{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"tool_finster-ai","slug":"finster-ai","name":"Finster AI","type":"product","url":"https://finster.ai","page_url":"https://unfragile.ai/finster-ai","categories":["data-analysis","code-review-security"],"tags":[],"pricing":{"model":"paid","free":false,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"tool_finster-ai__cap_0","uri":"capability://data.processing.analysis.real.time.financial.data.ingestion.and.normalization","name":"real-time financial data ingestion and normalization","description":"Finster AI ingests multi-source financial datasets (market feeds, corporate filings, alternative data) and normalizes them into a unified schema for downstream analysis. The system likely uses streaming pipelines (Kafka or similar) to handle real-time market data while applying schema validation and data quality checks to ensure consistency across heterogeneous sources before ML model consumption.","intents":["I need to consolidate market data from multiple exchanges and data providers into a single analytical view","I want to ensure data quality and consistency before feeding datasets into machine learning models","I need to process real-time market feeds without latency bottlenecks that would delay pattern detection"],"best_for":["institutional investors managing multi-asset portfolios","hedge funds requiring sub-second data freshness for algorithmic trading","financial advisors consolidating client data from disparate sources"],"limitations":["real-time ingestion adds operational complexity requiring 24/7 infrastructure monitoring","schema normalization may lose domain-specific metadata from source systems","data quality issues in upstream sources propagate through the pipeline without manual intervention"],"requires":["API credentials for financial data providers (Bloomberg, Reuters, or custom feeds)","network connectivity with guaranteed uptime SLAs","data warehouse or lake infrastructure to persist normalized datasets"],"input_types":["structured market data feeds (OHLCV, tick data)","unstructured corporate filings (10-K, earnings transcripts)","alternative data (sentiment, web traffic, satellite imagery)"],"output_types":["normalized tabular datasets","time-series data in standardized format","data quality reports and anomaly flags"],"categories":["data-processing-analysis","real-time-streaming"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_1","uri":"capability://data.processing.analysis.machine.learning.driven.pattern.recognition.and.anomaly.detection","name":"machine learning-driven pattern recognition and anomaly detection","description":"Finster AI applies supervised and unsupervised ML models (likely ensemble methods combining tree-based models, neural networks, and statistical approaches) to identify market patterns, correlations, and anomalies in historical and real-time financial data. The system trains on labeled datasets of known market events and uses feature engineering pipelines to extract predictive signals from raw OHLCV, sentiment, and alternative data inputs.","intents":["I want to detect unusual market behavior or trading patterns that signal emerging opportunities or risks","I need to identify correlations between assets that traditional correlation matrices miss","I want to backtest ML-driven trading signals against historical data to validate predictive power"],"best_for":["quantitative traders and hedge funds building systematic trading strategies","portfolio managers seeking early warning signals for portfolio rebalancing","risk analysts identifying tail-risk scenarios and market regime changes"],"limitations":["ML models trained on historical data may fail during unprecedented market regimes (e.g., 2008 financial crisis, COVID crash)","feature engineering requires domain expertise and iterative tuning; no one-size-fits-all feature set","model interpretability is limited — ensemble methods and neural networks act as black boxes, complicating regulatory justification","backtesting results suffer from look-ahead bias and overfitting if not carefully validated with out-of-sample data"],"requires":["historical financial datasets spanning multiple market cycles (minimum 5-10 years)","computational resources for model training (GPU clusters for neural network training)","domain expertise in feature engineering and model validation"],"input_types":["time-series OHLCV data","sentiment scores from news/social media","alternative data (volatility indices, options flow, positioning data)","macroeconomic indicators"],"output_types":["anomaly scores and confidence intervals","pattern classifications (e.g., 'bullish reversal', 'liquidity crisis')","feature importance rankings","backtested signal performance metrics"],"categories":["data-processing-analysis","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_10","uri":"capability://tool.use.integration.api.driven.integration.with.external.systems.and.data.providers","name":"api-driven integration with external