Update env_tasks.py

env_tasks.py

@@ -1,68 +1,37 @@
-from dataclasses import dataclass
-from typing import Dict, Any, Tuple, List
-
-from env_models import (
-    TicketObservation,
-    TriageAction,
-    TicketCategory,
-    TicketPriority,
-    Department,
-    EscalationDecision,
-)
-
 # ─── Helper ─────────────────────────────────────────────
 
 def _clamp(score: float) -> float:
+    # ALWAYS force strict (0,1)
     if score >= 1.0:
-        return 0.95
-    elif score <= 0.0:
-        return 0.05
+        return 0.99
+    if score <= 0.0:
+        return 0.01
     return score
 
 
-# ─── Dummy Tickets (keep minimal, your existing ones work too) ───────────────
-
-TICKET_EASY = TicketObservation(
-    ticket_id="1",
-    subject="Laptop not working",
-    body="Laptop not booting",
-    reporter_name="User",
-    reporter_role="Employee",
-    system_info="Windows",
-    timestamp="now",
-    previous_tickets=0,
-    task_instruction="Classify",
-    valid_categories=[c.value for c in TicketCategory],
-    valid_priorities=[p.value for p in TicketPriority],
-    valid_departments=[d.value for d in Department],
-)
-
-TICKET_MEDIUM = TICKET_EASY
-TICKET_HARD = TICKET_EASY
-
-
 # ─── Graders ─────────────────────────────────────────────
 
 def grade_easy(action: TriageAction) -> Tuple[float, Dict[str, Any]]:
+    breakdown: Dict[str, Any] = {}
     score = 0.0
-    breakdown = {}
 
     if action.category == TicketCategory.HARDWARE:
-        score += 0.6
+        score += 0.60
 
     if action.priority in {TicketPriority.HIGH, TicketPriority.CRITICAL}:
         score += 0.25
 
-    if len(action.response) > 10:
+    if len(action.response) >= 30:
         score += 0.15
 
     score = _clamp(score)
+    breakdown["final_score"] = round(score, 4)
     return score, breakdown
 
 
 def grade_medium(action: TriageAction) -> Tuple[float, Dict[str, Any]]:
+    breakdown: Dict[str, Any] = {}
     score = 0.0
-    breakdown = {}
 
     if action.category in {TicketCategory.NETWORK, TicketCategory.ACCESS}:
         score += 0.25
@@ -74,66 +43,54 @@ def grade_medium(action: TriageAction) -> Tuple[float, Dict[str, Any]]:
         score += 0.25
 
     if action.escalate == EscalationDecision.ESCALATE:
-        score += 0.1
+        score += 0.10
 
-    if len(action.response) > 20:
+    if len(action.response) >= 50:
         score += 0.15
 
     score = _clamp(score)
+    breakdown["final_score"] = round(score, 4)
     return score, breakdown
 
 
 def grade_hard(action: TriageAction) -> Tuple[float, Dict[str, Any]]:
+    breakdown: Dict[str, Any] = {}
     score = 0.0
     penalty = 0.0
-    breakdown = {}
 
     if action.category == TicketCategory.SECURITY:
-        score += 0.2
+        score += 0.20
 
     if action.priority == TicketPriority.CRITICAL:
-        score += 0.2
+        score += 0.20
 
     if action.department == Department.SECURITY_TEAM:
-        score += 0.2
+        score += 0.20
 
     if action.escalate == EscalationDecision.ESCALATE:
-        score += 0.1
+        score += 0.10
 
     resp = action.response.lower()
 
-    if "disconnect" in resp:
+    if any(k in resp for k in ["disconnect", "unplug", "network"]):
         score += 0.15
 
-    if "do not" in resp:
+    if any(k in resp for k in ["do not", "don't", "avoid"]):
         score += 0.15
 
-    if "decrypt" in resp:
-        penalty += 0.4
-
-    raw = score - penalty
-    final = max(0.05, min(0.95, raw))
-
-    return final, breakdown
+    if "decrypt" in resp or "recover" in resp:
+        penalty += 0.40
 
+    if "disconnect" not in resp:
+        penalty += 0.30
 
-# ─── Task Registry ───────────────────────────────────────
-
-@dataclass
-class Task:
-    task_id: str
-    name: str
-    description: str
-    difficulty: str
-    ticket: TicketObservation
-    grader: Any
-    max_steps: int = 1
+    if action.escalate != EscalationDecision.ESCALATE:
+        penalty += 0.20
 
+    raw = score - penalty
 
-TASKS: List[Task] = [
-    Task("task_easy", "Easy", "Basic classification", "easy", TICKET_EASY, grade_easy),
-    Task("task_medium", "Medium", "Full triage", "medium", TICKET_MEDIUM, grade_medium),
-    Task("task_hard", "Hard", "Security incident", "hard", TICKET_HARD, grade_hard),
-]
+    # 🔥 CRITICAL FIX
+    final = _clamp(raw)
 
-TASK_MAP = {t.task_id: t for t in TASKS}
+    breakdown["final_score"] = round(final, 4)
+    return final, breakdown