When an agent's content filter triggers and blocks a response, what is the typical 'finish_reason' returned in the API response?
check_circle
Correct AnswerC
The content_filter finish reason explicitly tells the developer that the output was suppressed due to safety policy violations.
Question 192
You need to handle a scenario where an agent times out. Which property of the Run object should you inspect to find the error details?
check_circle
Correct AnswerC
If a Run fails or expires, the last_error object provides a code and a human-readable message explaining the failure.
Question 193
You want to build a "Translator Agent" that works in real-time. Which multimodal model is best suited for this?
check_circle
Correct AnswerD
GPT-4o is a natively multimodal model optimized for real-time interaction across text, audio, and vision.
Question 194
Which component is responsible for converting a text query into a numerical vector (embedding)?
check_circle
Correct AnswerB
Embedding models are specialized LLMs that transform text into high-dimensional vectors for similarity searching.
Question 195
You want to check the specific Python code that the code_interpreter generated and executed during a Run. Where is this found?
check_circle
Correct AnswerC
Run Steps contain the tool_calls details, including the specific input code and the resulting output/logs from the interpreter.
Question 196
Which Azure service provides a central dashboard to view all "Content Safety" violations (e.g., Hate, Violence) across your AI applications?
check_circle
Correct AnswerA
Content Safety Studio provides a visual interface to configure filters and review flagged content and activity.
Question 197
You are creating a Run and want to override the agent's default System Prompt for just this one interaction. Which parameter should you use?
check_circle
Correct AnswerD
additional_instructions appends to the agent's existing prompt, while instructions completely replaces it for that specific run.
Question 198
Which feature allows you to audit "Who" accessed the Azure OpenAI service and "When"?
check_circle
Correct AnswerB
Enabling Diagnostic Settings allows you to send audit logs to Log Analytics to track all API calls and metadata.
Question 199
You need to detect an "Indirect Prompt Injection" where a malicious instruction is hidden inside a web page the agent is browsing. Which tool is responsible?
check_circle
Correct AnswerA
Prompt Shields are specifically designed to catch instructions hidden in external data sources (Indirect Attacks).
Question 200
You are building a "Chat with your Data" app. Which architecture pattern describes using an LLM to answer questions based on retrieved documents?
check_circle
Correct AnswerB
RAG is the architecture of retrieving external data and feeding it into the LLM as context for generating an answer.
