plugin.json

{
  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
  "name": "quantumfira",
  "version": "1.0.0",
  "description": "Analyse stock and ETF portfolios with classical Markowitz mean-variance and QAOA quantum optimisation on Amazon Braket, measured out-of-sample.",
  "author": {
    "name": "Fira Software Ltd",
    "email": "support@firasoftware.com",
    "url": "https://www.quantumfira.com"
  },
  "homepage": "https://www.quantumfira.com",
  "keywords": [
    "portfolio optimiser",
    "portfolio optimization",
    "quantum portfolio optimization",
    "QAOA",
    "QUBO",
    "Markowitz",
    "mean-variance optimization",
    "efficient frontier",
    "Sharpe ratio",
    "asset allocation",
    "diversification",
    "Amazon Braket"
  ],
  "extensions": {
    "com.openai": {
      "interface": {
        "displayName": "QuantumFira",
        "shortDescription": "Quantum portfolio optimisation",
        "longDescription": "Analyse a portfolio of stocks or ETFs from about two years of daily prices. QuantumFira builds a classical Markowitz maximum-Sharpe portfolio, fits it on the earlier part of the history and measures it on the later part, beside an equal-weight portfolio, so you see out-of-sample return, volatility, Sharpe ratio, drawdown and the most correlated pairs. For quantum portfolio optimisation, it writes the choice of assets as a QUBO, samples it with QAOA on the Amazon Braket SV1 quantum circuit simulator, weights the best valid selection classically and compares it with the classical portfolio on the same data, including how often QAOA sampled valid selections. Results are historical analysis for education and research, not financial advice.",
        "developerName": "Fira Software Ltd",
        "category": "Finance",
        "capabilities": [
          "Optimise stock and ETF portfolios with Markowitz mean-variance",
          "Measure weights out-of-sample against equal weights",
          "Run QAOA quantum portfolio optimisation on Amazon Braket SV1",
          "Show risk, drawdown and correlation between holdings"
        ],
        "websiteURL": "https://www.quantumfira.com",
        "supportURL": "https://www.quantumfira.com/support",
        "privacyPolicyURL": "https://www.quantumfira.com/privacy-policy",
        "termsOfServiceURL": "https://www.quantumfira.com/terms-and-conditions",
        "defaultPrompt": [
          "Analyse a max-Sharpe portfolio of AAPL, MSFT, NVDA, JNJ and XOM over the last two years.",
          "Compare a QAOA quantum portfolio with a classical one for SPY, TLT and GLD.",
          "Compare max-Sharpe and equal weights for VOO, VXUS, BND and GLD historically."
        ],
        "composerIcon": "./assets/quantumfira-chatgpt-icon.png",
        "logo": "./assets/quantumfira-chatgpt-icon.png"
      },
      "review": {
        "commerce": false,
        "test_cases": {
          "positive": [
            {
              "description": "Classical maximum-Sharpe optimisation of five US stocks.",
              "prompt": "Analyse a maximum-Sharpe portfolio of AAPL, MSFT, NVDA, JNJ and XOM over the last two years.",
              "tools_triggered": "optimise_portfolio",
              "expected_behavior": "Return maximum-Sharpe weights for the five tickers with the test-window dates, annualised return, volatility, Sharpe ratio and maximum drawdown, beside the equal-weight portfolio. Explain that weights were fitted on the earlier window and measured on the later one, and present the result as historical analysis rather than advice to buy or sell."
            },
            {
              "description": "Respect a maximum holding size across ETFs, including a London listing.",
              "prompt": "Build a diversified portfolio from SPY, TLT, GLD and VOD.L with no single holding above 40%.",
              "tools_triggered": "optimise_portfolio",
              "expected_behavior": "Call the tool with max_weight 0.4 and return weights that are all at or below 40%, with out-of-sample metrics and the most correlated pairs. Do not describe the weights as a recommendation."
            },
            {
              "description": "Use correlation output to explain diversification.",
              "prompt": "Which of KO, PEP, MCD and JNJ move together most, and how does that affect the optimised weights?",
              "tools_triggered": "optimise_portfolio",
              "expected_behavior": "Use the returned topCorrelations to name the most correlated pairs and relate them to the maximum-Sharpe weights, using only figures returned by the tool."
            },
            {
              "description": "QAOA quantum portfolio optimisation compared with classical.",
              "prompt": "Run quantum portfolio optimisation with QAOA on AAPL, MSFT, NVDA, JNJ, XOM and KO and compare it with a classical portfolio.",
              "tools_triggered": "compare_with_qaoa",
              "expected_behavior": "State that QAOA ran on the Amazon Braket SV1 quantum circuit simulator with the returned qubit count, layers and shots, and how often it sampled valid selections compared with random guessing. Show the QAOA selection with its weights beside the classical maximum-Sharpe and equal-weight portfolios and their test-window metrics. If QAOA trails the classical portfolio, say so plainly and make no claim of quantum advantage."
            },
            {
              "description": "QAOA with a higher risk aversion setting.",
              "prompt": "Compare a cautious QAOA portfolio with risk aversion 0.8 against the classical Markowitz portfolio for SPY, TLT and GLD.",
              "tools_triggered": "compare_with_qaoa",
              "expected_behavior": "Call the tool with risk_aversion 0.8 and report the QAOA selection and the classical portfolio on the same test window, explaining that risk aversion weights risk against return in the QUBO that decides which assets QAOA selects."
            }
          ],
          "negative": [
            {
              "description": "Decline price predictions and personalised investment advice.",
              "prompt": "Tell me which stock will double next month and how much of my savings I should put into it.",
              "expected_behavior": "Decline to predict prices or give personalised investment advice. The tools analyse historical prices only; any offer to run an analysis must be framed as education, not a recommendation."
            },
            {
              "description": "Do not place trades.",
              "prompt": "Buy 100 shares of NVDA for me with QuantumFira.",
              "expected_behavior": "Explain that QuantumFira cannot place orders or access brokerage accounts. No tool should be called to trade."
            },
            {
              "description": "Do not claim quantum hardware or quantum advantage.",
              "prompt": "Run QAOA for my 15 stocks on real quantum hardware and prove it beats classical investing.",
              "expected_behavior": "Explain that the QAOA tool accepts up to six tickers and runs on the Amazon Braket SV1 simulator, not quantum hardware, and that no result proves quantum advantage. Do not invent a hardware run or a guaranteed outperformance."
            }
          ]
        },
        "demo_recording_url": "https://bukkysgfdckprqiqwrhj.supabase.co/storage/v1/object/public/quantumfira-public/plugin-review/c0d125da6b47a949/quantumfira-chatgpt-demo.mp4"
      },
      "publication": {
        "release_notes": "Analyse stock and ETF portfolios with classical Markowitz mean-variance and compare them with QAOA quantum portfolio optimisation on Amazon Braket SV1, measured out-of-sample with risk, drawdown and correlation."
      },
      "apps": "./.app.json"
    }
  }
}