{"id":19422,"date":"2026-09-23T05:48:02","date_gmt":"2026-09-23T05:48:02","guid":{"rendered":"https:\/\/www.examlabs.com\/certification\/?p=19422"},"modified":"2026-09-23T05:48:02","modified_gmt":"2026-09-23T05:48:02","slug":"snowflake-snowpro-specialty-gen-ai-ges-c01-practice-test-questions-and-exam-dumps-part-20-q381-400","status":"publish","type":"post","link":"https:\/\/www.examlabs.com\/certification\/snowflake-snowpro-specialty-gen-ai-ges-c01-practice-test-questions-and-exam-dumps-part-20-q381-400\/","title":{"rendered":"Snowflake SnowPro Specialty Gen AI GES-C01 Practice Test Questions and Exam Dumps Part 20 Q381-400"},"content":{"rendered":"

View Full <\/b>Snowflake SnowPro Specialty Gen AI GES-C01<\/b> Exam Dumps<\/b><\/a> and Practice Test Dumps<\/b><\/p>\n

 <\/p>\n

Question 381: Which Snowflake capability is primarily intended to help users interact with structured business data using natural-language questions?<\/b><\/p>\n

    \n
  1. <\/b> Document AI<\/span><\/li>\n
  2. <\/b> Cortex Search<\/span><\/li>\n
  3. <\/b> Cortex Analyst<\/span><\/li>\n
  4. <\/b> AI_SUMMARIZE<\/span><\/li>\n<\/ol>\n

    Correct Answer: 3. Cortex Analyst<\/b><\/p>\n

    Explanation:<\/b>
    \n<\/b>Cortex Analyst is designed to support natural-language interaction with structured business data. It uses semantic information about business concepts, metrics, dimensions, and relationships to help interpret user questions and produce analytical responses. Cortex Search focuses on search and retrieval, while Document AI is designed for extracting information from documents. AI_SUMMARIZE focuses on summarizing content. Therefore, when users need to ask business questions in natural language and receive analytics based on structured enterprise data, Cortex Analyst is the relevant capability.<\/span><\/p>\n

    Question 382: What does a vector index primarily help a retrieval system accomplish?<\/b><\/p>\n

      \n
    1. <\/b> Efficiently search and compare vector representations<\/span><\/li>\n
    2. <\/b> Automatically validate business policies<\/span><\/li>\n
    3. <\/b> Generate natural-language responses<\/span><\/li>\n
    4. <\/b> Authenticate users<\/span><\/li>\n<\/ol>\n

      Correct Answer: 1. Efficiently search and compare vector representations<\/b><\/p>\n

      Explanation:<\/b>
      \n<\/b>A vector index provides data structures or indexing mechanisms that help retrieval systems efficiently locate vectors that are similar to a query vector. This is important in semantic-search workflows because embeddings represent queries and source content numerically. Rather than comparing a query against every stored representation in an inefficient manner, an appropriate indexing strategy can improve retrieval performance. Vector indexing does not generate responses, authenticate users, or enforce business policies by itself. Those responsibilities belong to other components of the application architecture.<\/span><\/p>\n

      Question 383: Why can chunk overlap improve retrieval quality for some documents?<\/b><\/p>\n

        \n
      1. <\/b> It eliminates the need for source attribution<\/span><\/li>\n
      2. <\/b> It prevents all duplicate content from being stored<\/span><\/li>\n
      3. <\/b> It preserves context that spans the boundary between adjacent chunks<\/span><\/li>\n
      4. <\/b> It guarantees that every chunk is independently complete<\/span><\/li>\n<\/ol>\n

        Correct Answer: 3. It preserves context that spans the boundary between adjacent chunks<\/b><\/p>\n

        Explanation:<\/b>
        \n<\/b>Chunk overlap repeats a portion of content between adjacent chunks so that information near a chunk boundary is less likely to become disconnected from its surrounding context. This can be useful when a sentence, explanation, or logical relationship crosses an otherwise artificial boundary. The amount of overlap should be selected based on the document structure and retrieval requirements. Excessive overlap can create unnecessary redundancy, so it should not be assumed that more overlap is always better. Chunk overlap supports contextual continuity but does not guarantee complete chunks or replace source attribution.<\/span><\/p>\n

