import random class SupportEnv: def __init__(self, cases): self.cases = cases self.index = 0 self.correct_count = 0 def reset(self): self.index = 0 self.correct_count = 0 return self._get_obs() def step(self, action): case = self.cases[self.index] correct = case["correct_action"] if action == correct: reward = 0.9 self.correct_count += 1 elif action == "ask_proof" and correct == "ask_proof": reward = 0.5 else: reward = 0.1 self.index += 1 done = self.index >= len(self.cases) obs = self._get_obs() if not done else "All cases completed." return obs, reward, done # TASK GRADER def get_score(self): total = len(self.cases) score = self.correct_count / total if total > 0 else 0.0 # ensure strict range return max(0.01, min(0.99, score)) def _get_obs(self): case = self.cases[self.index] return f""" Customer Query: {case['query']} Days since delivery: {case['days']} Proof provided: {case.get('has_proof', False)} Choose one action: refund / replace / reject / ask_proof """