AI Integration

LaraDashboard features an Agentic CMS with AI-powered content generation capabilities. This guide covers configuring AI providers, using the command center, and integrating AI into your workflows.

Overview#

What is Agentic CMS?#

LaraDashboard's Agentic CMS is an AI-powered command center that allows you to:

• Generate content using natural language commands
• Enhance and modify existing content
• Automate repetitive content tasks
• Get AI assistance for writing and editing

Supported Providers#

Provider	Default Model	Best For
OpenAI	GPT-6 Luna (gpt-6-luna)	General content generation, image generation (DALL-E 3)
Anthropic	Claude 3 Haiku	Long-form content, analysis, structured output
Google Gemini	Gemini 2.0 Flash	Fast responses, cost-effective
Ollama	Llama 3.2	Local/private deployment, no API key needed

OpenAI models available in Settings → Integrations: GPT-6 Luna (gpt-6-luna, fast and affordable, the default) and GPT-6 Sol (gpt-6-sol, most capable). The retired GPT-4o, GPT-4 Turbo and GPT-3.5 Turbo models were removed in v1.5.0; sites that had one selected fall back to the default.

Use outside AI agents (MCP)#

The command center runs AI inside the admin. To let an external agent — Claude.ai, Claude Desktop, Cursor or Claude Code — work with your site instead, turn on the built-in MCP server:

See MCP Server for setup and MCP tools for developers to add tools from your own modules.

Configuration#

Admin settings#

Choose the default provider, models, API keys and maximum tokens under Settings → Integrations. These values take priority over .env.

Environment Variables#

.env values are fallbacks, used when a key or model isn't set in the admin:

# OpenAI
OPENAI_API_KEY=sk-...
OPENAI_MODEL=gpt-6-luna

# Anthropic (Claude)
ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_MODEL=claude-3-haiku-20240307

# Google Gemini
GEMINI_API_KEY=...
GEMINI_MODEL=gemini-2.0-flash

# Ollama (local LLM — no API key needed)
OLLAMA_BASE_URL=http://localhost:11434
OLLAMA_MODEL=llama3.2

Config File#

config/ai.php holds the provider fallbacks. Keys and models saved under Settings → Integrations take priority over these values.

<?php

return [
    'openai' => [
        'api_key' => env('OPENAI_API_KEY'),
        'model' => env('OPENAI_MODEL', 'gpt-6-luna'),
    ],

    'anthropic' => [
        'api_key' => env('ANTHROPIC_API_KEY'),
        'model' => env('ANTHROPIC_MODEL', 'claude-3-haiku-20240307'),
    ],

    'gemini' => [
        'api_key' => env('GEMINI_API_KEY'),
        'model' => env('GEMINI_MODEL', 'gemini-2.0-flash'),
    ],

    'ollama' => [
        'base_url' => env('OLLAMA_BASE_URL', 'http://localhost:11434'),
        'model' => env('OLLAMA_MODEL', 'llama3.2'),
    ],
];

Using the Command Center#

Accessing the Command Center#

The AI Command Center is available when editing content:

1. Navigate to any post/page editor
2. Click the AI Assistant button (🤖)
3. The command center panel opens

Command Syntax#

Enter natural language commands:

Generate a 500-word blog post about Laravel best practices

Summarize this article in 3 bullet points

Rewrite this paragraph to be more engaging

Available Commands#

Content Generation#

Command	Description
Generate [description]	Create new content
Write [description]	Alias for generate
Create [type] about [topic]	Generate specific content type

Examples:

Generate a product description for a wireless mouse
Write an introduction for a blog about AI
Create a headline for an article about climate change

Content Enhancement#

Command	Description
Improve [selection]	Enhance selected text
Rewrite [selection]	Rewrite in different style
Expand [selection]	Make content longer
Summarize [selection]	Create a summary
Simplify [selection]	Make easier to read

Examples:

Improve the introduction paragraph
Rewrite this in a professional tone
Expand the features section
Summarize this article in 100 words

SEO Optimization#

Command	Description
Generate meta description	Create SEO description
Suggest keywords for [topic]	Get keyword ideas
Optimize title for SEO	Improve title

Translation#

Command	Description
Translate to [language]	Translate content
Localize for [region]	Adapt for region

Streaming Responses#

AI responses stream in real-time for better UX:

// Response appears progressively
"Generating... ▌"
"Here is your content: Lorem ipsum... ▌"
"Here is your content: Lorem ipsum dolor sit amet..."

