{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"neptune-ai","slug":"neptune-ai","name":"Neptune AI","type":"platform","url":"https://neptune.ai","page_url":"https://unfragile.ai/neptune-ai","categories":["model-training","deployment-infra"],"tags":[],"pricing":{"model":"free","free":true,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"neptune-ai__cap_0","uri":"capability://data.processing.analysis.experiment.metadata.tracking.with.hierarchical.versioning","name":"experiment metadata tracking with hierarchical versioning","description":"Captures and stores experiment metadata (hyperparameters, metrics, artifacts, environment configs) through SDK instrumentation that logs to a centralized metadata store with immutable versioning. Uses a hierarchical schema supporting nested parameter spaces, metric time-series, and artifact lineage tracking across thousands of concurrent experiments without requiring code refactoring.","intents":["I need to log hyperparameters, metrics, and artifacts from my training runs without modifying my existing training code","I want to track the full lineage of an experiment including environment, dependencies, and data versions","I need to compare metrics across 100+ experiments to identify which hyperparameter combinations performed best"],"best_for":["ML teams running distributed training across multiple machines","researchers iterating rapidly on model architectures and need audit trails","organizations managing thousands of concurrent experiments"],"limitations":["Metadata ingestion latency increases with experiment scale (>10k concurrent runs may see 500ms+ delays)","Artifact storage requires external cloud provider integration (S3, GCS, Azure) — Neptune stores references, not blobs","Real-time metric streaming has eventual consistency model (~5-10 second propagation delay)"],"requires":["Python 3.7+ or compatible ML framework (PyTorch, TensorFlow, scikit-learn, XGBoost)","Neptune API key and project credentials","Network connectivity to Neptune cloud or self-hosted instance","Artifact storage credentials if logging files (S3, GCS, etc.)"],"input_types":["structured metadata (dicts, JSON)","numeric scalars and arrays (metrics, losses)","file artifacts (models, plots, CSVs)","environment variables and system info"],"output_types":["structured experiment records with versioning","time-series metric data","artifact references with download URLs","experiment comparison matrices"],"categories":["data-processing-analysis","mlops-observability"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_1","uri":"capability://data.processing.analysis.multi.dimensional.experiment.comparison.with.custom.dashboards","name":"multi-dimensional experiment comparison with custom dashboards","description":"Provides a query engine that filters and compares experiments across arbitrary dimensions (hyperparameters, metrics, tags, date ranges) and renders interactive dashboards with scatter plots, parallel coordinates, and heatmaps. Uses columnar indexing on metadata to enable sub-second filtering across millions of metric points and supports custom dashboard templates with drag-and-drop widget composition.","intents":["I want to visualize how learning rate and batch size interact to affect final accuracy across 500 experiments","I need to create a dashboard showing model performance trends over the last month grouped by data version","I want to identify outlier experiments that achieved unexpectedly high performance and drill down into their configs"],"best_for":["ML practitioners performing hyperparameter optimization and sensitivity analysis","teams conducting model selection and comparison across architectures","stakeholders reviewing experiment results without direct code access"],"limitations":["Custom dashboard persistence is limited to 100 saved dashboards per project in free tier","Real-time dashboard updates require polling (no WebSocket push for metric changes)","Complex multi-level grouping (>3 dimensions) may require 2-5 seconds to render on large datasets"],"requires":["Experiments logged to Neptune with consistent metadata schema","Web browser with JavaScript enabled (Chrome, Firefox, Safari, Edge)","Read access to project containing experiments"],"input_types":["experiment metadata queries (filter expressions)","metric names and aggregation functions","custom dimension definitions"],"output_types":["interactive visualizations (scatter, parallel coordinates, heatmaps)","CSV exports of filtered experiment sets","shareable dashboard URLs with embedded filters"],"categories":["data-processing-analysis","search-retrieval"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_10","uri":"capability://automation.workflow.team.workspace.management.with.role.based.access.control","name":"team-workspace-management-with-role-based-access-control","description":"Organizes experiments into team workspaces with role-based access control (RBAC) supporting Owner, Editor, and Viewer roles. Enables fine-grained permissions (e.g., 'can promote models to production' vs. 