DocIntel — Document Intelligence API

Upload any document → get structured, machine-readable data back.

DocIntel is an end-to-end document intelligence pipeline that takes PDFs and images through layout detection, OCR, and named entity extraction — returning clean, structured JSON. Built for real-world use cases like digitizing land deeds, processing research papers, and extracting data from scanned forms.

---

Screenshots

Upload Interface

Drag-and-drop document upload with supported format indicators and processing capabilities overview.

Text Blocks — Layout Detection Results

Extracted text blocks with semantic type classification (Title, Paragraph, List, Table, Footer), confidence scores, and bounding box coordinates.

Entity Extraction

Named entities detected across pages — people, locations, dates, monetary values, and organizations — with type-coded badges and character offsets.

Structured JSON Output

Full API response with syntax highlighting — ready for downstream integration.

---

The Problem

Organizations worldwide — from smallholder farmers registering land deeds, to NGOs digitizing health records, to researchers processing paper archives — need to convert unstructured documents into machine-readable data. Existing tools are either expensive cloud APIs, or fragmented open-source libraries that require significant glue code.

DocIntel provides a single API endpoint that handles the entire pipeline: upload a document, get structured JSON back with text blocks, bounding boxes, and extracted entities.

Features

• Multi-format support — PDF, PNG, JPG, JPEG, TIFF, BMP
• Layout detection — Identifies titles, paragraphs, tables, lists, and figures using heuristic-based connected component analysis
• OCR extraction — Tesseract-powered text extraction with per-block confidence scores
• Named entity recognition — Extracts dates, monetary amounts, percentages, emails, phone numbers, and addresses via regex patterns; optional spaCy integration for PERSON, ORG, GPE entities
• Structured output — Clean JSON with bounding boxes, block types, and page-level organization
• Async processing — Submit large documents for background processing with status polling
• React dashboard — Upload documents and explore results with an interactive UI

Architecture

Tech Stack

Layer	Technology
API	FastAPI, Uvicorn, Pydantic v2
OCR	Tesseract (via pytesseract)
PDF	pdf2image (Poppler), pypdf
NLP	Regex patterns + spaCy (optional)
Image	Pillow, NumPy
Frontend	React 18, Vite, Tailwind CSS
Container	Docker, Docker Compose

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Tesseract OCR (brew install tesseract / apt install tesseract-ocr)
• Poppler (brew install poppler / apt install poppler-utils)

Option 1: Docker (Recommended)

git clone https://github.com/Jonathan-321/docintel.git
cd docintel
docker-compose up --build

The API will be available at http://localhost:8000 and the frontend at http://localhost:3000.

Option 2: Manual Setup

Backend:

cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt

# Optional: install spaCy model for enhanced NER
python -m spacy download en_core_web_sm

uvicorn app.main:app --reload --port 8000

Frontend:

cd frontend
npm install
npm run dev

API Reference

Process Document (Sync)

POST /api/v1/process
Content-Type: multipart/form-data

Parameter	Type	Description
file	File	Document to process

Response:

{
  "filename": "land_deed.pdf",
  "num_pages": 2,
  "pages": [
    {
      "page_number": 1,
      "width": 2550,
      "height": 3300,
      "blocks": [
        {
          "text": "CERTIFICATE OF TITLE",
          "confidence": 0.95,
          "bbox": { "x": 120, "y": 50, "width": 800, "height": 60 },
          "block_type": "title"
        }
      ],
      "entities": [
        {
          "text": "January 15, 2024",
          "label": "DATE",
          "start": 45,
          "end": 61
        },
        {
          "text": "$150,000",
          "label": "MONEY",
          "start": 120,
          "end": 128
        }
      ]
    }
  ],
  "metadata": {
    "ocr_engine": "tesseract",
    "language": "eng",
    "spacy_available": true
  },
  "processing_time_ms": 1234.56
}

Process Document (Async)

POST /api/v1/process/async
Content-Type: multipart/form-data

Returns a job_id for polling:

{
  "job_id": "abc123",
  "status": "processing",
  "progress": 0.0
}

Check Job Status

GET /api/v1/status/{job_id}

List Supported Formats

GET /api/v1/formats

Interactive Docs

FastAPI auto-generates interactive API documentation:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

Project Structure

docintel/
├── backend/
│   ├── app/
│   │   ├── api/
│   │   │   └── routes.py          # API endpoints
│   │   ├── core/
│   │   │   └── config.py          # Settings & configuration
│   │   ├── models/
│   │   │   └── schemas.py         # Pydantic data models
│   │   ├── pipeline/
│   │   │   ├── processor.py       # Main document processor
│   │   │   ├── layout.py          # Layout detection
│   │   │   ├── ocr.py             # OCR engine
│   │   │   └── entities.py        # Entity extraction
│   │   ├── utils/
│   │   │   └── file_utils.py      # File handling utilities
│   │   └── main.py                # FastAPI app entry point
│   ├── tests/
│   │   ├── test_health.py
│   │   ├── test_process.py
│   │   └── test_entities.py
│   ├── Dockerfile
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── FileUpload.jsx
│   │   │   ├── ResultsView.jsx
│   │   │   ├── ProcessingStatus.jsx
│   │   │   ├── Header.jsx
│   │   │   └── Sidebar.jsx
│   │   ├── api/
│   │   │   └── client.js
│   │   ├── App.jsx
│   │   └── main.jsx
│   ├── package.json
│   └── vite.config.js
├── docker-compose.yml
├── LICENSE
└── README.md

Development

Running Tests

cd backend
pip install pytest pytest-asyncio httpx
pytest -v

Environment Variables

Copy .env.example to .env in the backend directory:

Variable	Default	Description
DEBUG	false	Enable debug mode
OCR_LANGUAGE	eng	Tesseract language pack
TESSERACT_CMD	tesseract	Path to tesseract binary
MAX_FILE_SIZE	20971520	Max upload size in bytes (20MB)
UPLOAD_DIR	/tmp/docintel	Temporary upload directory

Use Cases

• Land administration — Digitize property deeds and extract parcel numbers, dates, and monetary values
• Healthcare — Process scanned medical records and extract patient information, dates, and diagnoses
• Research — Bulk-process academic papers and extract titles, authors, citations, and key findings
• Finance — Extract transaction data from scanned invoices, receipts, and bank statements
• Government — Digitize census forms, birth certificates, and other civil documents

Roadmap

• Table structure recognition (row/column detection)
• Handwriting recognition support
• Multi-language OCR (Arabic, Kinyarwanda, French)
• Document classification (invoice vs. letter vs. form)
• Batch processing endpoint
• Webhook notifications for async jobs
• Fine-tuned layout detection model (YOLO-based)

License

MIT License — see LICENSE for details.