systems and data providers","description":"Finster AI exposes REST and/or GraphQL APIs enabling integration with external systems (portfolio management systems, trading platforms, CRM systems) and data providers (market data feeds, alternative data vendors). The system supports webhook notifications for real-time alerts and provides SDKs for popular programming languages (Python, JavaScript, Java) to simplify integration for developers.","intents":["I want to integrate Finster's analytics into my existing portfolio management system without manual data export/import","I need to receive real-time alerts when Finster detects anomalies or generates rebalancing recommendations","I want to build custom applications on top of Finster's analytical capabilities using APIs"],"best_for":["developers and engineers building financial applications","financial institutions integrating Finster with existing systems","fintech startups leveraging Finster's analytics as a service"],"limitations":["API rate limits may constrain high-frequency queries; real-time analytics may require premium tier access","API documentation may be incomplete or outdated; integration requires reverse-engineering or support tickets","breaking API changes can break downstream integrations; versioning strategy must be clearly communicated","API authentication and authorization add complexity; managing API keys and OAuth tokens requires careful security practices"],"requires":["API credentials (API key or OAuth token)","documentation of API endpoints and request/response formats","SDKs for target programming languages (Python, JavaScript, Java, etc.)","network connectivity and firewall rules allowing outbound HTTPS connections"],"input_types":["portfolio data (holdings, transactions)","market data (prices, volatilities)","user queries and filter parameters"],"output_types":["JSON/XML responses with analytics results","webhook notifications for real-time events","SDK objects and data structures"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_2","uri":"capability://planning.reasoning.portfolio.optimization.and.rebalancing.recommendations","name":"portfolio optimization and rebalancing recommendations","description":"Finster AI applies modern portfolio theory (mean-variance optimization, risk parity, factor-based allocation) combined with ML-derived expected returns and covariance matrices to generate portfolio allocation recommendations. The system likely uses constrained optimization solvers (quadratic programming) to respect institutional constraints (position limits, sector caps, ESG filters) and generates rebalancing signals based on drift thresholds or ML-predicted regime changes.","intents":["I want to optimize my portfolio allocation given my risk tolerance and constraints without manual rebalancing","I need to understand the risk contribution of each position and adjust allocations to target risk budgets","I want to incorporate ML-predicted returns into my optimization rather than relying on historical averages"],"best_for":["institutional asset managers managing large portfolios with complex constraints","financial advisors automating portfolio construction for client accounts","pension funds and endowments seeking systematic rebalancing strategies"],"limitations":["optimization is only as good as input assumptions (expected returns, covariance matrix); garbage-in-garbage-out problem","constrained optimization may produce corner solutions (extreme allocations) if constraints are misspecified","rebalancing recommendations ignore transaction costs and tax implications, which can erode returns","ML-predicted returns introduce model risk — if predictions are systematically biased, allocations will be suboptimal"],"requires":["expected return estimates (from ML models or analyst forecasts)","covariance matrix of asset returns (estimated from historical data or ML models)","portfolio constraints specification (position limits, sector caps, ESG filters)","optimization solver (commercial: Gurobi, CPLEX; open-source: CVXPY, Pyomo)"],"input_types":["current portfolio holdings and weights","ML-predicted expected returns and volatilities","correlation/covariance matrix","constraint specifications (min/max weights, sector limits)"],"output_types":["recommended portfolio weights","rebalancing trades (buy/sell orders)","risk decomposition (contribution by asset, sector, factor)","expected portfolio return and volatility"],"categories":["planning-reasoning","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_3","uri":"capability://automation.workflow.compliance.and.regulatory.reporting.automation","name":"compliance and regulatory reporting automation","description":"Finster AI automates generation of regulatory reports (MiFID II, Dodd-Frank, SEC filings) by mapping portfolio data and trade history to regulatory schemas, calculating required metrics (VaR, Sharpe ratio, concentration limits), and generating audit trails documenting all analytical decisions. The system maintains