'can only view experiments'). Supports SSO integration (SAML, OAuth) for enterprise deployments and audit logging of all access and modifications.","intents":["I need to organize experiments by team and control who can modify or deploy models","I want to grant read-only access to stakeholders without giving them experiment modification rights","I need to integrate Neptune with our company's SSO system for centralized user management","I need an audit log of who accessed or modified experiments for compliance"],"best_for":["enterprise teams with formal access control requirements","organizations with compliance or regulatory needs","large teams needing workspace isolation"],"limitations":["RBAC limited to 3 predefined roles; no custom role definitions","SSO integration only available in paid tiers","Audit logs retained for 90 days; longer retention requires paid plan","No project-level permissions — access control is workspace-wide"],"requires":["Neptune workspace with multiple team members","User accounts with email addresses","SSO provider (optional, for enterprise deployments)"],"input_types":["user identity and email","role assignment (Owner, Editor, Viewer)","workspace configuration"],"output_types":["workspace access control lists","audit logs of access and modifications","user role assignments","SSO configuration"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_11","uri":"capability://data.processing.analysis.custom.dashboard.builder.with.widget.composition","name":"custom-dashboard-builder-with-widget-composition","description":"Allows users to create custom dashboards by composing widgets (charts, tables, metrics cards) that pull data from experiments. Widgets support dynamic filtering and drill-down to experiment details. Dashboards are shareable via links and can be embedded in external tools via iframes. Supports scheduled dashboard refreshes and email delivery of dashboard snapshots.","intents":["I want to create a custom dashboard showing key metrics from my active experiments","I need to share a dashboard with non-technical stakeholders showing model performance trends","I want to embed a Neptune dashboard in my internal wiki or reporting system","I need to receive daily email summaries of experiment metrics without logging into Neptune"],"best_for":["teams needing custom reporting and visualization","organizations with non-technical stakeholders needing experiment insights","teams integrating Neptune metrics into broader reporting systems"],"limitations":["Widget customization limited to predefined chart types; no custom visualization code","Dashboard refresh frequency limited to hourly minimum; real-time updates not supported","Email delivery limited to 1 email per day per dashboard","Embedded dashboards (iframes) require authentication; no public embedding option"],"requires":["Neptune workspace with experiments logged","Web browser for dashboard builder UI","Email address for scheduled delivery (optional)"],"input_types":["widget type and configuration","data source (experiment metrics, tags)","filtering and grouping criteria"],"output_types":["custom dashboards (HTML)","shareable dashboard URLs","dashboard snapshots (email, PDF)","embedded dashboard iframes"],"categories":["data-processing-analysis","text-generation-language"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_2","uri":"capability://memory.knowledge.model.registry.with.versioning.and.metadata.lineage","name":"model registry with versioning and metadata lineage","description":"Centralized model storage with semantic versioning, stage transitions (staging/production/archived), and full lineage tracking linking models to source experiments, training data versions, and deployment metadata. Implements a state machine for model lifecycle management with audit logging of all stage transitions and supports model comparison by metrics, parameters, and artifact checksums.","intents":["I need to promote a model from staging to production and have an audit trail of who approved it and when","I want to compare two model versions side-by-side to see which experiment produced better metrics","I need to track which data version and training config produced each model in production for debugging"],"best_for":["teams managing multiple models in production with compliance/audit requirements","ML platforms needing model governance and approval workflows","organizations tracking model lineage for reproducibility and debugging"],"limitations":["Model artifacts themselves are stored externally (S3, GCS, etc.) — Neptune stores metadata and references only","Stage transition workflows are linear (staging → production → archived) — no custom workflow