@whyshu/argent-engine 0.0.0-stage → 0.6.0

package/engine/scripts/wenyao-login.sh

@@ -0,0 +1,34 @@
+#!/bin/bash
+set -euo pipefail
+# 上海文鳐门户「人工过验证 + 保存登录态」入口（2026-09-10）
+#
+# 用法（本机或客户机都可用）：
+#   bash scripts/wenyao-login.sh              # 等 30 分钟，你在弹出的 Chrome 里过滑块 + 点登录
+#   bash scripts/wenyao-login.sh 600          # 等 10 分钟
+#   bash scripts/wenyao-login.sh --check      # 只检查登录态是否还有效（不弹验证）
+#
+# 背景：门户登录用阿里云行为验证码（滑块），合成鼠标过不了（实测 2 次失败）且反复试会触发风控 ⇒
+# 采用"人工过一次 → 复用登录态"：登录成功后把 storage_state + 持久 profile 存到
+# ~/.dsh/argent/（600 权限），之后的发票流程自动化直接复用，不再需要过验证。
+# 凭据来自 ~/.dsh/argent/wenyao.env（600，不进仓库）。
+
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+PY="${WENYAO_PY:-~/.dsh/argent/venv/bin/python3}"
+if [ ! -x "$PY" ]; then
+  PY="$(command -v python3 || true)"
+fi
+if [ -z "$PY" ]; then
+  echo "[wenyao-login] 找不到 python3" >&2
+  exit 1
+fi
+if ! "$PY" -c "import playwright" >/dev/null 2>&1; then
+  echo "[wenyao-login] 当前 python 无 playwright：请用 ~/.dsh/argent/venv/bin/python3 或安装 playwright" >&2
+  echo "               （可用 WENYAO_PY=/path/to/python3 指定）" >&2
+  exit 1
+fi
+if [ ! -f "$HOME/.dsh/argent/wenyao.env" ]; then
+  echo "[wenyao-login] 缺凭据文件 ~/.dsh/argent/wenyao.env（需含 WENYAO_URL/WENYAO_PHONE/WENYAO_PASSWORD，600 权限）" >&2
+  exit 1
+fi
+
+exec "$PY" "$HERE/wenyao_login.py" "$@"

package/engine/scripts/wenyao-template-map.json

@@ -0,0 +1,292 @@
+{
+  "_doc": "上海文鳐《发票模版》列映射真源（2026-09-11）。运营版表头在第 7 行、数据填第 8 行（说明行会被覆盖）。source 取值：field=直接取飞书字段；calc=按算式算；const=固定值；llm=需翻译；image=附件转浮动图片；storemap=店铺→卖家ID。 numeric=true 的列按数字写入（否则门户可能不认文本数字）；货件号/HS/SKU/卖家ID 一律文本（防科学计数法与丢前导零）。",
+  "sheet": "Sheet1",
+  "headerRow": 7,
+  "dataRow": 8,
+  "imageAnchor": "AA8",
+  "clearRowsBelow": 9,
+  "topBlock": {
+    "客户名称": {
+      "cell": "A2",
+      "source": "const",
+      "value": "武汉六禾园"
+    },
+    "下单时间": {
+      "cell": "B2",
+      "source": "const",
+      "value": "",
+      "note": "运营确认：非必须，留空"
+    },
+    "预计进仓时间": {
+      "cell": "C2",
+      "source": "const",
+      "value": "__TODAY__",
+      "format": "date",
+      "note": "运营确认：填当前日期值"
+    },
+    "委托编号": {
+      "cell": "D2",
+      "source": "const",
+      "value": "",
+      "note": "免填"
+    },
+    "起运港": {
+      "cell": "E2",
+      "source": "const",
+      "value": "CNYTN"
+    },
+    "目的港": {
+      "cell": "F2",
+      "source": "const",
+      "value": "USLGB"
+    },
+    "港前仓": {
+      "cell": "G2",
+      "source": "const",
+      "value": "华南仓",
+      "note": "2026-09-11 用户确认以运营填的华南仓为准"
+    },
+    "送货国家": {
+      "cell": "H2",
+      "source": "const",
+      "value": "墨西哥"
+    }
+  },
+  "columns": [
+    {
+      "col": "A",
+      "header": "我司关联单号（无需填写）",
+      "source": "const",
+      "value": ""
+    },
+    {
+      "col": "B",
+      "header": "产品中文品名",
+      "source": "field",
+      "field": "产品名称"
+    },
+    {
+      "col": "C",
+      "header": "产品英文品名",
+      "source": "llm",
+      "field": "产品名称",
+      "note": "中文品名 → 英文（LLM），如 42-Piece Professional Makeup Brush and Sponge Cosmetic Bag - Brown"
+    },
+    {
+      "col": "D",
+      "header": "货件号/唛头",
+      "source": "field",
+      "field": "货件号"
