The AI Interview - Master AI/ML Interviews

Implement a basic ReAct (Reasoning + Acting) pattern where an agent alternates between generating thoughts about the current situation and taking actions. The agent should:

Generate a thought based on the observation and history
Generate an action based on the thought
Determine if it should continue (steps < max_steps)

Constraints:

Must track history properly
Should handle empty histories gracefully
Must respect max_steps limit
Return structured output with all three fields

Examples

Example 1:

Input: react_agent('The user wants weather in Paris', [], [], max_steps=3)

Output:

{'thought': 'I need to get weather information for Paris', 'action': 'weather_api(city="Paris")', 'should_continue': True}

Explanation: Agent generates thought about needing weather, then appropriate action, and continues since steps < max_steps

Starter Code

def react_agent(observation, thought_history, action_history, max_steps=5):
    """
    Implement a basic ReAct (Reasoning + Acting) agent loop.
    The agent should alternate between Thought and Action.
    
    Args:
        observation: Current environment observation (string)
        thought_history: List of previous thoughts
        action_history: List of previous actions
        max_steps: Maximum reasoning steps allowed
    
    Returns:
        dict with 'thought', 'action', 'should_continue'
    """
    # Your implementation here
    pass

ReAct Pattern Implementation

Examples

Starter Code