实现 site_interact 加购与购物车校验（Playwright 通道）

探针实测发现 httpx 因 TLS/HTTP2 指纹被 Rakuten 拒认（cookie 有效但
站点不识别 session），site_interact 改为全程 Playwright + storage_state。
add_to_cart 与 verify_cart 已实测可用：从商品页 __INITIAL_STATE__ 抽
purchase 块、调用方补 variant_id 与 choice、POST basketDomain、用 cart
count JSONP API 与 SPA 渲染后的 cart 页校验。enter_checkout 及之后仍
NotImplementedError：Rakuten 对 checkout 要求 session upgrade（重输密码），
是自动 checkout 的硬墙，未实测过墙方案前留接口缝。

SiteInteractor 类持有 Playwright BrowserContext，由 WorkerRunner 实例
持有、container/main 接线、lifespan 管理生命周期。测试侧 23 个用例，
全套件 333 passed，架构边界守住。

附 5 个探针脚本（scripts/probe_*.py），记录实测路径与响应结构。

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

scripts/probe_checkout.py

@@ -0,0 +1,226 @@
+"""Rakuten 下单流程只读探针：摸清 cart / 加购响应 / 确认页结构
+
+只读模式：不发任何写请求。只 GET 现有购物车页（用户可能在历史会话里已有商品），
+分析页面结构、找出 item_id 与"レジに進む"链接的位置。
+
+要跑通完整流程（加购 → 确认页），加 --mutate 参数；加购后尝试自动清理。
+
+用法：
+    .venv/Scripts/python.exe scripts/probe_checkout.py              # 只读
+    .venv/Scripts/python.exe scripts/probe_checkout.py --mutate     # 含加购与清理
+    .venv/Scripts/python.exe scripts/probe_checkout.py --mutate --item-url https://item.rakuten.co.jp/...
+"""
+from __future__ import annotations
+
+import argparse
+import asyncio
+import json
+import re
+import sys
+from pathlib import Path
+
+# 允许 `python scripts/probe_checkout.py` 直接运行
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+import httpx
+
+from app.shared.config import get_settings
+from app.trading.core import auth_site
+
+PROBE_DIR = Path(__file__).resolve().parent.parent / ".probe" / "checkout"
+PROBE_DIR.mkdir(parents=True, exist_ok=True)
+
+# 默认测试商品：一个低单价、稳定的乐天市场直营商品（用户可覆盖）
+# 留空则只跑只读流程；--mutate 必须显式传 --item-url
+DEFAULT_ITEM_URL = ""
+
+MOBILE_UA = auth_site.RAKUTEN_USER_AGENT
+
+
+def load_cookies(state_path: Path) -> list[dict]:
+    state = json.loads(state_path.read_text(encoding="utf-8"))
+    return state.get("cookies", [])
+
+
+def build_client(cookies: list[dict]) -> httpx.AsyncClient:
+    client = httpx.AsyncClient(
+        headers=auth_site.PROFILES["rakuten"].headers(),
+        timeout=30.0,
+        follow_redirects=True,
+        http2=True,
+    )
+    for cookie in cookies:
+        name = cookie.get("name")
+        value = cookie.get("value")
+        if not name or value is None:
+            continue
+        client.cookies.set(
+            name, value, domain=cookie.get("domain") or "", path=cookie.get("path") or "/"
+        )
+    return client
+
+
+def save(name: str, content: str | bytes, *, is_bytes: bool = False) -> Path:
+    path = PROBE_DIR / name
+    if is_bytes:
+        path.write_bytes(content)  # type: ignore[arg-type]
+    else:
+        path.write_text(content, encoding="utf-8")  # type: ignore[arg-type]
+    print(f"  saved → {path}")
+    return path
+
+
+async def probe_readonly(client: httpx.AsyncClient) -> None:
+    print("\n=== 步骤 1：GET 购物车页 ===")
+    resp = await client.get(auth_site.RAKUTEN_CART_URL)
+    print(f"  status={resp.status_code} final_url={resp.url}")
+    save("01-cart.html", resp.text)
+    body = resp.text
+
+    logged_out = auth_site.RAKUTEN_LOGGED_OUT_MARKER in body
+    print(f"  logged_in={not logged_out}")
+
+    # 找页面里出现的 item_id（购物车里所有商品）
+    # 常见模式：data-item-id="..."、name="item_id" value="..."、/item/.../ 等
+    item_ids = set(re.findall(r'"item_id"\s*[:=]\s*"?(\d+)', body))
+    print(f"  cart item_ids={sorted(item_ids)}")
+
+    # 找 "レジに進む" 链接
+    checkout_links = re.findall(
+        r'href=["\']([^"\']*checkout[^"\']*)["\']', body, re.IGNORECASE
+    )
+    checkout_links += re.findall(
+        r'href=["\']([^"\']*step\.rakuten[^"\']*)["\']', body, re.IGNORECASE
+    )
+    print(f"  checkout links={sorted(set(checkout_links))[:5]}")
+
+    # 如果有 checkout 链接，尝试 GET
+    if checkout_links:
+        target = checkout_links[0]
+        if not target.startswith("http"):
+            target = "https://sp.cart.step.rakuten.co.jp" + target
+        print(f"\n=== 步骤 2：GET checkout 链接 {target} ===")
+        resp2 = await client.get(target)
+        print(f"  status={resp2.status_code} final_url={resp2.url}")
+        save("02-checkout.html", resp2.text)
+    else:
+        print("\n  购物车里没有可结账商品（或链接结构变化），跳过 checkout 探测")
+
+
+async def probe_with_mutation(client: httpx.AsyncClient, item_url: str) -> None:
+    """完整跑一遍：加购 → 校验 → 进确认页。完成后尝试清理"""
+    from app.scraping.parsers.item import parse_item_detail
+    from app.scraping.parsers.state import require_state_marker
+    import app.scraping.parsers.state as state_mod
+
+    print(f"\n=== 加载商品 purchase 块：{item_url} ===")
