{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"runpod","slug":"runpod","name":"RunPod","type":"platform","url":"https://runpod.io","page_url":"https://unfragile.ai/runpod","categories":["deployment-infra"],"tags":[],"pricing":{"model":"usage","free":false,"starting_price":null},"status":"active","verified":false},"capabilities":[{"id":"runpod__cap_0","uri":"capability://automation.workflow.on.demand.gpu.pod.provisioning.with.per.second.billing","name":"on-demand gpu pod provisioning with per-second billing","description":"Provisions isolated GPU compute environments (single or multi-GPU) on Community Cloud or Secure Cloud with per-second or per-hour billing models. Uses a containerized pod architecture where users SSH into fully-loaded environments with pre-installed CUDA, drivers, and framework support. Spins up in under 60 seconds by leveraging pre-warmed container images and rapid network attachment of persistent storage volumes.","intents":["I need to train a model for a few hours without committing to long-term infrastructure","I want to test GPU code locally before scaling to production","I need interactive access to a GPU environment for debugging and iteration"],"best_for":["researchers and solo developers running ad-hoc training jobs","teams prototyping models before committing to reserved capacity","users with bursty, unpredictable compute needs"],"limitations":["Per-second billing means idle time costs accumulate; no automatic shutdown on inactivity","No built-in autoscaling for Pods — manual provisioning required for multi-pod workflows","Cold-start latency of ~60 seconds may be prohibitive for real-time inference","Pricing not fully transparent in documentation (shows $0/s placeholders)"],"requires":["SSH client or web terminal access","API key or account authentication","Sufficient account balance or payment method on file","Network connectivity to RunPod regions (8+ worldwide)"],"input_types":["code (Python, CUDA, shell scripts)","model checkpoints (PyTorch, TensorFlow, ONNX)","datasets (via S3-compatible storage or direct upload)"],"output_types":["trained model weights","logs and metrics (real-time streaming)","inference results"],"categories":["automation-workflow","infrastructure-provisioning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_1","uri":"capability://automation.workflow.serverless.gpu.endpoint.auto.scaling.with.flex.and.active.worker.modes","name":"serverless gpu endpoint auto-scaling with flex and active worker modes","description":"Deploys inference APIs that auto-scale from 0 to 1000s of workers in seconds using two distinct billing models: Flex workers scale down to zero after job completion (pay-per-execution), while Active workers maintain always-on state with ~30% cost discount. Uses FlashBoot technology to achieve sub-200ms cold-start latency on Flex workers by pre-loading container images and model weights into memory. Handles request routing, load balancing, and worker lifecycle management transparently.","intents":["I want to deploy an inference API that scales automatically with traffic spikes without paying for idle capacity","I need consistent low-latency inference for always-on services like chatbots or recommendation engines","I want to compare cost-benefit of always-on vs. on-demand inference for my workload"],"best_for":["teams building event-driven inference pipelines (batch processing, webhooks)","startups deploying inference APIs with unpredictable traffic patterns","production services requiring sub-200ms latency with cost optimization"],"limitations":["Flex workers have sub-200ms cold-start but still incur per-second compute charges during execution; not suitable for ultra-latency-sensitive applications (<50ms)","Active workers require continuous billing even during idle periods; cost-effective only if average utilization >30%","No built-in request queuing or priority scheduling documented; high-traffic spikes may exceed scaling capacity","FlashBoot sub-200ms claim is unverified; actual latency depends on model size and worker initialization overhead"],"requires":["Containerized inference code (Docker image with HTTP server)","Model weights accessible via S3-compatible storage or bundled in container","API endpoint definition (input/output schema)","RunPod Serverless SDK or direct HTTP client"],"input_types":["HTTP POST requests (JSON, binary, multipart)","model inference inputs (tensors, text, images)"],"output_types":["HTTP JSON responses","inference results (predictions, embeddings, generated text)"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_10","uri":"capability://automation.workflow.automatic.failover.and.pod.recovery.with.transparent.restart","name":"automatic failover and pod recovery with transparent restart","description":"Automatically detects pod failures (hardware issues, OOM, crashes) and