feat/ Build WordPress REST API endpoints for AI client

[Ref34t]

WordPress REST API Endpoints for AI Client

This PR implements comprehensive REST API endpoints that expose all PromptBuilder capabilities through WordPress's native REST API architecture and part of Project MVP Road #1.

Overview

Adds a unified REST controller that creates individual endpoints for each PromptBuilder terminate method, enabling WordPress plugins to integrate AI functionality through familiar REST API patterns with manual registration for distributed ecosystem compatibility.

Key Features

• 10 REST endpoints mapping 1:1 to PromptBuilder terminate methods
• Manual registration only via REST_Route_Registrar::register_rest_routes('namespace')
• Custom namespace support for multi-plugin compatibility
• No auto-registration - prevents conflicts when multiple plugins bundle this library
• Full parameter support for all PromptBuilder options
• WordPress coding standards compliance (PHPCS + PHPStan max level)

Endpoints Added

POST /wp-json/{namespace}/prompt/generate-text
POST /wp-json/{namespace}/prompt/generate-texts
POST /wp-json/{namespace}/prompt/generate-image
POST /wp-json/{namespace}/prompt/generate-images
POST /wp-json/{namespace}/prompt/generate-speech
POST /wp-json/{namespace}/prompt/generate-speeches
POST /wp-json/{namespace}/prompt/generate-result
POST /wp-json/{namespace}/prompt/generate-text-result
POST /wp-json/{namespace}/prompt/generate-image-result
POST /wp-json/{namespace}/prompt/generate-speech-result

Usage

Manual Registration Required

// Plugin developers must manually register routes
add_action( 'rest_api_init', function() {
    REST_Route_Registrar::register_rest_routes( 'my-plugin/v1' );
} );

// Access via REST API
// POST /wp-json/my-plugin/v1/prompt/generate-text
// Body: {"prompt": "Write a story", "max_tokens": 100}

Architecture

• Prompt_Controller - Unified controller handling all 10 endpoints with WordPress conventions
• REST_Route_Registrar - Entry point for manual route registration
• No Auto-Registration - Library provides classes only, plugins must opt-in
• PHPStan Configuration - WordPress-specific static analysis with phpstan.neon.dist

Distributed Ecosystem Support

This implementation follows WordPress distributed ecosystem patterns:

• ✅ No automatic route registration - prevents namespace conflicts
• ✅ Plugin-controlled namespaces - each plugin chooses their own namespace
• ✅ Library-only approach - provides functionality without imposing registration

Quality Standards

• ✅ PHPCS - WordPress coding standards compliance
• ✅ PHPStan - Maximum level static analysis with WordPress stubs
• ✅ WordPress Patterns - Follows core WordPress REST API conventions

This implementation provides the foundation for WordPress plugins to integrate AI capabilities through manual REST API registration, maintaining compatibility in WordPress's distributed plugin ecosystem.

[felixarntz]

@Ref34t Thank you for getting this started!

One early feedback:

POST /wp-json/{namespace}/prompt/generate-text
POST /wp-json/{namespace}/prompt/generate-texts
POST /wp-json/{namespace}/prompt/generate-image
POST /wp-json/{namespace}/prompt/generate-images
POST /wp-json/{namespace}/prompt/generate-speech
POST /wp-json/{namespace}/prompt/generate-speeches
POST /wp-json/{namespace}/prompt/generate-result
POST /wp-json/{namespace}/prompt/generate-text-result
POST /wp-json/{namespace}/prompt/generate-image-result
POST /wp-json/{namespace}/prompt/generate-speech-result

I think we should only have endpoints for the generate-*-result terminate methods.

The other ones are shortcuts to get a specific part of the results, and that's easily something we can handle on the JavaScript API side. No need to have the other shortcut endpoints which are almost duplicates.

[felixarntz]

This message shouldn't manually define all the arguments. The PHP AI client SDK provides all the definitions (check the getJsonSchema() methods), and we can reuse those. Otherwise it's a lot of duplicate code, and it misses many arguments. No need to have validate_callbacks here either because it would be handled by default.

We need to support two arguments:

• prompt: I think we can make this a oneOf of:
  • either array of Message DTOs
  • or a single Message DTO
  • or a single MessagePart DTO
  • or a single string
• config: A single ModelConfig DTO

For all of those we can combine the schemas from the respective DTOs to get this overall args array.

[Ref34t]

@felixarntz handled here 1cf0a2a

[felixarntz]

We should also register the response schema for the routes here.

Per my previous comment, by only implementing the generate-*-result endpoints, this is easy: We can simply use GenerativeAiResult::getJsonSchema().

[Ref34t]

@felixarntz handled here 1cf0a2a