Laravel AI Assistant

Laravel AI Assistant is a modern, production-ready package for integrating OpenAI's powerful language models into your Laravel applications. Built on a Single Source of Truth (SSOT) architecture, it provides Ai::responses() as the unified entry point for all OpenAI operations—chat, streaming, audio transcription/translation/speech, and image generation/editing/variations. This design eliminates API confusion while providing a clean, fluent interface with excellent IDE support.

✨ What's New in v3.0

SSOT Architecture: Version 3.0 establishes Ai::responses() as the single, unified entry point for all OpenAI operations. This architectural shift provides:

🎯 Single Entry Point: One API to learn instead of multiple competing approaches
🔧 Unified Interface: Consistent builder pattern across chat, audio, and image operations
📚 Better DX: Excellent IDE autocompletion and type safety throughout
🚀 Future-Proof: Aligns with OpenAI's modern API design principles

Deprecated Components (see MIGRATION.md for migration guide):

❌ AiAssistant class and facade → Use Ai::responses() or Ai::chat()
❌ OpenAIClientFacade → Use Ai facade methods
❌ OpenAiRepository → Use Ai::responses() with adapters
❌ Compat directory (/src/Compat/) → Use SSOT API

Migration Path: All deprecated components will be removed in v4.0. See the Migration Guide section below for detailed examples.

🚀 Quick Start (≤5 Minutes)

Step 1: Install (1 min)

composer require creativecrafts/laravel-ai-assistant
php artisan ai:install

Step 2: Configure (1 min)

Set your OpenAI API key in .env:

OPENAI_API_KEY=your-openai-api-key-here

Step 3: Hello World Chat (1 min)

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Simple one-off request
$response = Ai::quick('Explain Laravel queues in simple terms');
echo $response->text;

Step 4: Streaming Response (2 min)

// Real-time streaming for better UX
foreach (Ai::stream('Tell me about Laravel') as $chunk) {
    echo $chunk;
    flush();
}

Step 5: Audio Transcription (1 min)

// Transcribe audio files to text
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/recording.mp3'),
        'action' => 'transcribe',
    ])
    ->send();

echo $response->text; // "Hello, this is a test recording..."

Step 6: Image Generation (1 min)

// Generate images from text descriptions
$response = Ai::responses()
    ->input()
    ->image([
        'prompt' => 'A futuristic Laravel logo with neon lights',
    ])
    ->send();

// Save generated images
$response->saveImages(storage_path('images'));

That's it! You now have AI-powered chat, streaming, audio, and image capabilities. See examples/ for more patterns.

📖 Core Concepts

Why SSOT Architecture?

Version 3.0 introduces a Single Source of Truth (SSOT) design where Ai::responses() serves as the unified entry point for all OpenAI operations:

🎯 Single Entry Point: One consistent API for chat, audio, and image operations
🔧 Intelligent Routing: Automatically routes requests to the correct OpenAI endpoint based on input type
📚 Better DX: Fluent builder pattern with excellent IDE autocompletion across all operations
🚀 Simplified Learning: Master one API instead of juggling multiple interfaces
Future-proof: Built on OpenAI's modern Responses API and Conversations API

Note: Legacy APIs like AiAssistant, OpenAIClientFacade, and OpenAiRepository are deprecated in v3.0 and will be removed in v4.0. See MIGRATION.md for migration instructions.

Available Methods

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// SSOT API - Primary entry point (recommended)
Ai::responses();                         // Unified API for all operations (chat, audio, images)
Ai::conversations();                     // Stateful conversation management

// Convenience methods built on SSOT
Ai::quick('Your prompt here');           // One-off requests
Ai::chat('System instructions');         // Multi-turn conversations
Ai::stream('Your prompt here');          // Streaming responses

// Low-level completion API (advanced)
$ai->complete(Mode::TEXT, Transport::SYNC, $request);

Comparison: Sync vs Stream

Feature	SYNC	STREAM
Response Time	Wait for complete response	First token arrives quickly
Memory Usage	Lower	Higher (buffering)
User Experience	Blocking	Progressive, real-time
Best For	Batch processing, APIs	Interactive UIs, long responses

Feature Capabilities

Capability	Description	Supported
Text Chat	Single-turn and multi-turn conversations	✅
Streaming	Real-time token-by-token responses	✅
Tool Calls	Function calling and external integrations	✅
Audio Transcription	Convert audio to text (speech-to-text)	✅
Audio Translation	Translate audio to English	✅
Text-to-Speech	Generate natural speech from text	✅
Image Generation	Create images from text prompts (DALL-E)	✅
Image Editing	Modify existing images with prompts	✅
Image Variations	Generate variations of existing images	✅
Webhooks	Asynchronous background processing	✅
Observability	Built-in logging, metrics, and monitoring	✅

🎯 Unified Completion API (Recommended)

The unified completion API provides a consistent interface for all AI operations with explicit control over mode and transport.

