修复报价上下文并完善订单节点和异常统计

AGENTS.md

@@ -16,6 +16,7 @@ Use `am.py` for CLI/WSGI, `logistics_chat/` for business code, and `static/css/`
 | Pricing | `logistics_chat/pricing.py`: `RATES`, `calc_freight`; `quote_intent.py`: `parse_intent`; `quote_service.py`: `answer_quote`; `quote_engine.py`: `quantity`, `render_rates`; `quote_answers.py`, `quote_fees.py`: local business terms and fees |
 | US quote data | `logistics_chat/quote_workbook.py`: local detail export; `quote_data.py`: versioned SQLite imports; `quote_engine.py`: matching/calculation; `quote_service.py`: conversational reply; `scripts/import_quote_data.py` |
 | Common customer questions | `logistics_chat/shipping_faq.py`: pre-shipment material; `quote_allocations.py`: per-warehouse weights; `quote_context.py`: chronological edits; `order_explanations.py`: business status, latest actual nodes and operation boundaries |
+| Reply boundaries / order summaries | `logistics_chat/customer_faq.py`: supported actions; `quote_scope.py`: US-only fallback; `quote_edits.py`: rejected conditions; `quote_material.py`: partial metal chargeable weight; `order_query.py`: bounded filters; `order_listing.py`: scoped five-row summaries and conversation pages; `order_tracking.py`: `current_stage` |
 | Tracking | `logistics_chat/tracking.py`: `extract_tracking_no`, `build_track`, `CARD_PREFIX` |
 | Order test data | `logistics_chat/order_data.py`: `lookup_order`, `count_orders`, `answer_query`, `import_records`; `scripts/import_order_test_data.py` |
 | Persistence | `logistics_chat/db.py`: `init_db`, `save_message`; `logistics_chat/history_service.py` |

