It seems RAG is one of their main purpose. I'm looking to do my first local RAG project and am looking for suitable 4b and 8b models. Also, which of the LLM benchmarks are important when considering the RAG application?
Just started using oMLX. Its great! But so far I’m serving it to my coding agents. I tried its Chat panel, but it doesn’t seem to do web search. Is it in the settings (that I might have missed) or not supported at all? If not supported, what app y’all are using for chat conversations?!
With local LLMs, space and power is a constraint, hence Pi is the least token hungry and hence seems to be the fastest by far (only behind IDE based tools such as Continue.dev). Hermes Agent is really appealing because of its self learning aspect. The more work initially but would pay off soon as the agent knows your style and preferences. So, for those who are knowingly choosing Open Code instead of these two, why?
My use case is scientific computing, BTW. M4 Pro with 48GB (recently bought. Wishing had gotten M5 with 128GB instead 😫)
I love how lightweight Pi is and have been using it for weeks. However, recently I've been experimenting with Hermes Agent (as a purely coding agent), and I really appreciate its self-improvement framework. I am not a dev, and my use case is mainly for scientific data analysis for my own domain, so I really appreciate the agent learning new skills catered to my workflow. I am wondering if the Pi extensions, such as persistent-memory, or total-recall, etc., get it to be on par with Hermes in this aspect?
I find Hermes an excellent chat assistant (similar to Claude Desktop/LM Studio/OpenWebUI, rather than the Messaging apps), but I've been using it in CLI. I am seeing YT videos on the new Hermes Desktop- but apparently, there is more than one? Also, Nous Research has its own UI 'Hermes Workspace'. Which one do y'all recommend?