restarts pods transparently, with claimed failover handling by RunPod infrastructure. Mechanism for failure detection and restart policy not documented. Persistent storage volumes remain attached across restarts, preserving checkpoint data and training progress.","intents":["I want my training job to automatically recover from transient hardware failures","I need long-running jobs to survive pod crashes without manual intervention","I want to resume training from the last checkpoint after a pod restart"],"best_for":["teams running long-duration training jobs (days/weeks) with high failure risk","production inference endpoints requiring high availability","users without manual monitoring and restart capabilities"],"limitations":["Failover mechanism not documented; unclear if restarts are automatic or require manual intervention","Restart latency not specified; unclear if recovery is sub-second or takes minutes","No restart count limit documented; risk of infinite restart loops on persistent failures","Checkpoint-based recovery requires application-level implementation; no automatic state preservation"],"requires":["Persistent storage for checkpoints (S3-compatible or network volume)","Application code that saves and loads checkpoints","Pod configuration with restart policy (if configurable)"],"input_types":["checkpoint data (model weights, optimizer state, training step)","failure detection signals (pod crash, OOM, timeout)"],"output_types":["restarted pod with preserved storage","resumed training from last checkpoint"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_11","uri":"capability://data.processing.analysis.cost.estimation.and.transparent.per.second.billing.with.no.hidden.fees","name":"cost estimation and transparent per-second billing with no hidden fees","description":"Provides per-second billing granularity for on-demand pods and serverless endpoints, enabling precise cost tracking and elimination of hourly minimum charges. Pricing calculator available on website (though actual rates show $0/s placeholders in documentation). No setup fees, data transfer fees (within RunPod), or hidden charges documented; egress fees apply only to data leaving RunPod infrastructure.","intents":["I want to estimate costs for short-lived training jobs without paying for unused hours","I need transparent pricing to compare RunPod vs. competitors for my workload","I want to track and optimize GPU spending across multiple pods"],"best_for":["cost-conscious teams running variable-duration workloads","startups with limited budgets requiring precise cost forecasting","users comparing cloud providers based on transparent pricing"],"limitations":["Actual per-second rates not visible in documentation (shows $0/s placeholders); pricing requires visiting website or contacting sales","Hybrid per-second + per-hour billing for Instant Clusters makes cost calculation complex","Reserved Cluster pricing requires sales negotiation; no transparent pricing for volume discounts","Egress fees to external cloud providers not documented; unclear if zero-egress applies only to RunPod storage"],"requires":["GPU type and region selection","Estimated job duration","Optional: pricing calculator access"],"input_types":["pod configuration (GPU type, vCPU, memory, region)","job duration estimate (hours, days)"],"output_types":["cost estimate (USD)","hourly/daily cost breakdown","comparison with competitor pricing"],"categories":["data-processing-analysis","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_12","uri":"capability://memory.knowledge.community.and.ecosystem.with.750.000.developers","name":"community and ecosystem with 750,000+ developers","description":"RunPod claims 750,000+ developers using the platform with 4.8-star rating (source unverified). Community features not documented; unclear if platform includes forums, Discord, GitHub discussions, or other collaboration mechanisms. Partnerships with OpenAI (Model Craft Challenge Series) and unnamed 'world's leading AI companies' suggest ecosystem maturity, but specific integrations and community contributions not detailed.","intents":["I want to learn from community examples and best practices for GPU workloads","I need support from other developers using RunPod for similar projects","I want to contribute templates or tools to the RunPod ecosystem"],"best_for":["developers seeking community support and shared knowledge","teams evaluating platform maturity based on user base","researchers interested in collaborating with other AI practitioners"],"limitations":["Community features not documented; unclear if forums, Discord, or GitHub discussions exist","No community contribution workflow documented; unclear how to share templates or tools","Community size (750,000+) unverified; no public metrics on active users or engagement","No public artifact