+    resp = await client.get(item_url)
+    print(f"  item page status={resp.status_code}")
+    save("00-item-page.html", resp.text)
+
+    # 抽 __INITIAL_STATE__
+    m = re.search(
+        r'<script[^>]*>window\.\_\_INITIAL_STATE\_\_\s*=\s*(\{.*?\});?\s*</script>',
+        resp.text, re.DOTALL,
+    )
+    if not m:
+        # 备用：脚本不一定有 window. 前缀
+        m = re.search(
+            r'\_\_INITIAL\_STATE\_\_\s*[:=]\s*(\{.*?\})\s*[;<]',
+            resp.text, re.DOTALL,
+        )
+    if not m:
+        print("  ✗ 抽不出 __INITIAL_STATE__，可能不是手机版模板或反爬被触发")
+        return
+
+    state = json.loads(m.group(1))
+    save("00-item-state.json", json.dumps(state, ensure_ascii=False, indent=2))
+
+    shop_code_match = re.search(r"/([^/]+)/([^/]+)/?$", item_url.rstrip("/"))
+    shop_code = shop_code_match.group(1) if shop_code_match else ""
+    detail = parse_item_detail(state, item_url=item_url, shop_code=shop_code, include_sku_variants=True)
+    pu = detail.purchase
+    print(f"  cart_url={pu.cart_url}")
+    print(f"  form_fields={pu.form_fields}")
+    print(f"  quantity_field={pu.quantity_field} variant_field={pu.variant_field} options_field={pu.options_field}")
+
+    if not pu.cart_url:
+        print("  ✗ basketDomain 为空，不能加购")
+        return
+
+    # 组装加购表单
+    payload = dict(pu.form_fields)
+    payload[pu.quantity_field or "units"] = "1"
+    # 单一库存商品 variant_id 已在 form_fields 里；多规格需调用方传
+
+    print(f"\n=== POST 加购 → {pu.cart_url} ===")
+    print(f"  payload={payload}")
+    resp = await client.post(pu.cart_url, data=payload)
+    print(f"  status={resp.status_code} final_url={resp.url}")
+    save("03-cart-add-response.html", resp.text)
+
+    print("\n=== GET 购物车页（校验）===")
+    resp = await client.get(auth_site.RAKUTEN_CART_URL)
+    print(f"  status={resp.status_code}")
+    save("04-cart-after-add.html", resp.text)
+    item_id = pu.form_fields.get("item_id", "")
+    found = item_id and item_id in resp.text
+    print(f"  item_id={item_id} in_cart={found}")
+
+    # 找 レジに進む 链接
+    checkout_links = re.findall(
+        r'href=["\']([^"\']*checkout[^"\']*)["\']', resp.text, re.IGNORECASE
+    )
+    if checkout_links:
+        target = checkout_links[0]
+        if not target.startswith("http"):
+            target = "https://sp.cart.step.rakuten.co.jp" + target
+        print(f"\n=== GET 确认页 {target} ===")
+        resp = await client.get(target)
+        print(f"  status={resp.status_code} final_url={resp.url}")
+        save("05-checkout-confirm.html", resp.text)
+
+        # 在确认页里找金额
+        amounts = re.findall(r"[¥￥]?\s*([\d,]+)\s*円", resp.text)
+        print(f"  金额候选（前 5）={amounts[:5]}")
+
+    print("\n=== 清理：尝试从购物车删除 ===")
+    # 找删除链接
+    del_matches = re.findall(
+        r'href=["\']([^"\']*(?:delete|del|remove)[^"\']*)["\']',
+        resp.text, re.IGNORECASE,
+    )
+    if del_matches:
+        print(f"  delete 链接={del_matches[:3]}")
+        # 真实删除需 POST + 各种 token，留给用户手动清
+
+
+async def main() -> int:
+    parser = argparse.ArgumentParser(description="Rakuten 下单流程探针")
+    parser.add_argument("--mutate", action="store_true", help="包含加购写操作（默认只读）")
+    parser.add_argument("--item-url", default=DEFAULT_ITEM_URL, help="测试商品 URL")
+    args = parser.parse_args()
+
+    settings = get_settings()
+    state_path = settings.auth_state_path / "rakuten_state.json"
+    if not state_path.exists():
+        print(f"✗ 找不到登录态文件：{state_path}")
+        return 1
+
+    cookies = load_cookies(state_path)
+    print(f"加载 {len(cookies)} 条 cookie 自 {state_path}")
+
+    async with build_client(cookies) as client:
+        await probe_readonly(client)
+        if args.mutate:
+            if not args.item_url:
+                print("\n✗ --mutate 需要搭配 --item-url")
+                return 1
+            await probe_with_mutation(client, args.item_url)
+
+    print(f"\n探针输出目录：{PROBE_DIR}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(asyncio.run(main()))

scripts/probe_checkout_playwright.py

@@ -0,0 +1,179 @@
+"""Rakuten 下单流程探针（Playwright 版）
+
+httpx 因 TLS/HTTP2 指纹问题被 Rakuten 拒认（cookie 有效但站点不识别 session）。
+这个探针改用 Playwright（headless Chromium + storage_state）跑 cart → 加购 → 校验
+→ 确认页，把每步 SPA 渲染后的 HTML 落到 .probe/checkout/，作为 site_interact 实现
+的事实依据。
+
+加购是写操作。默认商品是 fixture 里的小额商品（fafachai/10000033，2190 円），
+跑完会尝试从购物车删除。可用 --item-url 覆盖。
+
+用法：
+    .venv/Scripts/python.exe scripts/probe_checkout_playwright.py
+    .venv/Scripts/python.exe scripts/probe_checkout_playwright.py --item-url https://item.rakuten.co.jp/...
