Enterprise Document Q&A System

A lightweight RAG (Retrieval-Augmented Generation) system built from scratch without LangChain. Enables semantic search and intelligent question-answering over document collections using Claude AI and ChromaDB.

🚀 Live Demo

Coming Soon: Deploy link will be added after initial deployment

Screenshots

Document Upload Interface

File Processing

Query & AI-Powered Answers

Source Citations with Relevance Scores

Why I Built This

I built this while exploring production RAG patterns for enterprise applications. Every company is racing to unlock knowledge trapped in documents, and I wanted to understand the full stack—from chunking strategies to deployment—without relying on heavy frameworks like LangChain.

Key learnings:

• Chunking strategies significantly impact retrieval quality
• Source citation is critical for enterprise trust
• Direct API integration gives better control than abstraction layers
• Proper error handling matters more than perfect embeddings

What this demonstrates:

• Production-ready RAG from scratch (no LangChain)
• Custom chunking and retrieval pipeline
• Direct Claude API integration
• Clean, maintainable code patterns
• End-to-end deployment with Docker

Features

• 📄 PDF Document Processing: Upload and index PDF documents
• 🔍 Semantic Search: Find relevant information using natural language
• 🤖 AI-Powered Answers: Get accurate responses backed by your documents
• 📎 Source Citations: See exactly which documents and pages informed each answer
• 🎯 Relevance Scoring: Understand confidence levels for retrieved information
• 🚀 Fast Retrieval: Optimized vector search with ChromaDB

Architecture

Custom RAG Pipeline (No LangChain)

┌─────────────┐
│   User      │
│   Query     │
└──────┬──────┘
       │
       ▼
┌──────────────────────────────────────┐
│      Streamlit UI Layer              │
└──────┬───────────────────────────────┘
       │
       ▼
┌──────────────────────────────────────┐
│   Custom RAG Pipeline                │
│                                      │
│  ┌──────────┐    ┌──────────────┐  │
│  │ PyPDF    │───▶│ Custom       │  │
│  │ Loader   │    │ Chunker      │  │
│  └──────────┘    └──────┬───────┘  │
│                         │           │
│                         ▼           │
│              ┌──────────────────┐  │
│              │  ChromaDB        │  │
│              │  (embeddings +   │  │
│              │   vector store)  │  │
│              └──────────┬───────┘  │
│                         │           │
│         ┌───────────────┴──────┐   │
│         ▼                      │   │
│  ┌──────────────┐             │   │
│  │  Semantic    │             │   │
│  │  Search      │             │   │
│  │  (Top-K)     │             │   │
│  └──────┬───────┘             │   │
│         │                     │   │
│         ▼                     │   │
│  ┌──────────────┐             │   │
│  │  Direct      │             │   │
│  │  Claude API  │◀────────────┘   │
│  └──────┬───────┘                 │
└─────────┼───────────────────────┘
          │
          ▼
     ┌─────────┐
     │ Answer  │
     │   +     │
     │ Sources │
     └─────────┘

Tech Stack

Component	Technology	Purpose
LLM	Claude 3.5 Sonnet	Response generation
Vector Store	ChromaDB	Semantic search & embeddings
Document Processing	PyPDF	PDF text extraction
UI	Streamlit	Web interface
Language	Python 3.11+	Core implementation

Note: Built without LangChain - direct API integration for full control and minimal dependencies.

Quick Start

Prerequisites

• Python 3.11 or higher
• Anthropic API key (Get one here)
• (Optional) Voyage AI or OpenAI API key for embeddings

Installation

1. Clone the repository

   git clone https://github.com/yourusername/enterprise-doc-qa.git
   cd enterprise-doc-qa

2. Create virtual environment

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment variables

   cp .env.example .env
   # Edit .env and add your API keys

5. Run the application

   streamlit run src/ui/app.py

6. Open your browser Navigate to http://localhost:8501

Docker Deployment

# Build the image
docker build -t doc-qa-system .

# Run the container
docker run -p 8501:8501 --env-file .env doc-qa-system

Usage

Sample Questions to Try

Once you've uploaded documents, try questions like:

• "What are the key terms of the contract?"
• "Summarize the main findings from the research report"
• "What security measures are mentioned?"
• "Compare the pricing models discussed"
• "What are the project timelines?"

Best Practices

For best results:

• Upload well-structured PDFs (avoid scanned images without OCR)
• Keep documents focused on a specific domain
• Ask specific questions rather than broad queries
• Review source citations to verify accuracy

Project Structure

enterprise-doc-qa/
├── src/
│   ├── components/
│   │   ├── document_loader.py    # PDF processing
│   │   ├── chunking.py            # Text splitting logic
│   │   ├── embeddings.py          # Vector generation
│   │   └── retrieval.py           # RAG chain implementation
│   └── ui/
│       └── app.py                 # Streamlit interface
├── tests/
│   ├── test_chunking.py
│   └── test_retrieval.py
├── data/                          # Sample documents (gitignored)
├── docs/                          # Additional documentation
├── .env.example                   # Environment template
├── .gitignore
├── requirements.txt
├── Dockerfile
└── README.md

Configuration

Key environment variables in .env:

# Required
ANTHROPIC_API_KEY=your_claude_api_key

# Optional (for embeddings)
VOYAGE_API_KEY=your_voyage_key
OPENAI_API_KEY=your_openai_key

# Tuning parameters
CHUNK_SIZE=1000
CHUNK_OVERLAP=200
TOP_K_RESULTS=4

Limitations & Future Improvements

Current Limitations

• PDF only (no DOCX, TXT, HTML yet)
• No multi-document comparison
• Chat history not persisted across sessions
• English language only

Roadmap

• Add support for DOCX, TXT, Markdown files
• Implement conversation memory
• Add authentication and multi-user support
• Hybrid search (keyword + semantic)
• Export Q&A history
• Advanced chunking strategies (semantic splitting)
• Custom embedding fine-tuning

Testing

# Run all tests
pytest

# Run with coverage
pytest --cov=src tests/

# Run specific test file
pytest tests/test_chunking.py

Performance

Benchmarks (on M1 Mac, 100-page PDF):

• Document processing: ~15 seconds
• Query response time: ~2-3 seconds
• Embedding generation: ~5 seconds (cached afterward)

Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

License

MIT License - see LICENSE file for details

Acknowledgments

• Powered by Anthropic Claude
• Vector search by ChromaDB
• PDF processing by PyPDF
• UI framework by Streamlit

Contact

Sandeep Uppalapati

• LinkedIn: linkedin.com/in/sandeep-uppalapati
• GitHub: @sandeepuppalapati
• Project: github.com/sandeepuppalapati/enterprise-doc-qa

---

Note: This is a demonstration project. For production use, add proper authentication, rate limiting, and security measures.