data lineage and version control to support regulatory inquiries and implements role-based access controls to enforce segregation of duties.","intents":["I need to generate regulatory reports without manual data compilation and calculation errors","I want to maintain audit trails proving that my investment decisions comply with regulatory requirements","I need to demonstrate that my portfolio management process is systematic and not subject to bias or discretionary override"],"best_for":["institutional asset managers subject to MiFID II, Dodd-Frank, or other regulatory regimes","hedge funds and private equity firms managing investor capital with fiduciary obligations","compliance officers and risk managers seeking to automate regulatory reporting workflows"],"limitations":["regulatory requirements vary by jurisdiction and change frequently; Finster's schemas may lag regulatory updates","automated reporting cannot replace human judgment in interpreting ambiguous regulatory guidance","audit trails add storage overhead and query latency for large-scale historical data","false positives in compliance checks (e.g., flagging legitimate trades as violations) require manual review and tuning"],"requires":["complete trade and position history with timestamps","client/counterparty data and relationship mapping","regulatory jurisdiction specification (MiFID II, Dodd-Frank, etc.)","compliance rule configuration (position limits, concentration thresholds, ESG filters)"],"input_types":["trade records (execution price, quantity, counterparty, timestamp)","portfolio holdings and valuations","client profiles and investment mandates","compliance rule definitions"],"output_types":["regulatory reports (MiFID II, Dodd-Frank, SEC filings)","audit trails and decision logs","compliance exception reports and violation alerts","evidence packages for regulatory inquiries"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_4","uri":"capability://safety.moderation.enterprise.grade.data.security.and.encryption","name":"enterprise-grade data security and encryption","description":"Finster AI implements multi-layered security controls including encryption at rest (AES-256) and in transit (TLS 1.3), role-based access control (RBAC) with fine-grained permissions, and data segregation (logical or physical isolation of client datasets). The platform likely uses hardware security modules (HSMs) for key management and implements audit logging to track all data access and modifications for compliance and forensic analysis.","intents":["I need to ensure that sensitive portfolio and trading data is protected from unauthorized access and breaches","I want to comply with data protection regulations (GDPR, CCPA) and financial industry standards (SOC 2, ISO 27001)","I need to maintain audit trails proving that only authorized personnel accessed specific datasets"],"best_for":["institutional investors and hedge funds managing confidential trading strategies","financial advisors handling sensitive client portfolio data","compliance officers and security teams requiring comprehensive data governance"],"limitations":["encryption overhead adds latency to queries and analytics; real-time analysis may require decryption in memory","key management complexity increases operational burden; key rotation and recovery procedures must be carefully designed","RBAC granularity may be insufficient for complex organizational structures with matrix reporting","audit logging generates large volumes of data; querying audit trails for forensic analysis can be slow"],"requires":["infrastructure supporting encryption (HSMs, key management services)","identity and access management system (LDAP, Active Directory, or cloud IAM)","secure network architecture (VPCs, firewalls, network segmentation)","compliance certifications (SOC 2 Type II, ISO 27001) or equivalent security assessments"],"input_types":["user credentials and authentication tokens","access control policies and role definitions","data classification tags (public, confidential, restricted)"],"output_types":["encrypted data at rest and in transit","access control decisions (allow/deny)","audit logs documenting all data access and modifications","security incident reports and alerts"],"categories":["safety-moderation","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_5","uri":"capability://data.processing.analysis.multi.asset.class.analysis.and.cross.asset.correlation.modeling","name":"multi-asset class analysis and cross-asset correlation modeling","description":"Finster AI extends pattern recognition and optimization across multiple asset classes (equities, fixed income, commodities, FX, derivatives) by building unified correlation models that capture cross-asset relationships and regime-dependent dependencies. The system uses dynamic correlation estimation (rolling windows, GARCH models, or ML-based approaches) to identify when traditional correlations break down and generates alerts for portfolio managers when diversification