Quick Example

use CreativeCrafts\LaravelAiAssistant\Services\AiManager;
use CreativeCrafts\LaravelAiAssistant\Enums\{Mode, Transport};
use CreativeCrafts\LaravelAiAssistant\DataTransferObjects\CompletionRequest;

$ai = app(AiManager::class);

// Text completion (sync)
$result = $ai->complete(
    Mode::TEXT,
    Transport::SYNC,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'prompt' => 'Write a haiku about Laravel',
        'temperature' => 0.2,
    ])
);

echo (string) $result;

Mode Options

Mode::TEXT - Simple text completion (single prompt string)
Mode::CHAT - Chat with message history (conversation context)

Transport Options

Transport::SYNC - Wait for complete response (blocking)
Transport::STREAM - Stream tokens as generated (accumulated into final result)

Chat Example

$result = $ai->complete(
    Mode::CHAT,
    Transport::SYNC,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'messages' => [
            ['role' => 'system', 'content' => 'You are a helpful assistant'],
            ['role' => 'user', 'content' => 'Explain dependency injection'],
        ],
    ])
);

$data = $result->toArray();

Streaming Example

// Stream accumulates to final result
$finalResult = $ai->complete(
    Mode::TEXT,
    Transport::STREAM,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'prompt' => 'Tell me a story',
    ])
);

echo (string) $finalResult;

Note: For incremental streaming events (processing chunks as they arrive), use Ai::stream() instead.

See docs/API.md for complete API reference.

📚 Cookbook

Quick Calls (One-Off Requests)

Use Ai::quick() for stateless, single-turn requests.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Simple text response
$response = Ai::quick('What is dependency injection?');
echo $response->text;

// Force JSON output
$response = Ai::quick([
    'message' => 'Give me 3 fun facts about Laravel',
    'response_format' => 'json',
]);
$facts = json_decode($response->text, true);

// Custom model and temperature
$response = Ai::quick([
    'message' => 'Write a haiku about PHP',
    'model' => 'gpt-4o-mini',
    'temperature' => 0.9,
]);
echo $response->text;

Stateful Chat (Multi-Turn Conversations)

Use Ai::chat() for conversations that maintain context across multiple turns.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Create a chat session
$chat = Ai::chat('You are a helpful Laravel expert');

// First turn
$chat->message('What are service providers?');
$response = $chat->send();
echo $response->text;

// Continue the conversation (maintains context)
$chat->message('Can you show me an example?');
$response = $chat->send();
echo $response->text;

// Access conversation history
$messages = $chat->getMessages();

Customizing Chat Sessions

$chat = Ai::chat('You are a coding assistant')
    ->setModelName('gpt-4')
    ->setTemperature(0.2)
    ->setMaxTokens(500);

$response = $chat
    ->message('Explain middleware in Laravel')
    ->send();

Streaming Responses

Stream responses token-by-token for real-time UX (perfect for SSE, WebSockets, or Livewire).

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Simple streaming
foreach (Ai::stream('Tell me a story about Laravel') as $chunk) {
    echo $chunk; // Outputs each token as it arrives
}

// Streaming with chat sessions
$chat = Ai::chat('You are a storyteller');
$chat->message('Tell me a story about a developer');

foreach ($chat->stream() as $chunk) {
    echo $chunk;
}

// Streaming with callbacks
$fullText = '';
$stream = Ai::stream(
    'Explain design patterns',
    onEvent: function ($event) use (&$fullText) {
        $fullText .= $event;
        // Broadcast to WebSocket, emit to frontend, etc.
    }
);

foreach ($stream as $chunk) {
    // Process each chunk
}

HTTP Streaming Response

// routes/web.php
use CreativeCrafts\LaravelAiAssistant\Facades\Ai;
use CreativeCrafts\LaravelAiAssistant\Http\Responses\StreamedAiResponse;
use Illuminate\Http\Request;

Route::get('/ai/stream', function (Request $request) {
    $prompt = $request->input('q', 'Explain Laravel Eloquent');
    $generator = Ai::stream($prompt);
    
    return StreamedAiResponse::fromGenerator($generator);
});

Cancellation (Stopping Streams)

Stop streaming operations mid-flight using the shouldStop callback:

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Limit by chunk count
$maxChunks = 10;
$chunkCount = 0;

$stream = Ai::stream(
    'Write a very long essay',
    onEvent: null,
    shouldStop: function () use (&$chunkCount, $maxChunks): bool {
        $chunkCount++;
        return $chunkCount >= $maxChunks; // Stop after 10 chunks
    }
);

foreach ($stream as $chunk) {
    echo $chunk;
    if ($chunkCount >= $maxChunks) {
        break;
    }
}

// Limit by time
$startTime = microtime(true);
$maxDuration = 5.0; // 5 seconds

$stream = Ai::stream(
    'Generate content',
    shouldStop: function () use ($startTime, $maxDuration): bool {
        return (microtime(true) - $startTime) >= $maxDuration;
    }
);

foreach ($stream as $chunk) {
    echo $chunk;
}

// User-initiated cancellation
$cancelled = false; // Set to true from external signal