+    .venv/Scripts/python.exe scripts/probe_checkout_playwright.py --headful  # 调试时看浏览器
+"""
+from __future__ import annotations
+
+import argparse
+import asyncio
+import json
+import re
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from app.shared.config import get_settings  # noqa: E402
+from app.trading.core import auth_site  # noqa: E402
+
+PROBE_DIR = Path(__file__).resolve().parent.parent / ".probe" / "checkout"
+PROBE_DIR.mkdir(parents=True, exist_ok=True)
+
+# 默认测试商品：来自 tests/fixtures/item_state.json，2190 円小额商品
+DEFAULT_ITEM_URL = "https://item.rakuten.co.jp/fafachai/10000033/"
+
+
+def save(name: str, content: str) -> Path:
+    path = PROBE_DIR / name
+    path.write_text(content, encoding="utf-8")
+    print(f"  saved -> {path} ({len(content)} bytes)")
+    return path
+
+
+async def run(item_url: str, *, headful: bool) -> int:
+    from playwright.async_api import async_playwright
+
+    settings = get_settings()
+    state_path = settings.auth_state_path / "rakuten_state.json"
+    if not state_path.exists():
+        print(f"找不到登录态文件：{state_path}")
+        return 1
+
+    print(f"加载 storage_state: {state_path}")
+
+    async with async_playwright() as pw:
+        browser = await pw.chromium.launch(
+            headless=not headful,
+            channel=settings.browser_channel or None,
+            args=["--no-first-run", "--disable-blink-features=AutomationControlled"],
+        )
+        context = await browser.new_context(
+            storage_state=state_path,
+            user_agent=auth_site.RAKUTEN_USER_AGENT,
+            locale="ja-JP",
+            timezone_id="Asia/Tokyo",
+            viewport={"width": 390, "height": 844},
+            is_mobile=True,
+            has_touch=True,
+        )
+        page = await context.new_page()
+
+        # 抓所有 XHR，便于看到 SPA 调用了哪些 API
+        api_calls: list[tuple[int, str]] = []
+
+        def on_response(r):
+            url = str(r.url)
+            if any(kw in url for kw in ["cart", "step", "checkout", "basket", "member", "order"]):
+                if "r.r10s.jp" not in url:  # 排除静态资源
+                    api_calls.append((r.status, url))
+
+        page.on("response", on_response)
+
+        print("\n=== 步骤 1：先去 myrakuten 触发完整 session 建立 ===")
+        await page.goto("https://www.rakuten.co.jp/", wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_timeout(1500)
+
+        print("\n=== 步骤 2：商品详情页 ===")
+        await page.goto(item_url, wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_timeout(2500)
+        item_html = await page.content()
+        save("00-item-page.html", item_html)
+
+        # 从 __INITIAL_STATE__ 抽 purchase 块
+        m = re.search(r"window\.__INITIAL_STATE__\s*=\s*(.+?);\s*window\.", item_html, re.DOTALL)
+        if not m:
+            print("  ✗ 抽不出 __INITIAL_STATE__（可能 PC 模板或被反爬）")
+        else:
+            try:
+                state = json.loads(m.group(1))
+                pu = state.get("purchase", {}).get("sellType", {}).get("normalPurchase", {})
+                print(f"  basketDomain: {pu.get('basketDomain')}")
+                print(f"  minPrice: {pu.get('minPrice')}")
+                save("00-item-state.json", json.dumps(state, ensure_ascii=False, indent=2))
+            except json.JSONDecodeError as exc:
+                print(f"  state JSON 解析失败：{exc}")
+
+        print("\n=== 步骤 3：点 '買い物かごに入れる' ===")
+        # 商品页直接点按钮；fallback 是直接 POST basketDomain
+        btn = page.locator('input[type="submit"][value*="買い物かご"], button:has-text("買い物かごに入れる")').first
+        try:
+            await btn.wait_for(state="visible", timeout=5000)
+            print(f"  找到加购按钮，点击")
+            await btn.click()
+            await page.wait_for_timeout(3000)
+            after_add_url = page.url
+            print(f"  点击后落地：{after_add_url}")
+            after_add_html = await page.content()
+            save("03-cart-add-landing.html", after_add_html)
+        except Exception as exc:
+            print(f"  没找到加购按钮或点击失败：{type(exc).__name__}: {exc}")
+            # 直接 POST basketDomain
+            print("  fallback：直接构造表单 POST 到 basketDomain")
+            # 这块留给 site_interact 实现，探针只看页面行为
+
+        print("\n=== 步骤 4：访问购物车页 ===")
+        await page.goto(auth_site.RAKUTEN_CART_URL, wait_until="networkidle", timeout=30_000)
+        await page.wait_for_timeout(3000)
+        cart_html = await page.content()
+        save("04-cart-after-add.html", cart_html)
+        print(f"  final_url={page.url} body_len={len(cart_html)}")
+
+        # 检查 SPA 状态
+        for kw in ["レジに進む", "注文手続き", "小計", "商品の金額", "itemId", "shopName"]:
+            print(f"  contains {kw!r}: {kw in cart_html}")
+        m = re.search(r'"isLoggedIn"\s*:\s*(true|false)', cart_html)
+        if m:
+            print(f"  isLoggedIn={m.group(1)}")
+
+        # 抓 レジ / checkout 链接
+        checkout_links = re.findall(r'href=["\']([^"\']*(?:checkout|step1|order)[^"\']*)["\']', cart_html, re.IGNORECASE)
+        print(f"  checkout links: {sorted(set(checkout_links))[:3]}")
+
+        print("\n=== 步骤 5：捕获 SPA 调用的 API ===")
+        for status, url in api_calls[-30:]:
+            print(f"  {status} {url[:140]}")
+
+        # 不在探针里跑 checkout——避免真的进入付款流程
+        # 真实 site_interact 实现时，需要点 レジに進む 后看跳转
+
+        print("\n=== 步骤 6：尝试清理购物车 ===")
+        # 找删除按钮
