Invoice OCR (Laravel 12 + Inertia + React)

Extract invoice number and invoice date from PDF invoices using Tesseract OCR (native C/C++ library) for image-based PDFs and OpenAI for smart parsing.

🖼 Screenshots

Upload Invoice

OCR Results

✨ Features

• Upload PDF invoices
• Smart text extraction:
  • Direct PDF text extraction (no OCR) for machine-readable PDFs
  • OCR fallback using Tesseract OCR for scanned/image-based PDFs
• AI-powered parsing with OpenAI
• Extracts:
  • Invoice Number
  • Invoice Date
• Clean UI with Inertia + React
• Easy to extend for more fields later

🧱 Tech Stack

• Backend: Laravel 12
• Frontend: Inertia.js + React
• OCR Engine: Tesseract OCR (native C/C++ library)
  • Repo: https://github.com/tesseract-ocr/tesseract
• AI Parsing: OpenAI API
• Build Tooling: Vite

⚙️ Requirements

• PHP 8.3+
• Composer
• Node.js 18+
• Tesseract OCR (native C/C++ binary – required for scanned/image PDFs)
• OpenAI API Key (required)

⚠️ This project will not work without:

• Tesseract installed on your system
• A valid OpenAI API key

🛠 Project Setup

1. Clone the repo

git clone https://github.com/levintoo/invoice-scan.git
cd invoice-ocr

2. Install backend deps

composer install

3. Install frontend deps

npm install

4. Environment setup

cp .env.example .env
php artisan key:generate

Set your OpenAI key:

OPENAI_API_KEY=sk-xxxxxxxxxxxxxxxx

5. Install Tesseract (Native Binary)

Tesseract is a native C/C++ OCR engine. You must install it on your OS.

Linux (Ubuntu/Debian):

sudo apt install tesseract-ocr

Verify:

tesseract --version

6. Start dev servers

php artisan serve
npm run dev

Open:

http://localhost:8000

How It Works

1. User uploads a PDF invoice

2. App first attempts direct text extraction from the PDF

3. If meaningful text is found → ➜ Text is sent directly to OpenAI (no OCR used)

4. If the PDF is scanned / image-based or text extraction fails → ➜ The document is passed through Tesseract OCR

5. Extracted text is sent to OpenAI

6. OpenAI returns structured data:

   • invoice_number
   • invoice_date

7. Results are displayed in the UI

This approach avoids unnecessary OCR and speeds up processing for machine-readable PDFs.