Archivist AI - Obsidian Plugin

A RAG-powered AI assistant for your Obsidian vault. Chat with context from your notes, find connections, generate content, and more.

Features

Chat & Search

• RAG Chat - AI responses grounded in your vault content
• Semantic Search - Find notes by meaning, not just keywords
• Smart Connections - Discover related notes automatically
• Source Preview - Review and edit sources before sending

Writing Assistance

• Summarize Note - Generate concise summaries of any note
• Continue Writing - AI continues from your cursor position
• Expand Selection - Add detail to selected text
• Simplify Selection - Make text clearer and easier to understand
• Fix Grammar - Correct grammar and spelling errors

Organization

• Suggest Tags - AI-suggested tags based on content
• Suggest Backlinks - Find notes you should link to
• Find Duplicates - Discover semantically similar notes
• Find Orphans - Find notes with no connections

Generation

• Generate Note - Create new notes from a topic
• Generate Flashcards - Create Q&A cards from any note
• Generate Outline - Create structured outlines
• Save Chat History - Export conversations as notes

Providers

• OpenAI - GPT-5 family models
• Google Gemini - Free tier available
• Anthropic Claude - High quality responses
• Groq - Fast inference, free tier

Installation

From Community Plugins

1. Open Settings > Community Plugins
2. Search for "Archivist AI"
3. Install and enable

Manual Installation

1. Download the latest release
2. Extract to .obsidian/plugins/obsidian-scribe
3. Enable in Settings > Community Plugins

From Source

git clone https://github.com/MitchellTLarsen/obsidian-scribe.git
cd obsidian-scribe
npm install
npm run build

Copy main.js, manifest.json, and styles.css to your vault's .obsidian/plugins/obsidian-scribe folder.

Setup

1. Open Settings > Archivist AI
2. Add at least one API key:
   • OpenAI - Required for embeddings and GPT models
   • Gemini - Free tier available
   • Claude - High quality responses
   • Groq - Fast inference, free tier
3. Click "Index Vault" to create embeddings
4. Open Archivist AI from the ribbon icon

Usage

Commands

Access all commands via Command Palette (Cmd/Ctrl+P) and type "Scribe":

Command	Description
Open Archivist AI Chat	Open the chat panel
Open Scribe Connections	View related notes
Open Semantic Search	Search by meaning
Index vault for RAG	Re-index all notes
Summarize current note	Generate note summary
Summarize selection	Summarize selected text
Continue writing	AI continues from cursor
Expand selection	Add detail to selection
Simplify selection	Make text clearer
Fix grammar	Correct errors
Suggest tags	AI-suggested tags
Suggest backlinks	Find related notes to link
Find duplicates	Find similar notes
Find orphan notes	Find unconnected notes
Generate flashcards	Create Q&A cards
Generate note	Create note from topic
Generate outline	Create structured outline
Save chat history	Export chat as note

Ribbon Icons

• Chat (speech bubble) - Open Scribe Chat
• Connections (branch) - View note connections
• Search (magnifying glass) - Semantic search

Context Menu

Right-click in the editor to access:

• Summarize selection
• Expand selection
• Simplify selection
• Fix grammar
• Continue writing
• Suggest tags

Settings

Setting	Description
API Keys	Add keys for OpenAI, Gemini, Claude, Groq
Default Provider	Which AI to use by default
Models	Select model for each provider
Include Folders	Only index these folders
Excluded Files	Don't index these files
Context Size	Number of chunks to include (1-100)
Show Sources	Display source links in responses

How It Works

1. Indexing - Splits notes into chunks, creates embeddings via OpenAI
2. Search - Converts your query to an embedding, finds similar chunks
3. Context - Top matching chunks are included in the AI prompt
4. Generation - AI responds based on your notes

Cost Reference

Model	Input $/1M tokens
gpt-5-nano	$0.05
gpt-5-mini	$0.25
gpt-4o-mini	$0.15
text-embedding-3-small	$0.02
gemini-2.0-flash	FREE
groq (llama)	FREE

Privacy

• API keys are stored locally in your vault
• Your notes are sent to AI providers only when you use features
• Embeddings are cached locally for performance
• No data is sent to third parties beyond the AI providers you configure

License

MIT

Support

• GitHub Issues
• Documentation