+        try:
+            del_btn = page.locator('a:has-text("削除"), button:has-text("削除"), [data-testid*="delete"]').first
+            if await del_btn.count() > 0:
+                print("  找到削除按钮，点击")
+                await del_btn.click()
+                await page.wait_for_timeout(2000)
+                print(f"  清理后 url={page.url}")
+            else:
+                print("  没找到削除按钮（购物车可能本就为空或按钮选择器要调）")
+        except Exception as exc:
+            print(f"  清理失败：{type(exc).__name__}: {exc}")
+
+        await browser.close()
+
+    print(f"\n探针输出目录：{PROBE_DIR}")
+    return 0
+
+
+async def main() -> int:
+    parser = argparse.ArgumentParser()
+    parser.add_argument("--item-url", default=DEFAULT_ITEM_URL)
+    parser.add_argument("--headful", action="store_true")
+    args = parser.parse_args()
+    return await run(args.item_url, headful=args.headful)
+
+
+if __name__ == "__main__":
+    raise SystemExit(asyncio.run(main()))

scripts/probe_purchase_block.py

@@ -0,0 +1,260 @@
+"""Rakuten 加购探针：验证抓取端 purchase 块字段 + 官方旗舰店是否可加购
+
+要做的事：
+1. 从搜索结果取若干商品 URL（普通 + 旗舰店）
+2. 用 Playwright 打开商品页，抽 __INITIAL_STATE__.purchase（与抓取端同样的逻辑）
+3. 用同一份 cookie POST basketDomain，看响应
+4. 对比抓取端 contract 是否充分、官方店是否走同一端点
+
+输出：.probe/checkout/probe-purchase-block.txt
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import re
+import sys
+from pathlib import Path
+from urllib.parse import urlparse
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from app.shared.config import get_settings  # noqa: E402
+from app.trading.core import auth_site  # noqa: E402
+
+PROBE_DIR = Path(__file__).resolve().parent.parent / ".probe" / "checkout"
+PROBE_DIR.mkdir(parents=True, exist_ok=True)
+
+# 测试矩阵：
+# - 普通市场商品（来自搜索）
+# - 楽天ブックス（books.rakuten.co.jp）
+# - 楽天ブランドアベニュー（brandavenue.rakuten.co.jp）
+# - ビックカメラ（biccamera.rakuten.co.jp）
+NORMAL_URLS = [
+    "https://item.rakuten.co.jp/livingut/uni437951",
+    "https://item.rakuten.co.jp/waabbit/088-4p",
+]
+BOOKS_URLS = [
+    "https://books.rakuten.co.jp/rb/17427435/",  # 占位：实际跑时换当前有库存的
+]
+# brandavenue / biccamera 占位（探针跑时若搜索命中会自动用上）
+
+
+def save(name: str, content: str) -> Path:
+    path = PROBE_DIR / name
+    path.write_text(content, encoding="utf-8")
+    return path
+
+
+def parse_state(html: str) -> dict | None:
+    m = re.search(r"window\.__INITIAL_STATE__\s*=\s*(.+?);\s*window\.", html, re.DOTALL)
+    if not m:
+        return None
+    try:
+        return json.loads(m.group(1))
+    except json.JSONDecodeError:
+        return None
+
+
+def extract_purchase_fields(state: dict) -> dict:
+    """与 scraping/parsers/item.py _purchase_info 相同的契约"""
+    purchase = state.get("purchase") or {}
+    sell_type = (purchase.get("sellType") or {}).get("normalPurchase") or {}
+    raw_sku = purchase.get("sku") or {}
+    item = state.get("item") or {}
+    shop = (state.get("shop") or {}).get("information") or {}
+
+    basket_domain = sell_type.get("basketDomain") or ""
+    inventory_type = raw_sku.get("inventoryType")
+    inventory_flag = "2" if inventory_type == "multiple" else "1"
+    shop_id = shop.get("shopId")
+    item_id = item.get("itemId")
+    variant_id = item.get("variantId")
+
+    form_fields: dict[str, str] = {
+        "shop_bid": str(shop_id) if shop_id else "",
+        "item_id": str(item_id) if item_id else "",
+        "inventory_flag": inventory_flag,
+        "__event": "ES01_003_001",
+    }
+    if inventory_flag == "1" and variant_id:
+        form_fields["variant_id"] = str(variant_id)
+
+    return {
+        "basket_domain": basket_domain.replace("\\u002F", "/"),
+        "form_fields": form_fields,
+        "quantity_field": "units",
+        "variant_field": "variant_id",
+        "options_field": "choice" if (purchase.get("information") or {}).get("options") else "",
+        "min_price": sell_type.get("minPrice"),
+        "purchase_condition": sell_type.get("purchaseCondition"),
+        "shop_name": shop.get("shopName"),
+        "item_name": item.get("itemName"),
+    }
+
+
+async def probe_one(page, context, item_url: str, log: list[str]) -> bool:
+    log.append(f"\n=== {item_url} ===")
+    try:
+        await page.goto(item_url, wait_until="domcontentloaded", timeout=30_000)
+    except Exception as exc:
+        log.append(f"  goto FAIL: {exc}")
+        return False
+
+    try:
+        await page.wait_for_function(
+            "() => window.__INITIAL_STATE__ && window.__INITIAL_STATE__.purchase",
+            timeout=10_000,
+        )
+    except Exception:
+        log.append("  no __INITIAL_STATE__.purchase within 10s")
+        return False
+
+    html = await page.content()
+    state = parse_state(html)
+    if not state:
+        log.append("  state parse failed")
+        return False
+
+    fields = extract_purchase_fields(state)
+    log.append(f"  basket_domain: {fields['basket_domain']}")
+    log.append(f"  form_fields: {fields['form_fields']}")
+    log.append(f"  shop: {fields['shop_name']}")
+    log.append(f"  item: {(fields['item_name'] or '')[:60]}")
+    log.append(f"  price: {fields['min_price']}, condition: {fields['purchase_condition']}")
+
+    # 跳过卖完的
+    if fields["purchase_condition"] != "enabled":
+        log.append(f"  SKIP: condition={fields['purchase_condition']}")
+        return False
+
+    payload = dict(fields["form_fields"])
