README

Symfony 7 bundle for AI-powered content generation. Access multiple LLMs and image models through a single, consistent API powered by Pollinations.ai.

Why Pollinations?

This bundle uses Pollinations.ai as its AI gateway, providing:

Unified API - One integration for Claude, GPT, Gemini, Mistral, DeepSeek, and more
No vendor lock-in - Switch models with a config change, no code modifications
Generous free tier - Start generating content without API keys
Transparent pricing - Pay only for what you use with premium models
Open ecosystem - Built on open-source principles

Features

Text generation with multiple models (Claude, Gemini, GPT, Mistral, DeepSeek)
Image generation with multiple models (GPTImage, Flux, Seedream, Turbo)
Image subject generator - Two-step anchor extraction for unique, differentiated images
Task Types - Configure different models for different tasks (content, prompts, images)
Cost tracking - See estimated costs per model (pricing reference)
Style presets for consistent image generation
Configurable prompt templates with intelligent variant selection
Automatic fallback between models when one fails
SonataMedia integration (optional)
Sonata Admin integration with image regeneration UI (optional)

Requirements

PHP >= 8.2
Symfony >= 7.0
symfony/http-client

Optional Dependencies

For full functionality, consider these additional bundles:

Package	Version	Purpose
sonata-project/admin-bundle	^4.0	Admin panel with image regeneration UI
sonata-project/media-bundle	^4.0	Media storage for generated images

Quick Start

1. Installation

composer require xmon-org/ai-content-bundle

2. Configuration

The bundle works out of the box with zero configuration using free-tier models:

# config/packages/xmon_ai_content.yaml
xmon_ai_content: ~  # That's it! Uses free-tier defaults

Default models (no API key required):

Text: mistral → nova-micro → gemini-fast → openai-fast
Image: flux → zimage → turbo

For custom configuration:

xmon_ai_content:
    # Configure models per task type
    tasks:
        news_content:
            default_model: 'mistral'
            allowed_models: ['mistral', 'openai-fast', 'gemini-fast']
        image_prompt:
            default_model: 'openai-fast'
            allowed_models: ['openai-fast', 'mistral', 'nova-micro']
        image_generation:
            default_model: 'flux'
            allowed_models: ['flux', 'zimage', 'turbo']

    # Text generation (optional - these are the defaults)
    text:
        model: 'mistral'
        fallback_models: ['nova-micro', 'gemini-fast', 'openai-fast']

    # Image generation (optional - these are the defaults)
    image:
        model: 'flux'
        fallback_models: ['zimage', 'turbo']

Tip: Get a free API key at enter.pollinations.ai to unlock premium models like claude, gemini, and gptimage.

3. Generate Text

use Xmon\AiContentBundle\Enum\TaskType;
use Xmon\AiContentBundle\Service\AiTextService;

class MyService
{
    public function __construct(
        private readonly AiTextService $aiTextService,
    ) {}

    public function summarize(string $content): string
    {
        // Uses the model configured for NEWS_CONTENT task
        $result = $this->aiTextService->generateForTask(
            TaskType::NEWS_CONTENT,
            systemPrompt: 'You are a helpful assistant.',
            userMessage: "Summarize: {$content}",
        );

        return $result->getText();
    }
}

4. Generate Image

use Xmon\AiContentBundle\Service\AiImageService;

class MyService
{
    public function __construct(
        private readonly AiImageService $aiImageService,
    ) {}

    public function generateImage(): void
    {
        // Uses the model configured for IMAGE_GENERATION task
        $result = $this->aiImageService->generateForTask(
            prompt: 'A serene Japanese dojo with morning light',
        );

        file_put_contents('image.png', $result->getBytes());
    }
}

5. Override Options Per-Request

// Use specific model without fallback
$result = $this->aiTextService->generate($system, $user, [
    'model' => 'claude',
    'use_fallback' => false,
    'timeout' => 120,
]);

// Use specific model with fallback to configured fallback_models
$result = $this->aiTextService->generate($system, $user, [
    'model' => 'claude',
    'use_fallback' => true,
]);

Documentation

Installation & Setup

Installation Guide - Full setup instructions

Guides

Task Types - Configure models per task
Text Generation - Generate text with AI
Image Generation - Generate images with AI
Image Subject Generator - Two-step anchor extraction
Prompt Templates - Configurable prompts with variants
Admin Integration - Sonata Admin integration

Reference

Configuration - Full YAML reference
Providers - Available AI models and costs
Fallback System - How automatic fallback works
Architecture - Bundle structure

Development

Development Guide - Local setup and commands

Available Models

Powered by Pollinations.ai - Unified access to Claude, GPT, Gemini, Mistral, Flux, and more.

Query current models and pricing directly from the API:

# Text models with Pollen pricing
curl https://gen.pollinations.ai/text/models

# Image models with Pollen pricing
curl https://gen.pollinations.ai/image/models

API Documentation: enter.pollinations.ai/api/docs

Text Models

Prices updated: 2025-12-27 | These are approximate estimates for reference only. For current pricing, query the endpoints above.

Model	Tier	~Resp/$	Description
`openai-fast`	anonymous	~2,270	GPT-5 Nano - Ultra fast
`openai`	anonymous	~1,660	GPT-5 Mini - Balanced
`gemini-fast`	seed	~2,500	Gemini 2.5 Flash - Fast & cheap
`gemini`	seed	~330	Gemini 3 Flash - Pro-grade
`gemini-search`	seed	~330	Gemini 3 Flash with Search
`deepseek`	seed	~595	DeepSeek V3.2 - Reasoning
`mistral`	seed	~2,850	Mistral Small 3.2 - Efficient
`claude`	flower	~65	Claude Sonnet 4.5 - Premium

Image Models