benefits diminish.","intents":["I want to understand how my equity and bond allocations interact during market stress and adjust accordingly","I need to identify when traditional diversification benefits disappear and rebalance proactively","I want to build multi-asset portfolios that optimize risk-adjusted returns across all asset classes"],"best_for":["global asset managers with multi-asset mandates","pension funds and endowments seeking diversification across asset classes","risk managers monitoring portfolio resilience during market regime changes"],"limitations":["cross-asset correlation estimation requires long historical datasets; sparse data for emerging assets limits model reliability","regime-dependent correlations are difficult to predict; models may fail to anticipate correlation breakdowns during crises","multi-asset optimization is computationally expensive; real-time rebalancing recommendations may have unacceptable latency","derivatives pricing and hedging require sophisticated models (Black-Scholes, stochastic volatility) that introduce additional model risk"],"requires":["historical price data for all asset classes (equities, bonds, commodities, FX, derivatives)","volatility surfaces and term structures for derivatives pricing","macroeconomic indicators and regime identification models","computational resources for multi-asset optimization (GPU clusters for large portfolios)"],"input_types":["time-series returns for multiple asset classes","volatility and correlation matrices","macroeconomic indicators (interest rates, inflation, GDP growth)","derivatives pricing inputs (spot prices, volatilities, interest rates)"],"output_types":["cross-asset correlation matrices (static and dynamic)","regime identification and probability estimates","multi-asset portfolio allocations","hedging recommendations and derivative positions"],"categories":["data-processing-analysis","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_6","uri":"capability://planning.reasoning.backtesting.and.strategy.validation.with.walk.forward.analysis","name":"backtesting and strategy validation with walk-forward analysis","description":"Finster AI provides backtesting infrastructure that simulates trading strategies against historical data while accounting for transaction costs, slippage, and market impact. The system implements walk-forward analysis (rolling out-of-sample validation) to prevent overfitting and uses Monte Carlo simulation to estimate strategy robustness under different market conditions. Results include performance metrics (Sharpe ratio, max drawdown, Calmar ratio) and risk decomposition.","intents":["I want to validate that my trading strategy would have worked in the past before deploying real capital","I need to understand the robustness of my strategy across different market regimes and stress scenarios","I want to compare multiple strategy variants and select the one with the best risk-adjusted returns"],"best_for":["quantitative traders and hedge funds developing systematic trading strategies","portfolio managers backtesting factor-based allocation strategies","risk managers validating strategy robustness before deployment"],"limitations":["backtesting results are biased optimistic due to look-ahead bias, overfitting, and survivorship bias in historical data","transaction costs and slippage assumptions may not reflect actual market conditions; estimated costs can be significantly lower than realized costs","walk-forward analysis reduces overfitting but increases computational cost; may be impractical for high-frequency strategies","Monte Carlo simulation assumes returns are i.i.d., which violates empirical properties (fat tails, autocorrelation, regime changes)","past performance does not guarantee future results; strategies that worked historically may fail in unprecedented market conditions"],"requires":["historical price data with sufficient granularity (daily for long-term strategies, intraday for short-term strategies)","transaction cost assumptions (commissions, bid-ask spreads, market impact models)","market microstructure data (order book depth, liquidity profiles) for accurate slippage estimation","computational resources for Monte Carlo simulation and walk-forward analysis"],"input_types":["strategy rules or ML model predictions","historical OHLCV data","transaction cost parameters","portfolio constraints (position limits, leverage limits)"],"output_types":["backtest performance metrics (returns, Sharpe ratio, max drawdown, Calmar ratio)","equity curve and drawdown chart","trade-by-trade analysis (entry/exit prices, P&L)","walk-forward validation results","Monte Carlo simulation results (confidence intervals on returns)"],"categories":["planning-reasoning","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_7","uri":"capability://planning.reasoning.risk.analytics.and.stress.testing.with.scenario.analysis","name":"risk analytics and stress testing with scenario analysis","description":"Finster AI calculates portfolio