{"id":17511,"date":"2026-09-21T09:57:19","date_gmt":"2026-09-21T09:57:19","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=17511"},"modified":"2026-09-21T09:57:19","modified_gmt":"2026-09-21T09:57:19","slug":"microsoft-ai-300-practice-test-questions-and-exam-dumps-part17-q321-340","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/microsoft-ai-300-practice-test-questions-and-exam-dumps-part17-q321-340\/","title":{"rendered":"Microsoft AI-300 Practice Test Questions and Exam Dumps Part17 Q321-340"},"content":{"rendered":"

View Full <\/b>Microsoft AI-300 Exam Dumps<\/b><\/a> and Practice Test Dumps<\/b><\/p>\n

 <\/p>\n

Question 321. Which practice helps maintain traceability between a deployed model and the code used to create it?<\/b><\/p>\n

    \n
  1. <\/b> Deleting old source revisions<\/span>
    \n<\/span>2.<\/b> Deploying without metadata<\/span>
    \n<\/span>3.<\/b> Recording the source-control revision with the model version<\/span>
    \n<\/span>4.<\/b> Changing code directly in production<\/span><\/li>\n<\/ol>\n

    Correct Answer: 3. Recording the source-control revision with the model version<\/b><\/p>\n

    Explanation:<\/b><\/p>\n

    Recording the source-control revision associated with a model version creates a clear connection between the deployed artifact and the code that produced it. This information is valuable when troubleshooting unexpected behavior, auditing deployments, comparing releases, or reproducing a previous state. A model artifact by itself may not reveal which preprocessing logic, training code, or configuration was used. By preserving the source revision alongside model and deployment metadata, teams can identify the exact implementation associated with a release. This practice also supports rollback and controlled promotion because engineers can understand the complete lineage of the model rather than relying on undocumented assumptions.<\/span><\/p>\n

    Question 322. What is the main purpose of an automated data validation stage in an ML pipeline?<\/b><\/p>\n

      \n
    1. <\/b> To verify that input data meets defined quality requirements<\/span>
      \n<\/span>2.<\/b> To replace source control<\/span>
      \n<\/span>3.<\/b> To disable model testing<\/span>
      \n<\/span>4.<\/b> To automatically approve every model<\/span><\/li>\n<\/ol>\n

      Correct Answer: 1. To verify that input data meets defined quality requirements<\/b><\/p>\n

      Explanation:<\/b><\/p>\n

      Automated data validation checks whether data satisfies predefined requirements before it proceeds through important stages of an ML workflow. Checks may include schema conformity, required fields, data types, missing values, value ranges, duplicate records, and other quality conditions relevant to the model. Poor-quality or unexpected data can lead to unreliable training results or incorrect predictions. Automating these checks allows the pipeline to detect problems consistently and stop or flag processing when requirements are not met. This provides an important control before training, evaluation, or deployment and helps ensure that later pipeline stages operate on data that meets established expectations.<\/span><\/p>\n

      Question 323. Which Azure ML capability is useful for maintaining versions of registered models and their associated metadata?<\/b><\/p>\n

        \n
      1. <\/b> Model registry<\/span>
        \n<\/span>2.<\/b> Network security group<\/span>
        \n<\/span>3.<\/b> DNS zone<\/span>
        \n<\/span>4.<\/b> Virtual network gateway<\/span><\/li>\n<\/ol>\n

        Correct Answer: 1. Model registry<\/b><\/p>\n

        Explanation:<\/b><\/p>\n

        A model registry provides a centralized way to organize and manage model artifacts and their associated metadata. Versioning models in a registry allows teams to distinguish between different trained releases and identify which version has been evaluated or deployed. Metadata can include model descriptions, source information, evaluation results, tags, and other lineage details. This organization is especially important in MLOps because production systems need controlled promotion and rollback rather than relying on unstructured model files. A registry can therefore serve as an important reference point for managing the lifecycle of models from experimentation through validation, deployment, monitoring, and eventual replacement.<\/span><\/p>\n

