A user opened the Snowflake Cortex LLM Playground in the AI & ML Studio in Snowsight, chose gemma-7b in the Standard models drop-down list and typed "What is Gen AI?" in the dialog box. Which SQL statement will achieve the same result?
Answer(s): B
A ❌ (Incorrect):CLASSIFY_TEXT is used to categorize text into predefined labels. It does not generate open-ended answers like “What is Gen AI?”.B ✅ (Correct):SNOWFLAKE.CORTEX.COMPLETE('gemma-7b', 'What is Gen AI?') uses the specified language model( gemma-7b ) to generate a natural-language response to the prompt. This matches exactly what the LLMPlayground does when you select a model and type a question.C ❌ (Incorrect):EXTRACT_ANSWER is designed to extract answers from provided context, not to generate a free-form explanation from a prompt.D ❌ (Incorrect):SUMMARIZE is used to shorten existing text. Since no source text is being summarized, this function is not appropriate.
Which arguments are required when using the SNOWFLAKE.CORTEX.COMPLETE function? (Select TWO).
Answer(s): A,E
A: model ✅ (Correct):The model argument is required because it specifies which language model (for example, gemma-7b , mistral-large , etc.) Snowflake Cortex should use to generate the response. Without a model, the function cannot run.B: top_p ❌ (Incorrect):top_p is an optional parameter used to control randomness in the output. It is not required to call theCOMPLETE function.C: response_format ❌ (Incorrect):response_format is optional and is mainly used when structured output (such as JSON) is needed. It is not mandatory.D: max_tokens ❌ (Incorrect):max_tokens controls the maximum length of the generated response, but Snowflake provides a default value if it is not specified. Therefore, it is optional.E: prompt_or_history ✅ (Correct):The prompt_or_history argument is required because it contains the input prompt or conversation history that the model uses to generate a response.
A Gen AI Specialist has registered a new model in the Snowflake Model Registry and granted the READ privilege to a custom role, TRAINER_ROLE. When a new user with the TRAINER_ROLE attempted to download the model artifacts using the GET command, an error occurred. What would cause the error?
A: The user does not have the OWNERSHIP privilege on the model ❌ (Incorrect):OWNERSHIP is not required to download model artifacts. The READ privilege on the model allows users to access and download the model, provided other required privileges (like schema access) are in place.B: The user does not have the USAGE privilege on the schema ✅ (Correct):Even if a user has READ access to the model, they must also have the USAGE privilege on the schema where the model is registered. Without USAGE on the schema, Snowflake blocks access to objects within it, causing the GET command to fail.C: The user is not using the show models() download method ❌ (Incorrect):There is no requirement to use show models() to download model artifacts. Models can be accessed directly if proper privileges are granted.D: The user is not using Snowpark Python file_operation.get operation ❌ (Incorrect):While Snowpark provides helper methods to download files, they are not mandatory. The error is related to missing privileges, not the download method used.
Which command will use the Snowflake Document AI PREDICT method to extract structured data from a document?
Answer(s): D
A: SELECT paystubs!PREDICT('s3://bucket/invoice.pdf',1); ❌ (Incorrect):Snowflake Document AI cannot directly read files using an external S3 path. Document AI requires documents to be accessed from a Snowflake internal stage using a scoped file URL.B: SELECT paystubs!PREDICT('file:///local/path/invoice.pdf',1); ❌ (Incorrect):Local file system paths are not accessible to Snowflake. Document AI can only process files stored in Snowflake stages.C: SELECT paystubs!PREDICT(GET_PRESIGNED_URL(@my_stage, 'paystubs/invoice.pdf'),1); ❌ (Incorrect):GET_PRESIGNED_URL generates a temporary external URL intended for client-side access. Document AI does not accept presigned URLs for the PREDICT method.D: SELECT paystubs!PREDICT(BUILD_SCOPED_FILE_URL(@my_stage, 'paystubs/invoice.pdf'),1); ✅ (Correct):BUILD_SCOPED_FILE_URL creates a scoped file URL for a document stored in a Snowflake internal stage. This is the required and supported approach for passing documents to the Snowflake Document AI PREDICT method to extract structured data.
A team repeatedly requests that one filter be applied on the same two product grouping columns to run data analyses. Which field can be configured in Snowflake Cortex Analyst to apply the filter on column X whenever there is a filter applied on column Y?
