Local AI · Earlier experiment
Local models with LM Studio and AnythingLLM
I connected LM Studio and AnythingLLM on my M1 Max to explore local models and working with my own documents. The practical limits were as useful to learn as the possibilities.
In spring 2026, I spent time trying local AI on my MacBook Pro. I used LM Studio to run models and AnythingLLM as the interface, connecting the two so I could chat and explore using documents as context.
There was more to think about than which model had the best reputation. My Mac has 32GB of memory, and I had to leave room for the conversation and the rest of the computer. I explored different model formats and sizes, and learned that getting a model to load was only the beginning.
I also explored document retrieval: giving the assistant relevant material to search when answering. That led me towards choosing a small set of useful documents instead of putting my entire Obsidian vault into the system. This was an experiment, rather than a finished service searching everything I own.
I eventually simplified the setup. The local-model work gave me a better understanding of how the pieces fit together, while my Raspberry Pi and Hermes became the more lasting home for my personal assistant.
April 2026 setup record; 8 August 2026 retrospective capability profile
What came out of it
Connected a local model runner and document interface, explored model/memory trade-offs, then simplified the stack.