Design and Implement Retrieval-Augmented Generation (RAG) DP-800 Practice Questions (Page 4)

Questions 16–20

1. 16application · medium

   You need to call an Azure OpenAI endpoint from a T-SQL stored procedure. The endpoint requires a JSON payload with a 'prompt' field. You have a table of FAQs that you want to include in the prompt. What is the correct way to build the payload?

   • AUse FOR JSON to convert the FAQs to a JSON string, concatenate it with the prompt text, and pass it as the 'prompt' value.
   • BPass the FAQs as a separate field in the payload using sp_invoke_external_rest_endpoint.
   • CUse OPENJSON to convert the FAQs to JSON and then pass it as the prompt.
   • DUse a linked server to send the FAQs directly to the model.
   Reveal Solution Discussion Select an answer first
2. 17foundation · easy

   You need to send a prompt to an Azure OpenAI model from an Azure SQL database. Which statement correctly describes the use of sp_invoke_external_rest_endpoint?

   • AIt automatically formats the response as a JSON array of rows
   • BIt can be used to call the Azure OpenAI endpoint by providing the URL and a payload containing the prompt
   • CIt can only be used to call endpoints that are in the same virtual network as the database
   • DIt requires the endpoint to be an Azure Function, not an OpenAI service
   Reveal Solution Discussion Select an answer first
4. 18application · medium

   You have a stored procedure that calls an Azure OpenAI model using sp_invoke_external_rest_endpoint. The response is stored in a variable of type NVARCHAR(MAX). The response contains a JSON object with a 'choices' array, and each choice has a 'text' field. You need to extract the text from the first choice and insert it into a table. Which T-SQL statement should you use?

   • ASELECT JSON_QUERY(@response, '$.choices[0].text')
   • BSELECT @response.choices[0].text
   • CSELECT JSON_VALUE(@response, '$.choices[0].text')
   • DSELECT OPENJSON(@response, '$.choices[0].text')
   Reveal Solution Discussion Select an answer first
5. 19expert · hard

   A manufacturing company has a SQL database with sensor readings. They want to use an Azure OpenAI model to generate maintenance recommendations. The sensor data is stored in a table with millions of rows. The model has a token limit, and sending all data at once is not possible. You need to design a T-SQL solution that processes the data in batches and stores the recommendations. What should you do?

   • AUse a view that joins the sensor data with the model's response, and let the model access the database directly.
   • BUse a WHILE loop to process batches of sensor readings, convert each batch to JSON, call the model, and store the results.
   • CUse a single call with the entire table converted to JSON, relying on the model's large context window.
   • DUse a cursor to iterate over each sensor reading, call the model for each row, and store the result.
   Reveal Solution Discussion Select an answer first
6. 20application · medium

   You need to send a list of recent orders to a language model for summarization. The orders are stored in a relational table with columns OrderID, CustomerName, OrderDate, and TotalAmount. The model expects a JSON array of objects. What is the most efficient way to prepare the data for the prompt?

   • AUse FOR JSON PATH in a SELECT statement to generate the JSON array directly.
   • BManually concatenate strings in T-SQL to build the JSON array.
   • CUse the JSON_QUERY function to convert the table to JSON.
   • DExport the table to a CSV file and ask the model to parse it.
   Reveal Solution Discussion Select an answer first