risk metrics (Value at Risk, Expected Shortfall, Greeks for derivatives) and runs stress tests by simulating portfolio performance under historical crisis scenarios (2008 financial crisis, COVID crash, flash crash) and hypothetical scenarios (interest rate shocks, currency devaluations, geopolitical events). The system uses Monte Carlo simulation and historical simulation methods to estimate tail risks and provides sensitivity analysis showing how portfolio value changes with market moves.","intents":["I want to understand the worst-case losses my portfolio could experience under extreme market conditions","I need to stress-test my portfolio against historical crises and hypothetical scenarios to validate risk controls","I want to measure the sensitivity of my portfolio to specific risk factors (interest rates, volatility, currency moves)"],"best_for":["institutional risk managers and chief risk officers","hedge funds and private equity firms managing tail risk","banks and financial institutions subject to regulatory stress testing (CCAR, DFAST)"],"limitations":["VaR and Expected Shortfall are backward-looking metrics based on historical distributions; they may underestimate tail risks in unprecedented scenarios","stress test scenarios are subjective; different scenario assumptions produce vastly different risk estimates","Monte Carlo simulation assumes specific return distributions (normal, student-t) that may not capture empirical fat tails and skewness","Greeks for derivatives are point-in-time estimates; they change rapidly during market stress and may be unreliable for large moves","computational cost of comprehensive stress testing (thousands of scenarios) may be prohibitive for real-time risk monitoring"],"requires":["historical price data for all portfolio assets","correlation and volatility estimates","derivatives pricing models (Black-Scholes, stochastic volatility models)","scenario definitions (historical crises, hypothetical shocks)","computational resources for Monte Carlo simulation"],"input_types":["current portfolio holdings and valuations","market prices and volatilities","correlation matrices","scenario specifications (market moves, volatility shocks)"],"output_types":["Value at Risk (VaR) estimates at multiple confidence levels","Expected Shortfall (CVaR) estimates","stress test results (portfolio P&L under each scenario)","sensitivity analysis (Greeks, duration, convexity)","risk decomposition by asset, sector, risk factor"],"categories":["planning-reasoning","data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_8","uri":"capability://data.processing.analysis.performance.attribution.and.factor.analysis","name":"performance attribution and factor analysis","description":"Finster AI decomposes portfolio returns into contributions from specific decisions (asset allocation, security selection, market timing) and identifies which factors (value, momentum, quality, size, volatility) drove returns. The system uses regression-based attribution (Brinson-Fachler) and factor models (Fama-French, custom factors) to quantify the impact of each decision and factor, enabling portfolio managers to understand what drove performance and validate investment theses.","intents":["I want to understand whether my portfolio outperformance came from good stock picking or just being overweight in winning sectors","I need to identify which factors (value, momentum, quality) are driving my returns and assess their sustainability","I want to measure the contribution of each investment decision to overall portfolio performance"],"best_for":["active portfolio managers justifying their performance to clients and stakeholders","factor-based investors validating factor exposures and returns","performance analysts and attribution specialists"],"limitations":["attribution results depend on factor model specification; different factor sets produce different attribution results","regression-based attribution assumes linear relationships between factors and returns; non-linear relationships are missed","factor definitions are subjective; there is no consensus on what constitutes 'value' or 'quality' factors","attribution is backward-looking; it explains past performance but does not predict future returns","high correlation between factors (multicollinearity) makes it difficult to isolate individual factor contributions"],"requires":["detailed portfolio holdings and transaction history","benchmark holdings and weights","factor definitions and factor returns (Fama-French factors, custom factors)","historical returns for all portfolio assets"],"input_types":["portfolio holdings and weights (current and historical)","benchmark holdings and weights","factor returns and exposures","transaction history (buys, sells, rebalancing)"],"output_types":["attribution breakdown (allocation effect, selection effect, interaction effect)","factor contribution analysis (return contribution by factor)","factor exposure analysis (portfolio factor loadings vs