states","Lineage tracking requires experiments to be logged to Neptune — external models require manual metadata entry"],"requires":["Models logged from Neptune-tracked experiments or manual registration via API","External artifact storage (S3, GCS, Azure Blob) for model files","Project-level permissions configured for stage transition approvals"],"input_types":["model artifacts (pickle, SavedModel, ONNX, etc.)","model metadata (metrics, parameters, framework version)","experiment references linking to source training run"],"output_types":["versioned model records with stage and timestamp","lineage graphs showing experiment → model → deployment","model comparison reports with metric deltas","audit logs of all stage transitions"],"categories":["memory-knowledge","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_3","uri":"capability://tool.use.integration.collaborative.experiment.sharing.with.role.based.access.control","name":"collaborative experiment sharing with role-based access control","description":"Enables team members to view, comment on, and compare experiments with granular permission controls (viewer, editor, admin) at project and experiment level. Implements real-time collaboration features including experiment comments with threading, @mentions, and activity feeds showing who modified what and when, with audit logging of all access and modifications.","intents":["I want to share experiment results with my team lead for review without giving them write access to the project","I need to leave comments on specific experiments to discuss findings with colleagues","I want to see an audit trail of who accessed and modified each experiment for compliance"],"best_for":["distributed ML teams collaborating across time zones","organizations with compliance requirements for experiment audit trails","research groups needing lightweight experiment discussion without external tools"],"limitations":["Comments are stored in Neptune only — no integration with Slack/Teams for notifications","Role-based access control is project-level or experiment-level only — no fine-grained field-level permissions","Activity feed shows last 1,000 events per project — older events require manual audit log export"],"requires":["Team members with Neptune accounts in same workspace","Project-level permissions configured by project admin","Email notifications enabled (requires email verification)"],"input_types":["experiment IDs or URLs","comment text with optional @mentions","role assignments (viewer/editor/admin)"],"output_types":["shared experiment views with filtered data based on permissions","comment threads with timestamps and author info","activity feed with modification history","audit logs (CSV export)"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_4","uri":"capability://safety.moderation.production.monitoring.with.metric.alerts.and.anomaly.detection","name":"production monitoring with metric alerts and anomaly detection","description":"Monitors deployed models in production by ingesting live prediction metrics and comparing against baseline experiment metrics to detect performance degradation. Uses statistical anomaly detection (z-score, IQR, moving average) to identify metric drift and triggers configurable alerts via email, webhooks, or Slack when thresholds are breached, with root cause analysis linking degradation to data drift or model staleness.","intents":["I want to be alerted immediately if my model's accuracy drops below 95% in production","I need to detect when my model's performance degrades due to data drift and automatically trigger retraining","I want to compare production metrics against baseline experiment metrics to identify when the model is underperforming"],"best_for":["teams running models in production with SLA requirements","organizations needing automated drift detection and retraining triggers","ML platforms integrating model monitoring into MLOps pipelines"],"limitations":["Anomaly detection uses statistical methods only — no ML-based anomaly detection (e.g., isolation forests)","Alert routing requires manual webhook configuration — no native Slack/PagerDuty integration in free tier","Metric ingestion has 5-10 second latency, so real-time alerts may miss transient spikes","Root cause analysis is heuristic-based (data drift detection requires manual feature logging)"],"requires":["Production system logging metrics to Neptune via SDK or HTTP API","Baseline experiment metrics in Neptune for comparison","Webhook endpoint or email configured for alert delivery","Optional: Slack workspace token for direct Slack notifications"],"input_types":["production metrics (accuracy, precision, latency, etc.)","baseline metrics from reference experiments","alert threshold definitions (numeric or percentage)","feature data for drift detection (optional)"],"output_types":["alert notifications (email, webhook, Slack)","alert history with triggered conditions and timestamps","drift analysis reports comparing production vs. baseline","recommended actions (retrain, rollback, investigate)"],"categories":["safety-moderation","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_5","uri":"capability://tool.use.integration.sdk.based.experiment.logging.with.framework.integrations","name":"sdk-based experiment logging with framework integrations","description":"Provides language-specific SDKs (Python, JavaScript/TypeScript) that integrate with popular ML frameworks (PyTorch, TensorFlow, scikit-learn, XGBoost, Keras) via callbacks and decorators to automatically log metrics, hyperparameters, and artifacts without modifying training code. Implements lazy evaluation and batching to minimize logging overhead and supports both synchronous and asynchronous logging modes.","intents":["I want to log my PyTorch training metrics to Neptune without adding boilerplate code to my training loop","I need to automatically capture hyperparameters and model architecture from my TensorFlow model","I want to log artifacts (model checkpoints, plots) asynchronously so they don't slow down training"],"best_for":["ML engineers using standard frameworks (PyTorch, TensorFlow, scikit-learn) who want minimal code changes","teams running training on resource-constrained hardware where logging overhead matters","researchers prototyping models quickly without setting up custom logging infrastructure"],"limitations":["Framework integrations are limited to popular frameworks — custom training loops require manual SDK calls","Asynchronous logging may lose data if process crashes before flush — requires explicit flush() calls for safety","Callback-based logging has ~50-100ms overhead per batch in synchronous mode","JavaScript SDK has limited framework support compared to Python (no TensorFlow.js integration)"],"requires":["Python 3.7+ or Node.js 14+","Compatible ML framework (PyTorch 1.0+, TensorFlow 2.0+, scikit-learn 0.20+, XGBoost 1.0+)","Neptune API key and project credentials","Network connectivity to Neptune cloud"],"input_types":["training metrics (scalars, arrays, time-series)","hyperparameters (dicts, nested configs)","model artifacts (files, binary data)","system info (environment variables, dependencies)"],"output_types":["logged experiment records in Neptune","artifact references with download URLs","structured metadata for comparison and analysis"],"categories":["tool-use-integration","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_6","uri":"capability://automation.workflow.batch.experiment.execution.with.hyperparameter.sweep.orchestration","name":"batch experiment execution with hyperparameter sweep orchestration","description":"Orchestrates distributed hyperparameter sweeps by defining search spaces (grid, random, Bayesian) and automatically spawning training jobs across multiple machines with centralized result aggregation. Implements early stopping based on intermediate metrics and supports conditional parameter dependencies, enabling efficient exploration of high-dimensional hyperparameter spaces without manual job management.","intents":["I want to run a grid search over 100 hyperparameter combinations and have Neptune automatically manage job submission and result collection","I need to stop underperforming trials early to save compute resources while exploring the hyperparameter space","I want to define conditional hyperparameters (e.g., learning rate schedule only if optimizer is SGD) without manual filtering"],"best_for":["teams with access to compute clusters or cloud resources (AWS, GCP, Azure)","researchers optimizing models with expensive training (hours to days per trial)","organizations running regular hyperparameter optimization as part of MLOps pipelines"],"limitations":["Sweep orchestration requires integration with compute infrastructure (Kubernetes, Ray, or custom job submission) — Neptune doesn't provide compute itself","Early stopping is based on intermediate metrics only — requires metrics logged during training, not just final results","Bayesian optimization uses simple Gaussian process (no advanced acquisition functions like Expected Improvement)","Conditional parameters require manual definition — no automatic dependency inference"],"requires":["Compute infrastructure (Kubernetes, Ray, cloud VMs, or local machines)","Training script that logs metrics to Neptune","Sweep configuration (YAML or Python) defining search space","Integration code to submit jobs to compute infrastructure"],"input_types":["hyperparameter search space definitions (grid, random, Bayesian)","training script or Docker image","early stopping criteria (metric name, threshold, patience)","resource constraints (max parallel jobs, timeout)"],"output_types":["sweep execution logs with job status and metrics","aggregated