$stream = Ai::stream(
    'Long operation',
    shouldStop: fn() => $cancelled
);

foreach ($stream as $chunk) {
    echo $chunk;
    if ($cancelled) {
        break;
    }
}

Tool Calls (Function Calling)

Let the AI call functions/tools to perform actions or retrieve data.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Define tools
$tools = [
    [
        'type' => 'function',
        'function' => [
            'name' => 'get_weather',
            'description' => 'Get the current weather for a location',
            'parameters' => [
                'type' => 'object',
                'properties' => [
                    'location' => [
                        'type' => 'string',
                        'description' => 'City name, e.g. London, Paris',
                    ],
                    'unit' => [
                        'type' => 'string',
                        'enum' => ['celsius', 'fahrenheit'],
                    ],
                ],
                'required' => ['location'],
            ],
        ],
    ],
];

// Send request with tools
$response = Ai::responses()
    ->model('gpt-4')
    ->tools($tools)
    ->input()->message('What is the weather in London?')
    ->send();

// Check if AI wants to call a tool
if ($response->requiresAction) {
    foreach ($response->toolCalls as $toolCall) {
        if ($toolCall['function']['name'] === 'get_weather') {
            $args = json_decode($toolCall['function']['arguments'], true);
            $location = $args['location'];
            
            // Call your actual weather service
            $weatherData = getWeather($location);
            
            // Submit tool result back to AI
            // (Implementation depends on your flow)
        }
    }
}

Complete Tool Call Flow

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// 1. Initial request with tools
$conversation = Ai::conversations()->start();
$response = Ai::conversations()
    ->use($conversation)
    ->input()->message('What is the weather in Paris?')
    ->responses()
    ->tools($tools)
    ->send();

// 2. Execute tool if requested
if ($response->requiresAction) {
    foreach ($response->toolCalls as $toolCall) {
        $result = executeFunction($toolCall);
        
        // 3. Submit tool results
        $finalResponse = Ai::conversations()
            ->use($conversation)
            ->input()->toolResult($toolCall['id'], $result)
            ->send();
            
        echo $finalResponse->text;
    }
}

Conversations API (Advanced)

For explicit control over multi-turn conversations.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Create a new conversation
$conversationId = Ai::conversations()->start([
    'metadata' => ['user_id' => auth()->id()],
]);

// Send first turn
$response = Ai::conversations()
    ->use($conversationId)
    ->input()->message('Hello! What can you help me with?')
    ->responses()
    ->instructions('You are a helpful assistant')
    ->send();

echo $response->text;

// Continue conversation
$response = Ai::conversations()
    ->use($conversationId)
    ->input()->message('Tell me about Laravel')
    ->send();

// List conversation history
$items = Ai::conversations()
    ->use($conversationId)
    ->items();

foreach ($items as $item) {
    echo $item['role'] . ': ' . $item['content'] . "\n";
}

Advanced Conversation Management

// Create without auto-activation
$conversationId = Ai::conversations()->start(
    metadata: ['topic' => 'Laravel Help'],
    setActive: false
);

// Use existing conversation
$response = Ai::conversations()
    ->use($conversationId)
    ->input()->message('My question here')
    ->responses()
    ->model('gpt-4')
    ->instructions('You are an expert')
    ->send();

// Get a ResponsesBuilder bound to conversation
$responsesBuilder = Ai::conversations()
    ->use($conversationId)
    ->responses();

// Now customize and send
$response = $responsesBuilder
    ->model('gpt-4o-mini')
    ->input()->message('Another question')
    ->send();

Responses API (Low-Level)

For complete control over single-turn responses.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

$response = Ai::responses()
    ->model('gpt-4')
    ->instructions('You are a helpful coding assistant')
    ->input()->message('Explain SOLID principles')
    ->send();

echo $response->text;

// With conversation binding
$response = Ai::responses()
    ->inConversation($conversationId)
    ->model('gpt-4o-mini')
    ->input()->message('Continue')
    ->send();

Complex Input Items

$response = Ai::responses()
    ->model('gpt-4')
    ->withInput([
        ['role' => 'system', 'content' => 'You are a helpful assistant'],
        ['role' => 'user', 'content' => 'Hello'],
        ['role' => 'assistant', 'content' => 'Hi! How can I help?'],
        ['role' => 'user', 'content' => 'Explain Laravel routes'],
    ])
    ->send();

Webhooks (Background Processing)

Process AI requests asynchronously via webhooks.