+    payload[fields["quantity_field"]] = "1"
+
+    # POST 加购
+    resp = await context.request.post(
+        fields["basket_domain"],
+        form=payload,
+        max_redirects=5,
+        headers={"Referer": item_url, "Origin": "https://item.rakuten.co.jp"},
+    )
+    log.append(f"  POST add: final_status={resp.status} url={resp.url}")
+    body = await resp.text()
+    save(
+        f"add-{urlparse(item_url).path.replace('/', '_')}.html",
+        body,
+    )
+
+    if "/error" in str(resp.url):
+        log.append(f"  -> ERROR page (item may be stale or fields wrong)")
+        return False
+    # 成功标志
+    success_markers = ["レジに進む", "買い物かごに追加しました", "数量", "cart"]
+    is_success = any(m in body for m in success_markers)
+    log.append(f"  success_markers_hit={is_success}")
+    return is_success
+
+
+async def main() -> int:
+    settings = get_settings()
+    state_path = settings.auth_state_path / "rakuten_state.json"
+
+    from playwright.async_api import async_playwright
+
+    log: list[str] = []
+
+    async with async_playwright() as pw:
+        browser = await pw.chromium.launch(headless=True, args=["--no-first-run"])
+        context = await browser.new_context(
+            storage_state=state_path,
+            user_agent=auth_site.RAKUTEN_USER_AGENT,
+            locale="ja-JP",
+            timezone_id="Asia/Tokyo",
+            viewport={"width": 390, "height": 844},
+            is_mobile=True,
+            has_touch=True,
+        )
+        page = await context.new_page()
+
+        # 先去 sp.cart 建立 session
+        await page.goto(auth_site.RAKUTEN_CART_URL, wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_timeout(1500)
+
+        # 1. 搜普通市场商品（取前 2 个）
+        log.append("=== 搜索普通市场商品 ===")
+        await page.goto(
+            "https://search.rakuten.co.jp/search/mall/?max=500&min=100&s=1&p=1&v=2",
+            wait_until="domcontentloaded",
+            timeout=30_000,
+        )
+        await page.wait_for_timeout(2000)
+        normal_urls = re.findall(
+            r"https://item\.rakuten\.co\.jp/[\w-]+/[\w-]+/?",
+            await page.content(),
+        )
+        seen = set()
+        normal_urls = [u for u in normal_urls if not (u in seen or seen.add(u))][:3]
+        log.append(f"候选: {normal_urls}")
+
+        # 2. 搜 Books
+        log.append("\n=== 搜索楽天ブックス ===")
+        await page.goto(
+            "https://search.rakuten.co.jp/search/books/?max=500&min=100&s=1&p=1&v=2",
+            wait_until="domcontentloaded",
+            timeout=30_000,
+        )
+        await page.wait_for_timeout(2000)
+        books_html = await page.content()
+        books_urls = re.findall(
+            r"https://books\.rakuten\.co\.jp/rb/\d+/?",
+            books_html,
+        )
+        books_urls = [u for u in books_urls if not (u in seen or seen.add(u))][:2]
+        log.append(f"候选 books: {books_urls}")
+
+        # 3. 搜 brandavenue
+        log.append("\n=== 搜索楽天ブランドアベニュー ===")
+        await page.goto(
+            "https://search.rakuten.co.jp/search/?sid=brandavenue&max=2000&min=100&s=1&p=1",
+            wait_until="domcontentloaded",
+            timeout=30_000,
+        )
+        await page.wait_for_timeout(2000)
+        bv_urls = re.findall(
+            r"https://brandavenue\.rakuten\.co\.jp/[\w-]+/[\w-]+/?",
+            await page.content(),
+        )
+        bv_urls = [u for u in bv_urls if not (u in seen or seen.add(u))][:2]
+        log.append(f"候选 brandavenue: {bv_urls}")
+
+        # 4. 实际加购测试
+        log.append("\n========== 加购矩阵测试 ==========")
+        matrix = [
+            ("NORMAL", normal_urls),
+            ("BOOKS", books_urls),
+            ("BRANDAVENUE", bv_urls),
+        ]
+        results: dict[str, list[tuple[str, bool]]] = {}
+        for kind, urls in matrix:
+            results[kind] = []
+            for url in urls:
+                ok = await probe_one(page, context, url, log)
+                results[kind].append((url, ok))
+
+        # 5. 收尾
+        log.append("\n========== 汇总 ==========")
+        for kind, lst in results.items():
+            ok_count = sum(1 for _, ok in lst if ok)
+            log.append(f"{kind}: {ok_count}/{len(lst)} 成功")
+            for url, ok in lst:
+                log.append(f"  {'OK' if ok else 'FAIL'} {url}")
+
+        await browser.close()
+
+    save("probe-purchase-block.txt", "\n".join(log))
+    print(f"输出: {PROBE_DIR / 'probe-purchase-block.txt'}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(asyncio.run(main()))

scripts/probe_purchase_block_v2.py

@@ -0,0 +1,250 @@
+"""Rakuten 加购探针 v2：补 variant_id 与必填选项
+
+修复点：
+- 多规格商品（inventory_flag=2）：从 sku.variants[] 选第一个非售罄的，把 variant_id 加进 form
+- 必填选项：从 purchase.information.options[] 选第一个 value，拼成 "名:值" 加进 choice 字段
+- 加 Books 测试（普通书籍应该 inventory_flag=1，单一库存）
+
+目的：验证抓取端 PurchaseInfo 契约 + 调用方选 variant/option 后，能否成功加购。
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import re
+import sys
+from pathlib import Path
+from urllib.parse import urlparse
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from app.shared.config import get_settings  # noqa: E402
+from app.trading.core import auth_site  # noqa: E402
+
+PROBE_DIR = Path(__file__).resolve().parent.parent / ".probe" / "checkout"
+
+
+def save(name: str, content: str) -> Path:
+    path = PROBE_DIR / name
+    path.write_text(content, encoding="utf-8")
+    return path
+
+
+def parse_state(html: str) -> dict | None:
+    m = re.search(r"window\.__INITIAL_STATE__\s*=\s*(.+?);\s*window\.", html, re.DOTALL)
+    if not m:
+        return None
+    try:
+        return json.loads(m.group(1))
+    except json.JSONDecodeError:
+        return None
+
+
+def extract_full(state: dict) -> dict:
+    purchase = state.get("purchase") or {}
+    sell_type = (purchase.get("sellType") or {}).get("normalPurchase") or {}
+    raw_sku = purchase.get("sku") or {}
+    item = state.get("item") or {}
+    shop = (state.get("shop") or {}).get("information") or {}
+    information = purchase.get("information") or {}
+
+    basket_domain = (sell_type.get("basketDomain") or "").replace("\\u002F", "/")
+    inventory_type = raw_sku.get("inventoryType")
+    inventory_flag = "2" if inventory_type == "multiple" else "1"
+    shop_id = shop.get("shopId")
+    item_id = item.get("itemId")
+
+    form_fields: dict[str, str] = {
+        "shop_bid": str(shop_id) if shop_id else "",
+        "item_id": str(item_id) if item_id else "",
+        "inventory_flag": inventory_flag,
+        "__event": "ES01_003_001",
+    }
+
+    # 多规格：挑第一个非售罄 variant
+    chosen_variant = None
+    variants = raw_sku.get("variants") or []
+    if inventory_flag == "2" and variants:
+        for v in variants:
+            if not v.get("isSoldOut"):
+                chosen_variant = v
+                break
+        if chosen_variant is None and variants:
+            chosen_variant = variants[0]
+        if chosen_variant:
+            form_fields["variant_id"] = str(chosen_variant.get("variantId") or chosen_variant.get("id") or "")
+    elif inventory_flag == "1" and item.get("variantId"):
+        form_fields["variant_id"] = str(item.get("variantId"))
+
+    # 必填选项：拼成 "名:値"
+    choice_pairs: list[str] = []
+    options = information.get("options") or []
+    required_options = []
+    for opt in options:
+        if opt.get("isRequired"):
+            required_options.append(opt)
+            values = opt.get("values") or []
+            if values:
+                first = values[0]
+                choice_pairs.append(f"{opt.get('name')}:{first.get('name')}")
+    if choice_pairs:
+        form_fields["choice"] = ",".join(choice_pairs)
+
+    return {
+        "basket_domain": basket_domain,
+        "form_fields": form_fields,
+        "quantity_field": "units",
+        "shop_name": shop.get("shopName"),
+        "item_name": (item.get("itemName") or "")[:80],
+        "min_price": sell_type.get("minPrice"),
+        "purchase_condition": sell_type.get("purchaseCondition"),
+        "inventory_type": inventory_type,
+        "variant_count": len(variants),
+        "chosen_variant": chosen_variant.get("variantId") if chosen_variant else None,
+        "required_options": [o.get("name") for o in required_options],
+        "choice_value": form_fields.get("choice"),
+    }
+
+
+async def probe_one(page, context, item_url: str, log: list[str]) -> bool:
+    log.append(f"\n--- {item_url} ---")
+    try:
+        await page.goto(item_url, wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_function(
+            "() => window.__INITIAL_STATE__ && window.__INITIAL_STATE__.purchase",
+            timeout=10_000,
+        )
+    except Exception as exc:
+        log.append(f"  load fail: {exc}")
+        return False
+
+    state = parse_state(await page.content())
+    if not state:
+        log.append("  state parse fail")
+        return False
+
+    info = extract_full(state)
+    log.append(f"  shop: {info['shop_name']}")
+    log.append(f"  item: {info['item_name']}")
+    log.append(f"  price: {info['min_price']} condition: {info['purchase_condition']} inv: {info['inventory_type']} variants: {info['variant_count']} chosen_variant: {info['chosen_variant']}")
+    log.append(f"  required_options: {info['required_options']} choice: {info['choice_value']}")
+    log.append(f"  basket: {info['basket_domain']}")
+    log.append(f"  form: {info['form_fields']}")
+
+    if info["purchase_condition"] != "enabled":
+        log.append(f"  SKIP condition={info['purchase_condition']}")
+        return False
+    if info["inventory_type"] == "multiple" and not info["chosen_variant"]:
+        log.append("  SKIP: multi-inventory but no variant chosen")
+        return False
+    if info["required_options"] and not info["choice_value"]:
+        log.append("  SKIP: required options but no value")
+        return False
+
+    payload = dict(info["form_fields"])
+    payload[info["quantity_field"]] = "1"
+
+    resp = await context.request.post(
+        info["basket_domain"],
+        form=payload,
+        max_redirects=5,
+        headers={"Referer": item_url, "Origin": "https://item.rakuten.co.jp"},
+    )
+    body = await resp.text()
+    log.append(f"  POST status={resp.status} final_url={resp.url}")
+
+    # 错误页判断
+    if "/error" in str(resp.url):
+        # 抓错误页里的具体提示
+        m_msgs = re.findall(r'>([^<>]{20,200})<', body)
+        msgs = [s.strip() for s in m_msgs if s.strip() and not any(c in s for c in ['(', ')', '=', '{', '.', ':'])][:3]
+        log.append(f"  ERROR page messages: {msgs}")
+        save(f"err-{urlparse(item_url).path.replace('/', '_')}.html", body)
