AI Content Architect

AI Content Architect is an AI-assisted WordPress plugin that turns plain-English content ideas into reviewable, editable WordPress content models.

Instead of asking an AI model to generate PHP code, the plugin asks for structured JSON configuration. A WordPress admin can review that configuration, edit it, save it as a draft, and apply it only when it looks right.

What It Builds

Given a prompt like:

Create a job board with jobs, companies, locations, salary ranges, employment types, remote options, application deadlines, and application URLs.

AI Content Architect can propose:

• Custom post types
• Taxonomies
• Custom fields
• Admin columns
• Template suggestions
• Sample content
• Warnings and implementation notes

Applied models are stored as configuration and registered dynamically by the plugin. No generated PHP is executed.

Current Status

This plugin is intended for administrator-controlled content model prototyping and implementation inside WordPress.

It currently supports:

• A mock provider for local testing
• An OpenAI-compatible provider abstraction
• Generated model validation and sanitization
• Editable model review before applying
• Draft, applied, and disabled model states
• Import/export as JSON
• Runtime CPT and taxonomy registration
• Generated field meta boxes
• Generated admin columns
• Frontend field rendering
• Fallback frontend templates
• Model cleanup options on delete

Admin Screens

The plugin adds a WordPress admin menu:

AI Content Architect
├── Architect
├── Content Models
└── Settings

Architect

The Architect screen lets an administrator:

• Enter a natural-language content model prompt
• Use example prompt chips
• Generate a structured model
• Review and edit generated JSON-backed fields
• Save the model as a draft
• Apply the model to WordPress
• Optionally generate sample content

Generation never creates WordPress structures immediately. The admin must review and apply the model first.

Content Models

The Content Models screen lets an administrator:

• View saved models
• See model status: draft, applied, or disabled
• Edit models
• Apply models
• Disable models
• Export model JSON
• Import model JSON as draft
• Delete models

When deleting a model, a custom confirmation popover lets the admin choose:

• Delete the model only
• Delete the model plus generated posts and taxonomy terms

Media attachments are intentionally preserved during cleanup.

Settings

The Settings screen includes:

• Provider selection: mock, OpenAI, OpenAI-compatible, or custom
• API key storage without printing the key back to the browser
• Base URL
• Model name
• Strict validation
• Slug conflict prevention
• Review-before-apply enforcement
• Frontend field display toggle
• Fallback template toggle

Frontend Rendering

Generated CPT single pages can be rendered in two ways.

Theme Template + Plugin Injection

By default, the active theme renders the single post. AI Content Architect appends generated fields and taxonomies to the_content when:

• The view is singular
• The post type belongs to an applied AI Content Architect model
• Frontend field output is enabled
• The request is the main query and main loop

The output is scoped with classes such as:

.aica-content
.aica-content-fields
.aica-content-field
.aica-content-field-label
.aica-content-field-value
.aica-content-taxonomies
.aica-content-term

Plugin Fallback Templates

If fallback templates are enabled, the plugin can render generated CPT singles and archives using:

templates/single-dynamic-content.php
templates/archive-dynamic-content.php

The single template displays:

• Featured image
• Post title
• Main content
• Generated custom fields
• Generated taxonomy terms

The archive template displays:

• A responsive card grid
• Featured image or placeholder
• Title
• Excerpt
• A small set of generated field values
• Taxonomy term pills
• View details link

Theme Overrides

Themes can override plugin fallback templates by adding:

your-theme/ai-content-architect/single-dynamic-content.php
your-theme/ai-content-architect/archive-dynamic-content.php

The plugin checks the active theme first, then falls back to its bundled templates.

Field Formatting

Frontend generated fields are escaped and formatted by type.

Supported field types include:

• text
• textarea
• number
• email
• url
• date
• checkbox
• select
• radio
• image
• gallery
• wysiwyg

Frontend formatting behavior:

• Empty fields are skipped
• URLs are clickable links
• Email fields use mailto:
• Dates use the WordPress date format
• Numbers use localized number formatting
• Boolean fields render as Yes/No badges
• Images render as attachment images instead of raw IDs
• Galleries render as image grids
• WYSIWYG values use safe post HTML

Validation and Slug Conflicts

The validator sanitizes and validates generated or imported configurations before they are saved or applied.

It checks:

• Required model name
• Valid CPT keys and slugs
• Valid taxonomy keys and slugs
• Field post type references
• Supported field types
• Admin column references
• Template references
• Sample content references
• Reserved WordPress slugs
• Conflicts with WordPress, themes, plugins, and other AI Content Architect models

Generation uses softer conflict handling so the user can still review and edit the model. Applying a model uses stricter validation so real conflicts are blocked before registration.

When editing or revalidating an existing model, the validator excludes the current model ID from conflict detection so a model does not conflict with itself.

Model Lifecycle

Models are stored in the aica_content_models option.

Each model record contains:

• ID
• Name
• Status
• Config
• Created timestamp
• Updated timestamp
• Plugin version

Applied models register their generated CPTs and taxonomies on init.

When models are applied, disabled, or deleted, the plugin schedules rewrite flushing. This avoids stale 404s after deleting and recreating models with the same slug.

Delete and Cleanup Behavior

Deleting a model opens a confirmation popover.

The admin can choose:

Delete Model Only

This removes the model configuration but leaves existing posts, terms, media, and content in the database.

Delete Model + Content

This removes:

• Posts belonging to the model's generated CPTs
• Terms belonging to the model's generated taxonomies
• The model configuration

It does not delete media attachments.

This makes it easier to remove a generated model completely and later reuse the same CPT or rewrite slug.

