I was curious about what hermes is,
so I installed hermes on WSL and connected it to a local LLM, and I've tried using it several times today.
Agents are convenient,
but running them with a local LLM is quite challenging.
Even if I set the context to 128K, it seems like tokens will be consumed very quickly.
I immediately realized that this kind of thing can't be used with a local LLM.
Next time, I'll have to try using a free API.
Or maybe I should just endure the pain and save tokens.
My plan to add another GPU and use one for agent execution and the other for the necessary LLM execution so that I could run agents locally was scrapped after only one day.
For now, I've switched from a local LLM to a free LLM API (a kind of cloud that runs open LLMs).
I changed it from the local llm Gemma4-e4B free API GLM5.1 and made it work.
However, I don't know when the API key for this free API will expire.
And because it's free, there's a limit of 40 rpm. I wouldn't make more than 40 requests per minute anyway?