Programmatic Usage#

AiContentGeneratorService#

Use AI in your own code through App\Services\AiContentGeneratorService. It uses the default provider from Settings → Integrations unless you pick one with setProvider().

use App\Services\AiContentGeneratorService;

class ContentController extends Controller
{
    public function __construct(
        private AiContentGeneratorService $ai
    ) {}

    public function generate(Request $request)
    {
        if (! $this->ai->isConfigured()) {
            return response()->json(['message' => 'AI is not configured.'], 422);
        }

        // Returns the parsed result as an array (e.g. title/content for posts).
        $result = $this->ai->generateContent($request->input('prompt'), 'general');

        return response()->json($result);
    }
}

Available Methods#

// One-shot generation; $type selects the system prompt (e.g. 'general').
$result = $ai->generateContent('Write a product description for…', 'general');

// Rewrite existing text
$text = $ai->modifyText($existingContent, 'Make this more concise');

// Multi-turn chat (v1.4.5+) — OpenAI, Claude, Gemini and Ollama
$reply = $ai->chatCompletion([
    ['role' => 'system', 'content' => 'You are a helpful editor.'],
    ['role' => 'user', 'content' => 'Suggest three titles for a CRM guide.'],
], temperature: 0.7, maxTokens: 500);

// SEO title/description/keywords from page context
$seo = $ai->generateSeoMeta(['title' => $post->title, 'content' => $post->content]);

// Images (OpenAI DALL-E only)
if ($ai->canGenerateImages()) {
    $image = $ai->generateImage('A clean dashboard illustration', '1792x1024');
}

// Use a specific provider: openai, claude, gemini or ollama
$result = $ai->setProvider('claude')->generateContent('Analyze this changelog…');

// Introspection
$ai->getAvailableProviders();
$ai->getDefaultProvider();
$ai->getMaxTokens();

API Endpoints#

# Get available providers
GET /api/v1/ai/providers

# Generate content
POST /api/v1/ai/generate-content
{
    "prompt": "Write a blog post about...",
    "provider": "openai",
    "options": {
        "max_tokens": 1000
    }
}

# Modify text (admin session route, used by the editor)
POST /admin/ai/modify-text
{
    "text": "Original text here...",
    "instruction": "Make it shorter"
}

# Stream generation (admin session route, used by the command center)
POST /admin/ai/command/process-stream
{
    "command": "Generate article about AI",
    "context": {}
}

Command Logging#

Viewing Logs#

All AI commands are logged for auditing:

1. Navigate to Settings → AI → Command History
2. View command, response, user, and timestamp
3. Filter by date, user, or command type

Log Structure#

// ai_command_logs table
[
    'user_id' => 1,
    'command' => 'Generate blog post...',
    'provider' => 'openai',
    'model' => 'gpt-6-luna',
    'parameters' => ['max_tokens' => 1000],
    'response_tokens' => 850,
    'duration_ms' => 2500,
    'status' => 'success',
    'created_at' => '2024-01-15 10:30:00',
]

Cost Management#

Token Usage#

Monitor token usage:

// Get usage statistics
$stats = $aiService->getUsageStats([
    'user_id' => auth()->id(),
    'period' => 'month',
]);

// Returns
[
    'total_tokens' => 50000,
    'total_requests' => 150,
    'estimated_cost' => 1.50,
]

Rate Limiting#

Configure rate limits per user:

// config/ai.php
'rate_limits' => [
    'default' => 20,  // requests per minute
    'premium' => 100, // for premium users
],

Cost Estimation#

// Estimate before generating
$estimate = $aiService->estimateCost([
    'prompt' => $prompt,
    'max_tokens' => 1000,
]);

// Returns
[
    'input_tokens' => 150,
    'output_tokens' => 1000,
    'estimated_cost' => 0.045,
]

Prompt Engineering#

System Prompts#

Configure default system prompts:

// In service or config
$systemPrompts = [
    'content_writer' => "You are a professional content writer...",
    'seo_expert' => "You are an SEO specialist...",
    'translator' => "You are a professional translator...",
];

Context Injection#

Provide context for better results by putting it in the prompt:

$context = [
    'brand' => 'TechCorp',
    'product_type' => 'Wireless Mouse',
    'features' => ['ergonomic', 'long battery', 'multi-device'],
    'tone' => 'professional',
    'target_audience' => 'business professionals',
];

$result = $aiService->generateContent(
    "Write a product description.\n\nContext:\n" . json_encode($context, JSON_PRETTY_PRINT),
    'general'
);

Template Prompts#

Create reusable prompt templates:

// Store templates
$templates = [
    'blog_post' => [
        'system' => 'You are a skilled blog writer...',
        'template' => "Write a {length} word blog post about {topic}. Include {sections} sections. Tone: {tone}.",
    ],
    'product_description' => [
        'system' => 'You are an e-commerce copywriter...',
        'template' => "Write a compelling product description for {product}. Highlight: {features}. Target audience: {audience}.",
    ],
];