Mock Provider

The mock provider is intended for local testing without an API key.

It can return different sample models based on prompt keywords, including:

• Job board
• Real estate listings
• Restaurant menu
• Movie database

This prevents every local test prompt from returning the same job-board structure.

OpenAI-Compatible Provider

The OpenAI-compatible provider sends a chat-completions-style request to the configured endpoint and expects structured JSON back.

Settings include:

• API key
• Base URL
• Model name

The API key is saved but never printed back into the settings form.

The Settings screen includes provider connection testing and model discovery. For OpenAI and OpenAI-compatible providers, administrators can refresh the available model list from the provider API instead of typing model IDs manually. A curated fallback list is shown until a provider model list is refreshed, and advanced users can still enter a custom model ID.

Mock mode remains the recommended local testing path because it requires no API key and exercises the plugin workflow without external calls.

External Services

AI Content Architect does not contact external AI services while the mock provider is selected.

When an administrator selects OpenAI, OpenAI-compatible, or custom provider mode, the plugin sends the administrator's prompt, the selected model ID, and JSON-only generation instructions to the configured provider endpoint. The API key, when configured, is sent as an Authorization header. The plugin may also request a provider model list when an administrator clicks Refresh models or Test connection.

OpenAI service information:

• Service: OpenAI API
• Terms: https://openai.com/policies/terms-of-use
• Privacy policy: https://openai.com/policies/privacy-policy

Custom and OpenAI-compatible providers use the base URL configured by the site administrator. Site owners are responsible for reviewing the terms and privacy policy of any configured custom provider.

AI Content Architect is not affiliated with or endorsed by OpenAI.

Custom Providers

AI Content Architect includes a provider registry so integrations do not need to be hardcoded into the settings screen.

Built-in provider options include:

• Mock provider
• OpenAI
• OpenAI-compatible
• Custom provider

The custom provider option is intended for OpenAI-compatible local services, gateways, or third-party APIs. Developers can register additional providers with:

add_filter( 'aica_ai_providers', function ( array $providers ) {
    $providers['my_provider'] = array(
        'label'           => 'My Provider',
        'description'     => 'Custom provider for AI Content Architect.',
        'class'           => My_AICA_Provider::class,
        'supports_models' => true,
        'requires_key'    => true,
        'default_base_url'=> 'https://example.com/v1',
    );

    return $providers;
} );

Provider classes must implement AI_Provider_Interface. Providers that support model refresh and connection testing can also implement AI_Model_Provider_Interface.

Important Files

Bootstrap:

• ai-content-architect.php
• includes/class-plugin.php

Core:

• includes/class-model-store.php
• includes/class-content-model.php
• includes/class-schema-validator.php
• includes/class-model-cleaner.php
• includes/class-capabilities.php

AI providers:

• includes/class-ai-provider-interface.php
• includes/class-mock-provider.php
• includes/class-openai-provider.php

Registration:

• includes/class-cpt-registrar.php
• includes/class-taxonomy-registrar.php
• includes/class-field-registrar.php
• includes/class-admin-columns.php
• includes/class-frontend-display.php
• includes/class-template-manager.php

Admin:

• admin/views/page-architect.php
• admin/views/page-models.php
• admin/views/page-settings.php
• admin/views/partial-model-review.php
• admin/js/admin.js
• admin/css/admin.css

Frontend:

• frontend/css/frontend.css
• templates/single-dynamic-content.php
• templates/archive-dynamic-content.php

Import/export and samples:

• includes/class-export-import.php
• includes/class-sample-content-generator.php

Installation

1. Copy the plugin folder into:

wp-content/plugins/AI Content Architect

2. Activate the plugin in WordPress.
3. Go to:

AI Content Architect > Settings

4. Use the mock provider for local testing, or configure an OpenAI-compatible provider.

Local Testing Checklist

1. Go to AI Content Architect > Architect.
2. Generate a model from a prompt.
3. Review the generated CPTs, taxonomies, fields, columns, templates, and warnings.
4. Save as draft.
5. Apply the model.
6. Confirm the generated CPT appears in WordPress admin.
7. Add a generated CPT item and fill in custom fields.
8. View the single item on the frontend.
9. Confirm generated fields are styled and empty fields are hidden.
10. View the CPT archive.
11. Delete the model and choose whether to also delete generated content.
12. Recreate a model with the same slug and confirm frontend URLs do not 404.

Useful test prompts:

Create a job board with jobs, companies, locations, salary ranges, employment types, remote options, application deadlines, and application URLs.

Create a restaurant menu section with dishes, menu categories, ingredients, allergens, dietary labels, prices, spice levels, preparation time, dish images, featured dishes, and availability status.

Create a movie database with movies, actors, genres, release dates, ratings, trailers, posters, runtime, directors, and featured movies.

Security Notes

• Generated PHP is never executed.
• Generated configurations are validated before saving or applying.
• AJAX actions use nonces and capability checks.
• Field saving uses nonces and post editing permissions.
• Frontend output is escaped based on field type.
• API keys are stored but not printed back into admin forms.
• Cleanup actions are explicit and destructive content deletion requires a separate popover action.

Known Limitations

Good future improvements include:

• Visual add/remove controls for CPTs, taxonomies, and fields
• WordPress media picker support for image and gallery fields
• Model revision flow, such as "add a field" or "change taxonomy"
• Blueprint/template library
• ACF or Meta Box export
• Relationship fields
• Search/filter generation
• Schema.org output
• Rollback support

Asset Licensing

• Plugin code: GPLv3. See LICENSE.
• admin/images/ai-logo.png: Original plugin asset, licensed GPLv3 with the plugin.