1. Configure Environment

AI_ASSISTANT_WEBHOOK_SIGNING_SECRET=base64:YOUR_SECURE_SECRET_HERE
AI_BACKGROUND_JOBS_ENABLED=true
AI_QUEUE_NAME=ai-assistant

2. Register Webhook Route

// routes/web.php
use App\Http\Controllers\AiWebhookController;

Route::post('/ai/webhook', [AiWebhookController::class, 'handle'])
    ->middleware('verify.ai.webhook'); // Verifies webhook signature

3. Create Webhook Controller

namespace App\Http\Controllers;

use Illuminate\Http\Request;

class AiWebhookController extends Controller
{
    public function handle(Request $request)
    {
        $event = $request->input('event');
        $data = $request->input('data');
        
        match ($event) {
            'response.completed' => $this->handleResponseCompleted($data),
            'response.failed' => $this->handleResponseFailed($data),
            'tool.called' => $this->handleToolCall($data),
            default => null,
        };
        
        return response()->json(['status' => 'received']);
    }
    
    private function handleResponseCompleted(array $data): void
    {
        // Process completed AI response
        $text = $data['text'];
        $conversationId = $data['conversation_id'];
        
        // Store, broadcast, notify user, etc.
    }
    
    private function handleResponseFailed(array $data): void
    {
        // Handle failed request
        logger()->error('AI request failed', $data);
    }
    
    private function handleToolCall(array $data): void
    {
        // Execute tool and return result
    }
}

4. Trigger Background Request

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Dispatch to queue (webhook will be called when complete)
Ai::responses()
    ->model('gpt-4')
    ->input()->message('Generate a long report')
    ->sendAsync(webhookUrl: route('ai.webhook'));

Audio Processing

Process audio files with transcription, translation, and text-to-speech capabilities.

Audio Transcription

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Basic transcription
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/meeting.mp3'),
        'action' => 'transcribe',
    ])
    ->send();

echo $response->text; // Transcribed text

// Advanced transcription with language and context
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/interview.mp3'),
        'action' => 'transcribe',
        'language' => 'en',
        'prompt' => 'This is a technical interview about Laravel.',
        'temperature' => 0,
    ])
    ->send();

echo $response->text;

Audio Translation

// Translate audio to English
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/spanish_speech.mp3'),
        'action' => 'translate',
    ])
    ->send();

echo $response->text; // English translation

Text-to-Speech

// Generate speech from text
$response = Ai::responses()
    ->input()
    ->audio([
        'text' => 'Welcome to Laravel AI Assistant!',
        'action' => 'speech',
        'voice' => 'nova',
    ])
    ->send();

// Save audio file
$response->saveAudio(storage_path('audio/welcome.mp3'));

// Advanced speech with custom settings
$response = Ai::responses()
    ->input()
    ->audio([
        'text' => 'This is a professional announcement.',
        'action' => 'speech',
        'model' => 'tts-1-hd',
        'voice' => 'onyx',
        'format' => 'mp3',
        'speed' => 1.0,
    ])
    ->send();

$response->saveAudio(storage_path('audio/announcement.mp3'));

Image Generation and Editing

Create and manipulate images using DALL-E models.

Image Generation

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Basic image generation
$response = Ai::responses()
    ->input()
    ->image([
        'prompt' => 'A futuristic Laravel logo with neon lights',
    ])
    ->send();

// Save generated images
$paths = $response->saveImages(storage_path('images'));

// Advanced generation with DALL-E 3
$response = Ai::responses()
    ->input()
    ->image([
        'prompt' => 'A professional developer workspace with multiple monitors',
        'model' => 'dall-e-3',
        'size' => '1024x1024',
        'quality' => 'hd',
        'style' => 'vivid',
    ])
    ->send();

foreach ($response->images as $image) {
    echo $image['url'] . "\n";
    echo "Revised prompt: " . $image['revised_prompt'] . "\n";
}

Image Editing

// Edit existing image
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/photo.png'),
        'prompt' => 'Add a sunset background',
    ])
    ->send();

$response->saveImages(storage_path('images/edited'));

// Edit with mask
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/product.png'),
        'mask' => storage_path('images/mask.png'),
        'prompt' => 'Change the product color to blue',
        'n' => 3,
    ])
    ->send();

Image Variations

// Create variations of an existing image
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/logo.png'),
        'n' => 5,
    ])
    ->send();

$paths = $response->saveImages(storage_path('images/variations'));

🎵 Audio Features

Laravel AI Assistant provides comprehensive audio processing capabilities through OpenAI's Whisper and TTS models, all accessible via the unified Response API.

Audio Transcription

Convert audio files to text in their original language.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Basic transcription
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/recording.mp3'),
        'action' => 'transcribe',
    ])
    ->send();

echo $response->text;

// Advanced transcription with language specification and context
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/meeting.mp3'),
        'action' => 'transcribe',
        'model' => 'whisper-1',
        'language' => 'en',
        'prompt' => 'This is a business meeting discussing Q4 results.',
        'response_format' => 'json',
        'temperature' => 0,
    ])
    ->send();

echo $response->text;

// Access metadata
$duration = $response->metadata['duration'] ?? null;
$language = $response->metadata['language'] ?? null;

Supported audio formats: mp3, mp4, mpeg, mpga, m4a, wav, webm

Audio Translation

Translate audio from any supported language to English.

// Translate foreign language audio to English
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/spanish_presentation.mp3'),
        'action' => 'translate',
    ])
    ->send();

echo $response->text; // English translation

// With context for better translation
$response = Ai::responses()
    ->input()
    ->audio([
        'file' => storage_path('audio/french_interview.mp3'),
        'action' => 'translate',
        'prompt' => 'This is a formal technical presentation about software development.',
        'temperature' => 0.2,
    ])
    ->send();

echo $response->text;

Text-to-Speech

Generate natural-sounding speech from text using multiple voice options.

