Comment on [AIT] what is your selfhosted local LLM stack?
StrawberryPigtails@discuss.tchncs.de 2 days ago
Currently, I’m using Hermes and OpenCode mostly, backed by Ollama running a couple of different models. Usually I’m running the Gemma4 or Gwen3.5 models.
Memory management is done through Obsidian as every harness seems to have really good skills for it.
NormDeplume@lemmy.world 12 hours ago
Mind elaborating how you route them through obsidian?
StrawberryPigtails@discuss.tchncs.de 10 hours ago
My Hermes agent and OpenCode both can see the same Obsidian vault. I just started working with OpenCode and haven’t quite figured it out, however my Hermes agent has learned that everything I work on needs a file in Obsidian and that if it needs to look something up that it should look through Obsidian first. Prompts usually tell them to “look in the vault” or “write up what we’ve been doing in the vault, suitable for another agent to understand what is going on with no prior context”.
It’s not currently all that reliable, but I’ve only been working with Hermes for a few months and started playing with OpenCode a few weeks ago. I need to work a bit more on orginization, though. Last time I looked in there, I backed out quickly as it was a god awful mess.