Upload 8 files

Dockerfile

@@ -0,0 +1,9 @@
+FROM python:3.10
+
+WORKDIR /app
+
+COPY . .
+
+RUN pip install --no-cache-dir -r requirements.txt
+
+CMD ["python", "app.py"]

app.py

@@ -0,0 +1,55 @@
+import gradio as gr
+from env import EcommerceEnv
+from models import Action
+import random
+
+def simulate():
+    env = EcommerceEnv()
+    obs = env.reset()
+
+    log = ""
+    total_reward = 0
+    steps = 0
+    clicks = 0
+    purchases = 0
+
+    done = False
+
+    while not done:
+        # simple agent
+        action = Action(recommended_product=random.randint(1, 3))
+
+        obs, reward, done, _ = env.step(action)
+
+        steps += 1
+        total_reward += reward.score
+
+        if reward.score == 1.0:
+            purchases += 1
+        elif reward.score > 0:
+            clicks += 1
+
+        log += f"Step {steps} → Recommended: {action.recommended_product} | Reward: {reward.score}\n"
+
+    # Metrics
+    ctr = clicks / steps if steps else 0
+    conversion = purchases / steps if steps else 0
+
+    log += "\n--- SESSION SUMMARY ---\n"
+    log += f"Total Steps: {steps}\n"
+    log += f"Total Reward: {round(total_reward,2)}\n"
+    log += f"Clicks: {clicks}\n"
+    log += f"Purchases: {purchases}\n"
+    log += f"CTR: {round(ctr,2)}\n"
+    log += f"Conversion Rate: {round(conversion,2)}\n"
+
+    return log
+
+
+gr.Interface(
+    fn=simulate,
+    inputs=[],
+    outputs="text",
+    title="🛒 AI E-commerce Recommendation Simulator",
+    description="Simulates how an AI agent recommends products and optimizes user engagement & conversions."
+).launch(share=True)

baseline.py

@@ -0,0 +1,26 @@
+from env import EcommerceEnv
+from models import Action
+import random
+
+env = EcommerceEnv()
+
+total_reward = 0
+episodes = 3
+
+for ep in range(episodes):
+    obs = env.reset()
+    done = False
+
+    print(f"\nEpisode {ep+1}")
+
+    while not done:
+        # Simple agent (random recommendation)
+        action = Action(recommended_product=random.randint(1, 3))
+
+        obs, reward, done, _ = env.step(action)
+
+        print(f"Recommended: {action.recommended_product}, Reward: {reward.score}")
+
+        total_reward += reward.score
+
+print("\nFinal Total Reward:", total_reward)

env.py

@@ -0,0 +1,48 @@
+import random
+from models import Observation, Action, Reward
+
+class EcommerceEnv:
+    def __init__(self):
+        self.sessions = [
+            {"history": [1,2], "target": 3},
+            {"history": [2,3], "target": 1},
+            {"history": [1,3], "target": 2},
+        ]
+        self.max_steps = 5
+
+    def reset(self):
+        self.current_session = random.choice(self.sessions)
+        self.steps = 0
+        self.done = False
+        return self._get_obs()
+
+    def _get_obs(self):
+        return Observation(
+            user_id=1,
+            history=self.current_session["history"]
+        )
+
+    def step(self, action: Action):
+        self.steps += 1
+
+        target = self.current_session["target"]
+
+        # Reward shaping (IMPORTANT)
+        if action.recommended_product == target:
+            reward = 1.0   # purchase
+            self.done = True
+        elif action.recommended_product in self.current_session["history"]:
+            reward = 0.3   # click
+        else:
+            reward = -0.2  # irrelevant
+
+        if self.steps >= self.max_steps:
+            self.done = True
+
+        return self._get_obs(), Reward(score=reward), self.done, {}
+
+    def state(self):
+        return {
+            "steps": self.steps,
+            "target": self.current_session["target"]
+        }

grader.py

@@ -0,0 +1,4 @@
+def grade(total_reward, steps):
+    # normalize score between 0 and 1
+    score = max(0, min(1, total_reward / steps))
+    return score

models.py

@@ -0,0 +1,17 @@
+from pydantic import BaseModel
+from typing import List
+
+# What agent sees
+class Observation(BaseModel):
+    user_id: int
+    history: List[int]  # previously viewed products
+
+
+# What agent does
+class Action(BaseModel):
+    recommended_product: int  # product ID
+
+
+# Reward returned by environment
+class Reward(BaseModel):
+    score: float  # between -1 and 1

tasks.py

@@ -0,0 +1,8 @@
+def easy():
+    return "maximize clicks"
+
+def medium():
+    return "predict purchase"
+
+def hard():
+    return "maximize session reward"