Update app.py

app.py

@@ -4,30 +4,20 @@ import uvicorn
 import numpy as np
 import gradio as gr
 from fastapi import FastAPI
+from env import EmailTriageEnv, URGENCY_LABELS, ROUTING_LABELS, RESOLUTION_LABELS
 
 app = FastAPI()
 
-# Shared Dataset
-EMAIL_DATASET = [
-    {"difficulty": "easy", "description": "Spam promo", "correct_actions": (0, 0, 0)},
-    {"difficulty": "easy", "description": "Routine support", "correct_actions": (0, 1, 1)},
-    {"difficulty": "hard", "description": "IT password reset phish", "correct_actions": (2, 1, 2)},
-    {"difficulty": "hard", "description": "Ransomware threat", "correct_actions": (2, 2, 2)},
-    {"difficulty": "hard", "description": "Fake GDPR notice", "correct_actions": (2, 1, 2)},
-]
-
-# Labels from env
-URGENCY_LABELS = ["General", "Billing", "Security Breach"]
-ROUTING_LABELS = ["AI Auto-Reply", "Tech Support", "Legal"]
-RESOLUTION_LABELS = ["Archive", "Draft Reply", "Escalate to Human"]
-
-from env import EmailTriageEnv
-
-def classify_logic(desc):
+def smart_agent_logic(desc):
     desc = desc.lower()
-    if any(x in desc for x in ["password", "hacked", "breach", "phish"]):
-        if "threat" in desc or "ransom" in desc: return [2, 2, 2]
+    # Hard logic
+    if any(x in desc for x in ["password", "hacked", "breach", "phish", "ransomware", "threat"]):
+        if "threat" in desc or "ransomware" in desc: return [2, 2, 2]
         return [2, 1, 2]
+    # Medium logic (Billing)
+    if any(x in desc for x in ["billing", "refund", "dispute", "invoice"]):
+        return [1, 2, 2]
+    # Easy logic
     return [0, 0, 0]
 
 def run_demo(task):
@@ -37,25 +27,34 @@ def run_demo(task):
         results = []
         total_reward = 0
         
+        if not env._queue:
+            return f"No emails found for {task} difficulty."
+            
         for i, email in enumerate(env._queue):
-            action = classify_logic(email['description'])
+            action = smart_agent_logic(email['description'])
             _, reward, _, _, info = env.step(action)
             total_reward += reward
             
             status = "✅ MATCH" if reward >= 1.0 else "❌ MISMATCH"
-            results.append(f"#{i+1} {email['description'][:30]}...\n   Agent: {URGENCY_LABELS[action[0]]} | Status: {status}")
-
-        score = max(0, total_reward / len(env._queue)) if env._queue else 0
-        return "\n\n".join(results) + f"\n\nFINAL SCORE: {score:.3f}"
+            results.append(
+                f"#{i+1} [{task.upper()}] {email['description']}\n"
+                f"   Agent: {URGENCY_LABELS[action[0]]} | {ROUTING_LABELS[action[1]]} | {RESOLUTION_LABELS[action[2]]}\n"
+                f"   Status: {status}"
+            )
+
+        final_score = max(0.0, total_reward / len(env._queue))
+        return "\n\n".join(results) + f"\n\n--- FINAL EPISODE SCORE: {final_score:.3f} / 1.000 ---"
     except Exception as e:
-        return f"Error: {str(e)}"
+        return f"System Error: {str(e)}"
 
-with gr.Blocks() as demo:
+# UI Layout
+with gr.Blocks(title="Email Gatekeeper AI") as demo:
     gr.Markdown("# 📧 Email Gatekeeper AI")
-    diff = gr.Dropdown(["easy", "hard"], value="easy", label="Difficulty")
-    btn = gr.Button("Analyze Emails")
-    out = gr.Textbox(label="Logs", lines=15)
+    with gr.Row():
+        diff = gr.Dropdown(choices=["easy", "medium", "hard"], value="easy", label="Select Difficulty")
+        btn = gr.Button("Analyze Emails", variant="primary")
     
+    out = gr.Textbox(label="Logs", lines=15)
     btn.click(run_demo, inputs=diff, outputs=out)
 
 app = gr.mount_gradio_app(app, demo, path="/")