README

Turn any image or PDF into structured, actionable data.

A powerful, developer-friendly Laravel package that reads text from images and PDFs, understands the content, fixes scanning errors with AI, and delivers clean, structured data directly to your application.

Why this package? Most OCR tools just give you a dump of raw text. This package gives you typed DTOs, structured fields, and confidence scores. It knows the difference between an Invoice Number and a Phone Number.

📸 Demo

📑 Table of Contents

✨ Features
📐 Architecture
🚀 Installation
⚙️ Configuration
📖 Usage
🖥️ Artisan Commands
- Create a Template
- Process a Document
🎨 Blade Component
🗄️ Database Schema
🧩 OCR Driver Interface
- Supported Formats by Driver
- Building a Custom Driver
🤖 AI Cleanup Details
- Basic Rules (No API Required)
- AI-Powered Cleanup (via Laravel AI SDK)
🧪 Testing
📄 API Reference
📄 License

✨ Features

🧠 Multi-Driver OCR Engine: Seamlessly switch between Tesseract (offline/privacy-first), Google Cloud Vision, AWS Textract, or Azure Computer Vision drivers.
🤖 AI-Powered Cleanup: Uses laravel/ai SDK with the CleanupAgent to fix OCR typos (e.g., arnount → amount, nurnber → number) and normalize data formats. Supports OpenAI, Anthropic, Gemini, Ollama, DeepSeek, Groq, Mistral, and more.
📦 Typed DTOs: Returns an OcrResult data transfer object with text, confidence, bounds, and metadata properties — not just raw arrays.
📑 Advanced Invoice Extraction: Specialized algorithms to extract line items (quantity, description, unit price, total), subtotals, tax, shipping, and totals from complex invoice layouts.
🔍 Auto-Classification: Automatically detects document types (Invoice, Receipt, Contract, Purchase Order, Shipping, General) using keyword scoring.
📋 Reusable Templates: Define document templates with regex patterns and positional extraction to target specific fields. Supports template creation, import/export, and auto-matching.
⚡ Workflows: Define custom processing pipelines per document type in your config (e.g., "If Invoice → Extract Tables → Verify Required Fields").
🎨 Blade Component: Built-in <x-laravel-ocr::document-preview> component to visualize results with bounding box overlays, inline editing, and data export.
🖥️ Artisan Commands: CLI tools for template creation (laravel-ocr:create-template) and document processing (laravel-ocr:process).
💾 Database Persistence: Optionally save processed results to the ocr_processed_documents table with template associations, confidence scores, and processing times.
🔒 Enterprise Security: Encrypted storage options, MIME-type validation, malware scanning, and full offline support for sensitive data.

📐 Architecture

src/
├── Agents/
│   └── CleanupAgent.php          # Laravel AI SDK agent for OCR post-processing
├── Console/Commands/
│   ├── CreateTemplateCommand.php  # artisan laravel-ocr:create-template
│   └── ProcessDocumentCommand.php # artisan laravel-ocr:process
├── Contracts/
│   └── OCRDriver.php             # Interface: extract, extractTable, extractBarcode, extractQRCode, etc.
├── Drivers/
│   ├── TesseractDriver.php       # Offline OCR via thiagoalessio/tesseract_ocr
│   ├── GoogleVisionDriver.php    # Google Cloud Vision API
│   ├── AWSTextractDriver.php     # AWS Textract API
│   └── AzureOCRDriver.php        # Azure Computer Vision API
├── DTOs/
│   └── OcrResult.php             # Readonly DTO: text, confidence, bounds, metadata
├── Enums/
│   ├── DocumentType.php          # Invoice, Receipt, Contract, PurchaseOrder, Shipping, General
│   └── OcrDriver.php             # Tesseract, GoogleVision, AWSTextract, Azure
├── Exceptions/
│   ├── AICleanupException.php
│   ├── DocumentParserException.php
│   └── OCRException.php
├── Facades/
│   └── LaravelOcr.php            # Facade for OCRManager
├── Models/
│   ├── DocumentTemplate.php      # Template definition with fields relationship
│   ├── ProcessedDocument.php     # Stored OCR results with field accessors
│   └── TemplateField.php         # Individual field definition with validation rules
├── Services/
│   ├── OCRManager.php            # Driver manager (extends Illuminate\Support\Manager)
│   ├── AICleanupService.php      # AI cleanup + basic rule-based typo correction
│   ├── DocumentParser.php        # Main parsing engine with field/table/metadata extraction
│   └── TemplateManager.php       # Template CRUD, import/export, auto-matching
└── LaravelOcrServiceProvider.php  # Service provider: bindings, config, migrations, views, commands

