README

A comprehensive Drupal 10 recipe that creates an enterprise-grade AI coordination platform with multi-provider support, vector databases, intelligent workflows, and enterprise security.

Features

🤖 Multi-Provider AI Integration

LiteLLM (unified provider interface) - DEFAULT
Ollama (local, privacy-first)
OpenAI (GPT-4, GPT-3.5)
Anthropic (Claude 3)
LangChain (orchestration)
CrewAI (multi-agent systems)
HuggingFace (model hub)

LiteLLM Integration

The recipe now uses LiteLLM as the default AI provider, offering unified access to multiple AI services:

# Default configuration
default_provider: 'litellm'
base_url: 'http://localhost:4000'
default_model: 'ollama/llama3.2'

Benefits:

Single interface for all AI providers
Automatic failover between providers
Unified authentication and error handling
Reduced custom code maintenance

🔐 Enterprise Security

Built on secure_drupal foundation
FedRAMP, HIPAA, PCI DSS, SOC2 compliance ready
Advanced encryption and key management
Comprehensive audit logging

🔍 Intelligent Search

Vector embeddings with semantic search
Multiple vector database support
Search API integration
Faceted search capabilities

🤝 AI Agents & Automation

Multi-agent orchestration with CrewAI
Workflow automation with ECA
Content generation and moderation
Research and analysis agents

🔗 Model Context Protocol (MCP)

WebSocket, HTTP, and stdio transports
Community registry browser integration
Tool integration for enhanced capabilities
LangChain tools compatibility

Requirements

Drupal: 10.3+ or 11.0+
PHP: 8.1+
Node.js: 20.0.0+ (for npm packages)
Composer: 2.0+

Optional Requirements

Ollama: For local AI execution (recommended)
Redis: For enhanced caching and performance
Solr: For advanced search capabilities

Installation

Method 1: Using the Recipe (Recommended)

# Install the recipe and its dependencies
composer require bluefly/llm-platform-enterprise

# Apply the recipe to your Drupal site
php core/scripts/drupal recipe path/to/llm_platform

Method 2: With Input Variables

# Install with custom configuration
php core/scripts/drupal recipe path/to/llm_platform \
  --site_name="My AI Platform" \
  --admin_email="admin@mycompany.com" \
  --default_provider="ollama"

Method 3: Development Installation

# Clone the repository
git clone https://github.com/bluefly/llm-platform-recipe.git

# Install dependencies
cd llm-platform-recipe
composer install

# Apply the recipe
php core/scripts/drupal recipe . --site_name="Dev Platform"

Post-Installation Setup

1. Configure AI Providers

Ollama (Recommended for privacy)

# Install Ollama
curl -fsSL https://ollama.com/install.sh | sh

# Pull recommended models
ollama pull llama3.2:7b
ollama pull nomic-embed-text

# Verify in Drupal
Visit: /admin/config/ai/providers/test

Cloud Providers

Get API keys from providers (OpenAI, Anthropic, etc.)
Navigate to /admin/config/system/keys
Add your API keys securely
Configure providers at /admin/config/ai/providers

2. Complete Onboarding

Visit /admin/config/system/recipe-onboarding to:

Configure AI providers
Set up vector databases
Enable desired features
Test integrations

3. Security Configuration

Review Security Settings: /admin/config/security
Configure Encryption: /admin/config/system/encrypt
Set Up Key Management: /admin/config/system/keys
Run Security Review: /admin/reports/security-review

Configuration

Input Variables

The recipe accepts these input variables:

Variable	Type	Default	Description
`site_name`	string	"LLM Platform"	Name of the AI platform site
`admin_email`	email	system.site.mail	Administrator email address
`default_provider`	string	"ollama"	Default AI provider (ollama, openai, anthropic, langchain)

Module Configuration

Core AI Modules:

ai: Core AI framework
ai_logging: Request/response logging
llm: LLM Platform core functionality