        Question 324. Which metric can indicate that an online ML endpoint is experiencing more failures than expected?<\/b><\/p>\n

          \n
        1. <\/b> Training dataset size<\/span>
          \n<\/span>2.<\/b> Model file size<\/span>
          \n<\/span>3.<\/b> Source-code line count<\/span>
          \n<\/span>4.<\/b> Error rate<\/span><\/li>\n<\/ol>\n

          Correct Answer: 4. Error rate<\/b><\/p>\n

          Explanation:<\/b><\/p>\n

          Error rate measures the proportion or frequency of requests that fail compared with the total number of requests processed. Monitoring this metric is important for online ML endpoints because increases can indicate application failures, dependency problems, resource limitations, invalid requests, configuration issues, or deployment defects. Teams typically establish a baseline and define alert thresholds based on expected service behavior. When the error rate exceeds an acceptable level, engineers can investigate logs, recent deployment changes, resource metrics, and endpoint health information. Error-rate monitoring complements latency and availability measurements, providing a broader view of whether an inference service is operating normally.<\/span><\/p>\n

          Question 325. What is an important benefit of using managed identities for ML workloads?<\/b><\/p>\n

            \n
          1. <\/b> They require developers to share passwords<\/span>
            \n<\/span>2.<\/b> They provide identity-based access without embedding credentials in application code<\/span>
            \n<\/span>3.<\/b> They permanently grant administrator permissions<\/span>
            \n<\/span>4.<\/b> They remove the need for authorization controls<\/span><\/li>\n<\/ol>\n

            Correct Answer: 2. They provide identity-based access without embedding credentials in application code<\/b><\/p>\n

            Explanation:<\/b><\/p>\n

            Managed identities allow supported Azure resources and applications to authenticate to other services using an identity instead of storing passwords, keys, or connection secrets directly in application code. This reduces the need to manage long-lived credentials and can improve security when permissions are assigned according to least-privilege principles. For an ML workload, a managed identity might be granted access to a model registry, storage resource, monitoring service, or secret-management service. The identity itself does not automatically provide unrestricted access; appropriate role assignments are still required. Combining managed identities with carefully scoped permissions creates a more manageable authentication model for production workloads.<\/span><\/p>\n

            Question 326. Why should an ML deployment pipeline include automated security validation?<\/b><\/p>\n

              \n
            1. <\/b> To detect security-related problems before production promotion<\/span>
              \n<\/span>2.<\/b> To eliminate all logging<\/span>
              \n<\/span>3.<\/b> To prevent source-control usage<\/span>
              \n<\/span>4.<\/b> To guarantee that every model is automatically approved<\/span><\/li>\n<\/ol>\n

              Correct Answer: 1. To detect security-related problems before production promotion<\/b><\/p>\n

              Explanation:<\/b><\/p>\n

              Automated security validation provides an early control for identifying problems before a model or service reaches production. Depending on the environment, checks may examine configuration, permissions, secret handling, dependencies, exposed resources, or other security requirements. Running these checks consistently through the deployment pipeline reduces reliance on manual inspection and helps ensure that defined policies are applied to every release. If a security requirement fails, the pipeline can stop or pause promotion until the issue is investigated and resolved. Security validation works alongside testing, model evaluation, and operational checks, creating multiple safeguards before production exposure.<\/span><\/p>\n

              Question 327. Which approach allows teams to compare a new model with the existing model using a portion of production traffic?<\/b><\/p>\n

                \n
              1. <\/b> Source-code branching only<\/span>
                \n<\/span>2.<\/b> Offline archival<\/span>
                \n<\/span>3.<\/b> Traffic splitting<\/span>
                \n<\/span>4.<\/b> Manual configuration deletion<\/span><\/li>\n<\/ol>\n

