README

Status: Production Ready
Drupal: 10.1+ | 11
Type: Site Recipe

Security & Compliance: All security modules, compliance frameworks, and hardening are provided by the Secure Drupal Recipe, which is included as a dependency. The LLM Platform recipe focuses on AI/LLM features and depends on Secure Drupal for all security and compliance.

A comprehensive Drupal recipe that sets up a production-ready AI/LLM integration platform with unified orchestration, vector database management, and enterprise-grade monitoring.

Overview

This recipe installs and configures a complete AI platform for Drupal, providing:

Core AI Integration: LLM module with multi-provider support
Modern Admin Interface: Gin admin theme optimized for AI workflows
Vector Database Support: Semantic search and AI-powered content discovery
Enterprise Monitoring: Comprehensive analytics and performance tracking
Content Automation: AI-powered content workflows and moderation
Dependency Management: Proper module configuration and optimization
Security & Compliance: All security, encryption, and compliance are managed by the Secure Drupal recipe

Requirements

Drupal 10.1+ or Drupal 11
PHP 8.1+
Node.js 20+ (for complementary TypeScript services)
One or more AI provider API keys (OpenAI, Anthropic, etc.)

Architecture Overview

Unified AI Platform

                    🧠 AI LLM Platform - Unified
                                  |
┌─────────────────────────────────────────────────────────────┐
|                  UnifiedAiOrchestrator                     |
|                (Central Coordination Hub)                   |
└─────────────────────────────────────────────────────────────┘
                                  |
┌─────────────┬─────────────┬─────────────┬─────────────┬─────────────┐
|             |             |             |             |             |
│Vector       │Autonomous   │Monitor-     │Security     │Deploy-      │
│Database     │Agents       │ing          │(via         │ment         │
│Manager      │Manager      │Service      │Secure       │Manager      │
│             │             │             │Drupal)      │             │
└─────────────┘└─────────────┘└─────────────┘└─────────────┘└─────────────┘

Included Modules

Core Drupal Modules

Node, Field, Text, Filter, User, System, Path
Image, File, Media (for AI-generated content)

Contrib Modules

Gin - Modern admin theme optimized for AI workflows
ECA - Event-Condition-Action framework for AI workflows
AI Module Suite - Core AI framework and provider integrations
- ai
- ai_external_moderation
- ai_ckeditor
- ai_image
- ai_automators
Advanced Queue - Enterprise queue management

Custom AI Platform Modules

LLM - Unified AI orchestration and provider management
AI Agent Orchestra - Multi-agent workflow coordination
AI Provider Apple - Apple Intelligence integration
AI Provider LangChain - LangChain framework integration
MCP Client Extras - Model Control Plane support
Alternative Services - Service abstraction layer
API Normalization - Consistent API responses
Recipe Onboarding - Guided setup wizard
Gov Compliance - Government compliance and audit tools

Note: All security modules (key, encrypt, tfa, seckit, login_security, autologout, captcha, honeypot, password_policy, admin_toolbar, etc.) are managed by the Secure Drupal recipe and should not be listed or configured here.

Security & Compliance

All security, encryption, audit logging, access control, and compliance features are provided by the Secure Drupal Recipe, which is included as a dependency. This includes:

Field-level encryption
API key security and rotation
Audit logging and access control
GDPR, SOC 2, and government compliance
Content safety and moderation
All security configuration and validation

To update or customize security, edit the Secure Drupal recipe.

Performance Optimization

Caching Strategy

Response Caching: Intelligent AI response caching
Vector Caching: Optimized embedding storage
Result Memoization: Avoid duplicate processing
CDN Integration: Static asset optimization

Scalability Features

Queue Management: Async processing for heavy workloads
Load Balancing: Multi-provider failover
Resource Monitoring: Automatic scaling triggers
Database Optimization: Optimized queries and indexes

Troubleshooting

Common Issues

Provider API Errors
- Verify API keys are correctly configured
- Check rate limits and quotas
- Review error logs at /admin/reports/dblog
Vector Database Connection
- Verify database credentials and connectivity
- Check firewall and network settings
- Review vector database logs
Performance Issues
- Monitor resource usage in admin dashboard
- Check queue processing status
- Review caching configuration
Service Dependencies
- Run drush llm:service-check to validate services
- Check module dependencies and versions
- Review service definition conflicts

Debug Mode

Enable detailed logging:

drush config:set llm.settings debug_mode true
drush config:set llm.settings log_level debug

Support Resources

Module documentation at /admin/help/llm
Error logs at /admin/reports/dblog
System status at /admin/reports/status
Service health at /admin/config/llm/health

🏗️ MACH Alliance Compliance

The LLM Platform recipe is designed to enable MACH Alliance architecture principles for enterprise AI implementations:

Microservices Architecture

Service-Oriented Design: Each AI provider integration operates as an independent service
Container-First Deployment: Docker support for all AI services with health checks
Event-Driven Communication: ECA framework enables async communication between services
Independent Scaling: AI processing components can scale independently based on demand

API-First Implementation

Comprehensive API Coverage: REST, GraphQL, and JSON-RPC endpoints for all functionality
OpenAPI Documentation: Machine-readable API specifications for all AI services
SDK Generation: Automatic TypeScript SDK generation from OpenAPI specs
Headless Ready: Complete separation of AI processing from presentation layer

Cloud-Native Features

Kubernetes Support: Helm charts and deployment manifests for orchestration
Multi-Cloud Deployment: Platform-agnostic configuration supporting AWS, Azure, GCP
Auto-Scaling: Horizontal pod autoscaling based on AI workload metrics
Service Discovery: Automatic service registration and health monitoring
Circuit Breaker Pattern: Resilient AI provider integration with graceful degradation

Headless Architecture

Frontend Agnostic: AI capabilities accessible via APIs from any frontend framework
Mobile-First APIs: Optimized endpoints for mobile and progressive web applications
Edge Computing: Lightweight AI inference suitable for edge deployment
Multi-Channel Support: Same AI backend serving web, mobile, IoT, and voice interfaces

Enterprise Integration Patterns

Event Sourcing: Complete audit trail of AI operations for compliance
CQRS Implementation: Separate read/write models for AI data and analytics
Message Queues: Advanced queue management for AI workload distribution
Distributed Caching: Multi-layer caching strategy for AI responses and vectors

This recipe transforms Drupal into a cloud-native AI platform that supports modern enterprise architecture while maintaining the flexibility to operate in traditional environments.

Development & Testing

Test Suite

# Run full test suite (requires Node.js)
npm run test:full

# PHP unit tests only
vendor/bin/phpunit web/modules/custom/llm/tests

# Validate platform health
drush llm:validate

Development Mode

# Enable development settings
drush config:set llm.settings development_mode true

# Show detailed error messages
drush config:set system.logging error_level verbose

# Disable caching for development
drush config:set llm.settings cache_responses false

License

GPL-2.0-or-later

bluefly / llm_platform

Maintainers

Details