CLI personal knowledge base built on seekdb, featuring page management, hybrid search, timelines, tags, import/export, and MCP Server.

• Knowledge Graph Visualization - Interactive 3D graph with entity relationships, labels, and Ask Ex-Brain Q&A
• Intelligent Compilation - Semantic analysis with smart Compiled Truth updates
• Timeline Management - Automatic event extraction and history tracking
• Hybrid Search - Full-text search + vector semantic queries
• Entity Linking - Auto-detect entities and create linked pages

We recommend MarkSnip for data collection:

• One-click web clipping to Markdown format
• Supports code blocks, tables, math formulas
• Local processing, privacy-friendly
• Obsidian integration support

Use with ex-brain:

# After clipping with MarkSnip, import to knowledge base
cat article.md | ebrain put articles/slug --stdin

# Or import HTML directly; ex-brain extracts the readable article and converts it to Markdown
ebrain put articles/html-page --file article.html
cat article.html | ebrain put articles/html-page --stdin --format html

# Or intelligent compilation
ebrain compile companies/river-ai --file article.md --source web_clip

# Global installation (requires Bun or Node.js)
bun install -g ex-brain
# or
npm install -g ex-brain

ebrain --help

# Initialize (creates ~/.ebrain/data/ebrain.db automatically)
ebrain init

# Check active config
ebrain config

Configure LLM and embeddings in ~/.ebrain/settings.json if you want semantic search, AI compilation, and Ask Ex-Brain Q&A. Local hash embeddings work out of the box for lightweight use.

# Import Markdown or HTML directly
ebrain put my/note --file note.md
ebrain put articles/page --file article.html

# Pipe long content
cat article.md | ebrain put articles/slug --stdin
cat article.html | ebrain put articles/html-page --stdin --format html

# Batch import a directory
ebrain import ./docs --dry-run
ebrain import ./docs

# Compile + timeline + entity links in one command
ebrain smart-ingest companies/river-ai --file article.md

HTML input is extracted with Readability and converted to Markdown before storage, preserving readable structure, links, and images where possible.

ebrain graph                    # Start graph Web UI (http://localhost:3000)
ebrain graph --port 8080 --open # Custom port and auto-open browser

Use the graph Web UI to search entities, filter by type, inspect node details, rename or merge entities, and explore relationships. Open Ask Ex-Brain from the graph toolbar to ask natural-language questions with sourced answers.

You can also query from the CLI:

ebrain search "some topic"
ebrain query "some question"
ebrain query --llm "What is the main idea of River AI's product?"

Start the MCP server:

ebrain serve

Configure your agent or MCP client:

{
  "mcpServers": {
    "ebrain": {
      "command": "ebrain",
      "args": ["serve"]
    }
  }
}

Agents can then use eBrain tools to read and write pages, search the knowledge base, compile new facts, create entity links, extract timelines, and ingest PDF / Word / HTML / text documents.

Start the interactive 3D knowledge graph:

ebrain graph --open

The graph UI includes:

• Entity relationship visualization with type-based colors and clustering
• Search, type filters, fit view, node details, entity rename/merge actions
• Toggleable labels with performance-aware rendering for large graphs
• Important-node labels shown by default, plus nearby labels as you move the camera
• Lighter default links, with highlighted links preserved during focus/selection
• Ask Ex-Brain floating panel for sourced Q&A over the knowledge base
• Ask controls for answer font size, reset/re-ask, loading state, and draggable/resizable layout

Edit ~/.ebrain/settings.json:

{
  "db": { "path": "~/.ebrain/data/ebrain.db" },
  "embed": {
    "provider": "hash",          // or "openai_compatible"
    "baseURL": "...",
    "model": "...",
    "dimensions": 1024,
    "apiKey": "sk-..."
  },
  "llm": {
    "baseURL": "https://dashscope.aliyuncs.com/compatible-mode/v1",
    "model": "qwen-plus",
    "apiKey": "sk-..."
  },
  "extraction": {
    "confidenceThreshold": 0.7   // Entity extraction confidence (0~1)
  }
}

Run ebrain config to view active configuration. See docs/ebrain-cli.md for details.

Ask natural language questions and get answers based on your knowledge base:

# Basic Q&A
ebrain query --llm "What is the main idea of River AI's product?"

# Control context depth
ebrain query --llm "What happened in Q4?" --context-limit 3

How it works:

1. Semantic Search — Finds top matching pages for your question
2. Multi-Layer Context Collection — Builds rich context from:
   • Page Content — Compiled truth + timeline for each matched page
   • Raw Documents — Original imported documents (via raw set)
   • Linked Pages — Incoming and outgoing linked pages, filtered by semantic relevance to the question
3. LLM Synthesis — Generates a sourced answer with [[slug|title]] citations

Configure LLM in ~/.ebrain/settings.json:

{
  "llm": {
    "baseURL": "https://dashscope.aliyuncs.com/compatible-mode/v1",
    "model": "qwen-plus",
    "apiKey": "sk-..."
  }
}

bun install
bun run src/cli.ts --help
bun test