results ranked by performance","early stopping decisions with justification","best hyperparameters with confidence intervals"],"categories":["automation-workflow","planning-reasoning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_7","uri":"capability://data.processing.analysis.data.versioning.and.artifact.lineage.tracking","name":"data versioning and artifact lineage tracking","description":"Tracks data versions and artifact lineage by capturing dataset metadata (schema, row count, checksums), linking experiments to specific data versions, and enabling reproducibility by pinning training data versions. Implements content-addressable storage with checksums to detect silent data changes and supports querying experiments by data version to identify which models were trained on which datasets.","intents":["I want to know which data version was used to train each model so I can reproduce results or debug data issues","I need to detect when my training data changed (even silently) and identify which experiments were affected","I want to compare model performance across different data versions to understand data quality impact"],"best_for":["teams managing multiple data versions and needing reproducibility","organizations with strict data governance and audit requirements","ML platforms tracking end-to-end lineage from data to models to predictions"],"limitations":["Data versioning requires manual metadata logging — no automatic data version detection from file paths","Checksums are computed client-side — no server-side validation of data integrity","Lineage queries are limited to experiments in Neptune — external data sources require manual registration","Data schema tracking is optional — no automatic schema inference or validation"],"requires":["Manual data version metadata logging via Neptune SDK","Dataset files accessible from training environment","Optional: data schema definitions (JSON Schema or custom format)"],"input_types":["dataset metadata (name, version, row count, schema)","file checksums (MD5, SHA256)","data source information (path, database, API endpoint)"],"output_types":["data version records linked to experiments","lineage graphs showing data → experiment → model","data change detection alerts","experiment filtering by data version"],"categories":["data-processing-analysis","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_8","uri":"capability://tool.use.integration.api.first.architecture.with.rest.and.python.sdk","name":"api-first architecture with rest and python sdk","description":"Exposes all Neptune functionality via REST API and Python SDK, enabling programmatic access to experiments, models, and metrics for custom integrations and automation. Implements pagination, filtering, and sorting on all list endpoints with support for complex queries, and provides webhook support for triggering external actions on experiment events (completion, metric threshold crossed, etc.).","intents":["I want to query all experiments matching certain criteria (e.g., accuracy > 95%) and export results to a CSV","I need to integrate Neptune with my custom MLOps pipeline to automatically trigger model deployment when experiments complete","I want to build a custom dashboard that pulls data from Neptune and displays it in my internal tools"],"best_for":["teams building custom MLOps platforms or integrations","organizations with existing tools that need Neptune data integration","developers automating experiment management and model deployment workflows"],"limitations":["API rate limits are 100 requests/minute for free tier — batch operations may require throttling","Webhook delivery is at-least-once (no deduplication) — consumers must handle duplicate events","Complex queries require multiple API calls — no GraphQL for efficient data fetching","Pagination is cursor-based with max 1,000 results per page — large result sets require multiple requests"],"requires":["Neptune API key with appropriate permissions","HTTP client library (requests, httpx, etc.) or Python SDK","Understanding of REST API conventions and pagination"],"input_types":["query filters (experiment name, metric range, date range, tags)","sort criteria (metric value, creation date, etc.)","webhook event types (experiment.completed, metric.threshold_crossed, etc.)"],"output_types":["JSON experiment records with full metadata","paginated result sets with cursor tokens","webhook event payloads (JSON)","CSV/JSON exports of query results"],"categories":["tool-use-integration","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__cap_9","uri":"capability://data.processing.analysis.artifact.storage.and.versioning.with.deduplication","name":"artifact-storage-and-versioning-with-deduplication","description":"Stores experiment artifacts (model checkpoints, plots, CSVs, logs) in Neptune's cloud storage with content-based deduplication