🚀 Installation

Requires PHP 8.4+ and Laravel 10+ / 11 / 12.

1. Install via Composer

composer require mayaram/laravel-ocr

2. Publish Configuration & Assets

php artisan vendor:publish --tag=laravel-ocr-config
php artisan vendor:publish --tag=laravel-ocr-migrations
php artisan migrate

3. Publish Views (Optional)

php artisan vendor:publish --tag=laravel-ocr-views

4. Install Tesseract (Default Driver)

# macOS
brew install tesseract

# Ubuntu/Debian
sudo apt-get install tesseract-ocr

# Verify installation
tesseract --version

⚙️ Configuration

Set your preferred driver and credentials in your .env file.

OCR Driver Selection

# Options: tesseract, google_vision, aws_textract, azure
LARAVEL_OCR_DRIVER=tesseract

Tesseract (Offline / Privacy-First — Default)

All processing happens on your server. No data leaves your infrastructure.

LARAVEL_OCR_DRIVER=tesseract
TESSERACT_BINARY=/usr/bin/tesseract
TESSERACT_LANGUAGE=eng
TESSERACT_TIMEOUT=60

Google Cloud Vision

LARAVEL_OCR_DRIVER=google_vision
GOOGLE_VISION_KEY_FILE=/path/to/service-account.json
GOOGLE_VISION_PROJECT_ID=your-project-id

Requires: composer require google/cloud-vision

AWS Textract

LARAVEL_OCR_DRIVER=aws_textract
AWS_ACCESS_KEY_ID=your-key
AWS_SECRET_ACCESS_KEY=your-secret
AWS_DEFAULT_REGION=us-east-1

Included: aws/aws-sdk-php is a core dependency.

Azure Computer Vision

LARAVEL_OCR_DRIVER=azure
AZURE_OCR_ENDPOINT=https://your-resource.cognitiveservices.azure.com
AZURE_OCR_KEY=your-subscription-key
AZURE_OCR_VERSION=3.2

AI Cleanup (via Laravel AI SDK)

Enable AI-powered OCR post-processing to fix scanning errors and normalize data formats. The package uses laravel/ai with a dedicated CleanupAgent that supports multiple LLM providers.

LARAVEL_OCR_AI_CLEANUP=true
LARAVEL_OCR_AI_PROVIDER=openai
LARAVEL_OCR_AI_TIMEOUT=60
# Optional: Set a default custom prompt for all AI cleanup calls
# LARAVEL_OCR_AI_CUSTOM_PROMPT="Extract all amounts in INR. Format names in Title Case."

Supported AI Providers

Set the API key for your chosen provider:

Provider	Env Variable	Driver Key
OpenAI	`OPENAI_API_KEY`	`openai`
Anthropic	`ANTHROPIC_API_KEY`	`anthropic`
Google Gemini	`GEMINI_API_KEY`	`gemini`
Ollama (Local)	`OLLAMA_API_KEY`	`ollama`
DeepSeek	`DEEPSEEK_API_KEY`	`deepseek`
Groq	`GROQ_API_KEY`	`groq`
Mistral	`MISTRAL_API_KEY`	`mistral`
Azure OpenAI	`AZURE_OPENAI_API_KEY`	`azure`
Cohere	`COHERE_API_KEY`	`cohere`
OpenRouter	`OPENROUTER_API_KEY`	`openrouter`
xAI	`XAI_API_KEY`	`xai`
Jina	`JINA_API_KEY`	`jina`
VoyageAI	`VOYAGEAI_API_KEY`	`voyageai`
ElevenLabs	`ELEVENLABS_API_KEY`	`eleven`