+    },
+    {
+      "col": "E",
+      "header": "箱序",
+      "source": "calc",
+      "expr": "1-{箱数}",
+      "note": "运营批注：支持正整数/逗号/中横线/波浪线"
+    },
+    {
+      "col": "F",
+      "header": "SKU",
+      "source": "field",
+      "field": "SKU"
+    },
+    {
+      "col": "G",
+      "header": "卖家ID",
+      "source": "storemap",
+      "field": "店铺"
+    },
+    {
+      "col": "H",
+      "header": "单箱产品数量",
+      "source": "calc",
+      "expr": "{数量}/{箱数}"
+    },
+    {
+      "col": "I",
+      "header": "总产品数量",
+      "source": "field",
+      "field": "数量",
+      "numeric": true
+    },
+    {
+      "col": "J",
+      "header": "总箱数",
+      "source": "field",
+      "field": "箱数",
+      "note": "lip 表箱数是公式列，读值",
+      "numeric": true
+    },
+    {
+      "col": "K",
+      "header": "单箱总重",
+      "source": "field",
+      "field": "重量",
+      "numeric": true
+    },
+    {
+      "col": "L",
+      "header": "单箱产品净重",
+      "source": "field",
+      "field": "重量",
+      "numeric": true
+    },
+    {
+      "col": "M",
+      "header": "货箱总重",
+      "source": "calc",
+      "expr": "{重量}*{箱数}"
+    },
+    {
+      "col": "N",
+      "header": "货箱长度",
+      "source": "field",
+      "field": "长",
+      "numeric": true
+    },
+    {
+      "col": "O",
+      "header": "货箱宽度",
+      "source": "field",
+      "field": "宽",
+      "numeric": true
+    },
+    {
+      "col": "P",
+      "header": "货箱高度",
+      "source": "field",
+      "field": "高",
+      "numeric": true
+    },
+    {
+      "col": "Q",
+      "header": "单件方数",
+      "source": "calc",
+      "expr": "{长}*{宽}*{高}/1000000",
+      "numberFormat": "0.000000"
+    },
+    {
+      "col": "R",
+      "header": "总方数",
+      "source": "calc",
+      "expr": "({长}*{宽}*{高}/1000000)*{箱数}",
+      "numberFormat": "0.000000"
+    },
+    {
+      "col": "S",
+      "header": "产品材质",
+      "source": "const",
+      "value": "{材质}，{材质英文}",
+      "note": "运营示例为『塑料，plastics』⇒ 中英双语。材质英文走**三级解析**：① per-SKU 字典（<数据目录>/lip-sku-dict.json，键=SKU，优先 confirmed:true 的「材质英文」）→ ② 本映射的 `materialEn` 小词典 → ③ 会话传入（`--material-en` / 行内 `材质英文`，政策A 允许 LLM 译）。三级都没命中 ⇒ 填报**阻断**（不落盘）并指路「请翻译成英文后用 --material-en 重跑」；③ 命中时真跑会写回字典标 `confirmed:false`。⛔ HS 编码那条人工红线只管 HS。",
+      "perRow": true,
+      "perRowNote": "本列**随行变化**（每票材质不同，与货件号同级）⇒ 影子对比器按 `perRow` 放行、不再要求与黄金样本逐字相同；该列的判据改由填报器的三级解析 + 阻断兜住（⛔ 不是「英文为空就不管」）。"
+    },
+    {
+      "col": "T",
+      "header": "产品用途",
+      "source": "const",
+      "value": "家用"
+    },
+    {
+      "col": "U",
+      "header": "产品品牌",
+      "source": "const",
+      "value": "无"
+    },
+    {
+      "col": "V",
+      "header": "产品型号",
+      "source": "const",
+      "value": "无"
+    },
+    {
+      "col": "W",
+      "header": "产品申报单价",
+      "source": "const",
+      "value": 7
+    },
+    {
+      "col": "X",
+      "header": "产品申报总价",
+      "source": "calc",
+      "expr": "7*{数量}"
+    },
+    {
+      "col": "Y",
+      "header": "申报币种",
+      "source": "const",
+      "value": "USD"
+    },
+    {
+      "col": "Z",
+      "header": "产品海关编码",
+      "source": "field",
+      "field": "HS编码"
+    },