+        return False
+
+    # 成功标志
+    success_markers = ["レジに進む", "買い物かごに追加", "数量", "cart-item", "orderStep"]
+    is_success = any(m in body for m in success_markers)
+    log.append(f"  success_markers_hit={is_success}")
+    save(f"ok-{urlparse(item_url).path.replace('/', '_')}.html", body)
+    return is_success
+
+
+async def main() -> int:
+    settings = get_settings()
+    state_path = settings.auth_state_path / "rakuten_state.json"
+
+    from playwright.async_api import async_playwright
+
+    log: list[str] = []
+
+    async with async_playwright() as pw:
+        browser = await pw.chromium.launch(headless=True, args=["--no-first-run"])
+        context = await browser.new_context(
+            storage_state=state_path,
+            user_agent=auth_site.RAKUTEN_USER_AGENT,
+            locale="ja-JP",
+            timezone_id="Asia/Tokyo",
+            viewport={"width": 390, "height": 844},
+            is_mobile=True,
+            has_touch=True,
+        )
+        page = await context.new_page()
+
+        # 先去 cart 建立 session
+        await page.goto(auth_site.RAKUTEN_CART_URL, wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_timeout(1500)
+
+        # 搜普通市场商品
+        log.append("=== 搜 NORMAL ===")
+        await page.goto(
+            "https://search.rakuten.co.jp/search/mall/?max=300&min=100&s=1&p=1&v=2",
+            wait_until="domcontentloaded",
+            timeout=30_000,
+        )
+        await page.wait_for_timeout(2000)
+        normal_urls = list(dict.fromkeys(
+            re.findall(r"https://item\.rakuten\.co\.jp/[\w-]+/[\w-]+/?", await page.content())
+        ))[:3]
+        log.append(f"normal: {normal_urls}")
+
+        # 搜 Books
+        log.append("\n=== 搜 BOOKS ===")
+        await page.goto(
+            "https://search.rakuten.co.jp/search/books/?max=500&min=100&s=1&p=1&v=2",
+            wait_until="domcontentloaded",
+            timeout=30_000,
+        )
+        await page.wait_for_timeout(2000)
+        books_html = await page.content()
+        books_urls = list(dict.fromkeys(
+            re.findall(r"https://books\.rakuten\.co\.jp/rb/\d+/?", books_html)
+        ))[:3]
+        log.append(f"books: {books_urls}")
+
+        # 加购矩阵
+        log.append("\n========== 测试 ==========")
+        results: dict[str, list[tuple[str, bool]]] = {"NORMAL": [], "BOOKS": []}
+        for url in normal_urls:
+            ok = await probe_one(page, context, url, log)
+            results["NORMAL"].append((url, ok))
+        for url in books_urls:
+            ok = await probe_one(page, context, url, log)
+            results["BOOKS"].append((url, ok))
+
+        # 汇总
+        log.append("\n========== 汇总 ==========")
+        for kind, lst in results.items():
+            ok = sum(1 for _, s in lst if s)
+            log.append(f"{kind}: {ok}/{len(lst)} 成功")
+            for u, s in lst:
+                log.append(f"  {'OK' if s else 'FAIL'} {u}")
+
+        await browser.close()
+
+    save("probe-v2.txt", "\n".join(log))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(asyncio.run(main()))

scripts/probe_search_add.py

@@ -0,0 +1,226 @@
+"""Rakuten 探针：搜低价商品 → 实测 add-to-cart → 跨 cart/checkout
+
+策略（顺序）：
+1. 用 Playwright 打开 search.rakuten.co.jp 限定最低价筛选
+2. 从结果里逐个取商品 URL
+3. 用 Playwright 实际打开商品页，等 "買い物かごに入れる" 按钮
+4. 点按钮（不是手工 POST，确保 form 字段都是当前版本）
+5. 跟到 cart 后让 SPA 渲染够久
+6. 落证据，尝试清理（删除购物车里这件）
+
+避免 httpx（已被 TLS 指纹拦）。全程 Playwright。
+"""
+from __future__ import annotations
+
+import asyncio
+import json
+import re
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
+
+from app.shared.config import get_settings  # noqa: E402
+from app.trading.core import auth_site  # noqa: E402
+
+PROBE_DIR = Path(__file__).resolve().parent.parent / ".probe" / "checkout"
+PROBE_DIR.mkdir(parents=True, exist_ok=True)
+
+
+def save(name: str, content: str) -> Path:
+    path = PROBE_DIR / name
+    path.write_text(content, encoding="utf-8")
+    print(f"  saved -> {path} ({len(content)} bytes)")
+    return path
+
+
+async def collect_item_urls(page, limit: int = 10) -> list[str]:
+    """从搜索结果抽商品 URL（PC 与 SP 都试）"""
+    html = await page.content()
+    # 候选：/shop-code/item-code/、item.rakuten.co.jp、div[data-itemid]
+    urls: list[str] = []
+    for m in re.findall(r'https://item\.rakuten\.co\.jp/[\w-]+/[\w-]+/?', html):
+        if m not in urls:
+            urls.append(m)
+        if len(urls) >= limit:
+            break
+    return urls
+
+
+async def try_add_item(page, item_url: str) -> tuple[bool, str]:
+    """尝试加购一个商品。返回 (是否成功, 落地 URL)。"""
+    print(f"  open item page: {item_url}")
+    try:
+        await page.goto(item_url, wait_until="domcontentloaded", timeout=30_000)
+    except Exception as exc:
+        return False, f"goto failed: {exc}"
+
+    # 等 __INITIAL_STATE__ 出现
+    try:
+        await page.wait_for_function(
+            "() => window.__INITIAL_STATE__ && window.__INITIAL_STATE__.purchase",
+            timeout=10_000,
+        )
+    except Exception:
+        return False, "no __INITIAL_STATE__ (PC 模板/反爬)"
+
+    # 找加购按钮
+    candidates = [
+        'input[type="submit"][value*="買い物かご"]',
+        'input[type="submit"][value*="購入"]',
+        'button:has-text("買い物かご")',
+        'button:has-text("カートに入れる")',
+        '[data-testid*="add-cart"]',