Queue Configuration

Process documents asynchronously:

LARAVEL_OCR_QUEUE_ENABLED=false
LARAVEL_OCR_QUEUE_CONNECTION=default
LARAVEL_OCR_QUEUE_NAME=ocr-processing

Storage & Security

LARAVEL_OCR_STORAGE_DISK=local
LARAVEL_OCR_ENCRYPT_DATA=false
LARAVEL_OCR_SCAN_MALWARE=false

📖 Usage

1. Simple Text Extraction

The LaravelOcr facade provides a simple entry point via the OCRManager.

use Mayaram\LaravelOcr\Facades\LaravelOcr;

// Extract from a local file path
$result = LaravelOcr::extract('/path/to/document.png');
echo $result['text'];
// "INVOICE #1001..."

// Extract from an UploadedFile
$result = LaravelOcr::extract(request()->file('document'));

// Extract a table
$tableResult = LaravelOcr::extractTable('/path/to/invoice.png');
foreach ($tableResult['table'] as $row) {
    echo implode(' | ', $row) . "\n";
}

2. Smart Parsing (Structured Data via `OcrResult` DTO)

For powerful data extraction, use the DocumentParser. It returns a typed OcrResult DTO with text, confidence, bounds, and metadata properties.

use Mayaram\LaravelOcr\DTOs\OcrResult;

/** @var \Mayaram\LaravelOcr\Services\DocumentParser $parser */
$parser = app('laravel-ocr.parser');

$result = $parser->parse('storage/invoices/inv-2024.pdf', [
    'document_type' => 'invoice',
    'use_ai_cleanup' => true,
    'save_to_database' => true,
]);

// Access the OcrResult DTO properties
echo $result->text;                                  // Full extracted text
echo $result->confidence;                            // e.g., 0.98
echo $result->metadata['processing_time'];           // e.g., 1.2
echo $result->metadata['document_type'];             // "invoice"
echo $result->metadata['ai_cleanup_used'];           // true

// Access structured fields
$fields = $result->metadata['fields'];
$invoiceNumber = $fields['invoice_number']['value'];
$totalAmount = $fields['totals']['total']['amount'];

3. Batch Processing

Process multiple documents at once:

$parser = app('laravel-ocr.parser');

$results = $parser->parseBatch([
    'storage/invoices/inv-001.pdf',
    'storage/invoices/inv-002.pdf',
    'storage/invoices/inv-003.pdf',
], ['document_type' => 'invoice']);

foreach ($results as $result) {
    echo $result->text . "\n---\n";
}

4. Working with Line Items & Tables

The package includes an Advanced Invoice Extractor capable of parsing complex invoice tables into structured arrays.

$result = $parser->parse($invoicePath, [
    'document_type' => 'invoice',
]);

// Line items are extracted automatically for invoices
$lineItems = $result->metadata['fields']['line_items'] ?? [];

foreach ($lineItems as $item) {
    echo "{$item['description']}: {$item['quantity']} x \${$item['unit_price']} = \${$item['total']}\n";
}
// Output:
// Web Hosting: 12 x $10.00 = $120.00
// Domain Name: 1 x $15.00 = $15.00

// Invoice totals
$totals = $result->metadata['fields']['totals'] ?? [];
echo "Subtotal: " . ($totals['subtotal']['formatted'] ?? 'N/A');
echo "Tax: " . ($totals['tax']['formatted'] ?? 'N/A');
echo "Total: " . ($totals['total']['formatted'] ?? 'N/A');

5. Templates

Define reusable templates to target specific fields using regex patterns. Templates are stored in the database and support import/export.

use Mayaram\LaravelOcr\Facades\LaravelOcr;

// 1. Create a Template
$templateManager = app('laravel-ocr.templates');

$template = $templateManager->create([
    'name' => 'TechCorp Invoice',
    'type' => 'invoice',
    'description' => 'Template for TechCorp invoices',
    'fields' => [
        [
            'key' => 'order_id',
            'label' => 'Order ID',
            'pattern' => '/Order\s*ID:\s*([A-F0-9]+)/i',
            'type' => 'string',
            'validators' => ['required' => true],
        ],
        [
            'key' => 'total_amount',
            'label' => 'Total Amount',
            'pattern' => '/Total:\s*\$?([\d,]+\.?\d*)/i',
            'type' => 'currency',
        ],
    ],
]);

