Skip to content

README.md

Latest commit

 

History

History
124 lines (89 loc) · 3.25 KB

README.md

File metadata and controls

124 lines (89 loc) · 3.25 KB
Raw

🧠 RAG PDF Assistant | Production-Grade AI Document Q&A

Upload any PDF. Ask anything. Get grounded, accurate answers — backed by a production-ready pipeline.

🚀 What is this?

Most AI projects are great for learning but fall apart in production. This project bridges that gap — a fully production-grade RAG (Retrieval-Augmented Generation) application that lets users upload PDF documents and ask natural language questions about their content, with all the reliability features a real-world app demands.

Built with FastAPI + Inngest + LlamaIndex + Qdrant + Google Gemini + Streamlit.

✨ Features

  • 📄 PDF Ingestion — Upload and parse any PDF via LlamaIndex
  • 🔍 Vector Search — Embeddings stored and queried from Qdrant vector database
  • 🤖 Grounded Answers — Google Gemini LLM answers only from your document context, no hallucination
  • ⚙️ Production-Ready Backend — Async job processing via Inngest with full observability
  • 🔁 Automatic Retries — Failed jobs retry automatically, no silent failures
  • 🚦 Rate Limiting & Throttling — Concurrency controls to prevent abuse
  • 📊 Full Observability — Real-time job monitoring via Inngest dashboard
  • 🖥️ Streamlit Frontend — Clean, interactive UI for uploading and querying

🏗️ Architecture

User uploads PDF
      ↓
FastAPI endpoint → Inngest job queue
      ↓
LlamaIndex parses & chunks PDF
      ↓
Embeddings stored in Qdrant
      ↓
User asks question → Qdrant vector search → Gemini generates answer

🛠️ Tech Stack

Layer Technology
Backend API FastAPI
Job Queue Inngest
Document Parsing LlamaIndex
Vector Database Qdrant
LLM Google Gemini
Frontend Streamlit
Language Python 3.12+

📁 Project Structure

├── main.py              # FastAPI app & Inngest functions
├── data_loader.py       # PDF ingestion & chunking (LlamaIndex)
├── vector_db.py         # Qdrant vector store operations
├── custom_types.py      # Pydantic models
├── streamlit_app.py     # Frontend UI
├── test_embed.py        # Embedding tests
└── .env.example         # Environment variables template

🚀 Getting Started

Prerequisites

  • Python 3.12+
  • Qdrant instance (local or cloud)
  • Google Gemini API key
  • Inngest account

Installation

git clone https://github.com/GANGOUCH1/rag-pdf-assistant.git
cd rag-pdf-assistant
pip install uv
uv sync

Configure environment

cp .env.example .env
# Fill in your API keys in .env

Run

# Terminal 1 — Start FastAPI
uvicorn main:app --reload

# Terminal 2 — Start Inngest dev server
npx inngest-cli@latest dev

# Terminal 3 — Start Streamlit
streamlit run streamlit_app.py

💡 Why this project stands out

Unlike typical RAG tutorials, this app is built for the real world:

  • No silent failures — every job is tracked, retried, and observable
  • Scalable by design — rate limiting and concurrency controls built in
  • Grounded responses — Gemini only answers from your document context
  • Production patterns — async background processing with full observability

📜 License

MIT