Cloudflare Multi-Agent System

Status: 🚀 MVP Development - Multi-Agent Autonomous Build Timeline: 4-6 hours automated deployment Budget: $1000 Claude Code credits

Overview

Migration of 120-agent Content Forge system to Cloudflare Workers infrastructure, creating a generic, flexible, and portable multi-agent platform that can be consumed by any authenticated application.

Key Features

• Hierarchical Instance Management: Organization → Instance → Project
• Provider-Agnostic: Extensible framework for multiple AI providers
• Dynamic Model Configuration: Admin-managed AI models without code changes
• Unified Payload Mapping: Single interface across all providers and models
• Autonomous Development: 4 Team Leaders managing 16 agents building in parallel
• Production-Ready: Rate limiting, error handling, monitoring, CI/CD

Architecture

Organization (e.g., Acme Corp.)
  └── Instance (like a VM - "production", "staging")
      ├── API Keys (shared across projects)
      ├── Rate Limits (shared pool)
      ├── Workers (dedicated deployments)
      └── Projects (logical groupings)

Model Configuration System

The Model Configuration System enables flexible, admin-managed AI model configurations without requiring code changes.

Features

• Centralized Management: All model configs stored in D1 database
• Admin UI: Manage models through Admin Panel's Models page
• Provider Agnostic: Support any AI provider (OpenAI, Ideogram, Gemini, etc.)
• Dynamic Payload Mapping: Template-based transformation of inputs to provider formats
• Capability Tracking: Mark models for image, video, text, audio generation
• Pricing & Rate Limits: Built-in metadata for billing and throttling
• Lifecycle Management: Active, Beta, Deprecated status tracking

How It Works

1. Admin Configuration: Create model config in Admin Panel

   {
     "model_id": "ideogram-v2",
     "provider_id": "ideogram",
     "capabilities": {"image": true},
     "payload_mapping": {
       "endpoint": "/generate",
       "headers": {"Api-Key": "{api_key}"},
       "body": {"prompt": "{user_prompt}"}
     }
   }

2. User Selection: User selects model in Testing GUI dropdown

3. Dynamic Mapping: Worker fetches config and transforms request

   const config = await getModelConfig(modelId);
   const request = applyPayloadMapping(config.payload_mapping, userInputs, apiKey);

4. Provider Execution: Formatted request sent to provider API

Supported Providers

• Ideogram: V2, V2 Turbo
• OpenAI: DALL-E 3, DALL-E 2
• Gemini: Veo 3.1 (video), 2.5 Flash Image, Imagen 3
• Anthropic: Claude 3.5 Sonnet (text)
• Extensible: Add new providers via Admin Panel

Documentation

• Admin Guide - How to add/edit model configs
• User Guide - How to use models in Testing GUI
• API Documentation - Programmatic access
• Payload Mapping Spec - Template syntax details
• Database Schema - Schema documentation

MVP Scope

In Scope

• Config Service (D1 database + worker)
• Model Configuration System (admin-managed models)
• Image Generation Worker (Ideogram provider)
• Rate Limiting (Durable Objects)
• R2 Storage Integration
• Authentication & Authorization
• Deployment Automation (GitHub Actions)
• Testing GUI & Admin Interface

Out of Scope (Future)

• Advanced billing/usage tiers
• Multi-tenancy (single org for MVP)
• Real-time streaming generation
• Image editing/inpainting