        Question 384: Which retrieval approach combines semantic similarity with exact keyword matching?<\/b><\/p>\n

          \n
        1. <\/b> Hybrid search<\/span><\/li>\n
        2. <\/b> Classification<\/span><\/li>\n
        3. <\/b> Zero-shot prompting<\/span><\/li>\n
        4. <\/b> Document extraction<\/span><\/li>\n<\/ol>\n

          Correct Answer: 1. Hybrid search<\/b><\/p>\n

          Explanation:<\/b>
          \n<\/b>Hybrid search combines different retrieval signals, commonly semantic similarity and keyword-based matching. Semantic retrieval can identify content that is conceptually related even when terminology differs, while keyword matching is useful for exact terms, names, codes, and identifiers. Combining these signals can make retrieval more effective for enterprise use cases where both meaning and exact wording matter. Classification assigns inputs to categories, zero-shot prompting guides model behavior without examples, and document extraction focuses on extracting information from source documents. Hybrid search is therefore the approach that directly combines semantic and lexical retrieval signals.<\/span><\/p>\n

          Question 385: What should an application do when a retrieved document is highly relevant but the user is not authorized to access it?<\/b><\/p>\n

            \n
          1. <\/b> Include it because relevance has priority over security<\/span><\/li>\n
          2. <\/b> Provide the document only if the model requests it<\/span><\/li>\n
          3. <\/b> Exclude it from the user’s accessible context<\/span><\/li>\n
          4. <\/b> Include it but remove the document title<\/span><\/li>\n<\/ol>\n

            Correct Answer: 3. Exclude it from the user’s accessible context<\/b><\/p>\n

            Explanation:<\/b>
            \n<\/b>Authorization must be enforced independently of retrieval relevance. A document can be highly relevant to a query while still being restricted from the requesting user. Supplying unauthorized content to the model could expose sensitive information even if that information is not ultimately displayed directly. Retrieval and application logic should therefore incorporate appropriate access controls so that only authorized information enters the model context. Removing a title or relying on the model to decide access is not a sufficient security boundary. Authorization-aware retrieval is an essential part of governed enterprise AI architecture.<\/span><\/p>\n

            Question 386: Which metadata field can be particularly useful when retrieval should prioritize the most recent version of a document?<\/b><\/p>\n

              \n
            1. <\/b> Document update or publication timestamp<\/span><\/li>\n
            2. <\/b> Number of model parameters<\/span><\/li>\n
            3. <\/b> Prompt temperature<\/span><\/li>\n
            4. <\/b> Vector dimensionality<\/span><\/li>\n<\/ol>\n

              Correct Answer: 1. Document update or publication timestamp<\/b><\/p>\n

              Explanation:<\/b>
              \n<\/b>Publication or update timestamps provide metadata that can help retrieval systems identify the relative freshness of source documents. When current information is important, applications can use timestamps to filter or prioritize newer versions and reduce the chance of grounding responses in obsolete content. Other metadata such as document version, source authority, or department can also contribute to retrieval decisions. Model parameters, temperature, and vector dimensionality do not directly indicate document freshness. Maintaining useful document metadata is therefore an important consideration when designing enterprise retrieval workflows that require current information.<\/span><\/p>\n

              Question 387: What is the main purpose of prompt templates in an enterprise generative AI application?<\/b><\/p>\n

                \n
              1. <\/b> To provide a consistent structure for instructions and contextual information<\/span><\/li>\n
              2. <\/b> To replace the underlying language model<\/span><\/li>\n
              3. <\/b> To permanently store all generated responses<\/span><\/li>\n
              4. <\/b> To eliminate the need for evaluation<\/span><\/li>\n<\/ol>\n

                Correct Answer: 1. To provide a consistent structure for instructions and contextual information<\/b><\/p>\n

