Mechanism for enabling, distributing, and utilizing tools. (janhq blessed and community contributed) #1360
wolfspyre
started this conversation in
Feature Requests
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Given the way Jan / Cortex is laid out... it feels .... like an intuitive step to have a consistent mechanism to empower models to utilize tools.
I can see a few ways of doing this.... (and I'm sure there are others that peeps smarter than me have)
Essentially what I propose is a mechanism to empower models within cortex to perform actions via tools.
These tools should be downloadable,
These tools should be durably associated with the community members that have released/contributed to them.
(time/cpu/memory/steps/tokens/data/etc)
obviously, not all metrics are gonna be equal .... there's a lot of handwavey here... I'm trying REALLY HARD not to get into specific nuance ;)
I can see there being:
It seems reasonable to me that combining some set of performance metrics and some adoption/usage/rating metrics, could go a long way towards dissuading abuse, whilst allowing a collaboratively inclusive ecosystem with low (hopefully?)ish moderation overhead.
I feel that by having a consistent way of providing tools to cortex, alongside a robust suite of privacy-first telemetry reporting mechanisms, cortex can continue to evolve into a robust wrapper for AI/ML engines as capabilities evolve.
The AI/ML landscape of today is evolving at an incredible pace, thus allowing for the concept of 'engine/model to ....' interactions to be plug-in driven feels like one way to remain fluid in this landscape of shifting sands...
...but maybe I'm missing something?
Thoughts?
Beta Was this translation helpful? Give feedback.
All reactions