                Correct Answer: 3. Traffic splitting<\/b><\/p>\n

                Explanation:<\/b><\/p>\n

                Traffic splitting allows production requests to be distributed between different model deployments according to defined proportions. For example, a new model can receive a limited percentage of requests while the established model continues serving the majority. Teams can then compare operational and model-related metrics such as latency, error rate, prediction distributions, and quality indicators. This provides evidence about real production behavior without immediately moving all traffic to the new version. Traffic splitting must be implemented with appropriate monitoring and controls so that problems can be detected quickly. It can be particularly useful for gradual releases and controlled model comparisons.<\/span><\/p>\n

                Question 328. What should be retained to support investigation after a failed production deployment?<\/b><\/p>\n

                  \n
                1. <\/b> Only the deployment name<\/span>
                  \n<\/span>2.<\/b> Only the current model file<\/span>
                  \n<\/span>3.<\/b> Only the number of users<\/span>
                  \n<\/span>4.<\/b> Logs, configuration, model version, environment, and deployment metadata<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 4. Logs, configuration, model version, environment, and deployment metadata<\/b><\/p>\n

                  Explanation:<\/b><\/p>\n

                  A failed deployment can be difficult to investigate if important evidence is removed immediately. Retaining logs together with the model version, configuration, environment definition, deployment metadata, and relevant pipeline information provides a historical record of what happened. Engineers can use this information to identify errors, compare the failed release with a known-good version, and determine whether the issue originated in code, dependencies, configuration, infrastructure, or the model itself. Retaining deployment evidence also supports auditing and helps teams improve future pipeline controls. Automatic cleanup policies should therefore be designed carefully so that useful diagnostic information remains available for an appropriate period.<\/span><\/p>\n

                  Question 329. What is the primary purpose of a health probe for an online ML endpoint?<\/b><\/p>\n

                    \n
                  1. <\/b> To measure training accuracy<\/span>
                    \n<\/span>2.<\/b> To determine whether the service is responding and healthy<\/span>
                    \n<\/span>3.<\/b> To version source code<\/span>
                    \n<\/span>4.<\/b> To register every model automatically<\/span><\/li>\n<\/ol>\n

                    Correct Answer: 2. To determine whether the service is responding and healthy<\/b><\/p>\n

                    Explanation:<\/b><\/p>\n

                    A health probe checks whether a deployed service is functioning sufficiently to handle requests. For an online ML endpoint, health information can help determine whether the serving application is available, responding correctly, and able to operate as expected. Health checks are different from model-quality evaluation because a service can be technically available while still producing poor predictions. Combining health monitoring with latency, error rate, availability, and model-performance signals provides broader operational visibility. When a health check fails, automated deployment or recovery processes may stop traffic, restart resources, or redirect requests according to the organization\u2019s configured procedures.<\/span><\/p>\n

                    Question 330. Which practice makes infrastructure configuration easier to reproduce across ML environments?<\/b><\/p>\n

                      \n
                    1. <\/b> Infrastructure as Code<\/span>
                      \n<\/span>2.<\/b> Manual server changes<\/span>
                      \n<\/span>3.<\/b> Undocumented production edits<\/span>
                      \n<\/span>4.<\/b> Shared administrator accounts<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 1. Infrastructure as Code<\/b><\/p>\n

                      Explanation:<\/b><\/p>\n

                      Infrastructure as Code, commonly abbreviated IaC, represents infrastructure configuration through version-controlled definitions rather than relying entirely on manual changes. This can include resources, networking, permissions, service settings, and other infrastructure requirements needed by an ML workload. Version-controlled infrastructure definitions can be reviewed, tested, reused, and applied consistently across development, testing, and production environments. IaC also creates a historical record of changes, making troubleshooting and auditing easier. When infrastructure needs to be recreated after an incident or expanded for another environment, the documented definitions provide a repeatable starting point instead of requiring engineers to remember or manually reconstruct previous settings.<\/span><\/p>\n