docs/architecture.md

@@ -113,7 +113,7 @@ tests/                        # 按业务模块组织的回归
 - `chat_service.create_chat_stream` 的订单查询使用本地 `order_test_data`；忽略附件路径、重量、日期和三维尺寸，保留标识中的后缀，不再调用 `build_track` 生成模拟轨迹。旧模拟函数仅兼容旧导入，不进入聊天查询路径。未知单号、多条匹配和查表失败明确说明，禁止回退模型编造状态。
 - `CARD_PREFIX` 新增 `kind=order`：仅包含白名单概要、申报货物汇总和订单业务状态历史，不包含姓名、地址、内部备注或操作人。`order_tracking.py` 从 `order_test_data.tracking_json` 的已关联事实生成物流轨迹；存在有效事实时设置 `node_kind=logistics`，`nodes` 仅存实际节点，`estimated_nodes` 单独展示预计节点，`tracking_status` 只按实际节点序号推进。相同货柜/节点按最新观测修正时间，预计事实不覆盖实际事实；多柜独立展示。缺失的节点、地点和签收事实不得补造，提柜、还柜不等于签收。子单不继承父单重量或轨迹；COMPLETED 只表示头程订单完成。没有物流事实时继续展示业务状态历史。页面和模型上下文不附加测试副本声明或导入时间。订单及物流追问只在同一 `conv_id` 内复用最近单号，并重新查表取得修改后的值；新报价或其他话题不会沿用旧订单。
 - 订单表不提供管理页面或对外增删改接口；表数据在运行时 SQLite 内维护，导入脚本默认保留已有记录，只有显式 `--replace` 覆盖。匿名 HTTP 查单在读取订单前返回登录引导；`g.customer_tenant_id` 由可信客户认证适配器设置，不接受访客 key、请求参数或客户自报 Header。客户范围非空时必须传给 `lookup_order`，并与 `ORDER_DATA_TENANT_ID` 的额外限制取交集。源库连接凭据和快照在仓库外，业务库只读，不随服务查询访问源库。
-- 登录后的订单数量查询由 `parse_count_query` 识别，`count_orders` 使用与单号查询相同的 `tenant_filter` 交集规则，必须显式提供非空客户范围；SQL 直接计算订单票数和已登记入仓码的订单票数，按需过滤明确的业务状态。无额外筛选时统计授权范围全部订单，不复用上一票单号；无法识别的附加条件不静默丢弃，也不交给模型猜数。成功统计返回的内部 `_customer_data` 标记由 `chat_service` 消费并移除，使纯文本统计同样在保存/发送前绑定会话客户和范围；不增加公开 SSE 字段或管理接口。
+- 登录后的订单数量由 `parse_count_query` / `order_query.parse_query` 识别，汇总、筛选和列表共用 `order_listing.read_orders` 的服务端 `tenant_filter` 交集与同一只读快照；`count_orders` 保留原聚合调用契约。默认给总数、已登记入仓码数及最多5条按来源更新时间排序的摘要；支持业务状态、明确标记的异常和当前实际节点。未指定节点的轨迹统计按订单分组，每票只计一次，多柜不同进度单列，预计节点不推进进度，业务完成不当作签收。未知条件不默认为全部，也不交给模型猜数。内部 `_customer_data` 和 `_order_list_state` 在发送前移除，事实历史绑定客户与范围；无新增公开API。
 - 客户认证使用独立 `customer_users`、`customer_tenant_grants` 与 `customer_sessions` 表，由 `customer_service.attach_customer` 在访客恢复之后验证并设置 `g.customer`、单客户 `g.customer_tenant_id` 或多客户 `g.customer_tenant_ids`；只保存密码哈希与登录 token 哈希。默认账号首次关联当前本地订单的客户范围，之后不自动扩大，普通账号仍按明确关联过滤；`lookup_order` 的单客户参数及多客户白名单都与环境限制取交集。会话的 `customer_scope` 保存范围指纹。登录 Cookie 为 HttpOnly、SameSite=Lax，在 HTTPS 上为 Secure，绑定当前访客并在 8 小时后失效；退出、账号停用、密码或客户范围改变使旧会话失效。默认客户账号不赋予客服后台权限。
 - 匿名查单返回 SSE `login_required` 与正文，客户页提供登录按钮；`customer.js` 登录成功后仅在原会话版本未变化时继续原问题，保留期间输入的新草稿。登录过期导致界面清空时，只允许同一客户重新通过历史归属校验后恢复原会话并续查；换账号、访问范围变化、历史加载失败或切换话题均不重放旧问题。侧栏可独立登录/退出，登录状态在窗口恢复焦点时刷新。首次返回订单卡片时，会话绑定客户 ID 和客户范围；历史、人工会话及后续生成同时校验访客、客户和范围。退出立即清除当前可见订单内容和历史标题缓存，重新登录同一身份可恢复；匿名清空历史不删除被登录保护的会话。
 - 底部及欢迎卡片的查入仓码入口先确认客户登录，再打开 `OrderLookup`，不自动新建会话或发送缺少号码的问题。当前及历史订单卡片的号码也打开该查询框，保留真实的入仓码/订单号标签，支持单独复制和重新查询。提交使用既有 `/stream_chat` 并显示明确的用户问句，结果追加到当前会话；主输入框草稿及已上传附件不清空、不代发。人工接待、未加载历史或其他请求进行中时不抢占路由。切换话题/身份关闭弹窗，迟到的登录/状态/复制结果不得改动新上下文；弹窗跟随手机可视视口并将焦点还给触发按钮。
@@ -145,6 +145,20 @@ tests/                        # 按业务模块组织的回归
 
 `shipping_faq.py` 从激活版本回答交仓、标签、报关资料及查验费用问题；具体船期和计划发货日期需人工核对，不使用旧表日期承诺必达。`order_explanations.py` 只解释授权订单的业务状态和实际节点，不把预计节点当实际位置，不推断逐箱签收或亚马逊上架。文字事实回复与订单卡片同样绑定客户及范围；催单或拦截不声称已提交。本轮不扩展售后理赔规则。
 
