Smart Note Agent

An agentic AI assistant plugin for Obsidian that can read and — with your approval — modify vault notes, powered by your choice of LLM provider.

中文文档

Features

• Agentic chat — multi-turn conversations with an autonomous tool-calling loop
• Vault tools — full-text search, read notes, list folders, follow backlinks and outgoing links, get the active note and current selection
• Edit mode — create, edit (full replace or unified patch), delete, and move notes; every change is shown as a diff for your review before it is committed
• Multiple providers — OpenAI, Anthropic, DeepSeek, Qwen (Alibaba), Kimi (Moonshot), Zhipu (GLM), MiniMax, OpenRouter, Ollama (local), and any custom OpenAI- or Anthropic-compatible endpoint
• Three modes
  • Ask — read-only; for Q&A and research without touching your vault
  • Edit — full write access; all changes require your approval via a diff UI
  • Scheduled — automated background runs (daily summary, weekly review) with restricted write access
• Scheduled tasks — daily summaries and weekly reviews written automatically to configurable folders
• Auto-compaction — conversation history is compacted transparently when approaching the model's context limit
• Per-provider profiles — separate API key, base URL, and model saved per provider
• User profile — optional personal description injected into every system prompt
• i18n — English and Simplified Chinese UI, auto-detected from Obsidian's language setting

Installation

Via BRAT (recommended for pre-release)

1. Install the BRAT community plugin
2. In BRAT settings, add: Bin-Home/obsidian-note-agent
3. Enable the plugin in Settings → Community plugins

Manual

1. Download main.js, manifest.json, and styles.css from the latest release
2. Create the folder <your-vault>/.obsidian/plugins/smart-note-agent/
3. Copy the three files into that folder
4. Reload Obsidian and enable the plugin in Settings → Community plugins

Setup

1. Open Settings → Smart Note Agent
2. Select your LLM provider and enter your API key
3. Optionally set a custom base URL (for self-hosted or proxy endpoints) and model name
4. Choose a mode: Ask for read-only access, Edit for write access

Usage

• Click the bot icon in the left ribbon, or run Open Note Agent from the command palette
• Type your message and press Enter (or Shift+Enter for a new line)
• In Edit mode, the agent proposes note changes as unified diffs — approve or reject each one before it is written to disk
• Use New Chat to start a fresh conversation; previous conversations are saved and accessible via the history panel

Supported Providers

Provider	Notes
OpenAI	GPT-4o, GPT-4o-mini, o1, o3, etc.
Anthropic	Claude 3.5 / 4 series
DeepSeek	deepseek-v3, deepseek-r1
Qwen	Alibaba Cloud Dashscope (qwen-plus, qwen-max, etc.)
Kimi	Moonshot AI (moonshot-v1 series)
Zhipu	GLM-4 series
MiniMax	MiniMax-Text series
OpenRouter	Any model via openrouter.ai
Ollama	Local models — Llama, Mistral, Qwen, etc.
Custom	Any OpenAI-compatible or Anthropic-compatible endpoint

Disclosures

Account required

Using a remote LLM provider (OpenAI, Anthropic, DeepSeek, Qwen, Kimi, Zhipu, Z.ai, MiniMax, OpenRouter) requires an account with that provider and a valid API key. Local providers (Ollama, LM Studio) and custom self-hosted endpoints require no account.

Payment may be required

Remote LLM providers charge for API usage. You are billed directly by the provider you choose — this plugin has no subscription or in-app purchase of its own. Local providers (Ollama, LM Studio) are free.

Network use

When you send a message, the plugin transmits your message text and relevant vault context to the LLM provider you have configured. No data is sent to any service by default — the plugin is inert until you supply an API key and send a message. The following remote services may be contacted, depending on your provider selection:

Provider	Endpoint
OpenAI	https://api.openai.com
Anthropic	https://api.anthropic.com
DeepSeek	https://api.deepseek.com
Qwen (Alibaba Cloud)	https://dashscope.aliyuncs.com
Kimi (Moonshot AI)	https://api.moonshot.cn
Zhipu / Z.ai	https://open.bigmodel.cn, https://open.z.ai
MiniMax	https://api.minimax.chat
OpenRouter	https://openrouter.ai

Ollama and LM Studio communicate only with localhost — no data leaves your machine.

Development

npm install
npm run dev       # esbuild watch mode — rebuilds on save
npm run build     # tsc type-check + production bundle
npm test          # unit tests (Vitest)

For local testing, symlink or copy the dist/ folder into your vault:

<vault>/.obsidian/plugins/smart-note-agent/

The plugin artifacts (main.js, manifest.json, styles.css) are output to dist/ after each build.

License

MIT