Mindmap AI

Connect notes and ideas with local AI. Discover meaningful connections while you write, read, and explore your vault.

• Generate summaries, tags, and concepts from your notes.
• Explore related notes in a sidebar with a connection graph.
• Search your indexed notes with natural-language questions.
• Import Apple Books highlights and connect them to your ideas.
• Research notes and highlights with optional web research.

Mindmap is desktop-only and uses Ollama for note processing. Apple Books import and web research require macOS.

Requirements

• Obsidian Desktop 1.7.2 or newer.
• Ollama running with an embedding model and a text-generation model.

Install

1. Open Settings → Community plugins → Browse, find Mindmap AI, and install and enable it.
2. Install the default models, or choose models you already use:

ollama pull embeddinggemma-2:270m
ollama pull llama3.1:8b

First Run

1. In Mindmap settings, open Local AI, confirm your Ollama address and models, then choose Test draft → Apply.
2. Under Included folders, choose which folders Mindmap can process and select Save folders.
3. Choose Check setup, resolve any reported issues, then run Run Mindmap (all scopes) from the command palette to process your notes.

Everyday use

Open the Mindmap sidebar to explore connections or search across indexed notes. Connections refresh automatically; saved active notes are prioritized after typing settles, without interrupting your writing.

Summaries are expanded by default. Use a note's collapse button to toggle its summary, or Alt/Option-click it to expand or collapse all visible summaries.

Use the status menu to pause or resume processing. Under Automation, choose manual runs, periodic processing while Obsidian is open, or scheduled processing on macOS. Turn off Real-time summaries and tags to keep the local language model out of your writing sessions: the active note is still embedded for live connections right away, and summaries, tags, and concepts are added on the next scheduled or manual run. Connections shown sets how many connections (1 to 20) appear per note.

Summaries, tags, concepts, and related-note links are stored in note properties. To keep them out of the reading area, choose Obsidian Settings → Editor → Properties in document → Hidden. This hides the properties without deleting them.

Apple Books and research

Use the Apple Books import toggle in settings or the status menu to bring highlights into your vault as readable annotation notes. The first import lets you choose whether to process existing highlights; new highlights are processed automatically while import is enabled and Obsidian is open.

The separate Internet research toggle is off unless you enable it with consent. Enabling starts in Manual mode for selected text or an active note; choose Automatic for Apple Books in settings for eligible highlights. Research results for highlights are saved in separate companion notes, keeping the original annotations intact.

Privacy

• Note processing uses your configured Ollama server, which defaults to your own machine.
• Web research requires consent and sends derived search queries to Exa, not raw notes or highlights. The Exa API key is stored in macOS Keychain.
• Apple Books access is read-only; Mindmap never modifies the source database.
• Copied diagnostics redact credentials and other secret-shaped values.

Troubleshooting

Open Advanced and troubleshooting → Check setup for actionable checks. Confirm Ollama is running, both models are available, and your included folders are saved. Use Copy diagnostics when reporting a problem.

If an upgrade requires a migration, follow the status menu's guidance. Retry migration resumes interrupted progress. Applying a different embedding model or dimension automatically re-embeds notes into a staged index. The prior generation remains on disk until the replacement is verified and activated; ordinary processing waits so embedding spaces cannot mix. A failure leaves the prior generation intact and offers a retry. Run Mindmap full rebuild (all notes) compacts existing vectors; it does not change embedding models.

Technical details

Mindmap is written in TypeScript and bundled into the plugin. Runtime data stays under .obsidian/plugins/mindmap-ai/data/. Apple Books uses /usr/bin/sqlite3; Keychain access uses /usr/bin/security. The optional macOS LaunchAgent uses /bin/launchctl and /usr/bin/open to open the vault, not process notes while Obsidian is closed. These integrations use fixed arguments with shell: false.

Plugin version and compatibility are defined in manifest.json and versions.json. See Contributing for development instructions and Changelog for release history.