+`customer_faq.py` 在报价分流前处理下单、短信查单、开票、导出及附件读取等能力说明，不生成业务操作成功的假回复。美国常见费用/毛净重/议价/总价追问由当前价表直接回答；`quote_scope.py` 将尚未匹配的美国咨询保留为明确的核对回复，不送入旧欧洲/英国知识库。明确切换其他国家时仍走该地区原有分流，理赔规则未扩展。
+
+`quote_edits.py` 先移除被明确否定的仓码或交货仓，再解析替换条件。共享尺寸解析支持英寸转换和口语长宽高，“一个箱子”可形成单箱条件；总计立方米不作为尺寸单位“米”。
+
+`order_listing.py` 负责数量摘要、查看全部和翻页。数量摘要最多5票，直接翻页继续每页5票；明确列出订单或“查看全部”按每页20票。`chat_history.order_list_state` 保存筛选、页码、页大小和成功回复的消息ID，绑定客户与范围；只保存仍为最新消息的状态，迟到回复不能覆盖新话题。中间插问同条件总数保留列表页码，改筛选重置，新话题停止沿用。每次重新查询授权数据，不缓存数量；计数及分页来自同一只读快照，无有效上下文先引导查询，页码有界。
+
+`order_test_data.exception_reason` 是人工明确登记、允许客户查看的物流异常原因；非空表示该票有异常标记，空值表示未标记，不推断货物正常。取消、审核驳回和无轨迹均不自动标为物流异常。可直接维护此字段，不增加管理页面；旧快照未提供该字段时即使显式覆盖也保留已有原因，显式提供空字符串才清除。后续MCP适配返回相同事实即可。
+
+共享尺寸解析将单位与长宽高绑定，支持每边单位和混合单位，不受“12英寸显示器”等商品描述影响。报价沿当前会话回溯到最近话题边界，不按10轮截断；其他国家、新一票及订单话题隔离不变。“铁铝部分还是10KG”等口语仅在先前已明确计费重时更新该基数，既不覆盖整票总重，也不把未确认毛重直接当计费重。
+
+已确认海派计费重保留原小数，不向上取整，每箱最低12KG仍生效。原实重小数原样展示；无限小数的材积结果以“约”展示有限位数，金额计算仍使用原Decimal精度，不把显示近似当作计费进位。
+
+铁铝附加费只按铁铝货物计费重计算。`quote_material.py` 区分整票铁铝、部分铁铝与未明确材质的金属，并将明确的“铁铝部分计费重”从整票毛重解析中分离。部分货物未提供计费重时不按整票加费；提供明确基数后可将该项人民币费用加到公斤或按方运费中。更改整票重量、箱数或箱规后清除旧的部分计费重，防止沿用过期基数。未明确金属是否属于铁铝时先核实材质，不把所有金属都认作铁铝。
+
 先按 README 安装项目依赖。Windows 使用 `.\.venv\Scripts\python.exe -m pytest -v` 或同解释器的 `-m unittest discover -s tests -v`；Linux/macOS 使用 `.venv/bin/python`。业务测试依赖 Flask，不能用未安装依赖的全局 Python 运行整套测试。模型请求全部模拟，不代表真实 DeepSeek 联调。
 
 路径、函数或资产移动后同步更新根 `AGENTS.md`、本文、知识资料和 README 目录树。仅改导航时检查路径、符号、命令；改业务时补充相关成功和失败分支，前端改动检查实际页面。

logistics_chat/chat_service.py

@@ -91,14 +91,18 @@ def create_chat_stream(msg, conv_id):
                 return (sse(order_login_response()) + sse({'d':1}), {})
             return (sse({'e':'该会话已删除，请新建对话后重试。'}), {})
     order_answer = answer_query(msg, conv_id)
+    if order_answer is None:
+        from .customer_faq import answer_capability
+        order_answer = answer_capability(msg)
     if order_answer is None:
         order_answer = answer_quote(msg, conv_id)
     if order_answer is None and re.search(r'(?:转|联系|接入|找).*人工|人工客服',msg):
         order_answer = {'t':'点击“联系人工”可发起人工客服申请；非工作时间可留言。'}
-    if order_answer is None and re.search(r'照片|图片',msg) and re.search(r'报价|价格|判断|识别',msg):
-        order_answer = {'t':'可以通过输入框旁的“+”上传照片供人工查看。当前聊天不会自动识别图片，请同时用文字提供品名、重量、箱数、尺寸和目的仓码或邮编，再核算报价。'}
     if order_answer is None and re.fullmatch(r'\[图片\]/uploads/\S+', msg):
         order_answer = {'t':'收到您的图片，请描述一下您的问题。'}
+    if order_answer is None:
+        from .quote_scope import answer_us_fallback
+        order_answer = answer_us_fallback(msg,conv_id)
     ship = shipping_intent_with_context(msg, conv_id) if order_answer is None else None
     if ship and ship['region'] is None and (ship['weight'] is not None or ship['dims'] or
             re.search(r'\d+(?:\.\d+)?\s*(?:cbm|方|立方)', msg, re.I) or
@@ -111,6 +115,7 @@ def create_chat_stream(msg, conv_id):
         order_answer = {'t':'当前报价表未提供该目的地的按方费率，请提供目的仓或邮编、实际重量和尺寸，由人工客服确认报价。'}
     if order_answer is not None:
         customer_data = order_answer.pop('_customer_data',False) or 'card' in order_answer
+        order_list_state=order_answer.pop('_order_list_state',None)
 
