DocQuery AI – AI-powered Medical Document Q&A using RAG and FAISS

DocQuery AI is an intelligent chatbot that allows users to upload medical PDF documents and chat with them. It uses Retrieval-Augmented Generation (RAG), FAISS for vector search, and advanced language models to deliver accurate and context-aware answers.

---

🚀 Features

• ✅ Upload PDFs - Extract text using pypdf, if pages contain images then pytesseract (OCR) for text extraction.
• ✅ Smart Document Search – Uses FAISS for efficient vector-based retrieval.
• ✅ Context-Aware Answers – Retrieval-Augmented Generation ensures accurate responses.
• ✅ Free-Form Chat – Ask anything related to the uploaded documents.
• ✅ Scalable Architecture – Can be extended for multiple users and large datasets.

---

🔗 Live Demo

• 🌐 Streamlit App: Link

---

🏗️ Tech Stack

• Vector Database: FAISS
• Embedding Model: Hugging Face model (sentence-transformers/all-mpnet-base-v2)
• Language Model: OpenAI GPT (gpt-4.1-nano)
• Frontend: Streamlit
• PDF Parsing: pypdf + pytesseract (Image Parsing)

---

🔍 How It Works (Architecture)

1. Upload PDFs → Extract text using PDF parsers.
2. Text Chunking → Split text into chunks for better retrieval.
3. Embeddings Generation → Convert chunks into vector embeddings.
4. Store in FAISS → Efficient similarity search for queries.
5. Query Processing → Retrieve top matches using FAISS.
6. Generate Answer → Pass retrieved chunks to LLM for final response.

---

⚡ Installation

# Clone the repository
git clone https://github.com/tanishra/DocQuery-AI.git
cd docquery

# Create virtual environment
conda create -n venv 
conda activate venv

# Install dependencies
pip install -r requirements.txt

---

▶️ Usage

streamlit run main.py

• Upload your medical PDFs.
• Ask questions or chat with your documents.

---

📷 Screenshot

---

✅ Future Enhancements

• Source Linking (Citations) – Show document name and page number for each answer.
• Confidence Score – Display answer confidence based on FAISS similarity.
• Document Summarization – Generate quick summaries for uploaded PDFs.
• Hybrid Search – Combine keyword-based and semantic search for better results.
• Voice-based Interaction – Add Speech-to-Text and Text-to-Speech support.
• External Medical Knowledge Integration – Use PubMed or WHO data for extra context.
• Analytics Dashboard – Track user queries and most-referenced documents.

---

🤝 Contribution

I am open to feedback and contributions!

• Fork the repo and create a pull request.
• Feel free to use, test, and improve this project.
• If you like it, please give it a ⭐ on GitHub!