Multi-Agent Development Structure

Project Manager (Human)
├── Team Leader 1: Infrastructure (Phase 1 - Sequential)
│   ├── Agent 1.1: Database Schema
│   ├── Agent 1.2: Config Service Worker
│   ├── Agent 1.3: Authentication Middleware
│   └── Agent 1.4: Instance Lookup Logic
├── Team Leaders 2 & 3: Workers + Ops (Phase 2 - Parallel)
│   ├── Team 2: Worker Implementation
│   │   ├── Agent 2.1: Provider Adapter Framework
│   │   ├── Agent 2.2: Rate Limiter (Durable Objects)
│   │   ├── Agent 2.3: R2 Storage Manager
│   │   └── Agent 2.4: Image Generation Worker
│   └── Team 3: Operations
│       ├── Agent 3.1: Error Handling & Retries
│       ├── Agent 3.2: Logging System
│       ├── Agent 3.3: Deployment Scripts
│       └── Agent 3.4: GitHub Actions CI/CD
└── Team Leader 4: Interfaces (Phase 3 - Sequential)
    ├── Agent 4.1: Testing GUI
    ├── Agent 4.2: Admin Interface
    ├── Agent 4.3: Documentation
    └── Agent 4.4: Monitoring Dashboard

Quick Start

Prerequisites

• Node.js 18+
• Cloudflare account
• Wrangler CLI (npm install -g wrangler)

Setup

# Clone repository
git clone <your-repository-url>
cd cloudflare-multiagent-system

# Install dependencies
npm install

# Configure environment
cp .env.example .env
# Edit .env with your Cloudflare credentials

# Deploy infrastructure
npm run deploy-instance -- --config instances/production.json

Project Structure

/
├── docs/                    # Documentation and specifications
│   └── specs/              # Shared specs for all agents
├── infrastructure/          # Core infrastructure components
│   ├── database/           # D1 schema and migrations
│   ├── config-service/     # Central config management
│   ├── auth/               # Authentication middleware
│   └── lookup/             # Instance resolution
├── workers/                 # Cloudflare Workers
│   ├── shared/             # Shared utilities
│   │   ├── provider-adapters/
│   │   ├── rate-limiter/
│   │   ├── r2-manager/
│   │   ├── error-handling/
│   │   └── logging/
│   └── image-gen/          # Image generation worker
├── interfaces/              # User-facing interfaces
│   ├── testing-gui/        # Testing interface
│   ├── admin-panel/        # Instance management
│   └── monitoring/         # Dashboard
├── scripts/                 # Deployment automation
├── tests/                   # Test suites
└── prompts/                 # Multi-agent prompts

Development

Running Tests

npm test              # Run all tests
npm run test:watch    # Watch mode
npm run test:coverage # With coverage

Deploying Workers

npm run wrangler:dev    # Local development
npm run wrangler:deploy # Deploy to Cloudflare

Managing Instances

npm run deploy-instance -- --config config.json
npm run deploy-all      # Deploy all instances

Monitoring Progress

Track multi-agent development:

# Watch git commits from all agents
git log --all --oneline --graph

# Count completed agents
git log --all --grep="\[AGENT.*complete" | wc -l

# Check for escalations
git log --all --grep="ESCALATION"

Technical Stack

• Compute: Cloudflare Workers
• Database: D1 (SQLite)
• Storage: R2
• Cache: KV
• State: Durable Objects
• CI/CD: GitHub Actions
• Language: TypeScript
• Testing: Vitest

Success Criteria

• ✅ Config Service deployed and responding
• ✅ Model Configuration System operational
• ✅ Image Gen Worker functional with multiple providers
• ✅ Dynamic payload mapping working
• ✅ Rate limiting operational
• ✅ Testing GUI accessible with model selection
• ✅ Admin panel functional with Models page
• ✅ 2 instances deployed (production + development)
• ✅ All tests passing
• ✅ CI/CD pipeline working

License

MIT

Contributing

This project is built autonomously by AI agents. Human oversight for:

• Final approval before production merge
• Architectural decisions
• Credential management
• Monitoring and incident response

Adding New Workers/Services

When creating a new worker or service:

1. Add it to the Admin Panel Services page - See interfaces/admin-panel/ADDING_SERVICES.md
2. Follow the PR template - Complete the "New Service Checklist"
3. Document your API - Include endpoints, examples, and usage instructions
4. Create a Testing GUI (if user-facing) - Make it easy for others to try your service

This ensures all services are discoverable and properly documented for the team.

---

Built with Claude Code | Powered by Cloudflare Workers | Autonomous Multi-Agent Development