PDF Document Assistant

A full-stack AI-powered PDF document assistant with semantic search, question answering, and document summarization.

Features

• 📄 PDF Upload & Processing: Upload PDFs and automatically extract text
• 🔍 OCR Support: Automatic fallback to Tesseract OCR for scanned documents
• 🧠 Semantic Search: Find relevant information using AI embeddings
• 💬 Question Answering: Ask questions and get accurate answers with citations
• 📋 Document Summarization: Automatic generation of summaries and key points
• 🎯 Source Citations: Every answer includes page references and source information
• 🚀 Production-Ready: Clean architecture, proper error handling, and scalable design

Tech Stack

Backend

• FastAPI: Modern Python web framework
• PyMuPDF: PDF text extraction
• Tesseract OCR: Scanned document processing
• Sentence Transformers: Embedding generation
• ChromaDB: Vector database for semantic search
• LangChain: LLM integration
• Groq API: Large language model (Mixtral 8x7B)

Frontend

• Next.js: React framework
• TailwindCSS: Styling
• Axios: API client
• Lucide React: Icons

Project Structure

pdf-assistant/
├── backend/
│   ├── app.py                 # FastAPI application
│   ├── requirements.txt       # Python dependencies
│   ├── .env.example          # Environment variables template
│   ├── uploads/              # Uploaded PDFs
│   ├── chroma_db/            # Vector database
│   └── modules/
│       ├── parser.py         # PDF text extraction
│       ├── chunker.py        # Text chunking
│       ├── embeddings.py     # Embedding generation
│       ├── qa.py             # Question answering
│       └── summarizer.py     # Document summarization
└── frontend/
    ├── package.json
    ├── pages/
    │   └── index.tsx         # Main application
    ├── components/
    │   ├── FileUpload.tsx
    │   ├── Summary.tsx
    │   └── ChatInterface.tsx
    ├── lib/
    │   └── api.ts            # API client
    └── styles/
        └── globals.css

Installation

Prerequisites

• Python 3.9+
• Node.js 16+
• Tesseract OCR (for scanned PDF support)

Install Tesseract

macOS:

brew install tesseract

Ubuntu/Debian:

sudo apt-get install tesseract-ocr

Windows: Download from: https://github.com/UB-Mannheim/tesseract/wiki

Backend Setup

1. Navigate to backend directory:

cd pdf-assistant/backend

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. Create .env file:

cp .env.example .env

5. Configure environment variables in .env:

For Google Gemini:

GEMINI_API_KEY=your_gemini_api_key_here
LLM_PROVIDER=gemini

For OpenAI:

OPENAI_API_KEY=your_openai_api_key_here
LLM_PROVIDER=openai

6. Run the backend:

python app.py

The API will be available at http://localhost:8000

Frontend Setup

1. Navigate to frontend directory:

cd pdf-assistant/frontend

2. Install dependencies:

npm install

3. Create .env.local file:

cp .env.example .env.local

4. Run the development server:

npm run dev

The application will be available at http://localhost:3000

API Endpoints

Upload PDF

POST /upload
Content-Type: multipart/form-data

Response:
{
  "document_id": "string",
  "filename": "string",
  "pages": number,
  "chunks": number,
  "summary": {
    "summary": "string",
    "key_points": ["string"],
    "main_topics": ["string"],
    "success": boolean
  },
  "success": boolean
}

Ask Question

POST /ask
Content-Type: application/json

Request:
{
  "document_id": "string",
  "question": "string"
}

Response:
{
  "answer": "string",
  "citations": [
    {
      "source_id": number,
      "page_number": number,
      "chunk_id": "string"
    }
  ],
  "success": boolean
}

Get Summary

GET /summary/{document_id}

Response:
{
  "summary": "string",
  "key_points": ["string"],
  "main_topics": ["string"],
  "success": boolean
}

List Documents

GET /documents

Response:
{
  "documents": [
    {
      "document_id": "string",
      "filename": "string",
      "pages": number,
      "chunks": number
    }
  ]
}

Delete Document

DELETE /documents/{document_id}

Response:
{
  "success": boolean,
  "message": "string"
}

Health Check

GET /health

Response:
{
  "status": "healthy",
  "llm_provider": "string",
  "documents_loaded": number
}

How It Works

PDF Processing Pipeline

1. Upload: User uploads a PDF file
2. Text Extraction: PyMuPDF extracts text from each page
3. Scanned Detection: System detects if pages are scanned (low text content)
4. OCR Fallback: Tesseract OCR processes scanned pages
5. Text Cleaning: Extracted text is cleaned and normalized
6. Chunking: Text is split into 500-character chunks with 100-character overlap
7. Embedding: Chunks are converted to embeddings using sentence-transformers
8. Storage: Embeddings are stored in ChromaDB for semantic search
9. Summarization: LLM generates document summary and key points

Question Answering Pipeline

1. Query Embedding: User question is converted to embedding
2. Semantic Search: Top 5 similar chunks are retrieved from ChromaDB
3. Context Building: Retrieved chunks are formatted with citations
4. LLM Processing: Question and context are sent to LLM
5. Answer Generation: LLM generates grounded answer with source references
6. Response: Answer with citations is returned to user

Configuration

Chunking Strategy

• Chunk Size: 500 characters
• Overlap: 100 characters
• Preserves page numbers and chunk IDs in metadata

Embedding Model

• Model: all-MiniLM-L6-v2 (384-dimensional embeddings)
• Provider: Sentence Transformers
• Storage: ChromaDB with cosine similarity

LLM Configuration

• Temperature: 0.3 (for consistent, grounded answers)
• Max Tokens: Default (varies by model)
• Providers: Google Gemini or OpenAI

Performance Optimization

OCR Strategy

• Only OCR pages with minimal extracted text (< 50 characters)
• Skips OCR for text-based PDFs to save processing time

Caching

• Embeddings are cached in ChromaDB
• Summaries are cached after generation
• Document metadata is stored in memory

Scalability

• ChromaDB can be migrated to Pinecone for cloud deployment
• FastAPI supports async operations
• Frontend uses efficient React patterns

Troubleshooting

Backend Issues

"GEMINI_API_KEY not set"

• Ensure .env file exists in backend directory
• Check that GEMINI_API_KEY is set correctly

"Tesseract not found"

• Install Tesseract OCR for your OS
• On macOS: brew install tesseract

"ChromaDB connection error"

• Ensure chroma_db/ directory has write permissions
• Check disk space availability

Frontend Issues

"Cannot connect to backend API"

• Ensure backend is running on http://localhost:8000
• Check NEXT_PUBLIC_API_URL in .env.local
• Verify CORS is enabled in FastAPI

"PDF upload fails"

• Check file size (max 50MB)
• Ensure file is valid PDF
• Check backend logs for detailed error

Future Enhancements

• Support for multiple PDFs in single session
• Streaming responses for faster feedback
• Advanced filtering and search options
• User authentication and document management
• Export answers and summaries to PDF
• Multi-language support
• Custom embedding models
• Fine-tuned LLM models
• Document comparison features
• Batch processing

License

MIT

Support

For issues and questions, please open an issue on GitHub.