Skip to content

Getting Started

This guide walks you through setting up Pattern Vault on your local machine, configuring your preferred language model backend, and running your very first code ingestion pipeline.

Prerequisites

Before installing, ensure your environment meets the following specifications:

  • Python: Version 3.10 to 3.13 (Python 3.11 to 3.13 is required if you plan to run the legacy Chainlit interface).
  • Node.js: Version 20 or higher (required for the Vite + React dashboard UI).
  • Package Manager: We highly recommend uv for fast, predictable Python environment management.

Installation

Follow these steps to clone the repository and install the backend and frontend package dependencies.

1. Clone the Repository

git clone https://github.com/arunksingh16/pattern-vault.git
cd pattern-vault

2. Install Python Environment & Dependencies

Using uv, synchronize the virtual environment and install the required dependencies (including web tools for the API BFF): 

uv sync --extra web

3. Install React Dashboard Dependencies

Navigate to the frontend workspace and install the NPM packages: 

cd web
npm ci
cd ..

Configure Model Backends

Pattern Vault relies on an LLM to identify, classify, and summarize reusable patterns. You can configure any of the following four provider options by setting environment variables.

Uses Anthropic's Claude models directly. Provides the highest extraction quality.

export PATTERN_VAULT_BACKEND=anthropic
export ANTHROPIC_API_KEY="your-sk-ant-api-key"
# Optional model override (defaults to Claude 3.5 Sonnet)
export PATTERN_VAULT_MODEL="claude-3-5-sonnet-20241022"

Utilizes Claude models hosted inside your AWS account. Very secure and highly compliant.

export PATTERN_VAULT_BACKEND=bedrock
export AWS_REGION="us-east-1"
export AWS_ACCESS_KEY_ID="YOUR_ACCESS_KEY"
export AWS_SECRET_ACCESS_KEY="YOUR_SECRET_KEY"
# Optional session token for temporary credentials
export AWS_SESSION_TOKEN="YOUR_SESSION_TOKEN"

Proxies calls through a Bifrost endpoint wrapper. Useful for complex corporate proxy setups.

export PATTERN_VAULT_BACKEND=bifrost
export BIFROST_URL="http://localhost:8080/anthropic"
export PATTERN_VAULT_MODEL="bedrock/eu.anthropic.claude-haiku-4-5-20251001-v1:0"
export BIFROST_API_KEY="dummy"

Run completely local extraction using offline models. Ideal for offline work and private codebase compliance.

export PATTERN_VAULT_BACKEND=ollama
export OLLAMA_BASE_URL="http://localhost:11434/v1"
# We recommend high-performance coder models
export PATTERN_VAULT_MODEL="qwen2.5-coder:7b"

Let's test the installation by indexing a local folder (or even the Pattern Vault directory itself!) and performing an intent search.

1. Run a Dry-Run Indexing Job

A dry-run performs directory scanning and AST chunking without making any expensive LLM calls. Excellent for checking filters and file sizes: 

uv run pv index ./pattern_vault --dry-run

2. Ingest the Codebase

Now run a full indexing job. We will use the curated profile to focus on extraction quality rather than raw volume: 

uv run pv index ./pattern_vault --profile curated

3. Query the Database

Search the database by intent: 

uv run pv search "FastAPI app initialization"

Run the Custom React Dashboard

Pattern Vault comes with a premium multi-pane React dashboard. A simple unified helper script manages both the FastAPI backend and Vite frontend processes:

# Start backend (8001) and frontend (5173) in the background
./dev.sh start

Open http://localhost:5173 in your browser to view the interactive dashboard.

Helper Lifecycle Commands:

  • Status Check: Check background process health and PID allocations: 
    ./dev.sh status
  • Restart Server: Restart both processes: 
    ./dev.sh restart
  • Stop Server: Cleanly shutdown both servers: 
    ./dev.sh stop