Agentic AI (beginner)plain Mode
Review all 15 questions for Agentic AI (beginner).Plain Q&A (No XP)
#1Multiple Choice
What is the core structural difference between a plain LLM API call and an AI Agent?
#2Multiple Choice
How does an LLM signal its intent to invoke an external function during function calling?
#3Multiple Choice
What role string is used when appending a tool's execution result back into the chat completion thread?
#4Multiple Choice
Which framework paradigm interleaves Reasoning (Thought) and Acting (Tool Call) steps in an agent execution loop?
#5Multiple Choice
What is the primary danger of running an agent loop without a `max_iterations` circuit breaker limit?
#6Multiple Choice
What specification format is used to define tool capabilities for LLM function calling?
#7Scenario
Your agent gets stuck calling `get_server_status(server_id='42')` 15 times in a row. The tool returns `{'status': 'offline'}` every time. What 2 safeguards should be added to the agent code?
#8Scenario
An LLM agent calls a custom tool `query_sql(query='SELECT * FROM users')`. The tool handler executes the raw SQL on production DB and crashes with a syntax error. Why is giving an agent unconstrained SQL access dangerous?
#9Scenario
You define a tool `restart_pod(pod_name, namespace)`. In testing, the LLM hallucinates non-existent parameters `cluster_region='us-east-1'`. How do you fix tool call argument parsing?
#10Scenario
An agent needs to complete a task requiring 3 independent API queries (weather in Tokyo, London, NYC). In naive sequential ReAct, this takes 3 full LLM roundtrips (15s). How do modern function-calling APIs optimize this?
#11Code Fill
Complete the tool call message structure key that links a tool response back to the assistant's request:
messages.append({
'role': 'tool',
'___': tool_call.id,
'content': json.dumps(result)
})#12Code Fill
Complete the OpenAI API parameter used to pass an array of available function schemas to the model:
response = client.chat.completions.create(
model='gpt-4o',
messages=messages,
___=tools_schema
)#13Code Fill
Complete the `finish_reason` string returned by the API when the model requests function execution:
if response.choices[0].___ == 'tool_calls':
execute_tools()#14Open Ended
Deconstruct the 4 steps of a single ReAct agent iteration loop.
#15Open Ended