Skip to content

Pattern Vault

Pattern Vault Banner

Turn real code into a local, searchable library of reusable engineering patterns.

Pattern Vault scans your repositories, parses their source files into AST symbols using tree-sitter, classifies which components represent reusable software patterns using a large language model, and compiles them into a highly searchable local SQLite FTS5 database.

Once stored, you can query, inspect, and retrieve these patterns from your terminal (CLI), a custom React web dashboard, or directly within your IDE (like Cursor or Claude Code) via the Model Context Protocol (MCP).

Key Features

- :material-magnify: **Intent-Based Search** Search your entire codebase vault by intent, natural language, programming language, category, or source repository using fast BM25 ranking. - :material-robot: **Agentic Analysis Loop** An interactive agentic loop powered by Claude that can scan directories, read files, parse AST symbols, and auto-classify architecture patterns under your guidance. - :material-server: **Model Context Protocol (MCP)** An out-of-the-box MCP server supporting stdio and HTTP transport so Claude Code or Cursor can fetch real, tested implementation examples while you code. - :material-laptop: **Sleek React Dashboard** A modern, glassmorphic React 19 interface with workspace ingestion tracking, log playback, token usage graphs, and an interactive agent chat panel. - :material-shield-key: **Completely Local & Private** Keep everything secure. By default, your database lives in a local SQLite file (`~/.pattern-vault/patterns.db`) with zero external storage requirements. - :material-puzzle: **Multi-Backend Models** Works with Anthropic Claude API, AWS Bedrock, Bifrost, or local models using Ollama-compatible endpoints (e.g., Qwen 2.5 Coder).

Architecture Overview

Pattern Vault is structured into decoupled, robust layers:

flowchart TB
    user["Developer / Claude Code / Browser UI"]

    subgraph access["Access Layer"]
        cli["CLI\nindex / search / stats / chat / serve"]
        mcp["FastMCP Server\n8 vault tools"]
        web["React Dashboard\nVite + TanStack Query + Zustand"]
        api["FastAPI BFF\nchat / patterns / workspace / usage"]
    end

    subgraph agent["Agent Layer"]
        orchestrator["Claude Tool-Use Orchestrator\nscan -> read -> parse -> save"]
        tools["Agent Tools\nscan_directory / read_file / parse_symbols / save_pattern"]
    end

    subgraph indexing["Indexing Layer"]
        scanner["Repo Scanner\nlanguage + path filters"]
        chunker["Tree-sitter Chunker\nsymbols and fallback chunks"]
        extractor["Pattern Extractor\nmodel classification + quality score"]
        profiles["Indexing Profiles\ncurated / balanced / comprehensive"]
    end

    subgraph storage["Storage Layer"]
        sqlite["SQLite + FTS5\npatterns / chunks / insights / jobs / usage"]
    end

    user --> cli
    user --> mcp
    user --> web
    web --> api
    cli --> orchestrator
    api --> orchestrator
    mcp --> sqlite
    orchestrator --> tools
    tools --> scanner
    tools --> chunker
    tools --> sqlite
    cli --> scanner
    api --> scanner
    mcp --> scanner
    scanner --> chunker
    chunker --> profiles
    profiles --> extractor
    extractor --> sqlite

In Action

1. Ingest Patterns via CLI

uv run pv index /path/to/my-project --profile curated

2. Search Intent from Terminal

uv run pv search "retry with exponential backoff" --language python

3. Retrieve inside IDE using Claude Code

User: "Search my pattern vault for auth middleware."
Claude: "I found a pattern: JWT Auth Middleware (FastAPI, Quality: 0.92) in repo backend-services..."

Getting Started

Ready to build your personal codebase memory? Proceed to the Getting Started Guide to set up your environment, install dependencies, and run your first indexing job in under five minutes!