Update env_tasks.py

env_tasks.py

@@ -1,7 +1,18 @@
-# ─── Helper ─────────────────────────────────────────────
+from dataclasses import dataclass
+from typing import Dict, Any, Tuple, List
+
+from env_models import (
+    TicketObservation,
+    TriageAction,
+    TicketCategory,
+    TicketPriority,
+    Department,
+    EscalationDecision,
+)
+
+# ─── SAFE CLAMP ─────────────────────────────────────────
 
 def _clamp(score: float) -> float:
-    # ALWAYS force strict (0,1)
     if score >= 1.0:
         return 0.99
     if score <= 0.0:
@@ -9,29 +20,50 @@ def _clamp(score: float) -> float:
     return score
 
 
-# ─── Graders ─────────────────────────────────────────────
+# ─── TICKETS ───────────────────────────────────────────
+
+TICKET_EASY = TicketObservation(
+    ticket_id="1",
+    subject="Laptop not working",
+    body="Laptop won't boot",
+    reporter_name="User",
+    reporter_role="Employee",
+    system_info="Windows",
+    timestamp="now",
+    previous_tickets=0,
+    task_instruction="Classify issue",
+    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 += 0.60
+        score += 0.6
 
     if action.priority in {TicketPriority.HIGH, TicketPriority.CRITICAL}:
         score += 0.25
 
-    if len(action.response) >= 30:
+    if len(action.response) > 10:
         score += 0.15
 
     score = _clamp(score)
-    breakdown["final_score"] = round(score, 4)
+    breakdown["score"] = score
     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
@@ -43,54 +75,74 @@ def grade_medium(action: TriageAction) -> Tuple[float, Dict[str, Any]]:
         score += 0.25
 
     if action.escalate == EscalationDecision.ESCALATE:
-        score += 0.10
+        score += 0.1
 
-    if len(action.response) >= 50:
+    if len(action.response) > 20:
         score += 0.15
 
     score = _clamp(score)
-    breakdown["final_score"] = round(score, 4)
+    breakdown["score"] = score
     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 += 0.20
+        score += 0.2
 
     if action.priority == TicketPriority.CRITICAL:
-        score += 0.20
+        score += 0.2
 
     if action.department == Department.SECURITY_TEAM:
-        score += 0.20
+        score += 0.2
 
     if action.escalate == EscalationDecision.ESCALATE:
-        score += 0.10
+        score += 0.1
 
     resp = action.response.lower()
 
-    if any(k in resp for k in ["disconnect", "unplug", "network"]):
+    if "disconnect" in resp:
         score += 0.15
 
-    if any(k in resp for k in ["do not", "don't", "avoid"]):
+    if "do not" in resp:
         score += 0.15
 
-    if "decrypt" in resp or "recover" in resp:
-        penalty += 0.40
+    if "decrypt" in resp:
+        penalty += 0.4
 
     if "disconnect" not in resp:
-        penalty += 0.30
+        penalty += 0.3
 
     if action.escalate != EscalationDecision.ESCALATE:
-        penalty += 0.20
+        penalty += 0.2
 
     raw = score - penalty
-
-    # 🔥 CRITICAL FIX
     final = _clamp(raw)
 
-    breakdown["final_score"] = round(final, 4)
-    return final, breakdown
+    breakdown["score"] = final
+    return final, breakdown
+
+
+# ─── TASK STRUCTURE ───────────────────────────────────
+
+@dataclass
+class Task:
+    task_id: str
+    name: str
+    description: str
+    difficulty: str
+    ticket: TicketObservation
+    grader: Any
+    max_steps: int = 1
+
+
+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),
+]
+
+TASK_MAP = {t.task_id: t for t in TASKS}