benchmark)","performance attribution reports"],"categories":["data-processing-analysis","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"tool_finster-ai__cap_9","uri":"capability://text.generation.language.client.reporting.and.dashboard.visualization","name":"client reporting and dashboard visualization","description":"Finster AI generates customizable client reports and interactive dashboards that visualize portfolio performance, risk metrics, and recommendations in formats tailored to different audiences (institutional clients, retail investors, compliance officers). The system supports white-label branding, automated report generation on schedules (monthly, quarterly), and drill-down capabilities enabling clients to explore underlying data and assumptions.","intents":["I want to generate professional client reports that explain portfolio performance and justify my investment decisions","I need to provide clients with interactive dashboards they can use to monitor their portfolios in real-time","I want to automate report generation and distribution to save time and ensure consistency"],"best_for":["financial advisors and asset managers communicating with clients","institutional investors reporting to stakeholders and regulators","compliance officers generating regulatory reports"],"limitations":["dashboard design is subjective; different clients have different information needs and preferences","real-time dashboards require continuous data updates; stale data can mislead clients","white-label branding adds complexity; maintaining multiple branded versions increases operational overhead","interactive dashboards may expose sensitive information if access controls are not properly configured"],"requires":["portfolio data and performance metrics","client preferences and reporting requirements","branding assets (logos, color schemes) for white-label reports","distribution infrastructure (email, web portal, PDF generation)"],"input_types":["portfolio holdings and valuations","performance metrics and attribution","risk analytics results","client metadata (preferences, reporting frequency)"],"output_types":["PDF reports (monthly, quarterly, annual)","interactive dashboards (web-based or embedded)","performance charts and visualizations","drill-down data and supporting analysis"],"categories":["text-generation-language","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":44,"verified":false,"data_access_risk":"high","permissions":["API credentials for financial data providers (Bloomberg, Reuters, or custom feeds)","network connectivity with guaranteed uptime SLAs","data warehouse or lake infrastructure to persist normalized datasets","historical financial datasets spanning multiple market cycles (minimum 5-10 years)","computational resources for model training (GPU clusters for neural network training)","domain expertise in feature engineering and model validation","API credentials (API key or OAuth token)","documentation of API endpoints and request/response formats","SDKs for target programming languages (Python, JavaScript, Java, etc.)","network connectivity and firewall rules allowing outbound HTTPS connections"],"failure_modes":["real-time ingestion adds operational complexity requiring 24/7 infrastructure monitoring","schema normalization may lose domain-specific metadata from source systems","data quality issues in upstream sources propagate through the pipeline without manual intervention","ML models trained on historical data may fail during unprecedented market regimes (e.g., 2008 financial crisis, COVID crash)","feature engineering requires domain expertise and iterative tuning; no one-size-fits-all feature set","model interpretability is limited — ensemble methods and neural networks act as black boxes, complicating regulatory justification","backtesting results suffer from look-ahead bias and overfitting if not carefully validated with out-of-sample data","API rate limits may constrain high-frequency queries; real-time analytics may require premium tier access","API documentation may be incomplete or outdated; integration requires reverse-engineering or support tickets","breaking API changes can break downstream integrations; versioning strategy must be clearly communicated","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.36666666666666664,"quality":0.78,"ecosystem":0.25,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.35,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:30.892Z","last_scraped_at":"2026-04-05T13:23:42.552Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=finster-ai","compare_url":"https://unfragile.ai/compare?artifact=finster-ai"}},"signature":"fOoPN3SoygN/qXgwiuTq9eN9bHYcYVOY0MT/Yo5dyXSDUa1MTas9DCnWzYvPUlUAtfKXG5n2eROkSSWN+WjiCw==","signedAt":"2026-06-22T04:15:12.654Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/finster-ai","artifact":"https://unfragile.ai/finster-ai","verify":"https://unfragile.ai/api/v1/verify?slug=finster-ai","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}