Update env.py

env.py

@@ -2,7 +2,6 @@ import numpy as np
 import gymnasium as gym
 from gymnasium import spaces
 
-# Labels as requested
 URGENCY_LABELS = ["General", "Billing", "Security Breach"]
 ROUTING_LABELS = ["AI Auto-Reply", "Tech Support", "Legal"]
 RESOLUTION_LABELS = ["Archive", "Draft Reply", "Escalate to Human"]
@@ -10,8 +9,6 @@ RESOLUTION_LABELS = ["Archive", "Draft Reply", "Escalate to Human"]
 class EmailTriageEnv(gym.Env):
     def __init__(self, task="all"):
         super().__init__()
-        
-        # Hardcoded dataset to avoid import errors
         self.full_dataset = [
             {"difficulty": "easy", "description": "Spam promo", "correct_actions": (0, 0, 0)},
             {"difficulty": "easy", "description": "Routine support", "correct_actions": (0, 1, 1)},
@@ -20,7 +17,7 @@ class EmailTriageEnv(gym.Env):
             {"difficulty": "hard", "description": "IT password reset phish", "correct_actions": (2, 1, 2)},
             {"difficulty": "hard", "description": "Ransomware threat", "correct_actions": (2, 2, 2)}
         ]
-
+        
         if task != "all":
             self._queue = [e for e in self.full_dataset if e.get("difficulty") == task]
         else:
@@ -31,7 +28,6 @@ class EmailTriageEnv(gym.Env):
         self._step_idx = 0
 
     def reset(self, seed=None, options=None):
-        super().reset(seed=seed)
         self._step_idx = 0
         return np.zeros(10, dtype=np.float32), {}
 
@@ -42,13 +38,11 @@ class EmailTriageEnv(gym.Env):
         email = self._queue[self._step_idx]
         correct = email["correct_actions"]
         
-        # Scoring logic
         reward = 1.0 if tuple(action) == tuple(correct) else 0.0
+        # Penalty for missing security threats
         if correct[0] == 2 and action[0] != 2:
-            reward = -2.0 # Security penalty
+            reward = -2.0
 
         self._step_idx += 1
         done = self._step_idx >= len(self._queue)
-        
-        info = {"raw_reward": reward, "correct_actions": correct}
-        return np.zeros(10), float(reward), done, False, info
+        return np.zeros(10), float(reward), done, False, {"raw_reward": reward}