registry or model marketplace documented; community contributions may not be discoverable"],"requires":["RunPod account","Community platform access (Discord, forums, etc. — if available)","Optional: GitHub account for sharing code"],"input_types":["questions and discussions (text)","code examples and templates (GitHub)","feedback and feature requests"],"output_types":["community answers and advice","shared templates and tools","networking opportunities"],"categories":["memory-knowledge","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_2","uri":"capability://automation.workflow.multi.gpu.instant.cluster.provisioning.with.per.second.billing","name":"multi-gpu instant cluster provisioning with per-second billing","description":"Provisions temporary GPU clusters of 2-64 GPUs with per-second + per-hour hybrid billing, enabling distributed training and inference without long-term commitment. Uses cluster orchestration to attach multiple GPUs to a single network namespace with optimized inter-GPU communication (NVLink, PCIe). Supports frameworks like PyTorch Distributed Data Parallel, Horovod, and DeepSpeed out-of-the-box via pre-configured environments.","intents":["I need to train a large model across 8+ GPUs for a few days without buying hardware","I want to run distributed inference across multiple GPUs for batch processing","I need to benchmark distributed training performance before committing to reserved capacity"],"best_for":["ML teams training large models (70B+ parameters) with time-bounded budgets","researchers running distributed experiments with varying GPU counts","teams migrating from on-premise clusters to cloud without infrastructure expertise"],"limitations":["Hybrid per-second + per-hour billing model is less transparent than pure per-second; actual cost requires calculation","No automatic cluster scaling based on training metrics; manual resize required mid-job","Inter-GPU communication latency depends on physical proximity in data center; cross-region clusters not supported","Requires distributed training code; no managed orchestration for hyperparameter sweeps or multi-job workflows"],"requires":["Distributed training framework (PyTorch, TensorFlow, Horovod, DeepSpeed)","Model code compatible with distributed training (DDP, FSDP, or custom)","Cluster size specification (2-64 GPUs)","Persistent storage for checkpoints (S3-compatible or network volume)"],"input_types":["training code (Python with distributed framework)","model architecture and hyperparameters","training datasets (local or S3-compatible storage)"],"output_types":["trained model checkpoints","training logs and metrics","distributed inference results"],"categories":["automation-workflow","infrastructure-provisioning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_3","uri":"capability://automation.workflow.reserved.gpu.cluster.deployment.with.sla.backed.uptime.and.volume.discounts","name":"reserved gpu cluster deployment with sla-backed uptime and volume discounts","description":"Provisions dedicated GPU infrastructure with commitment terms (1-month to 12-month+) and SLA-backed uptime guarantees, enabling predictable costs and priority resource allocation. Uses dedicated hardware isolation to prevent noisy-neighbor effects and provides volume discounts for 10,000+ GPU scale. Requires sales contact for pricing; targets enterprise customers with sustained, high-volume compute needs.","intents":["I need guaranteed uptime and performance for production inference serving","I want to negotiate volume pricing for 10,000+ GPUs across multiple projects","I need dedicated infrastructure to meet compliance and data residency requirements"],"best_for":["enterprises running production AI services with SLA requirements","large-scale training operations with sustained, predictable compute needs","organizations requiring dedicated infrastructure for compliance (HIPAA, SOC2)"],"limitations":["Requires multi-month commitment; inflexible for workloads with uncertain duration","Pricing requires sales negotiation; no transparent pricing available for cost estimation","SLA details not documented (uptime percentage, incident response time, compensation terms unknown)","No auto-scaling within reserved capacity; manual resource adjustment requires contract modification"],"requires":["Minimum commitment term (1 month)","Sales contact and contract negotiation","Dedicated account management","Volume commitment (10,000+ GPUs for volume discounts)"],"input_types":["infrastructure requirements (GPU count, types, regions)","workload specifications (training, inference, batch processing)"],"output_types":["dedicated cluster configuration","SLA agreement","volume pricing quote"],"categories":["automation-workflow","infrastructure-provisioning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_4","uri":"capability://data.processing.analysis.s3.compatible.persistent.network.storage.with.zero.egress.fees","name":"s3-compatible