📖 Overview

Welcome to the Agentic Retrieval RAG System, an end-to-end, production-ready document Q&A application powered by PageIndex. This project demonstrates a paradigm shift in Retrieval-Augmented Generation (RAG).

Traditional Similarity-based RAG pipelines rely heavily on text chunking, embedding generation, and vector databases, which often lead to context loss and the "lost-in-the-middle" problem. This system utilizes Vectorless Reasoning-based RAG, allowing AI to natively read and comprehend entire documents—giving transparent, human-expert-level responses.

✨ Key Features

• Vectorless Agentic Retrieval: Process massive document contexts without manual chunking, vectors, or external embedding databases.
• Lightning-Fast Execution: Environment and dependency management powered by uv, the fastest package installer and resolver for Python.
• Real-Time Streaming: Tokens stream directly to the chat UI for an interactive, ChatGPT-like experience.
• Production-Ready Architecture: Cleanly separated modular design (app.py for UI, src/ for logic & API wrappers).
• Persistent File Logging: Beautiful console & rolling file logging via loguru.

📂 Project Structure

page-index-chat-rag/
├── src/
│   ├── __init__.py        
│   ├── logger.py          # Loguru configuration for console and persistent file logging
│   └── rag_api.py         # Backend logic & PageIndex API integration
├── logs/                  # Auto-generated application logs
├── .env.example           # Example environment variables
├── .gitignore             # Comprehensive python project git ignore
├── app.py                 # Streamlit Frontend application
├── pyproject.toml         # uv Project Configuration
├── uv.lock                # Lockfile for exact dependency reproducibility
└── README.md              # Project Documentation

🚀 Quick Start

1. Prerequisites

• Python 3.11 or higher
• uv installed (curl -LsSf https://astral.sh/uv/install.sh | sh or pip install uv)
• A PageIndex API Key (Get yours at Dash PageIndex)

2. Installation

Clone the repository and install dependencies seamlessly:

git clone https://github.com/your-username/page-index-chat-rag.git
cd page-index-chat-rag

# Sync dependencies and create an ultra-fast virtual environment
uv sync

3. Configuration

Copy the example environment variable file and add your API key:

cp .env.example .env

Edit .env and insert your API key:

PAGEINDEX_API_KEY=your_pageindex_api_key_here

4. Run the Application

Activate the virtual environment and launch Streamlit:

source .venv/bin/activate
streamlit run app.py

The interface will automatically open in your browser at http://localhost:8501.

🧠 How it Works

1. Upload: Users upload a PDF securely through the Streamlit UI.
2. Process: The backend (src/rag_api.py) transmits the file to PageIndex and polls the server until the document parsing is fully resolved.
3. Chat: Users submit highly complex, full-document reasoning queries.
4. Agentic Retrieval: The underlying PageIndex model completely skips naive top-K chunking. Instead, it natively reads the unstructured file, locates exact pages implicitly, and streams back perfect, context-aware answers.

---

Made With ❤️ By @muhammadadeelai.