// Use template
$result = $aiService->generateFromTemplate('blog_post', [
    'length' => 800,
    'topic' => 'Laravel Performance',
    'sections' => 5,
    'tone' => 'informative',
]);

Error Handling#

Common Errors#

try {
    $result = $aiService->generate($params);
} catch (AiProviderException $e) {
    // Provider API error
    Log::error('AI Provider Error: ' . $e->getMessage());
} catch (RateLimitException $e) {
    // Rate limit exceeded
    return response()->json([
        'error' => 'Too many requests. Please wait.',
        'retry_after' => $e->retryAfter,
    ], 429);
} catch (InsufficientCreditsException $e) {
    // No API credits
    return response()->json([
        'error' => 'API credits exhausted.',
    ], 402);
}

Fallback Providers#

Configure fallback when primary fails:

// config/ai.php
'fallback' => [
    'enabled' => true,
    'order' => ['openai', 'claude'],
    'on_errors' => ['rate_limit', 'timeout', 'server_error'],
],

Security Considerations#

API Key Security#

• Never commit API keys to version control
• Use environment variables
• Rotate keys periodically
• Use separate keys for production/development

Content Moderation#

Enable content filtering:

// config/ai.php
'moderation' => [
    'enabled' => true,
    'block_categories' => ['hate', 'violence', 'sexual'],
    'log_blocked' => true,
],

User Permissions#

Control AI access:

// Permissions
'ai.use'         // Can use AI features
'ai.settings'    // Can configure AI
'ai.view_logs'   // Can view command logs

Hooks#

AI Hooks#

use TorMorten\Eventy\Facades\Events as Eventy;

// Before AI generation
Eventy::action('ai.before_generate', function ($params, $user) {
    // Log or modify params
});

// After AI generation
Eventy::action('ai.after_generate', function ($result, $params, $user) {
    // Post-process result
});

// Filter AI prompt
$prompt = Eventy::filter('ai.prompt', $prompt, $context);

// Filter AI result
$content = Eventy::filter('ai.result', $content, $command);

Best Practices#

Prompt Design#

1. Be specific and clear
2. Provide context
3. Specify format requirements
4. Set appropriate length limits
5. Include examples when helpful

Performance#

1. Use streaming for long content
2. Cache repeated queries
3. Batch similar requests
4. Monitor token usage

Quality#

1. Review AI-generated content
2. Edit and personalize
3. Check facts and accuracy
4. Maintain brand voice

Troubleshooting#

No Response#

1. Check API key is valid
2. Verify network connectivity
3. Check rate limits
4. Review error logs

Poor Quality Output#

1. Improve prompt specificity
2. Provide more context
3. Adjust temperature (lower = more focused)
4. Try different model

Slow Generation#

1. Reduce max_tokens
2. Use a faster model (GPT-6 Luna instead of GPT-6 Sol)
3. Enable streaming
4. Check network latency

CRM AI Capabilities#

The CRM module includes extensive AI integration across all major entities:

Contact Intelligence#

• Lead Scoring — Score contacts 0-100 based on profile completeness and engagement
• Data Enrichment — Enrich contact profiles with AI-inferred data
• Duplicate Detection — Fuzzy matching to find potential duplicate contacts
• Smart Segmentation — AI-powered suggestions for contact segments
• Activity Summaries — Generate narrative summaries of contact interactions

Deal Intelligence#

• Deal Scoring — Health scoring and win probability for deals
• Pipeline Forecasting — Revenue predictions based on deal stages and velocity
• Stale Deal Detection — Identify deals that need attention
• Next Step Suggestions — AI recommendations for deal progression

Support Intelligence#

• Ticket Categorization — Auto-categorize incoming support tickets
• Sentiment Analysis — Detect customer sentiment in ticket messages
• Reply Suggestions — AI-generated reply drafts for support agents

Campaign & Invoice AI#

• Email Subject Generation — Generate high-converting subject lines
• Email Body Improvement — Enhance campaign email content
• Payment Prediction — Predict invoice payment likelihood
• Follow-up Suggestions — AI-generated follow-up email drafts

Dashboard Insights#

• Daily/Weekly Digest — AI-generated summaries of CRM activity
• Dashboard Insights — Pattern detection and trend analysis
• Follow-up Recommendations — Prioritized follow-up suggestions

CRM AI features are controlled individually via Settings → CRM → AI. Each capability can be toggled on or off independently.

Adding AI to Your Module#

LaraDashboard provides a modular AI architecture that any module can extend using the Capability / Action / Context pattern.

For a complete developer guide on integrating AI into your module, see:

• AI Architecture for Module Developers — Contracts, registries, implementation examples, and the CRM reference

Next Steps#

• AI Architecture for Modules - Build AI-powered modules
• Content Management - Use AI in content
• Email Templates - AI for emails
• Module Development - Build custom modules