persistent network storage with zero egress fees","description":"Provides S3-compatible object storage accessible from all GPU pods and serverless endpoints with no egress charges for data leaving RunPod storage to external destinations. Uses network-attached storage architecture to enable rapid model weight loading and dataset access without downloading to local pod storage. Integrates with standard S3 clients (boto3, AWS CLI, s3fs) via compatible API endpoints.","intents":["I want to store large model checkpoints and datasets without paying egress fees when downloading to local machines","I need fast access to training data from multiple GPU pods without duplicating storage","I want to use standard S3 tools and libraries without learning RunPod-specific APIs"],"best_for":["teams with large model artifacts (10GB-1TB+) requiring frequent access across pods","training workflows that iterate on datasets stored in cloud storage","users migrating from AWS S3 who want cost savings on egress"],"limitations":["Zero egress fees apply only to data leaving RunPod storage; egress to external cloud providers (AWS, GCP) may incur charges (not documented)","No built-in versioning, lifecycle policies, or access control documented; requires external tools for compliance","Storage pricing not transparent in documentation; only egress fee structure mentioned","No data encryption at rest or in transit documented; security posture unclear"],"requires":["S3-compatible client (boto3, AWS CLI, s3fs, rclone)","RunPod storage credentials (access key, secret key)","Network connectivity from GPU pods to storage endpoint"],"input_types":["model checkpoints (PyTorch .pt, TensorFlow .pb, ONNX)","datasets (parquet, CSV, images, raw binary)","code artifacts (Python wheels, Docker layers)"],"output_types":["object metadata (size, ETag, last-modified)","streamed data for training/inference"],"categories":["data-processing-analysis","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_5","uri":"capability://automation.workflow.real.time.pod.monitoring.and.logging.with.streaming.metrics","name":"real-time pod monitoring and logging with streaming metrics","description":"Provides real-time monitoring dashboards and log streaming for GPU pods, capturing metrics like GPU utilization, memory usage, temperature, and network throughput. Logs are streamed to the web console and accessible via API; no explicit log retention policy or query language documented. Enables developers to diagnose performance bottlenecks and resource contention without SSH-ing into pods.","intents":["I want to monitor GPU utilization and memory usage during training to optimize batch sizes","I need to debug why my pod is running slowly or crashing unexpectedly","I want to track resource consumption across multiple pods to estimate costs"],"best_for":["developers iterating on model training and debugging performance issues","teams monitoring production inference endpoints for anomalies","users optimizing resource allocation and cost efficiency"],"limitations":["Log retention policy not documented; unclear if logs persist after pod termination","No structured logging or query language documented; limited to real-time streaming view","No log export to external systems (CloudWatch, Datadog, ELK) documented","Metrics are pod-level only; no distributed tracing or cross-pod correlation"],"requires":["Active GPU pod or serverless endpoint","RunPod web console or API access","Network connectivity to monitoring endpoint"],"input_types":["pod runtime metrics (GPU, CPU, memory, network)","application logs (stdout, stderr)"],"output_types":["real-time metric dashboards","log streams (text)","metric time-series data (JSON via API)"],"categories":["automation-workflow","safety-moderation"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_6","uri":"capability://automation.workflow.template.marketplace.for.pre.configured.gpu.environments","name":"template marketplace for pre-configured gpu environments","description":"Provides a marketplace of pre-built container templates with frameworks, libraries, and model weights pre-installed, enabling one-click deployment of common AI workloads (LLM inference, image generation, training). Templates abstract away container configuration and dependency management; users select a template and customize hyperparameters. Specific template types, discovery mechanisms, and community contribution workflows not documented.","intents":["I want to deploy a popular open-source model (Llama, Stable Diffusion) without building a Docker image","I need a pre-configured environment for fine-tuning with all dependencies already installed","I want to explore different model architectures without managing container