Upload 4 files

README.md

@@ -1,13 +1,21 @@
----
-title: Email Triage Env
-emoji: 💻
-colorFrom: green
-colorTo: green
-sdk: gradio
-sdk_version: 6.10.0
-app_file: app.py
-pinned: false
-short_description: Email Triage AI Environment where an agent classifies emails
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# Email Triage AI Environment
+
+## Problem
+Email overload is a real-world problem. This project simulates an AI agent that classifies emails and generates replies.
+
+## Observation
+- Email text
+
+## Action
+- Label (important / spam / normal)
+- Reply
+
+## Reward
+- Correct label: +0.5
+- Correct reply: +0.5
+
+## Task
+Medium level: classify email and generate reply
+
+## How to Run
+Open baseline.py and press F5

app.py

@@ -0,0 +1,31 @@
+from env import EmailEnv
+
+env = EmailEnv()
+obs = env.reset()
+
+# simple rule-based agent
+def agent(obs):
+    text = obs["email_text"].lower()
+
+    if "meeting" in text:
+        return {
+            "label": "important",
+            "reply": "Sure, I will attend the meeting."
+        }
+    elif "lottery" in text:
+        return {
+            "label": "spam",
+            "reply": "This looks like spam."
+        }
+    else:
+        return {
+            "label": "normal",
+            "reply": "Sounds good!"
+        }
+
+action = agent(obs)
+next_obs, reward, done, _ = env.step(action)
+
+print("Observation:", obs)
+print("Action:", action)
+print("Reward:", reward)

env.py

@@ -0,0 +1,36 @@
+import random
+
+class EmailEnv:
+    def __init__(self):
+        self.emails = [
+            {
+                "text": "Meeting at 5 PM, please attend",
+                "label": "important",
+                "reply": "Sure, I will attend the meeting."
+            },
+            {
+                "text": "Congratulations! You won a lottery",
+                "label": "spam",
+                "reply": "This looks like spam."
+            },
+            {
+                "text": "Lunch tomorrow?",
+                "label": "normal",
+                "reply": "Sounds good!"
+            }
+        ]
+
+    def reset(self):
+        self.current = random.choice(self.emails)
+        return self.current
+
+    def step(self, action):
+        reward = 0
+
+        if action["label"] == self.current["label"]:
+            reward += 0.5
+
+        if action["reply"] == self.current["reply"]:
+            reward += 0.5
+
+        return reward