                Explanation:<\/b>
                \n<\/b>Prompt templates provide a reusable structure for instructions, user inputs, retrieved context, examples, and output requirements. Consistent templates can make application behavior easier to maintain and evaluate because the same basic prompting structure can be applied across requests. Templates do not replace the language model or eliminate the need for evaluation. They also do not inherently function as permanent storage for generated responses. In enterprise applications, well-designed prompt templates can support repeatability, versioning, testing, and controlled changes to the instructions provided to the model.<\/span><\/p>\n

                Question 388: Which prompting approach gives the model a task description without providing task-specific examples?<\/b><\/p>\n

                  \n
                1. <\/b> Few-shot prompting<\/span><\/li>\n
                2. <\/b> Zero-shot prompting<\/span><\/li>\n
                3. <\/b> Retrieval augmentation<\/span><\/li>\n
                4. <\/b> Hybrid search<\/span><\/li>\n<\/ol>\n

                  Correct Answer: 2. Zero-shot prompting<\/b><\/p>\n

                  Explanation:<\/b>
                  \n<\/b>Zero-shot prompting asks a model to perform a task using instructions without providing task-specific examples in the prompt. The model relies on its existing learned capabilities and the instructions supplied by the application. Few-shot prompting differs because it includes examples that demonstrate the desired behavior or output pattern. Retrieval augmentation supplies external context and is a separate architectural technique, while hybrid search combines retrieval signals. Zero-shot prompting can be useful when the task is sufficiently clear and examples are unnecessary or would consume additional context capacity.<\/span><\/p>\n

                  Question 389: What can happen when a RAG application includes excessive irrelevant context in the prompt?<\/b><\/p>\n

                    \n
                  1. <\/b> The response may become less focused on the user’s actual question<\/span><\/li>\n
                  2. <\/b> The model automatically gains additional training<\/span><\/li>\n
                  3. <\/b> Authorization is automatically strengthened<\/span><\/li>\n
                  4. <\/b> All hallucinations are eliminated<\/span><\/li>\n<\/ol>\n

                    Correct Answer: 1. The response may become less focused on the user’s actual question<\/b><\/p>\n

                    Explanation:<\/b>
                    \n<\/b>Excessive irrelevant context can make it harder for a language model to identify and use the information that matters most for the user’s request. It can consume context capacity and introduce competing or distracting information. Retrieval systems should therefore aim to provide context that is relevant, sufficient, and appropriately ranked rather than simply maximizing the amount of retrieved content. Additional context does not automatically retrain the model, strengthen authorization, or eliminate hallucinations. Careful retrieval filtering and ranking can help maintain a focused context and improve the usefulness of grounded responses.<\/span><\/p>\n

                    Question 390: Which Snowflake capability is intended to help process information from unstructured documents?<\/b><\/p>\n

                      \n
                    1. <\/b> Cortex Analyst<\/span><\/li>\n
                    2. <\/b> Document AI<\/span><\/li>\n
                    3. <\/b> Cortex Search<\/span><\/li>\n
                    4. <\/b> Semantic model<\/span><\/li>\n<\/ol>\n

                      Correct Answer: 2. Document AI<\/b><\/p>\n

                      Explanation:<\/b>
                      \n<\/b>Document AI is intended to help extract useful information from unstructured or semi-structured documents. Examples include invoices, forms, reports, and other business documents where important information may not already exist in structured database columns. Extracted information can subsequently support analytics, search, automation, or other AI workflows. Cortex Analyst is focused on natural-language analytics over structured data, while Cortex Search focuses on search and retrieval. A semantic model provides business meaning and relationships for analytics rather than directly performing general document extraction. Document AI is therefore the appropriate capability for this use case.<\/span><\/p>\n

                      Question 391: Why should retrieved content be clearly distinguished from trusted system instructions in a RAG prompt?<\/b><\/p>\n

                        \n
                      1. <\/b> Retrieved content may contain untrusted text that should not be treated as instructions<\/span><\/li>\n
                      2. <\/b> Retrieved content always has higher priority than system instructions<\/span><\/li>\n
                      3. <\/b> The distinction automatically increases vector similarity<\/span><\/li>\n
                      4. <\/b> System instructions are unnecessary when retrieval is enabled<\/span><\/li>\n<\/ol>\n