// Basic text-to-speech
$response = Ai::responses()
    ->input()
    ->audio([
        'text' => 'Hello! Welcome to Laravel AI Assistant.',
        'action' => 'speech',
    ])
    ->send();

// Save audio to file
$response->saveAudio(storage_path('audio/welcome.mp3'));

// Advanced speech generation
$response = Ai::responses()
    ->input()
    ->audio([
        'text' => 'This is a professional announcement with high-quality audio.',
        'action' => 'speech',
        'model' => 'tts-1-hd', // High-quality model
        'voice' => 'nova',      // Female, friendly voice
        'format' => 'mp3',
        'speed' => 1.0,
    ])
    ->send();

if ($response->isAudio()) {
    $path = storage_path('audio/announcement.mp3');
    $response->saveAudio($path);
}

Available voices:

alloy - Neutral, balanced (general purpose)
echo - Male, clear (announcements, narration)
fable - Warm, expressive (storytelling)
onyx - Deep, authoritative (professional content)
nova - Female, friendly (conversational)
shimmer - Soft, calming (meditation, relaxation)

Models:

tts-1 - Standard quality (faster, lower cost)
tts-1-hd - High quality (slower, higher quality)

See docs/API.md#audio-apis for complete audio API documentation.

🖼️ Image Features

Create, edit, and manipulate images using OpenAI's DALL-E models through the unified Response API.

Image Generation

Generate images from text descriptions.

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Basic image generation
$response = Ai::responses()
    ->input()
    ->image([
        'prompt' => 'A futuristic Laravel logo with neon lights',
    ])
    ->send();

// Access generated images
foreach ($response->images as $image) {
    $url = $image['url'];
    echo "Image URL: {$url}\n";
}

// Save all generated images
$paths = $response->saveImages(storage_path('images'));

// Advanced generation with DALL-E 3
$response = Ai::responses()
    ->input()
    ->image([
        'prompt' => 'A professional developer workspace with dual monitors showing Laravel code',
        'model' => 'dall-e-3',
        'size' => '1024x1024',
        'quality' => 'hd',
        'style' => 'vivid',
        'n' => 1,
    ])
    ->send();

if ($response->isImage()) {
    foreach ($response->images as $image) {
        echo "URL: " . $image['url'] . "\n";
        echo "Revised prompt: " . $image['revised_prompt'] . "\n";
    }
}

DALL-E Models:

dall-e-2 - Lower cost, supports multiple images (1-10), good quality
dall-e-3 - Higher cost, single image only, excellent quality and prompt following

Sizes:

DALL-E 2: 256x256, 512x512, 1024x1024
DALL-E 3: 1024x1024, 1792x1024, 1024x1792

Image Editing

Edit existing images using text prompts.

// Basic image editing
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/photo.png'),
        'prompt' => 'Add a sunset background',
    ])
    ->send();

$editedPaths = $response->saveImages(storage_path('images/edited'));

// Edit with transparency mask
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/product.png'),
        'mask' => storage_path('images/mask.png'),
        'prompt' => 'Change the product color to blue',
        'n' => 3,
        'size' => '1024x1024',
    ])
    ->send();

// Process multiple variations
foreach ($response->images as $index => $image) {
    $path = storage_path("images/variation_{$index}.png");
    file_put_contents($path, file_get_contents($image['url']));
}

Requirements:

Format: PNG only
Must be square (same width and height)
Max file size: 4MB
Only DALL-E 2 supports editing

Image Variations

Create variations of existing images without text prompts.

// Generate variations
$response = Ai::responses()
    ->input()
    ->image([
        'image' => storage_path('images/logo.png'),
        'n' => 5,
        'size' => '1024x1024',
    ])
    ->send();

// Save all variations
$paths = $response->saveImages(storage_path('images/variations'));

// Or download manually
foreach ($response->images as $index => $image) {
    $imageData = file_get_contents($image['url']);
    file_put_contents(
        storage_path("images/variation_{$index}.png"),
        $imageData
    );
}

Note: Image variations do not use prompts. If you need to modify images based on text descriptions, use Image Editing instead.

See docs/API.md#image-apis for complete image API documentation.

🔐 Security

Webhook Signature Verification

The verify.ai.webhook middleware automatically validates webhook signatures.

