or be upgraded without the replacement having to reconstruct the whole conversation.
I'm also thinking of extending this into a broader persistent context system:
User/Profile Context → Memory → Task State → Projects/Artifacts → Receipts/Audit
with per-user application-managed storage, worker-specific state views, and an exportable user-owned data layer. A local deployment could potentially put the persistent store on a NAS while clients/models remain on different machines.
The other reason I find it interesting is that it connects several agent patterns: specialist workers, multiple instances of the same specialist, sibling/critic workers, verification, temporary teams, model replacement, and long-running projects.
I've written the architecture out in more detail in the attached document.
(I have a friend who introduced me to coding and i thought might do some thinking about my-to-be project in advance. It has many ideas which currently exists. They just needs to be re-arranged. So, if it's not clear enough. I DON'T KNOW HOW TO ACTUALLY CODE. I am a bachelor, currently doing bachelors in both business administration and law, at the same time. With amateur future interest in coding and AI.
ZyroG2i_Master_Notes_vNext.docx )
No pressure to respond — I just thought the overlap with Odysseus was interesting.
@pewdiepie-archdaemon
All reactions