AI Summary

AI Summary is an Obsidian plugin that watches notes with ai_summary: pending, inserts a generated summary and key takeaways at the top of the note body, and then marks the note ai_summary: generated.

Built and maintained by Lab829.

What it does

• Supports ai_summary: none, ai_summary: pending, and ai_summary: generated as explicit workflow states.
• Inserts the generated block directly below YAML frontmatter so the original note body stays intact.
• Supports Gemini and OpenAI with your own API key.
• Lets you customize the summary and takeaway prompts for the existing output shape.
• Desktop only for v1.

Installation

Community Plugins

AI Summary is in the Obsidian community plugins directory. In Obsidian, open Settings -> Community plugins -> Browse, search for AI Summary, then install and enable it.

Manual install from GitHub releases

1. Download manifest.json, main.js, and styles.css from the latest release.
2. Copy them into <vault>/.obsidian/plugins/lab-ai-summary/.
3. Reload Obsidian and enable the plugin.

Setup

1. Open the plugin settings in Obsidian.
2. Choose Gemini or OpenAI.
3. Paste your API key.
4. Choose a model from the dropdown. If model discovery fails, enter the provider model ID manually.
5. Leave auto-process enabled if you want new pending notes to run automatically.

Workflow

Use this frontmatter to opt a note into enrichment:

---
ai_summary: pending
---

Successful processing inserts the generated block below frontmatter and flips the field to ai_summary: generated.

If you want a note to stay untouched until later, use ai_summary: none and change it to pending when you want enrichment to start.

Use with Obsidian Web Clipper

AI Summary works with the official Obsidian Web Clipper. Add one property to your clipper template, and every page you clip arrives as a pending note that the plugin summarizes.

1. Install Obsidian Web Clipper for your browser (Chrome, Firefox, Safari, Edge and others) from obsidian.md/clipper. The setup is the same in every browser.
2. Open the clipper's settings and select the template you use, for example Default.
3. Under Properties, add a property named ai_summary with the value pending.
4. Keep the template's behavior set to create a new note.
5. Clip a page. Once the note is in your vault and Obsidian is open on desktop, AI Summary adds the summary and takeaways and sets the property to generated.

Auto-process must stay on in the plugin settings. If Obsidian is closed when you clip, pending notes are processed the next time it starts.

Disclosures

Network use

This plugin makes network requests only to the AI provider you choose in settings, and only in these cases:

A note is summarized when it is created or edited while it has ai_summary: pending (after a short debounce), and once when the plugin loads for any notes that are still pending, as long as "Auto-process new notes" is on. Notes without ai_summary: pending are never sent.

What is sent: the body of the pending note (everything below the YAML frontmatter, excluding any previously generated summary block), your summary and takeaway prompts, and the model name. Frontmatter and other notes are not sent. Your use of the provider is governed by that provider's own terms and privacy policy.

Your own API key

• You must supply your own Gemini or OpenAI API key. The plugin does not provide one and does not require an account with Lab829.
• The key is stored in this plugin's settings file in your vault (<vault>/.obsidian/plugins/lab-ai-summary/data.json) in plain text, like other Obsidian plugin settings. Do not sync or share that file if you don't want to share the key.
• The key is sent only to the provider you selected: as a Bearer authorization header for OpenAI, and as the key query parameter for Gemini (Google's documented API format). It is never sent to Lab829 or any other server.

Paid features

The plugin is free and MIT licensed, and works with your own API key. Lab829 may offer an optional paid hosted service in the future (for example managed AI models). If that happens, it will be opt-in and clearly labeled, and using your own key will keep working without an account.

Privacy

• No telemetry, analytics, or usage tracking. The plugin does not contact Lab829 servers.
• No ads.
• Requests go directly from Obsidian to your chosen provider; nothing is proxied.
• The source code is not obfuscated and is MIT licensed.

Development

pnpm install
pnpm lint
pnpm type-check
pnpm test
pnpm build

If you want to test the built plugin locally inside a vault:

1. Copy .env.local.example to .env.local.
2. Set OBSIDIAN_PLUGIN_DIR to <vault>/.obsidian/plugins/lab-ai-summary/.
3. Run pnpm build:sync.

Contributing

Contributions are welcome. See CONTRIBUTING.md. First-time contributors are asked to sign a short Contributor License Agreement by commenting on their pull request.

Support

• Open a GitHub issue for bugs or feature requests.
• Include your Obsidian version, plugin version, provider, and any visible notice text when reporting a bug.

License

MIT © 2026 Lab829