🚀 ProRAG System: Local-First Intelligent RAG

A powerful, local-first Retrieval-Augmented Generation (RAG) system with a two-stage retrieval pipeline, OCR support for scanned documents, and professional observability.

✨ Key Features

• Two-Stage Retrieval: Vector search (Qdrant) + AI Re-ranking (Cross-Encoder).
• Intelligent OCR: Automatic fallback to Tesseract for scanned/image-based PDFs.
• Premium CLI: Highly verbose and organized output using rich.
• Pro Observability: Industry-standard dashboard with Arize Phoenix.
• Local-First: Complete privacy, running entirely on your machine via Docker and Pixi.

---

🛠 Prerequisites

• Pixi installed.
• Docker installed.

---

⚡️ Quick Start

1. Setup Infrastructure

Start the local Qdrant database and Phoenix dashboard:

pixi run up

2. Ingest Documents

Place your PDFs in the data/ folder and run:

pixi run ingest

3. Query the System

Search through your documents with AI re-ranking:

pixi run query "What is RAG?"

---

⌨️ Automation (Pixi Tasks)

Command	Description
pixi run up	Start Docker containers (Qdrant + Phoenix)
pixi run down	Stop all Docker containers
pixi run ingest	Process PDFs and store embeddings
pixi run query "..."	Search and re-rank results
pixi run test	Run unit tests
pixi run stats	Check collection statistics
pixi run dashboard	Open Arize Phoenix in browser
pixi run qdrant_ui	Open Qdrant Dashboard in browser

---

🏗 Project Structure

.
├── data/               # Source PDF documents
├── src/
│   └── prorag/         # Main package
│       ├── core/       # Config, Database, Model managers
│       ├── ingest/     # PDF processing & pipeline
│       ├── retrieval/  # Vector search & re-ranking
│       └── cli.py      # Unified CLI entry point
├── tests/              # Unit tests
├── docker-compose.yml  # Infrastructure as Code
└── pixi.toml          # Dependency & Task management