+        'button.add-cart',
+    ]
+    btn = None
+    for sel in candidates:
+        loc = page.locator(sel).first
+        try:
+            await loc.wait_for(state="visible", timeout=2000)
+            btn = loc
+            print(f"  found button via {sel!r}")
+            break
+        except Exception:
+            continue
+    if btn is None:
+        return False, "no add-to-cart button visible"
+
+    # 监听导航
+    try:
+        async with page.expect_navigation(timeout=15_000, wait_until="domcontentloaded"):
+            await btn.click()
+    except Exception as exc:
+        return False, f"click/navigation failed: {exc}"
+
+    final_url = page.url
+    body = await page.content()
+    # 成功标志：URL 含 cart/add-item 或 confirm；或正文含 レジに進む
+    success = (
+        "add-item" in final_url
+        or "cart" in final_url
+        or "レジに進む" in body
+        or "買い物かごに追加" in body
+    )
+    # 失败标志：URL 含 /error
+    if "/error" in final_url:
+        return False, f"error page: {final_url}"
+    return success, final_url
+
+
+async def main() -> int:
+    settings = get_settings()
+    state_path = settings.auth_state_path / "rakuten_state.json"
+    if not state_path.exists():
+        print(f"找不到 {state_path}")
+        return 1
+
+    from playwright.async_api import async_playwright
+
+    async with async_playwright() as pw:
+        browser = await pw.chromium.launch(
+            headless=True,
+            channel=settings.browser_channel or None,
+            args=["--no-first-run", "--disable-blink-features=AutomationControlled"],
+        )
+        context = await browser.new_context(
+            storage_state=state_path,
+            user_agent=auth_site.RAKUTEN_USER_AGENT,
+            locale="ja-JP",
+            timezone_id="Asia/Tokyo",
+            viewport={"width": 390, "height": 844},
+            is_mobile=True,
+            has_touch=True,
+        )
+        page = await context.new_page()
+
+        # 搜索：100~300 円筛选
+        print("=== 步骤 1：搜低价商品 ===")
+        search_url = "https://search.rakuten.co.jp/search/mall/?f=2&fs=1000&max=300&min=100&s=1&p=1"
+        # 备用关键词：邮费便宜 / 免邮的便宜商品
+        await page.goto(search_url, wait_until="domcontentloaded", timeout=30_000)
+        await page.wait_for_timeout(2000)
+        save("search-result.html", await page.content())
+        item_urls = await collect_item_urls(page, limit=10)
+        print(f"  候选商品 {len(item_urls)} 个")
+        for u in item_urls[:5]:
+            print(f"    {u}")
+
+        # 试加购，最多试 5 个
+        chosen: str | None = None
+        for i, url in enumerate(item_urls[:5]):
+            print(f"\n=== 步骤 2.{i+1}：试加购 {url} ===")
+            ok, info = await try_add_item(page, url)
+            print(f"  result: ok={ok} info={info}")
+            if ok:
+                chosen = url
+                save("03-add-success-landing.html", await page.content())
+                break
+            # 失败：保存页面用于复盘
+            save(f"03-add-fail-{i+1}.html", await page.content())
+
+        if not chosen:
+            print("\n✗ 5 个商品全部加购失败。")
+            await browser.close()
+            return 1
+
+        print(f"\n=== 步骤 3：访问 cart，让 SPA 充分渲染 ===")
+        # 先访问 cart
+        await page.goto(auth_site.RAKUTEN_CART_URL, wait_until="domcontentloaded", timeout=30_000)
+        # 等 SPA 把 cart 数据渲染出来：尝试等 "レジに進む" 出现，最多 15 秒
+        try:
+            await page.wait_for_selector(
+                'a:has-text("レジに進む"), button:has-text("レジに進む"), [data-testid*="checkout"]',
+                timeout=15_000,
+            )
+            print("  'レジに進む' 出现了！")
+        except Exception:
+            print("  'レジに進む' 15s 内没出现，继续抓现状")
+
+        cart_html = await page.content()
+        save("04-cart-after-add.html", cart_html)
+        print(f"  cart len={len(cart_html)} url={page.url}")
+
+        # 找 レジに進む href
+        checkout_links = re.findall(r'href=["\']([^"\']*(?:checkout|step1|order/step)[^"\']*)["\']', cart_html, re.IGNORECASE)
+        print(f"  checkout links: {sorted(set(checkout_links))[:3]}")
+
+        # 看 cart count API
+        resp = await context.request.get(
+            "https://cart-api.step.rakuten.co.jp/rms/mall/cart/count/all/jsonp/?sid=1010",
+            headers={"Referer": "https://sp.cart.step.rakuten.co.jp/cart"},
+        )
+        print(f"  count API: status={resp.status} body={(await resp.text())[:200]!r}")
+
+        print("\n=== 步骤 4：尝试清理购物车 ===")
+        # 找削除按钮
+        del_selectors = [
+            'a:has-text("削除")',
+            'button:has-text("削除")',
+            '[data-testid*="delete"]',
+            '[data-testid*="remove"]',
+        ]
+        for sel in del_selectors:
+            loc = page.locator(sel).first
+            try:
+                await loc.wait_for(state="visible", timeout=3000)
+                print(f"  找到删除按钮 {sel!r}，点击")
+                await loc.click()
+                await page.wait_for_timeout(2000)
+                break
+            except Exception:
+                continue
+        else:
+            print("  没找到删除按钮，请手动清理购物车")
+
+        save("05-cart-after-cleanup.html", await page.content())
+
+        await browser.close()
+
+    print(f"\n探针输出目录：{PROBE_DIR}")
+    print(f"测试用商品：{chosen}")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(asyncio.run(main()))