                        Correct Answer: 1. Retrieved content may contain untrusted text that should not be treated as instructions<\/b><\/p>\n

                        Explanation:<\/b>
                        \n<\/b>Retrieved documents are source data, not automatically trusted instructions. Enterprise content can potentially contain text that attempts to influence model behavior, including malicious or unintended instructions. Clearly separating system instructions, user input, and retrieved content helps establish their different roles within the application. This is especially important when considering prompt-injection risks. Retrieved content should be treated as information to evaluate and use rather than automatically obeyed as an instruction. Appropriate application controls, source handling, and validation should complement prompt design when building a secure RAG workflow.<\/span><\/p>\n

                        Question 392: Which practice is most useful for detecting weaknesses in retrieval across common and difficult user requests?<\/b><\/p>\n

                          \n
                        1. <\/b> Test only one simple query<\/span><\/li>\n
                        2. <\/b> Evaluate representative normal, difficult, and edge-case queries<\/span><\/li>\n
                        3. <\/b> Remove source documents before testing<\/span><\/li>\n
                        4. <\/b> Change the evaluation criteria for every query<\/span><\/li>\n<\/ol>\n

                          Correct Answer: 2. Evaluate representative normal, difficult, and edge-case queries<\/b><\/p>\n

                          Explanation:<\/b>
                          \n<\/b>Testing a representative range of queries helps reveal retrieval weaknesses that may not appear with simple examples. A useful evaluation set can include common requests, ambiguous questions, difficult cases, terminology variations, and other scenarios expected in production. Consistent evaluation criteria make results easier to compare over time. Testing only an easy query can create an incomplete view of system performance. A representative evaluation strategy is particularly useful when changing chunking, embeddings, ranking, metadata filters, or other retrieval components because it provides evidence about whether those changes improve or degrade behavior.<\/span><\/p>\n

                          Question 393: Which factor directly affects how much retrieved information can be supplied to a language model at one time?<\/b><\/p>\n

                            \n
                          1. <\/b> The model’s context-window capacity<\/span><\/li>\n
                          2. <\/b> The document’s file name<\/span><\/li>\n
                          3. <\/b> The user’s screen resolution<\/span><\/li>\n
                          4. <\/b> The number of database tables<\/span><\/li>\n<\/ol>\n

                            Correct Answer: 1. The model’s context-window capacity<\/b><\/p>\n

                            Explanation:<\/b>
                            \n<\/b>The model’s context window determines the amount of input information that can be processed within the model’s supported context limits. In a RAG workflow, this includes the user request, instructions, retrieved documents, conversation information, and other supplied context. Retrieval systems therefore need to balance the amount of retrieved information with relevance and available context capacity. File names, screen resolution, and the number of database tables do not directly determine the model’s context capacity. Understanding this constraint helps architects design effective chunking, ranking, and retrieval-result strategies.<\/span><\/p>\n

                            Question 394: What is a useful way to perform prompt regression testing?<\/b><\/p>\n

                              \n
                            1. <\/b> Compare different prompt versions using a consistent representative test set<\/span><\/li>\n
                            2. <\/b> Evaluate each prompt with completely different queries<\/span><\/li>\n
                            3. <\/b> Select the prompt that produces the longest answers<\/span><\/li>\n
                            4. <\/b> Test only responses that are already known to be correct<\/span><\/li>\n<\/ol>\n

                              Correct Answer: 1. Compare different prompt versions using a consistent representative test set<\/b><\/p>\n

                              Explanation:<\/b>
                              \n<\/b>Prompt regression testing compares a modified prompt with a previous version under consistent evaluation conditions. Using the same representative test set makes it easier to determine whether a change improved or degraded system behavior. The evaluation should use predefined criteria appropriate to the application’s requirements rather than relying on response length. Including only known successful cases can hide regressions and does not provide a realistic assessment. Consistent prompt testing is useful for maintaining predictable behavior as instructions, output requirements, or other parts of an AI workflow evolve.<\/span><\/p>\n

