> Zero Config, Auto-Indexing, Semantic Search, Infinite Evolution.
Index your local Markdown, TXT, and RST documents into seekdb. Provide MCP service for AI tool integration with real-time file watching and a full-featured web dashboard.
Everything you need to turn local documents into an AI-accessible knowledge base.
Vector-based semantic search understands query intent, not just keyword matching. Supports hybrid, vector, and fulltext modes.
Automatically watches directories for changes. New, modified, or deleted files are indexed in real time with 0.5s debounce.
Exposes MCP service via both stdio and SSE transports. Works with Claude Code, Cursor, and any MCP-compatible AI tool.
Intelligent semantic chunking: Markdown by headings, TXT by paragraphs, RST by sections. Handles oversized text automatically.
Each watched directory maps to an independent collection. Scoped search or cross-collection merge with score normalization.
Full-featured web UI: search, document management, configuration editing, and real-time log streaming via WebSocket.
Get up and running in three commands.
Clone and install with pip.
pip install -e .
Create config directory and database.
smm init
Index your documents.
smm import /path/to/docs
Start web UI + file watcher.
smm serve --daemon
# Install dependencies pip install -e ".[dev]" # Initialize configuration smm init # Import documents from a directory smm import ~/my-notes # Start web UI with file watcher (background) smm serve --daemon # Or start MCP server for AI tools smm mcp # Search from CLI smm search "how to deploy" --mode hybrid # Check daemon status smm status
CLI + Daemon architecture with clear component boundaries.
Built with modern, battle-tested tools.