📄 PDF Chatbot RAG

An AI-powered chatbot designed to answer queries from complex PDF documents using a full RAG-based pipeline.

---

🔧 Technologies Used

• FastAPI – REST API backend
• OpenAI – Language models for embeddings and response generation
• LangChain & LangGraph – RAG orchestration & knowledge graph capabilities
• Redis – Task queue for handling PDF processing
• PDF‑to‑Image – Layout parsing support
• QdrantDB – Vector store for semantic search
• Neo4j – Knowledge graph for structured context
• MongoDB – Metadata & chat history storage (for long‑term context)
• Docker & DevContainers – Containerized and reproducible environment

---

🚀 Features

1. Concurrent PDF Processing

   • Uploaded PDFs are queued via Redis and processed asynchronously.
   • Generates embeddings, images, semantic vectors, and knowledge graph entries.

2. Semantic QA

   • Queries are answered using semantic search with embeddings in Qdrant.
   • Responses are enriched using context from Neo4j graph.

3. Contextual Memory

   • MongoDB stores prior chats for long-term session context.

4. Real-world PDF Handling

   • Layout-based parsing using PDF-to-image conversion.
   • OCR-ready for structured documents.

5. Scalable, Deployable

   • Dockerized services (API, Redis, Qdrant, Neo4j, MongoDB).
   • Run in dev or production configurations via DevContainer.

---

🗂️ Repository Structure

---

🛠️ Setup & Run

Prerequisites

• Docker & Docker Compose
• OpenAI API key (set in .env)
• Cohere API Key (set in .env)
• MongoURL (set in .env)
• GPU server (optional for image processing and embeddings)

Quick Start (Docker)

git clone https://github.com/kratinsoni/Pdf-Chatbot-Rag.git
cd Pdf-Chatbot-Rag
cp .env.example .env             # Fill in `OPENAI_API_KEY`
docker compose up --build -d    # Launch all services

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt
python -m app.main.py
rq worker --with-scheduler --url redis://valkey:6379         # Start PDF ingestion manually