AgentOS — WASM Swarm Agent Operating System

A production-grade WASM-powered multi-agent system that optimizes prompts autonomously, executes across multiple AI models, scores best outputs, and evolves via feedback loops.

---

🚀 Quick Start

npm install
npm run dev        # http://localhost:3000

Open http://localhost:3000/dashboard to see the cyberpunk control panel.

---

✨ What is AgentOS?

AgentOS is an autonomous prompt optimization operating system — a self-evolving AI execution layer that:

• 🤖 Optimizes prompts through a 7-agent swarm pipeline
• ⚡ Executes across multiple AI models (OpenAI, Claude, Gemini, Llama)
• 📊 Scores and selects the best outputs via weighted evaluation
• 🔁 Evolves prompts via feedback loops (up to 10 iterations)
• 🛡️ Secures execution against prompt injection & jailbreaks
• 🧠 Stores and versions high-performing prompts
• 🌐 WASM execution layer for parallel, deterministic agent tasks

Backward compatibility: import { PromptOS } from '@/core' still works — PromptOS is aliased to SwarmOrchestrator.

---

🏗️ Architecture

/core
├── agents/          # 7 stateless typed agents
├── swarm/           # SwarmOrchestrator — pipeline coordinator
├── wasm/            # WasmRunner + WorkerPool
├── models/          # ModelRouter (OpenAI / Claude / Gemini / Llama)
├── evaluation/      # ScoringEngine (weighted formula)
├── memory/          # MemorySystem (short + long-term)
├── security/        # PromptSanitizer + InjectionDetector
└── prompt/          # Prompt DNA standard enforcement

Swarm Pipeline

Input → PromptArchitect → Security → Execute → Evaluate → [Optimize loop] → Deploy

Scoring Formula

score = (accuracy × 0.40) + (completeness × 0.25) + (determinism × 0.15)
      + (cost × 0.10)     + (latency × 0.10)

---

🧪 Testing

npm test               # run all tests
npm run test:coverage  # with coverage report

Coverage now includes:

• Unit tests for all swarm agents
• End-to-end swarm orchestration tests
• Determinism tests for WASM execution
• Worker-pool load and queue saturation tests

---

🔐 Enterprise CI/CD and Security

CI/CD Workflows

• CI (.github/workflows/ci.yml)
  • Deterministic npm ci
  • Node.js matrix validation (20, 22)
  • Type-check, lint, tests, build
  • Coverage artifact upload
• Release (.github/workflows/release.yml)
  • Triggered by semantic tags (v*.*.*)
  • Full pre-release validation
  • Reproducible package artifact + checksums
  • Automated GitHub release creation
• Deploy (.github/workflows/deploy.yml)
  • Human-triggered deployment path
  • Staging then production gating via environments

Security Governance

• Security workflow (.github/workflows/security.yml)
  • CodeQL analysis
  • Dependency review for pull requests
  • Scheduled and branch-based npm audit
• Prompt injection and XSS sanitization built into core runtime
• No embedded secrets in codebase

---

🚢 Deployment Playbooks

• Docker: docs/deployment/docker.md
• Kubernetes: docs/deployment/kubernetes.md
• Vercel: docs/deployment/vercel.md

Kubernetes base manifests are provided in deploy/k8s/base/ with secure defaults:

• Non-root execution
• Read-only root filesystem
• Capability drop (ALL)
• Liveness/readiness probes
• HorizontalPodAutoscaler

---

📈 Observability, Scaling, and Recovery

• Monitoring and alerting: docs/operations/monitoring-alerting.md
• Scaling strategy: docs/operations/scaling-playbook.md
• Disaster recovery: docs/operations/disaster-recovery.md

GET /api/status now returns health and operational metadata:

• status
• timestamp
• uptime
• model and score snapshot
• version count

---

⚙️ Configuration

Copy .env.example to .env.local and fill in your API keys:

OPENAI_API_KEY=
CLAUDE_API_KEY=
GEMINI_API_KEY=
LLAMA_API_KEY=
MARKETPLACE_ENABLED=true

The system works without any API keys — models can run through deterministic mock behavior.

---

🔧 Scripts

Script	Description
npm run dev	Start development server
npm run build	Production build
npm run type-check	TypeScript strict check
npm run lint	ESLint
npm test	Jest suite
npm run test:coverage	Jest coverage

---

📄 License

MIT — see LICENSE