tasks.py

from environment import CodeReviewEnv
from models import Action

def run_task(task_name: str, agent_func) -> float:
    env = CodeReviewEnv()
    obs = env.reset(task=task_name)
    done = False
    while not done:
        action = agent_func(obs)
        obs, reward, done, info = env.step(action)
    return reward.value