ReAct Agent Simulator

Problem Context

The ReAct (Reasoning + Acting) pattern is a prompting and agent design paradigm where an AI agent interleaves Thought (reasoning about the current state), Action (calling a tool or taking a step), and Observation (processing the result of that action) in a loop until a final answer is reached.

Your task is to implement a simplified ReAct Agent Simulator that processes a sequence of steps and determines the final answer based on a set of available "tools".

Problem Description

You are given:

• A query string representing the user's question.
• A list of steps, where each step is one of:
  • "Thought: <reasoning>" — the agent reflects on what to do next.
  • "Action: <tool>(<argument>)" — the agent calls a tool with an argument.
  • "Observation: <result>" — the result of the last action (provided externally).
  • "Answer: <final_answer>" — the agent concludes with a final answer.

You have access to the following tools:

• search(query) → returns the length of the query string (simulated search result count).
• calculate(expr) → evaluates a simple arithmetic expression (+, -, *, /) with two integer operands (e.g., "3+5" → 8).
• lookup(key) → returns the number of characters in key (simulated lookup).

Your simulator should:

1. Process each step in order.
2. When an Action step is encountered, execute the tool and record its output.
3. Verify that the following Observation step matches the tool's output. If it doesn't match, mark the trace as invalid.
4. Return the final answer string from the Answer step, or "NO_ANSWER" if no answer step exists, or "INVALID" if any observation mismatch is found.

Input / Output Format

Function signature: react_agent(query: str, steps: List[str]) -> str

• Input: A query string and a list of step strings.
• Output: A string — the final answer, "NO_ANSWER", or "INVALID".

Examples

Example 1

Input:

query = "What is 3 plus 5?"
steps = [
  "Thought: I need to calculate 3+5.",
  "Action: calculate(3+5)",
  "Observation: 8",
  "Answer: The result is 8."
]