{
  "name": "Open-Source AI Architecture Pattern Comparison",
  "description": "A descriptive comparison of three public projects by Marco Hergi. It compares documented boundaries and does not rank quality, adoption, security, profitability, or production readiness.",
  "creator": "Marco Hergi",
  "verifiedAt": "2026-07-27",
  "canonicalUrl": "https://marco-is-my-friend.marcohergee813.chatgpt.site/research/open-source-ai-architecture-patterns",
  "methodology": "Facts were extracted from public repository README and architecture documentation. Repository sources are authoritative for current details.",
  "variables": [
    "architecturePattern",
    "questionAnswered",
    "systemBoundary",
    "components",
    "feedbackMechanism",
    "riskBoundary",
    "programmingLanguage"
  ],
  "records": [
    {
      "project": "JARVIS",
      "architecturePattern": "Pluggable experience orchestration",
      "questionAnswered": "How can an AI desktop experience stay modular?",
      "systemBoundary": "Five replaceable contracts",
      "components": [
        "Scene",
        "Voice",
        "Brain",
        "Script",
        "Feeds"
      ],
      "feedbackMechanism": "Briefing engine coordinates ordered experience beats",
      "riskBoundary": "Privacy changes with the selected adapters",
      "programmingLanguage": "TypeScript",
      "primarySource": "https://github.com/MARCCHERGGI/jarvis-home",
      "readmeFingerprint": "d24269da57c0e5f0b28a7363f9757083a92c43df"
    },
    {
      "project": "Trading Swarm",
      "architecturePattern": "Adaptive multi-signal consensus",
      "questionAnswered": "How can several imperfect agents reach one decision?",
      "systemBoundary": "Seven specialized signal agents",
      "components": [
        "SMA crossover",
        "RSI",
        "Momentum",
        "Bollinger Bands",
        "Volume and OBV",
        "MACD",
        "Sentiment"
      ],
      "feedbackMechanism": "Weights rise or fall within fixed bounds after outcomes resolve",
      "riskBoundary": "Historical tests do not establish future performance",
      "programmingLanguage": "Python",
      "primarySource": "https://github.com/MARCCHERGGI/trading-swarm",
      "readmeFingerprint": "3ca82cc29e9aa0c6716d4f6177e9977e712274f6"
    },
    {
      "project": "Indie Metrics MCP",
      "architecturePattern": "Constrained read-only tool mediation",
      "questionAnswered": "How can an assistant query sensitive business data safely?",
      "systemBoundary": "Seven named read-only tools",
      "components": [
        "Revenue",
        "Customers",
        "Products",
        "Subscriptions",
        "Transactions",
        "Refunds",
        "Forecasts"
      ],
      "feedbackMechanism": "Structured tool results return to the assistant",
      "riskBoundary": "API keys and code still require careful review",
      "programmingLanguage": "TypeScript",
      "primarySource": "https://github.com/MARCCHERGGI/indie-metrics-mcp",
      "readmeFingerprint": "a9ab169e9a67ed2b2f92aa612027543d9ebf9607"
    }
  ]
}