📄 DocumentGPT Pro

AI-Powered Document Intelligence using Retrieval-Augmented Generation (RAG)

Build a private AI knowledge assistant that understands your documents using semantic search, vector databases, and local Large Language Models.

---

📖 Overview

DocumentGPT Pro is a production-inspired Retrieval-Augmented Generation (RAG) application that allows users to upload PDF, DOCX, and TXT documents and interact with them using natural language.

Instead of sending an entire document to a Large Language Model, the application retrieves only the most semantically relevant document chunks using Sentence Transformers and ChromaDB before generating responses with Llama 3.2 running locally through Ollama.

The entire pipeline runs completely offline, making the application privacy-friendly while demonstrating modern AI engineering concepts such as vector databases, embeddings, semantic search, prompt engineering, and conversational document intelligence.

---

✨ Features

🤖 AI	📄 Documents	⚡ User Experience
Retrieval-Augmented Generation	PDF	Streaming Responses
Semantic Search	DOCX	Suggested Questions
Ollama Local LLM	TXT	Copy Answer
AI Summarization	Multi-page Support	Export Chat
Context Relevance	Source References	Modern UI
Conversation Memory	Document Insights	Fully Offline

---

🎯 Why DocumentGPT Pro?

Large Language Models are powerful but cannot answer questions about your private documents unless those documents are supplied as context.

DocumentGPT Pro solves this using a complete Retrieval-Augmented Generation pipeline.

Instead of relying on the model's internal knowledge, the system:

• Parses uploaded documents
• Splits text into semantic chunks
• Generates vector embeddings
• Stores embeddings in ChromaDB
• Retrieves only the most relevant context
• Generates grounded answers using Ollama

This approach produces more accurate, explainable, and context-aware responses while keeping all data local.

---

🖼️ Application Preview

🏠 Home

The landing page provides a clean interface for uploading and processing documents.

---

🧠 AI Workspace

The workspace presents AI-generated summaries, suggested questions, document insights and statistics.

---

💬 Intelligent Conversation

Chat naturally with your documents while viewing retrieved source references.

---

📊 Document Insights

Automatically generated document statistics help users quickly understand uploaded content.

---

🏗️ System Architecture

The application follows a modular Retrieval-Augmented Generation (RAG) architecture.

Pipeline Overview

1. Upload document
2. Extract document text
3. Split into semantic chunks
4. Generate vector embeddings
5. Store embeddings in ChromaDB
6. Retrieve relevant document chunks
7. Construct contextual prompt
8. Generate answer using Ollama
9. Stream the response back to the user

---

🛠️ Technology Stack

Layer	Technology
Programming Language	Python 3.12
Frontend	Streamlit
AI Framework	LangChain
Large Language Model	Llama 3.2 (Ollama)
Embeddings	all-MiniLM-L6-v2
Embedding Framework	Sentence Transformers
Vector Database	ChromaDB
PDF Processing	PyPDF2
DOCX Processing	python-docx
PDF Export	ReportLab

---

📂 Project Structure

The project follows a modular architecture where each module has a single responsibility, making the application easier to maintain, extend, and test.

DocumentGPT-Pro/
│
├── assets/                 # Banner, architecture diagram & custom styling
├── chroma_db/              # ChromaDB vector database
├── components/             # Reusable Streamlit UI components
├── screenshots/            # README screenshots
├── tests/                  # Unit tests
├── uploads/                # Uploaded documents
├── utils/                  # Core AI, RAG and helper modules
│
├── app.py                  # Main application
├── config.py               # Application configuration
├── requirements.txt        # Python dependencies
├── README.md
└── .gitignore

---

🚀 Installation

1. Clone the Repository

git clone https://github.com/nakul85/DocumentGPT-Pro.git

cd DocumentGPT-Pro

---

2. Create a Virtual Environment

Windows

python -m venv venv

venv\Scripts\activate

Linux / macOS

python3 -m venv venv

source venv/bin/activate

---

3. Install Dependencies

pip install -r requirements.txt

---

4. Install Ollama

Download Ollama from:

https://ollama.com/download

Pull the required model:

ollama pull llama3.2:3b

Verify installation:

ollama list

---

5. Run the Application

streamlit run app.py

The application will automatically launch in your browser.

---

💻 Usage

Step 1

Upload a document.

Supported formats:

• PDF
• DOCX
• TXT

---

Step 2

Click 🚀 Process Document

The application automatically:

• Extracts document text
• Splits the content into semantic chunks
• Generates vector embeddings
• Stores vectors in ChromaDB
• Creates an AI-generated summary
• Generates suggested questions
• Calculates document insights

---

Step 3

Ask questions naturally.

Example prompts:

Summarize this document.

Explain the key concepts.

Generate interview questions.

What are the important findings?

Explain the conclusion.

Compare two topics.

What is Retrieval-Augmented Generation?

---

Step 4

The application retrieves the most relevant document chunks before querying the LLM.

The generated response includes:

• AI-generated answer
• Retrieved source references
• Context relevance score
• Streaming response
• Conversation history

---

🧠 RAG Pipeline

Document
    │
    ▼
Text Extraction
    │
    ▼
Recursive Chunking
    │
    ▼
Sentence Transformers
    │
    ▼
Vector Embeddings
    │
    ▼
ChromaDB
    │
User Question
    │
    ▼
Similarity Search
    │
    ▼
Context Builder
    │
    ▼
Ollama (Llama 3.2)
    │
    ▼
Streaming Response

---

🔒 Privacy

DocumentGPT Pro is designed with privacy as a core principle.

• ✅ No cloud APIs
• ✅ No external document uploads
• ✅ Local vector database
• ✅ Local embedding generation
• ✅ Local LLM inference

All processing is performed entirely on your machine.

---

🚀 Future Improvements

Although DocumentGPT Pro is fully functional, several enhancements can further extend its capabilities.

• 📂 Multi-document knowledge base
• 🔍 Hybrid Search (Keyword + Semantic Search)
• 🖼 OCR support for scanned PDFs
• 🎤 Voice-based document interaction
• 🌐 Cloud deployment (AWS, Azure or GCP)
• 👥 User authentication and document management
• 📑 Citation highlighting within documents
• 📱 Responsive mobile interface
• 🌍 Multi-language document support
• 🧠 Support for larger local language models

---

📌 Project Highlights

✅ Retrieval-Augmented Generation (RAG)

✅ Semantic Search using Sentence Transformers

✅ Local LLM Inference with Ollama

✅ Vector Database using ChromaDB

✅ AI-powered Document Summarization

✅ Suggested Question Generation

✅ Context Relevance Scoring

✅ Source-grounded Responses

✅ Conversation Export

✅ Modular Python Architecture

✅ Offline & Privacy-first Design

---

📜 License

This project is licensed under the MIT License.

Feel free to use, modify, and distribute this project under the terms of the MIT License.

---

👨‍💻 Author

Nakul Firodiya

AI Engineer | Python | Machine Learning | Generative AI | Retrieval-Augmented Generation (RAG)

Connect with Me

• GitHub: https://github.com/nakul85
• LinkedIn: https://www.linkedin.com/in/nakulfirodiya74/

---

🙏 Acknowledgements

This project was built using several outstanding open-source technologies.

Special thanks to the communities behind:

• Ollama
• LangChain
• ChromaDB
• Hugging Face
• Streamlit
• Sentence Transformers

Their work makes modern AI application development accessible to everyone.

---

⭐ If you found this project helpful, consider giving it a Star!