                      Question 331. Why should an ML environment definition include explicit dependency versions?<\/b><\/p>\n

                        \n
                      1. <\/b> To make deployments intentionally unpredictable<\/span>
                        \n<\/span>2.<\/b> To prevent model registration<\/span>
                        \n<\/span>3.<\/b> To improve reproducibility and reduce unexpected dependency changes<\/span>
                        \n<\/span>4.<\/b> To remove testing requirements<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 3. To improve reproducibility and reduce unexpected dependency changes<\/b><\/p>\n

                        Explanation:<\/b><\/p>\n

                        Explicit dependency versions help ensure that an ML workload uses the same library releases and runtime components across environments. If dependencies are left unconstrained, a package update can introduce behavior changes, compatibility problems, or differences between training and production environments. Versioned environment definitions provide a reproducible description of the software required by the workload. This makes it easier to recreate an earlier model state and investigate issues caused by dependency changes. Dependency management is especially important for production ML systems because even a small library difference can affect preprocessing, inference behavior, performance, or integration with other services.<\/span><\/p>\n

                        Question 332. What should a deployment pipeline do if a candidate model fails its predefined performance threshold?<\/b><\/p>\n

                          \n
                        1. <\/b> Automatically send it to all production users<\/span>
                          \n<\/span>2.<\/b> Delete all evaluation records<\/span>
                          \n<\/span>3.<\/b> Disable future validation<\/span>
                          \n<\/span>4.<\/b> Stop or pause promotion until the issue is addressed<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 4. Stop or pause promotion until the issue is addressed<\/b><\/p>\n

                          Explanation:<\/b><\/p>\n

                          A predefined performance threshold acts as a quality gate for candidate models. If a model does not meet the required threshold, the pipeline should prevent automatic promotion and preserve the evaluation results for investigation. The team may determine that the model needs additional training, improved data, different features, configuration changes, or further evaluation. Automatically promoting a failed candidate would bypass the control intended to protect production quality. Keeping the failed candidate and its validation evidence also helps maintain traceability. Once the issue has been addressed, the model can undergo the required validation process again before any controlled promotion occurs.<\/span><\/p>\n

                          Question 333. Which monitoring signal can help identify changes in the distribution of model predictions?<\/b><\/p>\n

                            \n
                          1. <\/b> Prediction distribution<\/span>
                            \n<\/span>2.<\/b> Source-control branch count<\/span>
                            \n<\/span>3.<\/b> Infrastructure template length<\/span>
                            \n<\/span>4.<\/b> Number of deployment comments<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 1. Prediction distribution<\/b><\/p>\n

                            Explanation:<\/b><\/p>\n

                            Monitoring prediction distributions involves observing how model outputs are distributed over time and comparing current behavior with historical or expected patterns. Significant changes can provide an early signal that the input population, model behavior, or operating environment has changed. Prediction-distribution monitoring does not by itself establish that a model is incorrect, but it can identify situations that warrant additional investigation. Teams can correlate these changes with data-drift signals, input characteristics, model versions, deployment changes, and quality measurements. This makes prediction monitoring a useful part of production observability, particularly when direct ground-truth labels are delayed or unavailable.<\/span><\/p>\n

                            Question 334. What is the purpose of an approval gate in a model promotion workflow?<\/b><\/p>\n

                              \n
                            1. <\/b> To remove all automated validation<\/span>
                              \n<\/span>2.<\/b> To provide a controlled authorization point before promotion<\/span>
                              \n<\/span>3.<\/b> To guarantee model accuracy<\/span>
                              \n<\/span>4.<\/b> To delete previous model versions<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 2. To provide a controlled authorization point before promotion<\/b><\/p>\n

