🤖 RAG-Powered PDF Question Answering Bot

Ask questions from any PDF — powered by LangChain · IBM Watsonx · ChromaDB

Upload any PDF. Ask questions in plain English. Get precise, context-grounded answers — powered by a production-grade RAG pipeline using IBM Watsonx enterprise LLM.

---

🎯 What This Project Does

Most LLMs hallucinate when asked about documents they haven't seen. This bot eliminates that problem by grounding every answer in retrieved content from your actual PDF — no hallucinations, no guesswork.

PDF Upload → Chunking → Embedding (Watsonx slate-30m) → ChromaDB
                                                              │
User Query → Semantic Retrieval → RetrievalQA → IBM Watsonx LLM → Answer

---

✨ Features

• 📄 Upload any PDF — no preprocessing required
• ✂️ Automatic chunking — splits documents into semantically meaningful segments
• 🔢 Enterprise embeddings — IBM Watsonx slate-30m for high-quality vector representations
• 🗄️ ChromaDB vector store — fast, local semantic search over document chunks
• 🧠 RetrievalQA pipeline — LangChain orchestrates retrieval → generation end-to-end
• 🖥️ Gradio UI — clean, real-time Q&A interface, zero frontend code

---

🏗️ Architecture

┌─────────────────────────────────────────────────────────┐
│                    INDEXING PIPELINE                     │
│                                                          │
│  PDF File → PyPDF Loader → Text Splitter → Chunks       │
│                                    │                     │
│                          Watsonx Embeddings              │
│                                    │                     │
│                             ChromaDB Store               │
└─────────────────────────────────────────────────────────┘

┌─────────────────────────────────────────────────────────┐
│                    QUERY PIPELINE                        │
│                                                          │
│  User Query → Embed Query → Semantic Search (Chroma)    │
│                                    │                     │
│                          Relevant Chunks                 │
│                                    │                     │
│                    IBM Watsonx LLM (RetrievalQA)        │
│                                    │                     │
│                         Grounded Answer                  │
└─────────────────────────────────────────────────────────┘

---

⚙️ Tech Stack

Layer	Technology
RAG Orchestration	LangChain
LLM	IBM Watsonx (enterprise)
Embeddings	IBM Watsonx slate-30m
Vector Store	ChromaDB
UI	Gradio
Runtime	Python 3.8+
Integration	HuggingFace Hub

---

🚀 Getting Started

1. Clone the repository

git clone https://github.com/amarskdev/RAG-based-PDF-Question-Answering-Bot.git
cd RAG-based-PDF-Question-Answering-Bot

2. Create and activate virtual environment

python -m venv venv

# macOS / Linux
source venv/bin/activate

# Windows
venv\Scripts\activate

3. Install dependencies

pip install -r requirements.txt

4. Configure IBM Watsonx credentials

Add your credentials to a .env file:

WATSONX_API_KEY=your_api_key_here
WATSONX_PROJECT_ID=your_project_id_here
WATSONX_URL=https://us-south.ml.cloud.ibm.com

5. Run the app

python app.py

Open your browser at http://127.0.0.1:7860, upload a PDF, and start asking questions.

---

📁 Project Structure

RAG-based-PDF-Question-Answering-Bot/
│
├── app.py                  # Main Gradio app + RAG pipeline
├── requirements.txt        # Python dependencies
├── .env.example            # Credential template (safe to commit)
├── .gitignore              # Excludes .env, venv, __pycache__
└── README.md

---

🌍 Real-World Use Cases

• 📚 Research assistance — query academic papers, reports, whitepapers
• ⚖️ Legal document review — extract clauses and obligations from contracts
• 🏢 Enterprise knowledge base — make internal docs instantly searchable
• 📋 Compliance workflows — audit documents against policies
• 🎓 Study tool — ask questions from textbooks and lecture notes

---

🔭 Roadmap

• Multi-PDF support (query across multiple documents)
• Conversational memory (multi-turn Q&A)
• Source citation with page numbers
• REST API via FastAPI + Docker deployment
• Fine-tuning embeddings on domain-specific corpora

---

Built with enterprise-grade AI (IBM Watsonx) — not just another OpenAI wrapper.

---

🤝 Connect With Me

👤 About the Author

Amar Kumar
Senior Backend Engineer · IBM Certified AI Engineer

If you found this project useful, consider giving it a ⭐ — it means a lot!