Use Obsidian with a local LLM, without sending notes to the cloud
Copilot for Obsidian can talk to the model on your GPU. One command sets it up and tells you which settings would still send notes to a cloud service.
Set it up
- In Obsidian, install and enable the Copilot community plugin.
- Start the model:
localllm serve - Run
localllm connect obsidian --path "PATH/TO/YOUR/VAULT", then reload Obsidian.
The local model becomes Copilot's default chat model; your other models stay in the list.
--undo --path ... restores the plugin's previous settings.
Check what still leaves your PC
The command lists two things to look at: a cloud embedding model for vault search (building the index sends chunks of your notes to it), and cloud chat models still enabled in the menu. Turn those off in Copilot's settings if nothing should leave the PC.
Local search over your notes
For vault search, run a small local embedding model next to the chat model. localllm already downloads one for its router, multilingual-e5-small (126 MB, about 100 languages). The exact commands are in the API docs.