Provider Modules:

ai_provider_ollama: Local Ollama integration
ai_provider_openai: OpenAI GPT models
ai_provider_anthropic: Anthropic Claude models
ai_provider_langchain: Multi-provider orchestration
ai_agent_crewai: Multi-agent systems
ai_agent_huggingface: HuggingFace model hub

Platform Modules:

mcp_registry: MCP server registry browser
alternative_services: Service discovery and failover
recipe_onboarding: Guided setup wizard

Usage Examples

Basic AI Chat

$ai_service = \Drupal::service('llm.chat');
$response = $ai_service->chat('Explain quantum computing', [
  'provider' => 'ollama',
  'model' => 'llama3.2:7b'
]);

Multi-Agent Workflow

$crew_service = \Drupal::service('ai_agent_crewai.manager');
$result = $crew_service->executeCrewWorkflow([
  'task' => 'research_and_write_article',
  'topic' => 'AI in Healthcare',
  'agents' => ['researcher', 'writer', 'editor']
]);

Vector Search

$search_service = \Drupal::service('llm.vector_search');
$results = $search_service->semanticSearch('machine learning algorithms', [
  'provider' => 'ollama',
  'model' => 'nomic-embed-text',
  'similarity_threshold' => 0.8
]);

Architecture

Provider Hierarchy

Ollama (Primary) - Local execution, privacy-first
OpenAI (Secondary) - Cloud fallback for advanced features
Anthropic (Tertiary) - Specialized tasks requiring Claude models
LangChain - Multi-provider orchestration layer

Security Foundation

Built on the secure_drupal recipe providing:

AES-256-GCM encryption
Multi-factor authentication
Comprehensive audit logging
Compliance frameworks (FedRAMP, HIPAA, PCI DSS, SOC2)

Performance Optimization

Redis caching for AI responses
Connection pooling for providers
Request batching and streaming
Tiered cache strategy

Variants

The recipe includes three variants for different use cases:

Minimal (`variants/minimal/`)

~34 modules
Essential AI features only
Single provider (Ollama)
Basic security

Standard (`variants/standard/`)

~76 modules
Full-featured AI platform
Multi-provider support
Enhanced security

Enterprise (`variants/enterprise/`)

~119 modules
Enterprise-grade with compliance
All providers and features
Maximum security

API Documentation

The platform provides comprehensive APIs:

REST API: /api/ai/* endpoints
GraphQL: /graphql with AI schema
OpenAPI: /admin/config/services/openapi
MCP Protocol: WebSocket on :3001/mcp

Monitoring & Health

Health Dashboard

Visit /service-health for real-time monitoring:

Provider availability
Response latency
Error rates
Token usage
Cost tracking

Logs & Debugging

AI Logs: /admin/reports/ai-logging
System Status: /admin/reports/status
Security Review: /admin/reports/security-review

Troubleshooting

Common Issues

Ollama Connection Failed

# Check Ollama is running
curl http://localhost:11434/api/version

# Restart Ollama service
ollama serve

Module Dependencies Missing

# Update Composer dependencies
composer update

# Clear Drupal cache
drush cr

Permission Denied

Ensure user has 'administer ai configuration' permission
Check security module restrictions
Verify API key access

Performance Issues

Enable Redis caching
Configure connection pooling
Adjust rate limiting settings
Monitor token usage

Contributing

Fork the repository
Create a feature branch: git checkout -b feature/new-feature
Follow Drupal coding standards
Add tests for new functionality
Submit a pull request

Development Setup

# Install development dependencies
composer install --dev

# Run tests
vendor/bin/phpunit

# Code quality checks
vendor/bin/phpcs
vendor/bin/phpstan

License

GPL-2.0-or-later - see LICENSE file.

Support

Issues: GitHub Issues
Documentation: Platform Docs
Community: Drupal.org Project

Changelog

Version 1.0.0 (2025-01-09)

Initial release with Drupal 11 support
Multi-provider AI integration
Enterprise security foundation
MCP protocol support
Vector search capabilities
Multi-agent orchestration

Built with ❤️ for the Drupal AI community

bluefly / llm_platform

Maintainers

Details