// 2. Apply a template during extraction
$result = LaravelOcr::extractWithTemplate($file, $template->id);

// 3. Auto-detect template based on content
$parser = app('laravel-ocr.parser');
$result = $parser->parse($file, ['auto_detect_template' => true]);

// 4. Export / Import templates
$json = $templateManager->exportTemplate($template->id);
$imported = $templateManager->importTemplate('/path/to/template.json');

// 5. Duplicate a template
$clone = $template->duplicate('TechCorp Invoice v2');

6. Workflows

Configure processing pipelines in config/laravel-ocr.php to standardize how different document types are handled.

// config/laravel-ocr.php
'workflows' => [
    'invoice' => [
        'options' => [
            'use_ai_cleanup' => true,
            'auto_detect_template' => true,
            'extract_tables' => true,
        ],
        'post_processors' => [
            ['class' => 'App\OCR\Processors\InvoiceProcessor'],
        ],
        'validators' => [
            ['type' => 'required_fields', 'fields' => ['invoice_number', 'total']],
        ],
    ],

    'receipt' => [
        'options' => [
            'use_ai_cleanup' => true,
            'extract_line_items' => true,
        ],
        'post_processors' => [
            ['class' => 'App\OCR\Processors\ReceiptProcessor'],
        ],
    ],
],

// Usage
$parser = app('laravel-ocr.parser');
$result = $parser->parseWithWorkflow($file, 'invoice');

7. Switching Drivers at Runtime

use Mayaram\LaravelOcr\Facades\LaravelOcr;

// Use Tesseract (default)
$result = LaravelOcr::driver('tesseract')->extract($document);

// Switch to AWS Textract for this request
$result = LaravelOcr::driver('aws_textract')->extract($document);

// Switch to Google Vision
$result = LaravelOcr::driver('google_vision')->extract($document);

8. Document Metadata Extraction

$parser = app('laravel-ocr.parser');

$metadata = $parser->extractMetadata('/path/to/document.pdf');

// Returns:
// [
//     'file_name' => 'document.pdf',
//     'file_size' => 102400,
//     'mime_type' => 'application/pdf',
//     'created_at' => '2024-01-15 10:30:00',
//     'modified_at' => '2024-01-15 10:30:00',
//     'pdf_pages' => 3,
//     'pdf_author' => 'John Doe',
//     'pdf_title' => 'Q4 Invoice',
//     'pdf_creator' => 'Microsoft Word',
// ]

9. Working with Processed Documents

use Mayaram\LaravelOcr\Models\ProcessedDocument;

// Query processed documents
$documents = ProcessedDocument::where('document_type', 'invoice')
    ->where('confidence_score', '>=', 0.7)
    ->latest()
    ->get();

foreach ($documents as $doc) {
    // Get a specific field value
    $invoiceNo = $doc->getFieldValue('invoice_number');

    // Get all field values as a flat array
    $allFields = $doc->getAllFieldValues();

    // Check if the result is valid (status=completed & confidence >= 0.7)
    if ($doc->isValid()) {
        // Process the document
    }
}

🖥️ Artisan Commands

Create a Template

# Basic
php artisan laravel-ocr:create-template "My Invoice" invoice

# Interactive (prompts for fields, patterns, validators)
php artisan laravel-ocr:create-template "My Invoice" invoice --interactive

Process a Document

# Basic processing
php artisan laravel-ocr:process /path/to/document.pdf

# With options
php artisan laravel-ocr:process /path/to/invoice.pdf \
    --type=invoice \
    --template=1 \
    --ai-cleanup \
    --save \
    --output=json

Flags:

Flag	Description
`--template=ID`	Use a specific template
`--type=TYPE`	Set document type (invoice, receipt, etc.)
`--ai-cleanup`	Enable AI-powered cleanup
`--save`	Save results to database
`--output=FORMAT`	Output format: `json` or `table` (default)

🎨 Blade Component

Preview extracted documents and data directly in your UI with the built-in Alpine.js component.