builds"],"best_for":["non-technical users and researchers unfamiliar with Docker and containerization","teams rapidly prototyping with popular models","users exploring model capabilities before committing to custom implementations"],"limitations":["Template discovery and search mechanisms not documented; unclear how to find relevant templates","Community contribution workflow unknown; unclear if users can publish custom templates","Template versioning and update policies not documented; risk of stale dependencies","Customization depth unknown; unclear if templates support arbitrary code modifications or only parameter tuning"],"requires":["RunPod account with sufficient credits","Selection of compatible template","Optional: custom model weights or datasets"],"input_types":["template selection (marketplace UI)","hyperparameter overrides (JSON or web form)","custom datasets or model weights (optional)"],"output_types":["deployed GPU pod with pre-configured environment","inference API endpoint (if template includes server)"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_7","uri":"capability://automation.workflow.global.multi.region.pod.deployment.with.low.latency.performance","name":"global multi-region pod deployment with low-latency performance","description":"Enables deployment of GPU pods across 8+ worldwide regions with claimed low-latency performance and global reliability. Specific regions not documented; deployment mechanism (manual region selection vs. automatic geo-routing) unclear. Supports persistent storage access across regions via S3-compatible API, enabling data locality optimization for distributed workloads.","intents":["I want to deploy inference endpoints in regions close to my users to minimize latency","I need to run training jobs in specific geographic regions for data residency compliance","I want to replicate models across regions for redundancy and failover"],"best_for":["global applications requiring low-latency inference in multiple regions","organizations with data residency requirements (GDPR, data sovereignty)","teams building disaster recovery and failover strategies"],"limitations":["Specific regions not documented; unclear which geographic areas are covered","No automatic geo-routing or latency-based region selection documented; manual region selection required","Cross-region data transfer costs not documented; unclear if egress between regions incurs charges","No multi-region failover or load balancing documented; users must implement their own routing logic"],"requires":["Region selection (manual via API or console)","Network connectivity to selected region","Optional: S3-compatible storage for cross-region data access"],"input_types":["region specification (API parameter or console dropdown)","pod configuration (GPU type, vCPU, memory)"],"output_types":["pod endpoint (IP address or hostname)","region-specific performance metrics"],"categories":["automation-workflow","infrastructure-provisioning"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_8","uri":"capability://automation.workflow.ssh.and.web.terminal.access.to.gpu.pods.for.interactive.development","name":"ssh and web terminal access to gpu pods for interactive development","description":"Provides SSH and browser-based web terminal access to GPU pods, enabling interactive development, debugging, and experimentation without containerization expertise. Users can install packages, run ad-hoc commands, and modify code in real-time. Supports standard Linux tools (git, pip, conda, nvcc) pre-installed in pod environments.","intents":["I want to interactively debug training code and inspect intermediate outputs","I need to install custom dependencies or patch libraries during development","I want to run Jupyter notebooks on a GPU pod for exploratory analysis"],"best_for":["researchers and data scientists preferring interactive development over containerized workflows","teams debugging complex distributed training issues","users prototyping models with frequent code iterations"],"limitations":["Interactive access encourages non-reproducible environments; changes made via terminal are not persisted to container images","No session persistence; terminal disconnection may lose unsaved work","SSH key management not documented; unclear if keys are auto-generated or user-provided","No built-in IDE or development environment; users must use command-line editors (vim, nano) or forward X11 for GUI tools"],"requires":["SSH client or web browser","Pod running and network-accessible","SSH credentials (key or password)"],"input_types":["shell commands (bash, Python, CUDA)","file uploads (code, data, configs)"],"output_types":["command output (stdout, stderr)","file downloads (logs, checkpoints, results)"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__cap_9","uri":"capability://code.generation.editing.framework.agnostic.gpu.compute.with.no.custom.framework.requirements","name":"framework-agnostic