What is the state of the art right now regarding running local LLMS and connecting local agents to them?
Which models are suitable and good, what hardware (at a minimum) and software are required, and which open source agents are good to drive those models?