// Middleware is registered automatically
// Just add it to your webhook route:
Route::post('/ai/webhook', [AiWebhookController::class, 'handle'])
    ->middleware('verify.ai.webhook');

Rate Limiting

// config/ai-assistant.php
'rate_limiting' => [
    'enabled' => true,
    'max_requests_per_minute' => 60,
],

Content Filtering

$response = Ai::quick('Your prompt')
    ->withContentFilter([
        'hate' => 'medium',
        'violence' => 'medium',
        'sexual' => 'medium',
    ]);

⚙️ Configuration

Environment Variables

# Required
OPENAI_API_KEY=sk-...

# Optional
AI_ASSISTANT_MODEL=gpt-4o-mini
AI_ASSISTANT_TEMPERATURE=0.7
AI_ASSISTANT_MAX_TOKENS=1000
AI_ASSISTANT_TIMEOUT=30

# Streaming
AI_ASSISTANT_STREAMING_ENABLED=true
AI_ASSISTANT_STREAMING_BUFFER_SIZE=8192
AI_ASSISTANT_STREAMING_CHUNK_SIZE=1024

# Background Jobs
AI_BACKGROUND_JOBS_ENABLED=false
AI_QUEUE_NAME=ai-assistant

# Webhooks
AI_ASSISTANT_WEBHOOK_SIGNING_SECRET=base64:...

# Caching
AI_ASSISTANT_CACHE_ENABLED=true
AI_ASSISTANT_CACHE_TTL=3600

Presets

Choose a configuration preset for common use cases:

AI_ASSISTANT_PRESET=simple  # Options: simple, balanced, creative, precise, efficient

Presets:

simple: Fast responses with gpt-4o-mini, higher temperature
balanced: Standard gpt-4 with moderate settings (default)
creative: Higher temperature, more creative outputs
precise: Lower temperature, more deterministic
efficient: Optimized for speed and cost

🎯 Advanced Features

Caching Responses

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

$response = Ai::quick('Explain Laravel')
    ->cache(ttl: 3600); // Cache for 1 hour

// Cached responses return instantly
$cachedResponse = Ai::quick('Explain Laravel'); // Hits cache

JSON Mode

$response = Ai::quick([
    'message' => 'List 5 PHP frameworks with descriptions',
    'response_format' => 'json',
]);

$frameworks = json_decode($response->text, true);

Temperature Control

// Creative writing (high temperature)
$creative = Ai::chat('You are a creative writer')
    ->setTemperature(0.9)
    ->message('Write a poem about Laravel')
    ->send();

// Factual answers (low temperature)
$factual = Ai::chat('You are a technical expert')
    ->setTemperature(0.1)
    ->message('What is Laravel?')
    ->send();

Token Limits

$response = Ai::chat('Be concise')
    ->setMaxTokens(100)
    ->message('Explain Laravel in one sentence')
    ->send();

🧪 Testing

Mocking AI Responses

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;
use CreativeCrafts\LaravelAiAssistant\DataTransferObjects\ChatResponseDto;

// In your tests
Ai::fake();

Ai::shouldReceive('quick')
    ->once()
    ->with('Test prompt')
    ->andReturn(new ChatResponseDto([
        'text' => 'Mocked response',
        'model' => 'gpt-4',
        'usage' => ['total_tokens' => 10],
    ]));

$response = Ai::quick('Test prompt');
$this->assertEquals('Mocked response', $response->text);

Testing Streaming

Ai::fake();

Ai::shouldReceive('stream')
    ->andReturn(new \ArrayIterator(['chunk1', 'chunk2', 'chunk3']));

$chunks = [];
foreach (Ai::stream('Test') as $chunk) {
    $chunks[] = $chunk;
}

$this->assertCount(3, $chunks);

Integration Tests

This package includes comprehensive integration tests that verify the unified Response API with real OpenAI endpoints (audio, image, chat completion).

Running All Tests

# Run all tests (excludes integration tests by default)
composer test

# Run only unit tests
./vendor/bin/pest tests/Unit

# Run only feature tests
./vendor/bin/pest tests/Feature

Running Integration Tests

Integration tests are marked with @group integration and are automatically skipped unless you have a valid OpenAI API key configured.

# Run integration tests (simulated adapter tests - no API key required)
./vendor/bin/pest tests/Integration/AudioAdaptersIntegrationTest.php
./vendor/bin/pest tests/Integration/ImageAdaptersIntegrationTest.php

# Run real API integration tests (requires OPENAI_API_KEY)
./vendor/bin/pest tests/Integration/RealApiIntegrationTest.php

# Run all integration tests
./vendor/bin/pest --group=integration

# Exclude integration tests from CI
./vendor/bin/pest --exclude-group=integration

Real API Integration Tests

Real API tests make actual calls to OpenAI endpoints and require:

API Key: Set OPENAI_API_KEY environment variable
Cost Awareness: Full suite costs < $1 (using minimal test data)
Optional Execution: Tests are automatically skipped without API key

Cost Optimization:

Audio tests use minimal 483-byte MP3 file
Image tests use 256x256 size with DALL-E 2 (cheapest)
Text-to-speech uses minimal text (2 words)
Total cost: ~$0.075 for complete suite

Enable real API tests:

# Option 1: Set environment variable
export OPENAI_API_KEY="sk-..."
./vendor/bin/pest tests/Integration/RealApiIntegrationTest.php

# Option 2: Configure in your .env
OPENAI_API_KEY=sk-...

Test Coverage:

