When implementing stateful orchestration for agentic workflows using LangGraph, which memory management approach provides the best balance of performance and context retention?

• A. Use session-ID based checkpointer with user-defined schema for selective state persistence
• B. Implement rolling window memory with fixed conversation length limits
• C. Store complete conversation history in memory with periodic database syncing

You are designing a virtual assistant that helps users check weather updates via external APIs. During testing, the agent frequently calls the incorrect tools, often hallucinating endpoints or returning incorrect formats. You suspect the prompt structure might be the root cause of these failures.
Which prompt design best supports consistent tool invocation in this agent?

• A. Use structured prompt templates with few-shot tool usage examples
• B. Rely on the agent's internal knowledge to infer tool usage
• C. Include tool names in natural language but without parameter examples
• D. Provide only a generic system instruction with no examples

You're managing an agentic AI responsible for customer support ticket triage. The agent has been consistently accurate in routing tickets to the appropriate departments. However, a team leader has noticed a significant increase in the number of tickets requiring "escalation" - cases where the agent initially misclassified a complex issue as a simple, routine one, leading to delays and frustrated customers.
What would be an appropriate first step in resolving this issue?

• A. Adjusting the agent's reward function to prioritize speed of resolution over accuracy, as a first step in analysis of the problem.
• B. Conducting a "red-teaming" exercise, having human agents deliberately create complex and ambiguous scenarios to analyze the agent's robustness.
• C. Analyzing the agent's decision-making process, focusing on the specific criteria it uses to classify tickets, and identifying potential biases or blind spots.
• D. Increasing the agent's autonomy, granting it more decision-making power during triage to improve its efficiency.

When analyzing suboptimal agent response quality after deployment, which parameter tuning evaluation methods effectively identify the optimal configuration adjustments? (Choose two.)

• A. Use production traffic directly for parameter experiments, enabling real-world insights and faster identification of impactful settings.
• B. Implement A/B testing frameworks comparing temperature, top-k, and top-p variations while measuring task-specific quality metrics and user satisfaction scores.
• C. Randomly adjust all parameters simultaneously, allowing for broader exploration of the parameter space in a shorter time frame.
• D. Design ablation studies systematically varying individual parameters while holding others constant to isolate each parameter's impact on agent behavior and performance.
• E. Apply identical parameter settings across all agent types and tasks, promoting consistency and simplifying comparison across different use cases.