MCP 服务器

lore-mcp

Indexes local Markdown/text files into a SQLite database with vector embeddings and provides MCP tools for semantic search without cloud dependencies.

README

lore-mcp

LORE — Local Offline Retrieval Engine for MCP

An MCP server for semantic search over your local technical documents. No cloud, no external database — just a single .db file on your workstation.

What it does

Indexes a directory of Markdown/text files into a portable SQLite database using vector embeddings
Exposes two MCP tools (search_docs, list_sources) for any MCP client (Claude Code, Claude Desktop, Cursor, etc.)
Runs locally with automatic GPU/API/CPU fallback for embedding generation

Status

Early development. See the backlog for planned features.

Quickstart

Prerequisites

Python ≥ 3.10
(Optional) NVIDIA GPU with CUDA for faster embeddings

Installation

pip install lore-mcp

Note: not yet published on PyPI. For now, install from source:
git clone https://github.com/romainsc/lore-mcp.git
cd lore-mcp
pip install -e .

Configure your MCP client

Add to your MCP client configuration (e.g. .claude/settings.json):

{
  "mcpServers": {
    "lore": {
      "command": "lore-mcp",
      "args": []
    }
  }
}

Index your documents

lore-mcp index /path/to/your/docs/

Search

Once configured, your MCP client can use:

search_docs("your query", top_k=5) — semantic search over indexed documents
list_sources() — list all indexed files with chunk counts

Environment variables

Variable	Role	Default
`LORE_DB_PATH`	SQLite database file path	`./lore.db`
`LORE_MODEL`	Embedding model name	`BAAI/bge-m3`
`LORE_EMBED_MODE`	Mode: `auto`, `gpu`, `api`, `cpu`	`auto`
`LORE_API_URL`	Remote `/v1/embeddings` URL	(required if mode=api)
`LORE_API_MODEL`	Model name for remote API	same as `LORE_MODEL`

Architecture

lore-mcp uses BAAI/bge-m3 for embeddings (1024 dimensions, multilingual) and sqlite-vec for vector storage in a single .db file.

Embedding generation falls back automatically: local GPU (CUDA) → remote API (OpenAI-compatible) → local CPU.

See docs/ for detailed documentation (coming with MVP).

Roadmap

MVP (v0.1.0)

[ ] SQLite + sqlite-vec storage backend
[ ] Embedding with GPU/API/CPU fallback
[ ] MCP server (search_docs, list_sources)
[ ] CLI ingestion tool
[ ] Tests (TDD)

Post-MVP

[ ] Example corpus and sample database
[ ] Incremental re-indexing
[ ] Metadata filtering
[ ] Hybrid search (vector + keyword)
[ ] Image captioning during ingestion
[ ] Docker image

AI-assisted development

This project is developed with AI assistance (Claude, Anthropic). All AI-assisted content is marked with Co-Authored-By trailers in commits. Every contribution — human or AI-assisted — is reviewed, tested, and validated by a human before being committed.

See docs/ai-guidelines.md for the full guidelines.