<x-laravel-ocr::document-preview
    :document="$processedDocument"
    :show-overlay="true"
    :show-actions="true"
    :show-image="true"
/>

Props:

Prop	Type	Default	Description
`document`	`array`	required	Document data with `url`, `documentId`, and `fields`
`show-overlay`	`bool`	`false`	Show bounding box overlays on the document
`show-actions`	`bool`	`true`	Show Save/Export/Reprocess action buttons
`show-image`	`bool`	`true`	Use `<img>` tag (`true`) or `<iframe>` (`false`)

Expected Document Structure:

$processedDocument = [
    'url' => asset('storage/documents/invoice.pdf'),
    'documentId' => $document->id,
    'fields' => [
        [
            'key' => 'invoice_number',
            'label' => 'Invoice Number',
            'value' => 'INV-2024-001',
            'confidence' => 0.95,
            'bounds' => ['x' => 100, 'y' => 50, 'width' => 200, 'height' => 30],
        ],
        // ...more fields
    ],
];

Features:

Click fields to highlight them on the document
Inline editing of extracted values
Confidence badges: 🟢 High (≥80%), 🟡 Medium (≥60%), 🔴 Low (<60%)
Export extracted data as JSON
Reprocess documents with one click

🗄️ Database Schema

The package creates three tables:

`ocr_templates`

Column	Type	Description
`id`	bigint	Primary key
`name`	string	Template name
`description`	text	Optional description
`type`	string(50)	Document type (invoice, receipt, etc.)
`layout`	json	Layout configuration
`is_active`	boolean	Whether template is active
`version`	string(10)	Template version (default: 1.0)

`ocr_template_fields`

Column	Type	Description
`id`	bigint	Primary key
`template_id`	foreignId	References `ocr_templates`
`key`	string(50)	Field identifier (snake_case)
`label`	string	Human-readable label
`type`	string(30)	Field type: string, numeric, date, currency, email, phone
`pattern`	text	Regex pattern for extraction
`position`	json	Positional extraction config (line, start, end)
`validators`	json	Validation rules (required, regex, length, type)
`default_value`	string	Fallback value
`order`	integer	Display order

`ocr_processed_documents`

Column	Type	Description
`id`	bigint	Primary key
`original_filename`	string	Original file name
`document_type`	string(50)	Detected or specified type
`extracted_data`	json	Full extraction result
`template_id`	foreignId	Template used (nullable)
`confidence_score`	decimal(3,2)	Overall confidence
`processing_time`	decimal(8,3)	Time in seconds
`user_id`	bigint	User who processed (nullable)
`status`	string(20)	completed, failed, etc.
`error_message`	text	Error details if failed

🧩 OCR Driver Interface

All drivers implement the Mayaram\LaravelOcr\Contracts\OCRDriver interface:

interface OCRDriver
{
    public function extract($document, array $options = []): array;
    public function extractTable($document, array $options = []): array;
    public function extractBarcode($document, array $options = []): array;
    public function extractQRCode($document, array $options = []): array;
    public function getSupportedLanguages(): array;
    public function getSupportedFormats(): array;
}

Supported Formats by Driver

Format	Tesseract	Google Vision	AWS Textract	Azure
JPG/JPEG	✅	✅	✅	✅
PNG	✅	✅	✅	✅
PDF	✅	✅	✅	✅
TIFF	✅	✅	❌	✅
BMP	✅	✅	❌	✅
GIF	❌	✅	❌	❌
WebP	❌	✅	❌	❌

Building a Custom Driver

use Mayaram\LaravelOcr\Contracts\OCRDriver;

class MyCustomDriver implements OCRDriver
{
    public function extract($document, array $options = []): array
    {
        // Your implementation
        return [
            'text' => 'extracted text',
            'confidence' => 0.95,
            'bounds' => [],
            'metadata' => ['engine' => 'custom'],
        ];
    }

    public function extractTable($document, array $options = []): array { /* ... */ }
    public function extractBarcode($document, array $options = []): array { /* ... */ }
    public function extractQRCode($document, array $options = []): array { /* ... */ }
    public function getSupportedLanguages(): array { return ['en' => 'English']; }
    public function getSupportedFormats(): array { return ['jpg', 'png', 'pdf']; }
}

