A data team is designing a new Cortex Analyst application and wants to ensure optimal performance, accuracy, and user experience for text-to-SQL conversions. They are particularly interested in how custom instructions interact with other semantic model features and LLM functionalities. Which of the following statements about using in Cortex Analyst are accurate?

• A. Custom instructions are primarily used to define new logical tables or dimensions within the semantic model, effectively extending the data model at runtime.
• B. When both 'custom_instructionS and a highly relevant 'verified_query' exist for a user's question, Cortex Analyst will always prioritize the directives from the 'custom_instructions' over the SQL provided in the 'verified_querv'.
• C. The presence of 'custom_instructions' in a semantic model can potentially increase the token count for Cortex Analyst requests, as the instructions are passed as additional context to the LLM.
• D. Using detailed 'custom_instructions' can help mitigate issues where the LLM might struggle with domain-specific terminology or complex business logic not explicitly defined in column descriptions.
• E. The 'custom_instructions' in a semantic model directly influence the underlying Large Language Model (LLM) to generate SQL queries that align with specified business context or formatting preferences.

A Gen AI Specialist in Snowflake Cortex aims to fine-tune an LLM for enhanced task-specific performance. When creating a fine-tuning job using SNOWFLAKE. CORTEX. FINETUNE( 'CREATE', ... ) , which statement accurately describes the required training data format and a supported base model, aligning with Snowflake's Gen AI principles for leveraging LLMs?

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

A developer is building a conversational application using Snowflake Cortex Analyst and is interacting with it via the REST API. The application needs to support multi-turn conversations where users can ask follow-up questions. Which of the following statements accurately describe how to maintain conversational state and interact with the Cortex Analyst REST API for a multi-turn experience?

• A. To maintain state, the application must construct a 'messages' array for each new prompt, including all prior 'user' prompts and 'analyst' responses in chronological order, and send it with every API call.
• B. A special 'RESET_CONVERSATION' API endpoint should be called periodically to clear the LLM's internal context and manage token consumption for long conversations.
• C. For performance optimization, only the most recent user prompt and the immediate preceding analyst response should be passed in the 'messages' array for follow-up questions.
• D. The argument, when using the 'COMPLETE function (or its API equivalent), must be a string for single-turn conversations but an array of objects for multi-turn conversations, with each object containing 'role' and 'content'.
• E. The Cortex Analyst REST API automatically retains the full conversation history for each user session, eliminating the need to explicitly pass previous messages.

A Gen AI specialist is designing a RAG pipeline utilizing Cortex Search for an application that queries a large repository of unstructured text documents. To optimize the quality of retrieval and subsequent LLM responses, what are the critical best practices and understanding of Cortex Search's mechanisms that the specialist should consider regarding text processing and tokenization?

• A. Embedding models with larger context windows, such as snowflake-arctic-embed-1-v2.e-8k (8000 tokens), are always superior as they allow the RAG system to process the entire document as a single, highly relevant chunk.
• B. The SNOWFLAKE .CORTEX. COUNT TOKENS function is a helper function that can be used to accurately determine the token count for a given string based on a specified model, aiding in adherence to context window limits.
• C. For best search results, text in the search column should be split into chunks of no more than 512 tokens, as smaller chunks generally lead to more precise retrieval and relevant LLM context.
• D. Cortex Search operates solely on vector embeddings for semantic search; keyword-based retrieval is handled by a separate, less efficient mechanism outside the core search service.
• E. When text input exceeds an embedding model's context window, Cortex Search truncates the text for both semantic embedding and keyword-based retrieval, potentially losing critical information.