                              Question 395: Which statement correctly distinguishes retrieval from generation in a RAG architecture?<\/b><\/p>\n

                                \n
                              1. <\/b> Retrieval finds relevant source information, while generation uses context to produce a response<\/span><\/li>\n
                              2. <\/b> Retrieval creates the final response, while generation searches the source index<\/span><\/li>\n
                              3. <\/b> Retrieval controls user permissions, while generation stores metadata<\/span><\/li>\n
                              4. <\/b> Retrieval and generation are identical processes<\/span><\/li>\n<\/ol>\n

                                Correct Answer: 1. Retrieval finds relevant source information, while generation uses context to produce a response<\/b><\/p>\n

                                Explanation:<\/b>
                                \n<\/b>Retrieval and generation perform different functions in a RAG architecture. The retrieval stage searches available source information and identifies content that is relevant to the user’s request. The generation stage then uses the user request, instructions, and selected context to produce a natural-language response. Security, metadata, indexing, and validation may support these stages but are separate responsibilities. Keeping retrieval and generation conceptually distinct also allows organizations to evaluate each stage independently and determine whether a quality problem originates from poor source selection or from the generated response.<\/span><\/p>\n

                                Question 396: Why can changing the embedding model affect an existing semantic retrieval system?<\/b><\/p>\n

                                  \n
                                1. <\/b> Different embedding models can produce different numerical representations of the same content<\/span><\/li>\n
                                2. <\/b> Embedding models automatically change user permissions<\/span><\/li>\n
                                3. <\/b> Embedding models determine the model’s response temperature<\/span><\/li>\n
                                4. <\/b> Embedding models replace document metadata<\/span><\/li>\n<\/ol>\n

                                  Correct Answer: 1. Different embedding models can produce different numerical representations of the same content<\/b><\/p>\n

                                  Explanation:<\/b>
                                  \n<\/b>An embedding model converts content into numerical representations, and different models may represent the same text differently. Because semantic retrieval compares query and document representations, changing the embedding model can change similarity relationships and therefore alter which documents are retrieved or how they are ranked. Such a change should be evaluated using representative retrieval queries before being adopted in production. Embedding models do not directly control permissions, temperature, or metadata. Their primary impact is on the numerical representation used by semantic retrieval and the resulting search behavior.<\/span><\/p>\n

                                  Question 397: Which approach is most appropriate when a generated answer must comply with strict application-specific business rules?<\/b><\/p>\n

                                    \n
                                  1. <\/b> Rely solely on the language model’s instructions<\/span><\/li>\n
                                  2. <\/b> Use application-level validation after generation<\/span><\/li>\n
                                  3. <\/b> Increase temperature<\/span><\/li>\n
                                  4. <\/b> Retrieve more unrelated documents<\/span><\/li>\n<\/ol>\n

                                    Correct Answer: 2. Use application-level validation after generation<\/b><\/p>\n

                                    Explanation:<\/b>
                                    \n<\/b>Application-level validation is useful when generated output must satisfy strict business, structural, security, or operational requirements. A language model may follow instructions most of the time but can still produce output that violates a required rule or format. Validation logic can inspect the generated result and determine whether it meets predefined requirements before the application accepts or presents it. Depending on the failure-handling policy, invalid output can be rejected, corrected, regenerated, or escalated. Increasing temperature or retrieving unrelated documents does not enforce deterministic business rules.<\/span><\/p>\n

                                    Question 398: What is one benefit of separating retrieval evaluation from generation evaluation?<\/b><\/p>\n

                                      \n
                                    1. <\/b> It helps identify whether quality problems originate in source retrieval or response generation<\/span><\/li>\n
                                    2. <\/b> It guarantees that all retrieved sources are authoritative<\/span><\/li>\n
                                    3. <\/b> It eliminates the need for monitoring<\/span><\/li>\n
                                    4. <\/b> It prevents all prompt-injection attacks<\/span><\/li>\n<\/ol>\n

