AiWire

AiWire — AI Integration Module for ProcessWire

Connect your ProcessWire site to AI providers: Anthropic (Claude), OpenAI (GPT), Google (Gemini), xAI (Grok), and OpenRouter (400+ models).

Manage multiple API keys per provider, test connections from admin, and use AI in your templates with a clean PHP API.

Full documentation with 25 real-world examples → DOCUMENTATION.md

Features

• 5 providers — Anthropic, OpenAI, Google, xAI, OpenRouter
• Multiple API keys per provider with enable/disable toggle
• Automatic fallback — if one key/provider fails, the next one takes over
• Connection testing — verify each key with one click in admin
• Test Chat — interactive chat with parameter controls (temperature, tokens, model)
• File cache — TTL-based caching (day/week/month/year) with page context
• Field storage — save AI content to page fields for permanent, editable storage
• Content blocks — generate() method for multi-block pages (overview, FAQ, pairings…)
• Conversation history — multi-turn chat support
• Cost optimization — route tasks to different providers/models by complexity
• Logging — standard + debug logging via ProcessWire log system

Requirements

• PHP 8.1+
• ProcessWire 3.0.210+
• cURL extension enabled

Installation

1. Download or clone into site/modules/AiWire/
2. Admin → Modules → Refresh → Install AiWire
3. Configure → add API keys → click test button
4. Use $modules->get('AiWire') in your templates

site/modules/AiWire/
├── AiWire.module.php       # Main module
├── AiWireProvider.php       # API client for all providers
├── AiWireCache.php          # File-based cache system
├── models.json              # Editable provider model list
├── README.md                # This file
├── DOCUMENTATION.md         # Full API reference + 25 examples
└── LICENSE

Updating Models

Provider model IDs and labels are loaded from models.json. Edit that file when a provider adds, renames, or retires a model:

{
  "openai": {
    "defaultModel": "gpt-5.4",
    "models": {
      "gpt-5.4": "GPT-5.4",
      "gpt-5.4-mini": "GPT-5.4 Mini"
    }
  }
}

If models.json is missing or invalid, AiWire falls back to the built-in model list.

OpenAI and OpenRouter also support model refresh from the module settings. Refreshed models are stored in module config separately from models.json and are used first; if no refreshed list exists, AiWire falls back to models.json, then to built-in defaults.

Each API key can also define a custom model. The custom model value is optional, but when present it overrides the selected model for that key. This is useful for private, preview, account-gated, proxy, or newly released model IDs.

API Key Storage

Keys can be stored directly in the module settings, or as environment references such as env:OPENAI_API_KEY. Environment references avoid saving the real secret in ProcessWire module config.

Getting API Keys

Quick Start

$ai = $modules->get('AiWire');

// Simple — returns text only
echo $ai->chat('What is ProcessWire CMS?');

// Full response with metadata
$result = $ai->ask('Explain quantum computing in simple terms');
if ($result['success']) {
    echo $result['content'];              // AI response text
    echo $result['usage']['total_tokens']; // tokens used
}

// Fallback — tries Anthropic → OpenAI → Google
$result = $ai->askWithFallback('Summarize this article...', [
    'provider'          => 'anthropic',
    'fallbackProviders' => ['openai', 'google'],
]);

// Generate multiple AI blocks for a product page
$results = $ai->generate($page, [
    ['field' => 'ai_overview',    'prompt' => "Write overview of {$page->title}..."],
    ['field' => 'ai_food_pairing','prompt' => "Suggest pairings for {$page->title}..."],
    ['field' => 'ai_brand_story', 'prompt' => "Share 3 facts about {$page->brand->title}..."],
], ['cache' => 'M', 'temperature' => 0.7]);

API Methods

→ Full API reference: DOCUMENTATION.md

Result Format

// Successful response from ask(), askWithFallback(), askAndSave(), generate()
[
    'success' => true,
    'content' => 'The AI response text...',
    'message' => 'OK',
    'usage'   => ['input_tokens' => 25, 'output_tokens' => 148, 'total_tokens' => 173],
    'raw'     => [ /* full API response */ ],
    'cached'  => false,       // true if served from file cache
    'source'  => 'ai',        // askAndSave/generate only: 'ai', 'field', or 'error'
]

Options

Supported Models

The default editable list in models.json currently includes:

Documentation & Examples

Full documentation with detailed API reference, 25 real-world examples (with ProcessWire setup, code, and expected output), cache strategies, field storage, multi-provider pipelines, and best practices:

→ DOCUMENTATION.md

Content generation — product pages, brand enrichment, category descriptions, cocktail recipes, region guides, tasting notes, review summaries

SEO & translations — auto-generate meta descriptions, OG titles, multi-language content

Chatbots & APIs — AI sommelier, gift recommendations, product comparison, form routing

Infrastructure — multi-provider cost optimization, fallback chains, key rotation, cache strategies, provider monitoring, team key separation, bulk LazyCron generation

License

MIT — free for personal and commercial use.

Author

Maxim Alex — smnv.org — maxim@smnv.org

Built for the ProcessWire community.