A telecom company has a database of call center transcripts. They want to use an Azure OpenAI model to generate a summary of each call. The transcripts are stored in a table with a column of type NVARCHAR(MAX). Some transcripts are very long and exceed the model's token limit. You need to design a solution that handles these long transcripts. What should you do?

Use a fine-tuned model that has been trained on the transcripts.

Use a model with a larger context window and send the entire transcript.

Use a recursive CTE to split the transcript into chunks and call the model for each chunk, then combine the summaries.

Use a single call with the entire transcript, relying on the model's ability to handle long text.

You are building a RAG solution in Azure SQL Database. You need to send a prompt that includes a JSON array of product reviews to an Azure OpenAI model. The response should be a sentiment label. What is the correct way to construct and send the prompt?

Build the prompt as a string, convert the reviews to JSON with FOR JSON, and pass both to sp_invoke_external_rest_endpoint.

Use OPENJSON to convert the reviews to JSON and then call the endpoint.

Use sp_invoke_external_rest_endpoint to send the reviews as a raw table without converting to JSON.

Use a linked server to send the table directly to the model.

You have a RAG solution that calls a language model and receives a response with multiple fields, including 'text', 'confidence', and 'metadata'. You need to store all fields in separate columns. What is the most efficient way to do this?

Use OPENJSON to parse the response and insert the fields into the columns.

Use JSON_QUERY to extract each field.

Use JSON_VALUE multiple times to extract each field and insert them.

Store the entire response as a JSON string in one column.

You have a SQL query that returns a JSON string containing a prompt. You need to send this JSON to an external language model endpoint. Which step is required?

Use OPENJSON to convert the JSON string into a table before sending

Use sp_send_dbmail to email the JSON to the endpoint

Use BULK INSERT to load the JSON into a file and then send it

Call sp_invoke_external_rest_endpoint with the endpoint URL and the JSON payload

A customer support system stores troubleshooting guides in a SQL database. They want to build a chatbot that answers customer questions by referencing the most relevant guide sections. The chatbot must provide answers that are traceable to the guide content. Which design pattern should you use?

Fine-tune a model on the troubleshooting guides so it can answer from memory.

Retrieval-augmented generation: retrieve relevant guide sections from the database and include them in the prompt to the model.

Use a rule-based system that matches keywords to guide sections.

Use a model with a large context window and send all guides in every prompt.

Microsoft Certified:SQL AI Developer Associate

Domain 3Objective 3

Design and Implement Retrieval-Augmented Generation (RAG) DP-800 Practice Questions (Page 6)

Part of the Implement AI capabilities in database solutions domain, which accounts for 25–30% of the DP-800 exam. Microsoft does not publish an official question count, but from its 120-minute exam (~50–80 total, ~13–24 in this domain), expect 4–8 from this objective — we provide 33 practice questions to prepare you well beyond it. (estimate)

33questions here

7free pages

5concepts

25–30%of the exam

Questions 26–30

26expert · hard
A telecom company has a database of call center transcripts. They want to use an Azure OpenAI model to generate a summary of each call. The transcripts are stored in a table with a column of type NVARCHAR(MAX). Some transcripts are very long and exceed the model's token limit. You need to design a solution that handles these long transcripts. What should you do?
Select an answer first
27application · medium
You are building a RAG solution in Azure SQL Database. You need to send a prompt that includes a JSON array of product reviews to an Azure OpenAI model. The response should be a sentiment label. What is the correct way to construct and send the prompt?
Select an answer first
28expert · hard
You have a RAG solution that calls a language model and receives a response with multiple fields, including 'text', 'confidence', and 'metadata'. You need to store all fields in separate columns. What is the most efficient way to do this?
Select an answer first
29foundation · easy
You have a SQL query that returns a JSON string containing a prompt. You need to send this JSON to an external language model endpoint. Which step is required?
Select an answer first
30application · medium
A customer support system stores troubleshooting guides in a SQL database. They want to build a chatbot that answers customer questions by referencing the most relevant guide sections. The chatbot must provide answers that are traceable to the guide content. Which design pattern should you use?
Select an answer first

Finished these 5 questions?

Review the revealed explanations, or continue through the curriculum.

Next 5 questions →Back to DP-800

Free Basic Practice is a study aid with revealable answers — not a scored exam. Examers.io is independent and not affiliated with or endorsed by Microsoft. “DP-800” is a trademark of its owner, used for identification only.