import numpy as np, gymnasium as gym, ogbench
env = gym.make("cube-single-v0")  # goal-conditioned, multiple tasks
u = env.unwrapped
print("task_infos count:", len(getattr(u,"task_infos",[])) )
for i,ti in enumerate(getattr(u,"task_infos",[])[:6]):
    print("  task",i, {k:ti[k] for k in ti if k in ("task_name","task_id")})
ob, info = env.reset(options=dict(task_id=1, render_goal=False))
print("reset ob shape", np.asarray(ob).shape, "info keys", list(info.keys()))
goal = info.get("goal");
print("goal shape", None if goal is None else np.asarray(goal).shape)
succ=0; steps=0
for t in range(1001):
    a = env.action_space.sample()
    ob, r, term, trunc, info = env.step(a)
    steps+=1
    if info.get("success"): succ=1
    if term or trunc: break
print("episode: steps",steps,"reward_last",float(r),"term",term,"trunc",trunc,"success_seen",succ,"info_keys",list(info.keys()))
# singletask reward env
env2=gym.make("cube-single-singletask-task2-v0"); o,i=env2.reset()
o,r,te,tr,i=env2.step(env2.action_space.sample())
print("singletask step reward",float(r),"info",list(i.keys()))