Connecting a private Obsidian vault to Ollama on macOS

What they built

The author wanted AI summaries and answers about content in Obsidian, but the vault holds lots of private meeting notes, so connecting it to public services was out. They couldn't find an easy guide, so they wrote one.

They first installed Private GPT, then found the Obsidian plugins could talk to Ollama directly.

Setup

• macOS with Homebrew
• Ollama with llama3.1 and nomic-embed-text
• Obsidian plugins: Local GPT and AI Providers
• (Optional) Private GPT, which needs Python 3.11, not 3.13

Gotchas

• Private GPT turned out to be unnecessary for Obsidian
• Local GPT has no context menu out of the box; bind a hotkey to 'Local GPT: Show context menu'

Do it yourself

1. Install Ollama and pull the models
2. In AI Providers add an Ollama provider at http://127.0.0.1:11434 and pick llama3.1:latest (plus nomic-embed-text)
3. In Local GPT set the Main and Embedding providers
4. Bind a hotkey (author used OPT+CMD+G) to the Local GPT context menu
5. Run Ollama persistently with brew services

brew install ollama
ollama pull llama3.1
ollama pull nomic-embed-text
brew services start ollama