gpu compute with no custom framework requirements","description":"Supports arbitrary GPU workloads without framework restrictions; users can run PyTorch, TensorFlow, JAX, CUDA C++, or custom code. Pods come pre-installed with CUDA toolkit, cuDNN, and common frameworks, but users can install any framework via pip, conda, or source compilation. No proprietary APIs or framework-specific abstractions required.","intents":["I want to run custom CUDA kernels or C++ code on GPUs","I need to experiment with multiple frameworks (PyTorch, TensorFlow, JAX) in the same environment","I want to use niche frameworks or research code not supported by managed services"],"best_for":["researchers exploring novel architectures or custom kernels","teams using multiple frameworks across projects","users with legacy code or proprietary frameworks"],"limitations":["No managed framework support; users responsible for dependency management and version compatibility","No automatic framework optimization (e.g., TensorFlow XLA, PyTorch JIT compilation) documented","CUDA version and cuDNN version not specified; unclear if multiple versions are available","No framework-specific monitoring or profiling tools provided; users must use native tools (nvidia-smi, nvprof)"],"requires":["Framework installation (pip, conda, or source)","CUDA 11.x or 12.x compatible code","Knowledge of framework-specific APIs and optimization"],"input_types":["Python code (PyTorch, TensorFlow, JAX)","CUDA C++ code","compiled binaries"],"output_types":["trained models","inference results","profiling data"],"categories":["code-generation-editing","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"runpod__headline","uri":"capability://deployment.infra.on.demand.gpu.cloud.platform.for.ai.inference.and.training","name":"on-demand gpu cloud platform for ai inference and training","description":"RunPod is a flexible GPU cloud platform that provides on-demand and spot instances for AI inference and training, featuring competitive pricing and serverless endpoints.","intents":["best GPU cloud platform","GPU cloud for AI training","affordable GPU instances for AI","on-demand GPU for machine learning","cloud platform for AI inference"],"best_for":["AI developers","data scientists","machine learning engineers"],"limitations":[],"requires":[],"input_types":[],"output_types":[],"categories":["deployment-infra"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":56,"verified":false,"data_access_risk":"high","permissions":["SSH client or web terminal access","API key or account authentication","Sufficient account balance or payment method on file","Network connectivity to RunPod regions (8+ worldwide)","Containerized inference code (Docker image with HTTP server)","Model weights accessible via S3-compatible storage or bundled in container","API endpoint definition (input/output schema)","RunPod Serverless SDK or direct HTTP client","Persistent storage for checkpoints (S3-compatible or network volume)","Application code that saves and loads checkpoints"],"failure_modes":["Per-second billing means idle time costs accumulate; no automatic shutdown on inactivity","No built-in autoscaling for Pods — manual provisioning required for multi-pod workflows","Cold-start latency of ~60 seconds may be prohibitive for real-time inference","Pricing not fully transparent in documentation (shows $0/s placeholders)","Flex workers have sub-200ms cold-start but still incur per-second compute charges during execution; not suitable for ultra-latency-sensitive applications (<50ms)","Active workers require continuous billing even during idle periods; cost-effective only if average utilization >30%","No built-in request queuing or priority scheduling documented; high-traffic spikes may exceed scaling capacity","FlashBoot sub-200ms claim is unverified; actual latency depends on model size and worker initialization overhead","Failover mechanism not documented; unclear if restarts are automatic or require manual intervention","Restart latency not specified; unclear if recovery is sub-second or takes minutes","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.7,"quality":0.9,"ecosystem":0.15000000000000002,"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":"2026-05-24T12:16:25.061Z","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=runpod","compare_url":"https://unfragile.ai/compare?artifact=runpod"}},"signature":"jzMsW1c82DLEvsIIejGm2QsYJS1JvnxYwbvo9j+hVaAJbRiN9Fz3d/lvxVKNowKhQnIsZhwZNjCSk8+5eKBNBQ==","signedAt":"2026-06-20T22:07:37.350Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/runpod","artifact":"https://unfragile.ai/runpod","verify":"https://unfragile.ai/api/v1/verify?slug=runpod","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"}}