                                      Correct Answer: 1. It helps identify whether quality problems originate in source retrieval or response generation<\/b><\/p>\n

                                      Explanation:<\/b>
                                      \n<\/b>Separating retrieval and generation evaluation provides a clearer way to diagnose AI application quality problems. If relevant information is not retrieved, improving the language model’s generation settings may not address the underlying issue. Conversely, if retrieval is accurate but the generated response is poor, the problem may involve prompting, model selection, output handling, or validation. Independent evaluation can therefore guide optimization efforts more effectively. It does not guarantee source authority, eliminate monitoring, or prevent every security threat, but it provides useful diagnostic information about different stages of the RAG workflow.<\/span><\/p>\n

                                      Question 399: Which governance practice is particularly important when deploying generative AI against enterprise data?<\/b><\/p>\n

                                        \n
                                      1. <\/b> Ignore source permissions because the model can decide what users should see<\/span><\/li>\n
                                      2. <\/b> Combine access controls, evaluation, monitoring, and defined governance processes<\/span><\/li>\n
                                      3. <\/b> Allow every application to retrieve every enterprise document<\/span><\/li>\n
                                      4. <\/b> Disable validation to maximize model flexibility<\/span><\/li>\n<\/ol>\n

                                        Correct Answer: 2. Combine access controls, evaluation, monitoring, and defined governance processes<\/b><\/p>\n

                                        Explanation:<\/b>
                                        \n<\/b>Enterprise generative AI systems require governance controls that address security, data access, quality, operational behavior, and responsible use. Access controls help restrict information to authorized users, while evaluation measures system behavior against defined requirements. Monitoring provides visibility after deployment, and governance processes establish ownership, policies, review procedures, and ongoing oversight. Allowing unrestricted access or relying on the model itself to enforce permissions creates security risks. A combination of technical and organizational controls is therefore important for managing enterprise AI applications throughout their lifecycle.<\/span><\/p>\n

                                        Question 400: Which sequence best represents a mature enterprise generative AI lifecycle?<\/b><\/p>\n

                                          \n
                                        1. <\/b> Deploy immediately \u2192 evaluate later \u2192 define requirements if problems occur<\/span><\/li>\n
                                        2. <\/b> Define requirements \u2192 build and evaluate \u2192 deploy \u2192 monitor and reassess<\/span><\/li>\n
                                        3. <\/b> Generate responses \u2192 remove governance \u2192 deploy \u2192 ignore monitoring<\/span><\/li>\n
                                        4. <\/b> Retrieve unrestricted data \u2192 generate responses \u2192 establish access controls afterward<\/span><\/li>\n<\/ol>\n

                                          Correct Answer: 2. Define requirements \u2192 build and evaluate \u2192 deploy \u2192 monitor and reassess<\/b><\/p>\n

                                          Explanation:<\/b>
                                          \n<\/b>A mature enterprise AI lifecycle begins by defining the intended task, requirements, risks, and success criteria. The system can then be built and evaluated using representative data and use cases before deployment. After deployment, monitoring helps identify operational, security, retrieval, quality, or usage changes. The system can subsequently be reassessed and improved as requirements, data, models, or business conditions change. Establishing governance and access controls only after deployment is inappropriate because security and responsible-use requirements should be considered throughout the lifecycle.<\/span><\/p>\n

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

                                          View Full Snowflake SnowPro Specialty Gen AI GES-C01 Exam Dumps and Practice Test Dumps   Question 381: Which Snowflake capability is primarily intended to help users interact with structured business data using natural-language questions? Document AI Cortex Search Cortex Analyst AI_SUMMARIZE Correct Answer: 3. Cortex Analyst Explanation: Cortex Analyst is designed to support natural-language interaction […]<\/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\/19422"}],"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=19422"}],"version-history":[{"count":1,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/19422\/revisions"}],"predecessor-version":[{"id":19423,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/posts\/19422\/revisions\/19423"}],"wp:attachment":[{"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/media?parent=19422"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/categories?post=19422"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.examlabs.com\/certification\/wp-json\/wp\/v2\/tags?post=19422"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}