RE:
You are viewing a single comment's thread:
The main issue with having Hermes do it is that I work with only free inference providers, so I have to make sure not to make too many requests within too short a time, or I get rate limited, which is partly why I try to stay focused on one thing at a time, even though Hermes itself could do many things simultaneously. It's been going pretty well over the last few days at least. 😁🙏💚✨🤙
!ALIVE
!BBH
!HEARTBEAT
!PIXY
!UNICOIN
0
0
0.000
I thought that you use your Hermes agentic AI locally as much as possible !INDEED. 🤔🤯 More local generally means more freedom to use it, though you would be highly responsible for the hardware. 🧘♂️🤯🤓
!MMB !PIZZA !WINEX !HOPE
Yes, I was exclusively using local models for a while, but getting the larger models to run for extended periods was a bit of a problem, so I started using a few free LLM-inference providers, mostly OpenCode (that sadly just discontinued their free tier) and Nvidia, which is my new go-to, as it has my three favorite models, GLM5.3-Flash, Kimi-K3, and their own Nemotron-3-Ultra, all of which are very capable models with which I've made massive amounts of progress within very short amounts of time. I am still running local models, but they have specialized tasks now. 😁🙏💚✨🤙
!ALIVE
!BBH
!HEARTBEAT
!MMB
!PIZZA
We can say that both cloud-based and local models should complement each other, where local models should be used for tasks that require heavy oversight (such as automated monetary transactions and local file management) and cloud-based models should be used for everyday online searches !INDEED. 🤔🧘♂️🤖🤓
!MMB !BRN !STRIDE !HOPE
Yes, indeed, that's very true, though because I needed solidly-capable models to connect to Hive, and to build and refine my liquidity-pool agent, among other things, I've found it necessary to use cloud-based models for a bit of everything, so I haven't been able to maintain that sharp boundary. 😁🙏💚✨🤙
!ALIVE
!BBH
!MMB
!STRIDE
!ZOMBIE