scripts/check_hard_mode.py

#!/usr/bin/env python3
"""
Sanity-check hard-task strategy separation.
"""

import json
import random
import sys
from pathlib import Path

PROJECT_ROOT = Path(__file__).resolve().parents[1]
if str(PROJECT_ROOT) not in sys.path:
    sys.path.insert(0, str(PROJECT_ROOT))

from agents.baseline import run_episode as run_baseline_episode
from agents.random_agent import run_episode as run_random_episode
from env.environment import SynapseXEnvironment
from env.grader import TASK_REGISTRY, TASK_SEEDS, grade
from env.models import Action, ActionPayload


def run_policy(name: str, runner) -> dict:
    env = SynapseXEnvironment(task_config=TASK_REGISTRY["hard"], seed=TASK_SEEDS["hard"])
    actions = runner(env)
    result = grade("hard", actions)
    return {"name": name, **result.model_dump()}


def always_execute(env: SynapseXEnvironment) -> list[ActionPayload]:
    obs = env.reset()
    actions: list[ActionPayload] = []
    for _ in range(env.MAX_STEPS):
        if obs.episode_done:
            break
        active = [task for task in obs.tasks if task.is_active]
        if not active:
            break
        action = Action(action_type="execute", task_id=sorted(active, key=lambda task: task.id)[0].id)
        actions.append(action.model_dump())
        obs = env.step(action).observation
    return actions


def always_delay(env: SynapseXEnvironment) -> list[ActionPayload]:
    obs = env.reset()
    actions: list[ActionPayload] = []
    for _ in range(env.MAX_STEPS):
        if obs.episode_done:
            break
        active = [task for task in obs.tasks if task.is_active]
        if not active:
            break
        action = Action(action_type="delay", task_id=sorted(active, key=lambda task: task.id)[0].id)
        actions.append(action.model_dump())
        obs = env.step(action).observation
    return actions


def main():
    results = [
        run_policy("random", lambda env: run_random_episode(env, rng=random.Random(42))),
        run_policy("always_execute", always_execute),
        run_policy("always_delay", always_delay),
        run_policy("baseline", run_baseline_episode),
    ]

    print(f"Hard-mode strategy comparison (seed={TASK_SEEDS['hard']})")
    for result in results:
        print(f"{result['name']:15s} score={result['score']:.4f} reward_score={result['reward_score']:.4f}")
    print()
    print(json.dumps(results, indent=2))


if __name__ == "__main__":
    main()