🤖 AI Cleanup Details

The AICleanupService provides two modes:

Basic Rules (No API Required)

Set provider to basic to use built-in typo correction and field normalization without any AI provider:

$result = $parser->parse($document, [
    'use_ai_cleanup' => true,
    'provider' => 'basic',
]);

Built-in corrections:

Common OCR typos: invOice → invoice, arnount → amount, nurnber → number, custorner → customer, payrnent → payment
OCR pattern fixes: rn → m, O before numbers → 0, l before numbers → 1
Field normalization by type: numeric, date (→ Y-m-d), currency, email, phone

AI-Powered Cleanup (via Laravel AI SDK)

Uses the CleanupAgent with your configured LLM provider for intelligent correction:

$result = $parser->parse($document, [
    'use_ai_cleanup' => true,
    // Uses LARAVEL_OCR_AI_PROVIDER from config
]);

The agent is instructed to:

Fix typos and OCR errors contextually
Standardize formats (dates to YYYY-MM-DD, currency to decimal)
Preserve values it can't confidently fix
Return valid JSON matching the input structure

Custom Prompt

You can pass additional instructions to the AI cleanup agent to customize its behavior:

// Per-call custom prompt
$result = $parser->parse($document, [
    'use_ai_cleanup' => true,
    'custom_prompt' => 'Extract all amounts in INR. Normalize phone numbers to +91 format.',
]);

Or set a default custom prompt in config/laravel-ocr.php:

'ai_cleanup' => [
    'custom_prompt' => 'Always extract Hindi text. Format dates as DD/MM/YYYY.',
],

The custom prompt is appended as "Additional Instructions" to the default cleanup prompt, so the base OCR correction behavior is always preserved.

🧪 Testing

The package uses Pest for testing with three test suites:

# Run all tests
composer test

# Or directly with Pest
./vendor/bin/pest

# Run specific suite
./vendor/bin/pest --testsuite=Unit
./vendor/bin/pest --testsuite=Feature
./vendor/bin/pest --testsuite=Integration

Tests use SQLite in-memory database automatically via phpunit.xml configuration.

📄 API Reference

`LaravelOcr` Facade (via `OCRManager`)

Method	Returns	Description
`extract($document, $options)`	`array`	Extract text from document
`extractWithTemplate($document, $templateId, $options)`	`array`	Extract and apply template
`extractTable($document, $options)`	`array`	Extract table data
`extractBarcode($document, $options)`	`array`	Extract barcode
`extractQRCode($document, $options)`	`array`	Extract QR code
`driver($name)`	`OCRDriver`	Switch to a specific driver

`DocumentParser` (via `app('laravel-ocr.parser')`)

Method	Returns	Description
`parse($document, $options)`	`OcrResult`	Full parsing pipeline
`parseBatch($documents, $options)`	`OcrResult[]`	Process multiple documents
`parseWithWorkflow($document, $workflow)`	`OcrResult`	Parse using a named workflow
`extractMetadata($document)`	`array`	Extract file & PDF metadata

`OcrResult` DTO

Property	Type	Description
`text`	`string`	Full extracted text
`confidence`	`float`	Overall confidence score (0–1)
`bounds`	`array`	Bounding box / layout data
`metadata`	`array`	Processing time, document type, fields, template used, etc.

`TemplateManager` (via `app('laravel-ocr.templates')`)

Method	Returns	Description
`create($data)`	`DocumentTemplate`	Create a template with fields
`applyTemplate($extractedData, $templateId)`	`array`	Apply template to extracted data
`findTemplateByContent($text)`	`?DocumentTemplate`	Auto-detect matching template
`importTemplate($filePath)`	`DocumentTemplate`	Import from JSON file
`exportTemplate($templateId)`	`string`	Export as JSON string

📄 License

The MIT License (MIT). Please see License File for more information.

mayaram / laravel-ocr

Maintainers

Details