Skip to content
AI Red vs Blue Arena

Architecture Overview

Architecture Overview

The AI Red vs Blue Arena is built on a modular architecture with clear separation of concerns.

System Components

┌─────────────────────────────────────────────────────────────┐
│                      Frontend (Next.js)                      │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐   │
│  │ Dashboard│  │  Agents  │  │  Matches │  │ Settings │   │
│  └────┬─────┘  └────┬─────┘  └────┬─────┘  └────┬─────┘   │
└───────┼─────────────┼─────────────┼─────────────┼──────────┘
        │             │             │             │
        └─────────────┴─────────────┴─────────────┘
                         │ HTTP/WS
        ┌────────────────┴────────────────┐
        │      Backend (Express.js)        │
        │  ┌──────────────────────────┐   │
        │  │      API Routes          │   │
        │  │  - Agents               │   │
        │  │  - Matches               │   │
        │  │  - Events                │   │
        │  │  - Config                │   │
        │  └──────────┬───────────────┘   │
        │             │                   │
        │  ┌──────────▼───────────────┐   │
        │  │   Match Runner           │   │
        │  │  - Orchestrates rounds   │   │
        │  │  - Manages state          │   │
        │  └──────────┬───────────────┘   │
        │             │                   │
        │  ┌──────────▼───────────────┐   │
        │  │   Agent Executor          │   │
        │  │  - LLM provider management│   │
        │  │  - Response generation     │   │
        │  └──────────┬───────────────┘   │
        │             │                   │
        │  ┌──────────▼───────────────┐   │
        │  │   Cline Integration       │   │
        │  │  - Tool execution         │   │
        │  │  - Sandboxing             │   │
        │  └──────────┬───────────────┘   │
        │             │                   │
        │  ┌──────────▼───────────────┐   │
        │  │   WebSocket Server        │   │
        │  │  - Real-time updates      │   │
        │  └──────────────────────────┘   │
        └────────────┬────────────────────┘
                     │
        ┌────────────▼────────────┐
        │   Database (lowdb)      │
        │  - Agents               │
        │  - Matches              │
        │  - Events               │
        │  - Rounds                │
        └─────────────────────────┘

Data Flow

Match Execution Flow

  1. User creates match via Frontend/API
  2. Match Runner loads agents from database
  3. For each round:
    • Red Agent generates attack via Agent Executor
    • Target Agent processes attack
    • Blue Agent generates defense
    • Scoring system evaluates round
    • Events stored in database
    • WebSocket broadcasts updates
  4. Match completes with final scores

Tool Execution Flow

  1. Agent generates response with tool calls
  2. Cline checks permissions
  3. Tools executed in sandbox
  4. Results returned to agent
  5. Agent response enhanced with tool results
  6. Execution logged for audit

Component Responsibilities

Frontend

  • User interface and visualization
  • Real-time event display
  • Agent and match management
  • Settings configuration

Backend API

  • RESTful API endpoints
  • Request validation
  • Business logic orchestration
  • Database operations

Match Runner

  • Round orchestration
  • Agent coordination
  • State management
  • Event generation

Agent Executor

  • LLM provider management
  • Response generation
  • Model fallback handling
  • API key management

Cline Integration

  • Tool execution
  • Permission checking
  • Sandboxing
  • Audit logging

Database

  • Data persistence
  • Query operations
  • Schema management

WebSocket Server

  • Real-time event broadcasting
  • Client subscription management
  • Match updates

Integration Points

Kestra

  • Workflow orchestration
  • Scheduled matches
  • Batch processing
  • Dataset generation

Oumi

  • Training dataset export
  • Fine-tuning job submission
  • Model improvement pipeline

LLM Providers

  • Groq (primary)
  • OpenAI
  • Anthropic
  • Mock (fallback)

Security Architecture

  • Sandboxed execution for all tool calls
  • Permission-based access control
  • Input validation at all layers
  • Audit logging for all operations
  • CORS protection for API
  • Rate limiting via model fallback

Scalability Considerations

  • Stateless API design
  • File-based database (can be migrated to SQL)
  • WebSocket for real-time updates
  • Modular architecture for easy extension

Next Steps