to reduce storage costs. Each artifact is versioned and linked to its source experiment; supports retrieval by experiment ID or artifact name. Integrates with training frameworks to automatically capture checkpoints and logs without explicit code changes.","intents":["I want to store model checkpoints and training logs without managing cloud storage buckets","I need to retrieve a specific artifact version from an experiment without downloading the entire experiment","I want to reduce storage costs by deduplicating identical artifacts across experiments","I need to track which experiments produced which artifacts for reproducibility"],"best_for":["teams without dedicated cloud storage infrastructure","researchers needing artifact versioning and lineage tracking","organizations with storage cost constraints"],"limitations":["Free tier limited to 100 GB total artifact storage; paid tiers required for production-scale use","No direct access to underlying storage (e.g., S3 bucket); artifacts only accessible via Neptune API","Artifact upload latency increases with file size (large model checkpoints may take minutes)","No built-in artifact compression; large artifacts consume full storage quota"],"requires":["Neptune API key with artifact storage permissions","Network connectivity for artifact upload/download","Artifacts in supported formats (any binary or text format)"],"input_types":["binary files (model checkpoints, serialized objects)","text files (logs, CSVs, JSON)","images (plots, visualizations)"],"output_types":["versioned artifact entries","artifact URLs for download","artifact metadata (size, hash, upload timestamp)","artifact lineage (source experiment)"],"categories":["data-processing-analysis","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"neptune-ai__headline","uri":"capability://data.processing.analysis.mlops.metadata.management.platform","name":"mlops metadata management platform","description":"Neptune AI is a comprehensive metadata store designed for MLOps teams, enabling them to track experiments, models, and workflows at scale with powerful comparison dashboards and collaboration tools.","intents":["best MLOps platform","MLOps tool for tracking experiments","comparison dashboards for ML models","model registry solutions","collaboration tools for machine learning teams"],"best_for":["MLOps teams","data scientists","machine learning engineers"],"limitations":[],"requires":[],"input_types":[],"output_types":[],"categories":["data-processing-analysis"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":57,"verified":false,"data_access_risk":"high","permissions":["Python 3.7+ or compatible ML framework (PyTorch, TensorFlow, scikit-learn, XGBoost)","Neptune API key and project credentials","Network connectivity to Neptune cloud or self-hosted instance","Artifact storage credentials if logging files (S3, GCS, etc.)","Experiments logged to Neptune with consistent metadata schema","Web browser with JavaScript enabled (Chrome, Firefox, Safari, Edge)","Read access to project containing experiments","Neptune workspace with multiple team members","User accounts with email addresses","SSO provider (optional, for enterprise deployments)"],"failure_modes":["Metadata ingestion latency increases with experiment scale (>10k concurrent runs may see 500ms+ delays)","Artifact storage requires external cloud provider integration (S3, GCS, Azure) — Neptune stores references, not blobs","Real-time metric streaming has eventual consistency model (~5-10 second propagation delay)","Custom dashboard persistence is limited to 100 saved dashboards per project in free tier","Real-time dashboard updates require polling (no WebSocket push for metric changes)","Complex multi-level grouping (>3 dimensions) may require 2-5 seconds to render on large datasets","RBAC limited to 3 predefined roles; no custom role definitions","SSO integration only available in paid tiers","Audit logs retained for 90 days; longer retention requires paid plan","No project-level permissions — access control is workspace-wide","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7,"quality":0.9,"ecosystem":0.25,"match_graph":0.25,"freshness":0.75,"weights":{"adoption":0.3,"quality":0.25,"ecosystem":0.15,"match_graph":0.25,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"active","updated_at":"2026-05-24T12:16:23.328Z","last_scraped_at":null,"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=neptune-ai","compare_url":"https://unfragile.ai/compare?artifact=neptune-ai"}},"signature":"zA+GVWQ6Cc2koAaJGd5XJ1GT7fauU9/5cd9puADUmFADeyCkRbuQYqrfsCR8tLknB1WtXhJpdZxGdFcodInnDA==","signedAt":"2026-06-22T19:05:30.231Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/neptune-ai","artifact":"https://unfragile.ai/neptune-ai","verify":"https://unfragile.ai/api/v1/verify?slug=neptune-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"}}