📄 DocuQuery AI – AI-Powered Document Intelligence System

DocuQuery AI is an end-to-end AI-powered document understanding and question-answering system that allows users to upload documents (PDFs, images, or text files), automatically extract their content using PDF parsing and OCR, and ask natural language questions using a Retrieval-Augmented Generation (RAG) workflow powered by a Large Language Model (LLM).

This project focuses on building a practical, production-style AI pipeline by integrating document processing, OCR, backend APIs, and LLM-based reasoning into a single system.

🚀 Key Features

• Upload and process:
  • Digital PDFs
  • Scanned PDFs
  • Images (JPG, PNG, WEBP)
  • Text and code files
• Automatic file type detection
• Text extraction from PDFs using pdfplumber
• OCR fallback for scanned documents using OpenAI vision
• PDF page preview using pdf2image
• AI-powered question answering using LLM-based RAG
• REST API-based backend using Flask
• Interactive and responsive web interface

🧠 Core AI & Engineering Concepts

• Optical Character Recognition (OCR)
• Document parsing and preprocessing
• Retrieval-Augmented Generation (RAG)
• Prompt engineering
• LLM API integration
• REST API design
• Full-stack AI system integration
• Error handling and fallback pipelines

🛠️ Tech Stack

Backend

• Python
• Flask
• pdfplumber
• pdf2image
• OpenAI vision
• Pillow
• python-dotenv
• Werkzeug

AI

• OpenAI API (LLM-based question answering)

Frontend

• HTML
• CSS
• JavaScript (Fetch API)

🏗️ System Workflow

1. User uploads a document from the web interface.
2. Backend detects the file type automatically.
3. Based on file type:
   • Digital PDF → Text extracted using pdfplumber
   • Scanned PDF → Converted to image → OCR using OpenAI vision
   • Image file → OCR applied directly
   • Text file → Direct decoding
4. Extracted text is stored as contextual knowledge.
5. User submits a natural language question.
6. Document text + user query are sent to the LLM through a RAG-style prompt.
7. The AI generates a grounded answer based strictly on the document.
8. The response is displayed in the UI in real time.

📁 Project Structure

DocuQuery_AI/
│
├── app.py
├── config.py
├── pdf_utils.py
├── ocr_utils.py
├── rag_utils.py
├── file_utils.py
├── graph.py
├── requirements.txt
├── runtime.txt
├── README.md
│
├── templates/
│   └── index.html
│
└── static/
    ├── app.js
    └── style.css

🔌 API Endpoints

Endpoint	Method	Description
/	GET	Home page
/upload	POST	Upload and process a file
/preview_pdf_page	POST	Preview selected PDF page
/extract_pdf	POST	Extract PDF text (page/range/full)
/ocr	POST	Run OCR on image
/chat	POST	Ask questions using RAG

🧪 Example Use Cases

• Ask questions from research papers
• Extract text from scanned notes
• Summarize PDF reports
• Query invoices or bills
• Educational document analysis

---

📊 Learnings

• Designed a full multi-step AI workflow
• Integrated OCR + LLM reasoning
• Built REST APIs for document processing
• Handled real-world document edge cases
• Implemented RAG-style question answering
• Learned practical deployment challenges

🔮 Future Improvements

• Add vector embeddings + FAISS for advanced RAG
• Use EasyOCR / PaddleOCR for better accuracy
• Multi-document querying
• Automatic summarization and key-entity extraction
• Session memory and chat history
• Full cloud deployment (Render / AWS / Streamlit Cloud)

📹 Demo Video

https://drive.google.com/file/d/1d-gUCekkUgPPC9ay5JFjKJeG5n1fy5xv/view?usp=drive_link

📎 GitHub Repository

https://github.com/omprakash0702/DocuQuery_AI

✅ Conclusion

DocuQuery AI is a complete Document Intelligence and Question-Answering system that demonstrates how OCR, PDF parsing, and LLM-based reasoning can be combined into a real-world AI application. The project highlights strong applied AI skills, system design thinking, and clean backend–frontend integration.

Author: Omprakash
Domain: Applied AI | Document Intelligence | RAG Systems