                              Explanation:<\/b><\/p>\n

                              An approval gate introduces an explicit authorization point in the promotion process. Depending on organizational requirements, a responsible person or team can review model evaluation results, security checks, deployment configuration, monitoring plans, and other evidence before allowing a release to proceed. Approval gates can complement automated quality checks rather than replacing them. They are particularly useful when a deployment has operational, security, compliance, or business considerations that require documented review. The gate should be connected to a traceable workflow so that the approval decision and supporting evidence can be associated with the specific model and deployment version being promoted.<\/span><\/p>\n

                              Question 335. Which deployment approach can reduce the risk of immediately exposing all users to a new ML model?<\/b><\/p>\n

                                \n
                              1. <\/b> Direct full replacement<\/span>
                                \n<\/span>2.<\/b> Uncontrolled manual deployment<\/span>
                                \n<\/span>3.<\/b> Gradual traffic rollout<\/span>
                                \n<\/span>4.<\/b> Removing the previous model<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 3. Gradual traffic rollout<\/b><\/p>\n

                                Explanation:<\/b><\/p>\n

                                A gradual traffic rollout introduces a new model to production in controlled stages rather than immediately sending all traffic to it. The initial traffic percentage can be kept small while teams monitor operational and model-specific metrics. If the new model performs according to expectations, the traffic allocation can be increased in subsequent stages. If problems appear, the rollout can be paused or traffic can be redirected to the existing version. This approach provides an opportunity to observe real production behavior while limiting exposure. It works particularly well when combined with canary deployment, automated monitoring, defined thresholds, and a reliable rollback process.<\/span><\/p>\n

                                Question 336. Which practice helps protect sensitive credentials used by an ML deployment?<\/b><\/p>\n

                                  \n
                                1. <\/b> Storing passwords directly in source code<\/span>
                                  \n<\/span>2.<\/b> Sharing secrets through configuration files<\/span>
                                  \n<\/span>3.<\/b> Publishing credentials in deployment logs<\/span>
                                  \n<\/span>4.<\/b> Using a managed secret store with controlled access<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 4. Using a managed secret store with controlled access<\/b><\/p>\n

                                  Explanation:<\/b><\/p>\n

                                  A managed secret store provides a dedicated location for securely storing credentials, keys, certificates, and other sensitive configuration values. Applications can retrieve required secrets through authorized identity-based access rather than embedding them directly in source code or deployment scripts. This approach also supports centralized management, access control, auditing, and credential rotation. Production logs should not expose secret values, even when troubleshooting. Combining a managed secret store with managed identities and least-privilege permissions creates a stronger security model than distributing credentials across source files or configuration documents. These controls are particularly important for automated ML pipelines that interact with multiple cloud resources.<\/span><\/p>\n

                                  Question 337. What is a key advantage of automated CI testing for ML pipeline code?<\/b><\/p>\n

                                    \n
                                  1. <\/b> It can detect code-related problems before deployment<\/span>
                                    \n<\/span>2.<\/b> It guarantees perfect model predictions<\/span>
                                    \n<\/span>3.<\/b> It removes the need for production monitoring<\/span>
                                    \n<\/span>4.<\/b> It eliminates version control<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 1. It can detect code-related problems before deployment<\/b><\/p>\n

                                    Explanation:<\/b><\/p>\n

                                    Continuous integration testing automatically checks changes to pipeline or application code before those changes are incorporated into later deployment stages. Tests can identify syntax errors, incorrect logic, broken interfaces, and other problems that might otherwise reach later environments. For ML systems, CI can also test data-processing functions, pipeline components, configuration handling, and integration points. Successful CI testing does not guarantee that the model will perform correctly in production because model quality and runtime behavior require additional validation. However, it provides an important early safeguard and supports a repeatable development process in which changes are tested consistently before controlled promotion.<\/span><\/p>\n

