A lightweight Laravel package that provides a unified way to call AI providers (like OpenAI and Anthropic) using simple HTTP connectors and provider stubs — no vendor SDKs required.

This package focuses on:

• Simple configuration per provider
• Minimal surface area to call common AI endpoints (e.g., chat/completions)
• Extensibility to add new providers without pulling their full SDKs

• PHP 8.2+
• Laravel 10 or 11
• Composer

composer require debjit/php-ai-hub

Laravel package discovery should auto-register the service provider.

php artisan vendor:publish --provider="Debjit\PhpAiHub\Installer" --tag=ai-hub-config

This will publish:

• config/ai-hub.php — main hub config
• config/ai-hub/openai.php — OpenAI-specific config
• config/ai-hub/anthropic.php — Anthropic-specific config

If these files already exist, review and merge as needed.

Set the required keys for the providers you intend to use.

OpenAI:

OPENAI_API_KEY=sk-...
OPENAI_BASE_URL=https://api.openai.com
OPENAI_ORG_ID=org_...(optional)
OPENAI_PROJECT_ID=proj_...(optional)

Anthropic:

ANTHROPIC_API_KEY=sk-ant-...
ANTHROPIC_BASE_URL=https://api.anthropic.com
ANTHROPIC_VERSION=2023-06-01

You can override the base URLs to route through proxies/gateways if needed.

Main hub configuration: config/ai-hub.php

• Controls default provider and maps provider identifiers to resolver classes and stub implementations.

Example (high-level):

return [
    'default' => env('AI_HUB_DEFAULT_PROVIDER', 'openai'),

    'providers' => [
        'openai' => [
            'config' => base_path('config/ai-hub/openai.php'),
            'provider' => \Debjit\PhpAiHub\Stubs\OpenAI\Provider::class,
        ],
        'anthropic' => [
            'config' => base_path('config/ai-hub/anthropic.php'),
            'provider' => \Debjit\PhpAiHub\Stubs\Anthropic\Provider::class,
        ],
    ],
];

Per-provider configuration:

• config/ai-hub/openai.php contains base URL, auth header strategy, default model, timeouts, etc.
• config/ai-hub/anthropic.php contains base URL, version header, default model, timeouts, etc.

How config is resolved:

• Debjit\PhpAiHub\Support\ConfigResolver reads the main hub config, then loads the targeted provider config file.
• Environment variables are read via Laravel env() in those configs.
• Missing or invalid configurations will throw descriptive exceptions early.

• Registry: Debjit\PhpAiHub\Registry
  • Entry point to get a provider instance by key (e.g., "openai", "anthropic") or the default.
• Provider Stub: Debjit\PhpAiHub\Stubs\{Provider}\Provider
  • Facade-like API that exposes high-level operations (e.g., chat).
• Client/Connector: Debjit\PhpAiHub\Stubs\{Provider}\Connectors\HttpConnector
  • Responsible for HTTP requests to the provider with proper headers, endpoints, and payload shape.

You don’t install SDKs for each provider: we model the minimal HTTP required.

Resolve the default provider (configured via AI_HUB_DEFAULT_PROVIDER or config/ai-hub.php):

use Debjit\PhpAiHub\Registry;

$provider = app(Registry::class)->provider(); // default from config

Or explicitly:

$openai = app(Registry::class)->provider('openai');
$anthropic = app(Registry::class)->provider('anthropic');

use Debjit\PhpAiHub\Registry;

$openai = app(Registry::class)->provider('openai');

$response = $openai->chat()->create([
    'model' => 'gpt-4o-mini',
    'messages' => [
        ['role' => 'system', 'content' => 'You are a helpful assistant.'],
        ['role' => 'user', 'content' => 'Write a haiku about Laravel.'],
    ],
    'temperature' => 0.7,
]);

$content = $response['choices'][0]['message']['content'] ?? null;

Notes:

• The stub normalizes the request to OpenAI’s /v1/chat/completions.
• Auth header Authorization: Bearer {OPENAI_API_KEY} is added by the connector.

use Debjit\PhpAiHub\Registry;

$anthropic = app(Registry::class)->provider('anthropic');

$response = $anthropic->chat()->create([
    'model' => 'claude-3-5-sonnet-20240620',
    'messages' => [
        ['role' => 'user', 'content' => 'Summarize Laravel service providers.'],
    ],
    'max_tokens' => 512,
]);

$content = $response['content'][0]['text'] ?? null;

Notes:

• The stub targets Anthropic’s /v1/messages with headers:
  • x-api-key: {ANTHROPIC_API_KEY}
  • anthropic-version: {ANTHROPIC_VERSION}

If a provider supports streaming, the stub may expose a streaming method:

$stream = $openai->chat()->stream([
    'model' => 'gpt-4o-mini',
    'messages' => [/*...*/],
]);

foreach ($stream as $event) {
    // handle tokens/chunks
}

Refer to the specific provider stub for supported features.

You can type-hint the registry wherever needed:

use Debjit\PhpAiHub\Registry;

class MyController
{
    public function __construct(private Registry $registry) {}

    public function __invoke()
    {
        $ai = $this->registry->provider(); // default
        // ...
    }
}

1. Create a provider config: config/ai-hub/{provider}.php

   • Base URL, auth headers, defaults, timeouts/retries.

2. Implement a connector:

   • Example: src/Stubs/FooAI/Connectors/HttpConnector.php
   • Use Laravel HTTP client, apply headers and base URL from config.
   • Provide methods like chatCompletions(array $payload): array.

3. Implement a provider stub:

   • Example: src/Stubs/FooAI/Provider.php
   • Expose simplified methods, like chat()->create($payload) that internally calls the connector.
   • Keep the public API consistent with existing providers when possible.

4. Register in main config:

   • In config/ai-hub.php add the mapping:

     'providers' => [
         // ...
         'fooai' => [
             'config' => base_path('config/ai-hub/fooai.php'),
             'provider' => \Debjit\PhpAiHub\Stubs\FooAI\Provider::class,
         ],
     ],

5. Use it:

   $foo = app(\Debjit\PhpAiHub\Registry::class)->provider('fooai');
   $foo->chat()->create([...]);

Guidelines:

• Do not use the provider’s official SDK. Prefer HTTP with minimal dependencies.
• Normalize payloads where possible to keep your app code portable between providers.

• Missing API key or base URL

  • Ensure .env variables are set and the provider config reads them via env().

• Invalid provider key

  • Confirm the provider key exists in config/ai-hub.php under providers.

• HTTP errors (4xx/5xx)

  • The connector will return the response or throw exceptions based on your implementation. Check your timeouts and retry logic in the provider config.

• Config resolution issues

  • Debjit\PhpAiHub\Support\ConfigResolver loads config/ai-hub.php, validates provider entry, and then loads the provider’s own config. If paths or classes are wrong, you’ll get descriptive exceptions.

• Mock the connector classes and assert payload shapes.
• For integration tests, set fake API keys and use Laravel’s HTTP fake to simulate provider responses.

• Never log raw API keys or entire request/response bodies containing sensitive data.
• Prefer setting keys in .env and referencing via config files.

Follow semantic versioning (SemVer). Breaking changes will bump the major version.

MIT License. See LICENSE file.

Crafted for teams that prefer HTTP-first integrations with AI providers in Laravel without heavyweight SDKs.