AdamK29/Meta-OpenENV-Hackathon

1import gradio as gr2import asyncio3import matplotlib.pyplot as plt4 5from client import EmailEnvClient6from core.models import EmailAction7 8import ssl9import certifi10 11ssl._create_default_https_context = ssl._create_unverified_context12 13def decide(email):14    email = email.lower()15 16    if "free" in email or "win" in email:17        return "spam"18    if "deadline" in email or "asap" in email:19        return "urgent"20    return "important"21 22 23async def run_episode(task):24    env = EmailEnvClient(base_url="https://adamk29-meta-openenv-hackathon.hf.space")25    await env.__aenter__()26 27    rewards = []28    logs = []29 30    try:31        result = await env.reset(task=task)32 33        step = 034 35        while True:36            step += 137 38            email = result.observation.email_text39            action = decide(email)40 41            result = await env.step(42                EmailAction(action_type="classify", content=action)43            )44 45            reward = result.reward or 0.046            done = result.done47 48            rewards.append(reward)49 50            logs.append(51                f"""52[STEP {step}]53State (Email): {email}54Action: {action}55Reward: {reward:.2f}56Cumulative Reward: {sum(rewards):.2f}57"""58            )59 60            if done:61                break62 63        # Plot reward curve64        fig = plt.figure()65        plt.plot(rewards)66        plt.title("Reward Over Episode")67 68        score = sum(rewards)69 70        return "\n".join(logs), score, fig71 72    finally:73        await env.__aexit__(None, None, None)74 75 76def run(task):77    return asyncio.run(run_episode(task))78 79 80with gr.Blocks() as demo:81    gr.Markdown("# 🤖 RL Email Triage Environment")82    gr.Markdown("Trajectory-based RL interaction")83 84    task = gr.Dropdown(["easy", "medium", "hard"], value="easy")85 86    run_btn = gr.Button("▶ Run Episode")87 88    logs = gr.Textbox(lines=25, label="Trajectory Logs")89    score = gr.Number(label="Total Reward")90    graph = gr.Plot(label="Reward Curve")91 92    run_btn.click(run, inputs=task, outputs=[logs, score, graph])93 94demo.launch()