                                    Question 338. Why is model lineage important in an MLOps workflow?<\/b><\/p>\n

                                      \n
                                    1. <\/b> It prevents all future model changes<\/span>
                                      \n<\/span>2.<\/b> It connects models with their source data, code, environments, and evaluation information<\/span>
                                      \n<\/span>3.<\/b> It removes deployment history<\/span>
                                      \n<\/span>4.<\/b> It makes rollback impossible<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 2. It connects models with their source data, code, environments, and evaluation information<\/b><\/p>\n

                                      Explanation:<\/b><\/p>\n

                                      Model lineage provides a record of how a model was created and how it moved through the development and deployment lifecycle. Useful lineage can connect a model to its training data references, source-code revision, dependencies, environment definition, configuration, evaluation results, and deployment information. This allows teams to understand which inputs and processes contributed to a particular model version. Lineage is valuable for troubleshooting because engineers can compare releases and identify relevant changes. It also supports auditing and reproducibility. When a production issue occurs, strong lineage makes it easier to identify the exact artifacts associated with the affected deployment.<\/span><\/p>\n

                                      Question 339. Which operational metric measures whether an ML service remains available to receive requests?<\/b><\/p>\n

                                        \n
                                      1. <\/b> Training duration<\/span>
                                        \n<\/span>2.<\/b> Model artifact size<\/span>
                                        \n<\/span>3.<\/b> Availability or uptime<\/span>
                                        \n<\/span>4.<\/b> Dataset column count<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 3. Availability or uptime<\/b><\/p>\n

                                        Explanation:<\/b><\/p>\n

                                        Availability and uptime indicate whether a service remains operational and accessible over a defined period. For an online ML endpoint, availability monitoring helps determine whether applications can successfully reach the inference service when requests are made. A service may have acceptable latency during successful requests but still have an availability problem if it frequently becomes unreachable. Monitoring availability alongside error rate, latency, resource utilization, and health checks provides a broader picture of service reliability. When availability falls below an established threshold, alerts or recovery procedures can be triggered according to the organization\u2019s operational policies. This information is separate from model accuracy and should be monitored independently.<\/span><\/p>\n

                                        Question 340. Which set of practices provides a structured approach to managing ML models in production?<\/b><\/p>\n

                                          \n
                                        1. <\/b> Manual deployment, undocumented changes, and no monitoring<\/span>
                                          \n<\/span>2.<\/b> Versioning, automated validation, controlled promotion, monitoring, and rollback<\/span>
                                          \n<\/span>3.<\/b> Direct workstation deployment and shared credentials<\/span>
                                          \n<\/span>4.<\/b> Untracked artifacts and permanent administrator access<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 2. Versioning, automated validation, controlled promotion, monitoring, and rollback<\/b><\/p>\n

                                          Explanation:<\/b><\/p>\n

                                          A structured production MLOps process combines multiple controls across the model lifecycle. Versioning ensures that code, models, dependencies, configurations, and relevant metadata can be traced. Automated validation checks whether candidates satisfy predefined requirements before promotion. Controlled deployment techniques limit exposure while production behavior is evaluated. Monitoring then provides visibility into operational metrics and model-related signals after release. Finally, maintaining known-good versions and deployment configurations supports recovery when a release does not behave as expected. These practices work together rather than independently: versioning supports traceability, validation supports quality control, controlled promotion limits exposure, monitoring provides feedback, and rollback provides a recovery mechanism.<\/span><\/p>\n

                                           <\/p>\n","protected":false},"excerpt":{"rendered":"

                                          View Full Microsoft AI-300 Exam Dumps and Practice Test Dumps   Question 321. Which practice helps maintain traceability between a deployed model and the code used to create it? Deleting old source revisions 2. Deploying without metadata 3. Recording the source-control revision with the model version 4. Changing code directly in production Correct Answer: 3. […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":[],"categories":[1648,1647],"tags":[],"_links":{"self":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17511"}],"collection":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/comments?post=17511"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17511\/revisions"}],"predecessor-version":[{"id":17512,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/17511\/revisions\/17512"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=17511"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=17511"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=17511"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}