+    {
+      "col": "AA",
+      "header": "产品图片",
+      "source": "image",
+      "field": "产品图",
+      "note": "lip 附件 → 下载 → 以浮动图片插入本格（锚点 AA8 = 数据行「*产品图片」列）。**必须缩放到装得进单元格**：AA 列宽 13 ≈ 96px、行高 40pt ≈ 53px，原先写死 120px 会溢出（用户 2026-09-11 实测反馈）。",
+      "anchor": "AA8"
+    },
+    {
+      "col": "AB",
+      "header": "是否带电",
+      "source": "field",
+      "field": "是否带电",
+      "default": "否"
+    },
+    {
+      "col": "AC",
+      "header": "是否带磁",
+      "source": "field",
+      "field": "是否带磁",
+      "default": "否"
+    },
+    {
+      "col": "AD",
+      "header": "危险品",
+      "source": "const",
+      "value": "否"
+    },
+    {
+      "col": "AE",
+      "header": "是否单证报关",
+      "source": "const",
+      "value": "否"
+    },
+    {
+      "col": "AF",
+      "header": "是否投保",
+      "source": "const",
+      "value": "否",
+      "note": "2026-09-11 用户确认填否；门户模板有此列，运营版没有"
+    },
+    {
+      "col": "AG",
+      "header": "仓库编码",
+      "source": "const",
+      "value": "MXRC03",
+      "note": "2026-09-11 用户确认 MXRC03（门户显示 + 鑫中航 mapping 一致；运营说明里的 MXCR03 视为笔误）"
+    }
+  ],
+  "materialEn": {
+    "塑料": "plastics",
+    "牛津布": "oxford fabric",
+    "不锈钢": "stainless steel",
+    "硅胶": "silicone",
+    "玻璃": "glass",
+    "纸": "paper",
+    "木": "wood",
+    "金属": "metal"
+  },
+  "knownDivergence": {
+    "是否投保": "门户自带《跨境电商-下单模板》有『*是否投保』列，运营版表头没有；若导入时门户要求，需确认以哪版为准（用户 2026-09-11 确认：填『否』）",
+    "仓库编码列位": "门户模板：是否投保在 AF、仓库编码在 AG（运营版没有是否投保、仓库编码在 AF）⇒ 以门户模板为准"
+  },
+  "templateSource": "门户下载《跨境电商-下单模板.xlsx》（~/.dsh/argent/wenyao-downloads/）；运营版《上海文鳐发票模版.xlsx》仅作填法说明用",
+  "_storeIdsNote": "店铺→卖家ID 已**移出本文件**（2026-09-19）：真源唯一为 `scripts/ml-store-ids.json`，由 `scripts/ml_store_ids.py` 解析。原先三处各存一份（本文件裸键 '1店' / invoice-mapping.json 短名带后缀 / lip-po-mapping.json 完整店名），鑫中航那条线因匹配不上而静默回退成紫鸟 store_id，6 张发票写错。"
+}

package/engine/scripts/wenyao_attach.py

@@ -0,0 +1,131 @@
+#!/usr/bin/env python3
+"""wenyao_attach.py — 连到「长驻文鳐浏览器」(CDP) 做预填 / 登录态检查 / 截图（2026-09-10）
+
+配合 wenyao-browser.sh：Chrome 以独立 profile + `--remote-debugging-port=9333` 长驻，
+本脚本用 Playwright `connect_over_cdp` **接到那个浏览器**（不新开、不关它），因此：
+  · 你人工过一次滑块后，登录态在浏览器与 profile 里都留着，可**重复使用**；
+  · 后续自动化直接驱动同一个窗口，你能看着它操作；出问题也能人工接手。
+
+只读：不导入订单、不导出文件、不改对方任何业务数据。
+用法：
+  python3 wenyao_attach.py --prefill     # 预填手机号/密码（便于人工只滑滑块）
+  python3 wenyao_attach.py --status       # 打印当前 URL / 是否已登录 / 页面顶部文案
+  python3 wenyao_attach.py --shot a.png   # 截图（供人工/AI 查看）
+"""
+import argparse
+import json
+import os
+import sys
+from pathlib import Path
+from playwright.sync_api import sync_playwright
+
+HOME = Path.home()
+# 数据目录：**唯一判据**在 `scripts/lib/argent_datadir.py`（三级链，与 JS 内核同语义）
+sys.path.insert(0, str(Path(__file__).resolve().parent / 'lib'))
+from argent_datadir import resolve_data_dir  # noqa: E402  —— 勿在此复制判据
+DATA = Path(resolve_data_dir()['dir'])
+CDP = os.environ.get("WENYAO_CDP", "http://127.0.0.1:9333")
+
+
+def load_env():
+    env = {}
+    f = DATA / "wenyao.env"
+    if f.exists():
+        for line in f.read_text(encoding="utf-8").splitlines():
+            line = line.strip()
+            if line and not line.startswith("#") and "=" in line:
+                k, v = line.split("=", 1)
+                env[k.strip()] = v.strip()
+    env.setdefault("WENYAO_URL", "https://my.ffl-tech.com")
+    return env
+
+
+def pick_page(browser, url):
+    pages = [p for c in browser.contexts for p in c.pages]
+    for p in pages:
+        if "ffl-tech" in (p.url or ""):
+            return p
+    return pages[0] if pages else None
+
+
+def body(page):
+    try:
+        return page.evaluate("() => (document.body.innerText||'').replace(/\\s+/g,' ')")
+    except Exception:
+        return ""
+
+
+def logged_in(page):
+    t = body(page)
+    if "请按住滑块" in t or "拖动到最右" in t:
+        return False
+    if "账号密码登录" in t and "短信验证登录" in t:
+        return False
+    return any(k in t for k in ("退出登录", "个人中心", "导入订单", "订单管理", "我的订单", "头程", "发票"))
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--prefill", action="store_true")
+    ap.add_argument("--status", action="store_true")
+    ap.add_argument("--shot")
+    ap.add_argument("--entries", action="store_true", help="打印页面上的入口文本（找模板下载入口）")
+    a = ap.parse_args()
+    env = load_env()
+    out = {"cdp": CDP}
+    with sync_playwright() as p:
+        try:
+            browser = p.chromium.connect_over_cdp(CDP, timeout=8000)
+        except Exception as e:
+            print(json.dumps({"ok": False, "why": f"连不上 CDP({CDP})：{str(e)[:120]}",
+                              "hint": "先跑 bash scripts/wenyao-browser.sh start"}, ensure_ascii=False))
+            return 1
+        page = pick_page(browser, env["WENYAO_URL"])
+        if page is None:
+            c = browser.contexts[0] if browser.contexts else browser.new_context()
+            page = c.new_page()
+            page.goto(env["WENYAO_URL"], wait_until="domcontentloaded", timeout=45000)
+            page.wait_for_timeout(3000)
+        out["url"] = page.url
+        if a.prefill:
+            # 门户是 Vue SPA：CDP 连上时表单可能还没渲染 ⇒ 先等密码框出现（最多 25s，重试 3 轮）
+            for _ in range(3):
+                try:
+                    page.wait_for_selector("input[type=password]", timeout=25000)
+                    break
+                except Exception:
+                    page.wait_for_timeout(1500)
+            filled = {"user": False, "pwd": False}
+            for el in page.query_selector_all("input"):
+                try:
+                    if not el.is_visible():
+                        continue
+                    t = (el.get_attribute("type") or "text").lower()
+                    if t == "password" and not filled["pwd"]:
+                        el.fill(env.get("WENYAO_PASSWORD", "")); filled["pwd"] = True
+                    elif t in ("text", "tel", "number", "") and not filled["user"]:
+                        el.fill(env.get("WENYAO_PHONE", "")); filled["user"] = True
+                except Exception:
+                    continue
+            out["prefill"] = filled
+        if a.shot:
+            page.screenshot(path=a.shot, full_page=True)
+            out["shot"] = a.shot
+        if a.entries:
+            try:
+                ents = page.evaluate("""() => Array.from(document.querySelectorAll('a,button,li,span,div'))
+                  .map(e=>(e.innerText||'').trim()).filter(t=>t&&t.length<28)