✅ Audio transcription (Whisper API)
✅ Audio translation (Whisper API)
✅ Audio speech generation (TTS API)
✅ Image generation (DALL-E API)
✅ Image editing (DALL-E API)
✅ Image variations (DALL-E API)
✅ Unified API end-to-end flow
✅ Error handling across services

📊 Monitoring & Observability

Observability Integration

Track all AI operations with correlation IDs, metrics, and structured logging:

use CreativeCrafts\LaravelAiAssistant\Facades\{Ai, Observability};
use Illuminate\Support\Str;

// Set correlation ID for request tracing
Observability::setCorrelationId(Str::uuid()->toString());

// Track operation with memory monitoring
$startTime = microtime(true);
$checkpoint = Observability::trackMemory('ai-chat');

try {
    // Log request
    Observability::log('ai-request', 'info', 'Processing AI request');
    
    // Make AI request
    $response = Ai::quick('Hello AI');
    
    $duration = microtime(true) - $startTime;
    
    // Log success metrics
    Observability::logApiResponse(
        'ai-completion',
        true,
        ['text_length' => strlen($response->text)],
        $duration
    );
    
    // Record performance metrics
    Observability::recordApiCall(
        '/ai/complete',
        $duration * 1000,
        200
    );
    
    // End memory tracking
    $metrics = Observability::endMemoryTracking($checkpoint);
    
    Observability::logPerformanceMetrics(
        'ai-chat',
        $duration * 1000,
        $metrics
    );
    
} catch (\Exception $e) {
    Observability::endMemoryTracking($checkpoint);
    Observability::report($e, ['operation' => 'ai-chat']);
    throw $e;
}

Streaming with Observability

Observability::setCorrelationId(request()->header('X-Request-ID'));

$checkpoint = Observability::trackMemory('ai-stream');
$chunkCount = 0;

try {
    Observability::log('ai-stream', 'info', 'Starting stream');
    
    foreach (Ai::stream('Tell a story') as $chunk) {
        $chunkCount++;
        echo $chunk;
    }
    
    $metrics = Observability::endMemoryTracking($checkpoint);
    Observability::logPerformanceMetrics(
        'ai-stream',
        $duration * 1000,
        array_merge($metrics, ['chunks' => $chunkCount])
    );
} catch (\Exception $e) {
    Observability::endMemoryTracking($checkpoint);
    Observability::reportApiError(
        'ai-stream',
        '/stream',
        500,
        $e->getMessage(),
        [],
        ['error' => $e->getMessage()]
    );
    throw $e;
}

See docs/OBSERVABILITY.md for complete documentation on correlation IDs, logging, metrics, error reporting, and memory tracking.

Built-in Metrics

use CreativeCrafts\LaravelAiAssistant\Services\MetricsService;

$metrics = app(MetricsService::class);

// Get request metrics
$requestCount = $metrics->getRequestCount();
$avgResponseTime = $metrics->getAverageResponseTime();
$tokenUsage = $metrics->getTotalTokenUsage();

// Get error rates
$errorRate = $metrics->getErrorRate();

Logging

// Automatic logging of all requests (when enabled)
// config/ai-assistant.php
'logging' => [
    'enabled' => true,
    'channel' => 'stack',
    'level' => 'info',
],

🚀 Performance Optimization

Connection Pooling

// config/ai-assistant.php
'connection_pool' => [
    'enabled' => true,
    'max_connections' => 10,
    'timeout' => 30,
],

Request Batching

use CreativeCrafts\LaravelAiAssistant\Facades\Ai;

// Process multiple requests concurrently
$prompts = [
    'Explain Laravel',
    'Explain Symfony',
    'Explain CodeIgniter',
];

$promises = [];
foreach ($prompts as $prompt) {
    $promises[] = Ai::quick($prompt)->async();
}

$responses = Promise\Utils::unwrap($promises);

🔄 Migration Guide

Migrating to SSOT API (v3.0)

If you're using legacy APIs like OpenAiRepository, AiAssistant, or the Compat Client, please see MIGRATION.md for comprehensive migration instructions to the new unified SSOT API (Ai::responses()).

The migration guide includes:

Migration paths from all legacy APIs
8+ detailed code examples (chat, streaming, audio, images)
Before/after comparisons for every operation
Troubleshooting common issues

Version Upgrades

For general version upgrade information and breaking changes, see UPGRADE.md.

📝 Requirements

PHP 8.2+
Laravel 10.x, 11.x, or 12.x
OpenAI API key

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for details.

🔒 Security

If you discover any security-related issues, please email [email protected] instead of using the issue tracker.

