"""验三处修复：eta 跟帧数走、队列满返回 429、派发仍然分散。

跑之前要把调度层的 H3_MAX_QUEUE 调成 1，否则填满默认上限 4 要投 16 条、跑 6 分半。
用 22 帧的短任务，路由和限流逻辑与长任务完全一样。
"""
import json, urllib.request, urllib.error
from collections import Counter
from concurrent.futures import ThreadPoolExecutor

GW = "http://127.0.0.1:8000"
CASES = [
    ("832x480x124f 4步（基准，实测 767s）", {"width": 832, "height": 480, "frames": 124, "steps": 4}),
    ("1280x736x56f 4步（实测 736s）", {"width": 1280, "height": 736, "frames": 56, "steps": 4}),
    ("832x480x22f 4步（实测 96s）", {"width": 832, "height": 480, "frames": 22, "steps": 4}),
]
SPEC = {"prompt": "一只橘猫趴在洒满阳光的窗台上，慢慢眨眼", "turbo": True}


def post(body):
    req = urllib.request.Request(GW + "/v1/videos", data=json.dumps(body).encode(),
                                 headers={"Content-Type": "application/json"})
    try:
        with urllib.request.urlopen(req, timeout=60) as r:
            return r.status, json.load(r)
    except urllib.error.HTTPError as e:
        return e.code, e.read().decode()


print("=== 1. eta 是否跟帧数走（改之前三档全报 691）")
results = [post(dict(SPEC, seed=200 + i, **spec)) for i, (_, spec) in enumerate(CASES)]
for (name, _), (code, body) in zip(CASES, results):
    print(f"  {name:38} -> eta={body['eta_seconds']}s  {body['backend'][-4:]}")

print("\n=== 2. 队列满是否 429（H3_MAX_QUEUE=1，前 3 条已占 3 张卡）")
with ThreadPoolExecutor(max_workers=4) as pool:
    burst = list(pool.map(lambda i: post(dict(SPEC, seed=300 + i, frames=22, steps=4)), range(4)))
for i, (code, body) in enumerate(burst):
    detail = body if isinstance(body, str) else f"{body['backend'][-4:]} queue_ahead={body['queue_ahead']}"
    print(f"  #{i} HTTP {code}  {detail}")

accepted = [b for c, b in results + burst if c == 202]
print(f"\n=== 3. 派发分布: {dict(Counter(b['backend'][-4:] for b in accepted))}")
print(f"接受 {len(accepted)} 条，拒绝 {len(results) + len(burst) - len(accepted)} 条")