A development team is constructing a Gen AI application using Snowflake Cortex LLM functions, particularly for conversational and text generation tasks. They are concerned about potential high costs due to token consumption. Which of the following strategies would most effectively help minimize token usage and optimize costs when working with these Cortex LLM functions?

• A. To encourage more succinct LLM responses and reduce completion_tokens, configure the temperature option to a higher value (e.g., 0.7) in COMPLETE function calls.
• B. In multi-turn conversations within Cortex Analyst, integrate a dedicated LLM summarization agent to rephrase follow-up questions, which reduces the total conversational history passed as context to the main LLM.
• C. Utilize the COUNT_TOKENS (SNOWFLAKE .CORTEX) helper function to pre-validate the prompt length against the model's context window, thereby preventing truncation errors and subsequent re-runs.
• D. For multi-turn conversational experiences using SNOWFLAKE .CORTEX.COMPLETE, only send the most recent user prompt in each API call, as the model automatically retains previous context.
• E. When employing AI_COMPLETE for structured output tasks, providing concise and highly descriptive explanations for each field within the JSON schema will reduce the input tokens required for the LLM to understand and adhere to the schema accurately.

A data operations team is attempting to scale up their Document AI automated pipeline. They are using a Snowflake Task to process a large volume of daily scanned invoices and receipts, which are stored in an internal stage 'financial_docs stage'. The current processing involves documents that are frequently around 75 MB each, and often there are batches exceeding 1 ,200 documents in a single day. The pipeline is failing consistently. Which of the following factors could be contributing to the failures in this Document AI automated pipeline? (Select all that apply.)

• A. The account role executing the task lacks the 'SNOWFLAKCORTEX USER database role, which is a prerequisite for Document AI functions.
• B. The number of documents in a single daily batch (exceeding 1 ,200) surpasses Document AI's processing limit per query.
• C. The individual document size (75 MB) exceeds the maximum supported file size for Document AI.
• D. The was created without specifying 'ENCRYPTION = (TYPE = 'SNOWFLAKE SSE')'.
• E. The Snowflake Task is configured as a 'SERVERLESS TASK, which is not supported by Document AI.

A data engineering team is implementing a Document AI solution to automate the extraction of vendor invoice details. They have already published a Document AI model build, 'vendor_invoice_model', located in New invoices are uploaded to an internal stage, 'fin db.invoice_processing_schema.raw invoices_stage'. The 'invoice_pipeline role' is responsible for executing the
! PREDICT
method on these staged documents as part of an automated task. Which of the following USAGE privileges are essential for to successfully execute the
! PREDICT
method and process documents from the specified stage, assuming the required SNOWFLAKE.DOCUMENT_INTELLIGENCE_CREATOR role is already granted?

• A.
  
• B.
  
• C.
  
• D.
  
• E.
  

An operations team is investigating an issue with a generative AI application powered by Snowflake Cortex Analyst, where users reported unexpected behavior in generated SQL. To diagnose the problem, they examine the detailed event logs captured by Snowflake AI Observability. Which categories of information can they expect to find in these event tables to assist their investigation?

• A. Any error messages or warnings that occurred during the processing of the request.
• B. The exact SQL queries that Cortex Analyst generated in response to user questions.
• C. The complete request and response bodies associated with the application's execution steps.
• D. Real-time CPU and memory usage statistics for the Snowflake virtual warehouse executing the LLM inference.
• E. The full text of the natural language questions submitted by the users.