📄 License

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

🙏 Credits

📚 Additional Documentation

Core Documentation

docs/API.md - Complete API reference with all methods, modes, and examples
- Audio APIs - Audio transcription, translation, and text-to-speech documentation
- Image APIs - Image generation, editing, and variation documentation
- SSOT Architecture - Understanding the unified API design
MIGRATION.md - Comprehensive migration guide to SSOT API from legacy components
UPGRADE.md - Version upgrade guide and breaking changes
docs/OBSERVABILITY.md - Comprehensive observability guide with correlation IDs, metrics, and logging
CHANGELOG.md - Version history and changes

Advanced Guides

docs/ARCHITECTURE.md - Package architecture and design patterns
docs/ENVIRONMENT_VARIABLES.md - Complete environment variable reference
docs/PERFORMANCE_TUNING.md - Performance optimization guide
docs/PRODUCTION_CONFIGURATION.md - Production deployment guide
docs/SCALING.md - Scaling strategies and best practices

Examples

See examples/ for runnable code samples demonstrating:

Hello world chat completion
Streaming responses with cancellation
Unified completion API usage
Observability integration
Advanced patterns

Unified Completion Entrypoint (New)

Use a single public entrypoint to cover text/chat × sync/stream via AiManager::complete.

Text (sync):

use CreativeCrafts\LaravelAiAssistant\Enums\{Mode, Transport};
use CreativeCrafts\LaravelAiAssistant\DataTransferObjects\{CompletionRequest};
use CreativeCrafts\LaravelAiAssistant\Services\AiManager;

$ai = app(AiManager::class);

$result = $ai->complete(
    Mode::TEXT,
    Transport::SYNC,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'prompt' => 'Write a haiku about Laravel',
        'temperature' => 0.2,
    ])
);

echo (string) $result; // final text

Chat (sync):

$result = $ai->complete(
    Mode::CHAT,
    Transport::SYNC,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'messages' => [
            ['role' => 'user', 'content' => 'Summarize this: ...'],
        ],
    ])
);

$data = $result->toArray(); // normalized message array

Stream (accumulate to final result):

// Text stream → accumulated final string
$final = (string) $ai->complete(
    Mode::TEXT,
    Transport::STREAM,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'prompt' => 'Stream a short message',
    ])
);

// Chat stream → accumulated final array
$finalArray = $ai->complete(
    Mode::CHAT,
    Transport::STREAM,
    CompletionRequest::fromArray([
        'model' => 'gpt-4o-mini',
        'messages' => [ ['role' => 'user', 'content' => 'Hello'] ],
    ])
)->toArray();

// If you need incremental events, prefer ChatSession::stream(...)

Deprecation notice:

Legacy methods on AssistantService (textCompletion, chatTextCompletion, streamedCompletion, streamedChat) are deprecated and will be removed in a future release. They now log a deprecation warning once per process. Prefer AiManager::complete.

Name		Name	Last commit message	Last commit date
Latest commit History 202 Commits
.github		.github
config		config
database/migrations		database/migrations
docs		docs
examples		examples
resources/views/components/ai		resources/views/components/ai
src		src
stubs		stubs
tests		tests
.editorconfig		.editorconfig
.gitattributes		.gitattributes
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE.md		LICENSE.md
MIGRATION.md		MIGRATION.md
README.md		README.md
UPGRADE.md		UPGRADE.md
composer.json		composer.json
phpstan-baseline.neon		phpstan-baseline.neon
phpstan.neon.dist		phpstan.neon.dist
phpunit.xml.dist		phpunit.xml.dist
pint.json		pint.json

License

CreativeCrafts/laravel-ai-assistant

Folders and files

Latest commit

History

Repository files navigation

Laravel AI Assistant

✨ What's New in v3.0

🚀 Quick Start (≤5 Minutes)

Step 1: Install (1 min)

Step 2: Configure (1 min)

Step 3: Hello World Chat (1 min)

Step 4: Streaming Response (2 min)

Step 5: Audio Transcription (1 min)

Step 6: Image Generation (1 min)

📖 Core Concepts

Why SSOT Architecture?

Available Methods

Comparison: Sync vs Stream

Feature Capabilities

🎯 Unified Completion API (Recommended)

Quick Example

Mode Options

Transport Options

Chat Example

Streaming Example

📚 Cookbook

Quick Calls (One-Off Requests)

Stateful Chat (Multi-Turn Conversations)

Customizing Chat Sessions

Streaming Responses

HTTP Streaming Response

Cancellation (Stopping Streams)

Tool Calls (Function Calling)

Complete Tool Call Flow

Conversations API (Advanced)

Advanced Conversation Management

Responses API (Low-Level)

Complex Input Items

Webhooks (Background Processing)

1. Configure Environment

2. Register Webhook Route

3. Create Webhook Controller

4. Trigger Background Request

Audio Processing

Audio Transcription

Audio Translation

Text-to-Speech

Image Generation and Editing

Image Generation

Image Editing

Image Variations

🎵 Audio Features

Audio Transcription

Audio Translation

Text-to-Speech

🖼️ Image Features

Image Generation

Image Editing

Image Variations

🔐 Security

Webhook Signature Verification

Rate Limiting

Content Filtering

⚙️ Configuration

Environment Variables

Presets

🎯 Advanced Features

Caching Responses

JSON Mode

Temperature Control

Token Limits

🧪 Testing

Mocking AI Responses

Testing Streaming

Integration Tests

Running All Tests

Running Integration Tests

Real API Integration Tests

📊 Monitoring & Observability

Packages