Answer(s): C
A: ❌ Incorrect relationship_type This defines the type of relationship (e.g., one-to-many), not automatic filter behavior.B: ❌ Incorrect relationship columns These define how tables relate, but do not enforce conditional filter propagation.C: ✅ Correct custom_instructions Custom instructions allow you to define logic such as applying a filter on one column when another is filtered.D: ❌ Incorrect cortex_search_service This is used for search and retrieval, not for defining filter logic.
Which APIs or functions can be used to query the Snowflake Cortex Search Service? (Select THREE).
Answer(s): A,B,E
A: Snowflake Python APIs ✅ (Correct):Snowflake provides Python APIs (for example, through Snowpark and Cortex integrations) that can be used to programmatically query the Cortex Search Service. This is commonly used when building applications, chatbots, or agentic workflows.B: Snowflake REST APIs ✅ (Correct):The Cortex Search Service can be queried using Snowflake REST APIs, enabling external applications to interact with search indexes and retrieve relevant results over structured and unstructured data.C: The CORTEX.COMPLETE function ❌ (Incorrect):COMPLETE is a generative LLM function used for text generation. It does not directly query the Cortex Search Service.D: The EMBED_TEXT_768 function ❌ (Incorrect):This function generates vector embeddings for text. While embeddings may be used internally by search, EMBED_TEXT_768 itself does not query the Cortex Search Service.E: The SEARCH_PREVIEW function ✅ (Correct):SEARCH_PREVIEW is a SQL function specifically designed to query and test Cortex Search indexes directly from Snowflake.F: The CORTEX.SUMMARIZE function ❌ (Incorrect):SUMMARIZE is used to condense text content and does not perform search queries.
A Gen AI Specialist needs a solution that will support asking questions in natural language and receive answers based on structured data in Snowflake without the use of SQL How can this be accomplished? (Select TWO).
Answer(s): C,D
A: Create a custom data model, apply the data model to a database, and ask questions from the Snowflake Document AI option in the AI & ML Studio in Snowsight ❌ (Incorrect):Snowflake Document AI is used to extract data from unstructured documents such as PDFs and images. It does not support asking natural-language questions over structured Snowflake tables.B: Create a custom data model, enhance the model with cross-region inference, and ask questions in a custom-built application that calls the Snowflake Cortex Agent API ❌ (Incorrect):Cortex Agents and cross-region inference are not required for answering natural-language questions on structured data without SQL. This approach is unnecessarily complex for the stated requirement.C: Create a semantic model, upload the semantic model to a stage, and ask questions from the Snowflake Cortex Analyst option in the AI & ML Studio in Snowsight ✅ (Correct):Cortex Analyst is specifically designed to allow users to ask natural-language questions over structured data without writing SQL . Using a semantic model in Snowsight directly fulfills this requirement.D: Create a semantic model, upload the semantic model to a stage, and ask questions in a custom-built application that calls the Snowflake Cortex Analyst API ✅ (Correct):This option enables the same capability as option C but through a custom application . The Cortex Analyst API allows programmatic access to natural-language analytics over structured data without SQL.E: Create a Snowflake Cortex Agent in Snowsight, extract data from the database, and ask questions by fine-tuning an LLM with the extracted data ❌ (Incorrect):Fine-tuning an LLM is not necessary for querying structured data in natural language. Cortex Analyst already provides this functionality without model training.
When using the SNOWFLAKE.CORTEX.FINETUNE function, which column names must be included in the SQL statement in the training data query parameter? (Select TWO)
Answer(s): B,D
A: BASE MODEL ❌ (Incorrect):BASE MODEL is not a required column in the training data query. The base model is specified as a parameter to the SNOWFLAKE.CORTEX.FINETUNE function, not as a column in the training dataset.B: COMPLETION ✅ (Correct):The COMPLETION column is required because it contains the expected output that the model should learn to generate. During fine-tuning, the model learns to map prompts to these completions.C: INFERENCE ❌ (Incorrect):INFERENCE is not a valid or required column name for fine-tuning training data in Snowflake Cortex.D: PROMPT ✅ (Correct):The PROMPT column is required because it contains the input text that will be provided to the model. Fine-tuning uses prompt–completion pairs to train the model.E: TRAINING ❌ (Incorrect):TRAINING is not a recognized column name for the training data query. The function expects specific column names, not a generic training flag.
Post your Comments and Discuss Snowflake GES-C01 exam prep with other Community members:
Our website is free, but we have to fight against AI bots and content theft. We're sorry for the inconvenience caused by these security measures. You can access the rest of the GES-C01 content, but please register or login to continue.