+                  .filter((v,i,arr)=>arr.indexOf(v)===i).slice(0,120)""")
+            except Exception:
+                ents = []
+            out["entries"] = ents
+            out["interesting"] = [e for e in ents if any(k in e for k in ("模板", "模版", "下载", "导入", "导出", "订单", "发票"))][:20]
+        t = body(page)
+        out["logged_in"] = logged_in(page)
+        out["text_head"] = t[:200]
+        print(json.dumps(out, ensure_ascii=False))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())

package/engine/scripts/wenyao_download_template.py

@@ -0,0 +1,144 @@
+#!/usr/bin/env python3
+"""wenyao_download_template.py — 从文鳐门户下载「导入模版」并解析其结构（2026-09-10）
+
+背景：用户提示"登录后进系统找找能否有模板下载入口"。实测**有**：登录 → 跨境电商订单页 →
+「导入订单」弹窗里有「下载导入模版 / 跨境电商-下单模板.xlsx」。本脚本把它下载到本机固定目录
+并输出结构（sheet 名 + 表头列），用于确定"按货代路由的模板填写规则"。
+
+安全边界：只**下载模板**；不选任何文件、不点「导入」、不点导出/取消订单；下载后关闭弹窗。
+
+用法：python3 scripts/wenyao_download_template.py [--out-dir ~/.dsh/argent/wenyao-downloads]
+"""
+import argparse
+import json
+import os
+import sys
+from pathlib import Path
+from playwright.sync_api import sync_playwright
+
+HOME = Path.home()
+# 数据目录：**唯一判据**在 `scripts/lib/argent_datadir.py`（三级链，与 JS 内核同语义）
+sys.path.insert(0, str(Path(__file__).resolve().parent / 'lib'))
+from argent_datadir import resolve_data_dir  # noqa: E402  —— 勿在此复制判据
+DATA = Path(resolve_data_dir()['dir'])
+DEFAULT_OUT = DATA / "wenyao-downloads"
+CDP = os.environ.get("WENYAO_CDP", "http://127.0.0.1:9333")
+
+
+def summarize_xlsx(path):
+    try:
+        import openpyxl
+    except Exception as e:
+        return {"err": f"openpyxl 不可用: {e}"}
+    try:
+        wb = openpyxl.load_workbook(path, data_only=True, read_only=True)
+        out = {"sheets": wb.sheetnames, "detail": {}}
+        for ws in wb.worksheets:
+            rows = []
+            for i, r in enumerate(ws.iter_rows(values_only=True)):
+                if i > 4:
+                    break
+                rows.append([str(c)[:24] if c is not None else "" for c in r][:18])
+            out["detail"][ws.title] = {"dims": f"{ws.max_row}x{ws.max_column}", "head": rows}
+        return out
+    except Exception as e:
+        return {"err": str(e)[:200]}
+
+
+def main():
+    ap = argparse.ArgumentParser()
+    ap.add_argument("--out-dir", default=str(DEFAULT_OUT))
+    a = ap.parse_args()
+    out_dir = Path(a.out_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+    os.chmod(out_dir, 0o700)
+    res = {}
+    with sync_playwright() as p:
+        browser = p.chromium.connect_over_cdp(CDP, timeout=8000)
+        pages = [pg for c in browser.contexts for pg in c.pages]
+        page = next((pg for pg in pages if "ffl-tech" in (pg.url or "")), pages[0])
+        # 关掉可能残留的弹窗，保证从干净状态开始
+        try:
+            page.keyboard.press("Escape")
+            page.wait_for_timeout(600)
+        except Exception:
+            pass
+        opened = False
+        for sel in ["button:has-text('导入订单')", "text=导入订单"]:
+            try:
+                el = page.query_selector(sel)
+                if el and el.is_visible():
+                    el.click(); opened = True; break
+            except Exception:
+                continue
+        page.wait_for_timeout(2500)
+        res["dialog_opened"] = opened
+        # 找下载入口（文本可能是"下载导入模版"或"跨境电商-下单模板.xlsx"）
+        target = None
+        for sel in ["text=下载导入模版", "text=下载导入模板", "text=跨境电商-下单模板.xlsx",
+                    "a:has-text('模板')", "button:has-text('模板')"]:
+            try:
+                el = page.query_selector(sel)
+                if el and el.is_visible():
+                    target = sel
+                    break
+            except Exception:
+                continue
+        res["download_selector"] = target
+        if target:
+            # 关键：浏览器是 open -na 启动的（非 Playwright 启动）⇒ 收不到 Playwright 的 download 事件。
+            # 直接用 CDP 设置下载行为，把文件落到我们的目录，再轮询目录取新文件。
+            before = {f.name for f in out_dir.iterdir()} if out_dir.exists() else set()
+            try:
+                try:
+                    sess = page.context.new_cdp_session(page)
+                    sess.send("Browser.setDownloadBehavior",
+                              {"behavior": "allow", "downloadPath": str(out_dir), "eventsEnabled": True})
+                    res["cdp_download_behavior"] = True
+                except Exception as e:
+                    res["cdp_download_behavior"] = f"err: {str(e)[:100]}"
+                page.click(target)
+                import time as _t
+                dest = None
+                for _ in range(40):
+                    page.wait_for_timeout(800)
+                    now = {f.name for f in out_dir.iterdir()}
+                    new_files = [f for f in out_dir.iterdir() if f.name not in before and not f.name.endswith('.crdownload')]
+                    if new_files:
+                        dest = max(new_files, key=lambda f: f.stat().st_mtime)
+                        break
+                if dest:
+                    os.chmod(dest, 0o600)
+                    res["saved"] = {"path": str(dest), "bytes": dest.stat().st_size, "name": dest.name}
+                    if dest.suffix.lower() in (".xlsx", ".xls", ".xlsm"):
+                        res["structure"] = summarize_xlsx(dest)
+                else:
+                    res["download_err"] = "点击后 32s 内目录里没有新文件（可能弹了新标签页/需登录态刷新/是邮件发送而非下载）"
+            except Exception as e:
+                res["download_err"] = str(e)[:200]
+        # 关闭弹窗（只关不提交）
+        closed = False
+        for sel in ["button:has-text('取消')", ".next-dialog-close", "text=取消"]:
+            try:
+                el = page.query_selector(sel)
+                if el and el.is_visible():
+                    el.click(); closed = True; break
+            except Exception:
+                continue
+        if not closed:
+            try:
+                page.keyboard.press("Escape")
+            except Exception:
+                pass
+        res["dialog_closed"] = closed
+        page.wait_for_timeout(800)
+        try:
+            page.screenshot(path=str(DATA / "wenyao-probe" / "62-after-template-download.png"), full_page=True)
+        except Exception:
+            pass
+        print(json.dumps(res, ensure_ascii=False, indent=1))
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())