Obsidian Integration

Markdown knowledge-base app with a plugin ecosystem that can connect notes to AI assistants, embeddings, and local workflows.

Start local setup ↓Check model fit

Beginner to intermediateSetup

PartialLocal runtime

3Related paths

Recommended path

Obsidian local AI quick start

1. Install a community AI plugin deliberatelyInstall Obsidian Copilot or another reviewed community plugin, read its permissions and release notes, and keep a vault backup before enabling model-driven actions.
2. Connect and test the local chat providerIn Copilot BYOK settings, choose Ollama or LM Studio, enter the local /v1 base URL, and use Test. Enable the documented CORS option if chat is blocked after a successful test.
3. Configure retrieval as a separate model jobChoose an embedding model for vault search, rebuild the index when that model changes, and test retrieval before asking a chat model to answer from the returned notes.

Open official setup docs Check what fits your hardware

Best for

Individuals building personal knowledge bases, writing workflows, and local-first research systems.

Model support

AI in Obsidian comes from community plugins rather than the app. Copilot supports bring-your-own-key with a provider dropdown that names Ollama and LM Studio alongside custom OpenAI-compatible endpoints. Smart Connections ships a local embedding model and runs its semantic search privately and offline by default.

Choose models by role

Vault chat and writing

Instruction-following chat model

Use a chat model for drafting, summarising, and answers over retrieved notes. It should be sized for the note context the plugin sends.

Semantic vault search

Multilingual E5 Large

Use a dedicated embedding model to index and retrieve notes. Changing embedding models requires rebuilding the index; this model does not write the final response.

Open the model profile →

How to use this integration

Ask questions across a private vault with a model that never sends notes to a vendor
Run semantic search over years of notes using an embedding model on your own machine
Draft and refactor long-form writing locally, keeping unpublished work off third-party servers
Compare local and hosted models against the same vault from one settings panel

Connecting a local model

In Copilot, open Settings, Copilot, BYOK, choose the provider and set the base URL: http://localhost:11434/v1 for Ollama or http://localhost:1234/v1 for LM Studio, then use the Test button. If Test succeeds but chat still fails, enable the Enable CORS toggle, which is the documented fix for browser-origin restrictions.

Tradeoffs

The CORS toggle catches many local setups: a passing connection test does not guarantee that chat can make the browser-origin request. Copilot now documents mobile support, but mobile capability still depends on the plugin version, provider reachability and whether a feature needs a process that only exists on the desktop. Some multi-agent and project-context features need a paid subscription. Smart Connections is source-available under its own licence rather than OSI open source, with paid Pro features, and semantic search ranks meaning rather than exact words.

Source

Obsidian Copilot documentation