         def track_stream():
             if customer_data:
@@ -121,7 +126,10 @@ def create_chat_stream(msg, conv_id):
             save_message(conv_id, 'user', msg)
             content = (CARD_PREFIX + json.dumps(order_answer['card'], ensure_ascii=False)
                        if 'card' in order_answer else order_answer['t'])
-            save_message(conv_id, 'assistant', content)
+            message_id=save_message(conv_id, 'assistant', content)
+            if order_list_state is not None:
+                from .order_listing import save_state
+                save_state(conv_id,order_list_state,message_id)
             yield sse(order_answer)
             yield sse({'d': 1})
         return (track_stream(), {'Cache-Control': 'no-cache', 'X-Accel-Buffering': 'no'})

logistics_chat/customer_faq.py

@@ -0,0 +1,21 @@
+"""Describe supported customer actions without inventing operational integrations."""
+
+import re
+
+
+def answer_capability(message):
+    if re.search(r'照片|图片',message) and re.search(r'报价|价格|判断|识别|尺寸',message):
+        return {'t':'可以通过输入框旁的“+”上传照片供人工查看。当前聊天不会自动识别图片，请同时用文字提供品名、重量、箱数、尺寸和目的仓码或邮编，再核算报价。'}
+    if re.search(r'(?:手机号|手机号码|电话号码).*(?:查|订单|运单)|(?:查|查询).*手机号',message):
+        return {'t':'当前请使用客户账号登录后，按入仓码查询订单；暂不支持手机号或短信验证码查单。没有入仓码可联系人工核实订单归属。'}
+    if re.search(r'下单(?!号|时间|日期)|订舱|生成订单|创建订单|预约提货|安排提货|\b(?:place|create)\s+(?:an?\s+)?order\b',message,re.I):
+        return {'t':'下单、订舱和安排提货需要联系人工办理。当前可查询报价及已关联订单，尚未创建订单或安排提货；请提供目的仓、品名、箱数、重量和尺寸供人工核对。'}
+    if re.search(r'开(?:发票|票)|发票.*(?:税点|几个点|怎么开|申请)|月结|账期|付款账号|收款账号|银行账号|打款|(?:怎么|如何)付款',message):
+        return {'t':'开票税点、月结资格及付款方式需要由专属业务员确认。请联系人工核对对应订单和结算信息；当前聊天不提供未经确认的税点或收款账号。'}
+    if re.search(r'导出.*(?:订单|运单|Excel)|(?:订单|运单).*导出',message,re.I):
+        return {'t':'当前聊天暂不提供订单Excel导出，可联系人工核对导出需求。可先查看账号内订单列表，或发送入仓码查询单票详情。'}
+    if re.search(r'Excel|表格',message,re.I) and re.search(r'直接读|自动读|读取|识别|读里面',message):
+        return {'t':'当前聊天不会自动读取Excel附件里的箱规。请将每箱重量、长宽高及目的仓码粘贴为文字，我可据此核算；批量表格可交人工整理。'}
+    if re.search(r'换.*手机|换.*设备|跨设备|清除.*(?:浏览器|缓存)',message) and re.search(r'聊天|记录|历史',message):
+        return {'t':'当前聊天记录与原浏览器的访客身份关联，换手机或清除浏览器数据后不会自动同步恢复。请先在原设备查看；客户账号目前用于订单权限校验，尚未提供跨设备聊天同步。'}
+    return None

logistics_chat/db.py

@@ -54,6 +54,8 @@ def init_db():
         cur.execute('ALTER TABLE chat_history ADD COLUMN visitor_id TEXT')
     if 'customer_id' not in cols:
         cur.execute('ALTER TABLE chat_history ADD COLUMN customer_id INTEGER')
+    if 'order_list_state' not in cols:
+        cur.execute('ALTER TABLE chat_history ADD COLUMN order_list_state TEXT')
     if 'customer_scope' not in cols:
         cur.execute('ALTER TABLE chat_history ADD COLUMN customer_scope TEXT')
         if 'customer_tenant_id' in cols:
@@ -232,6 +234,7 @@ def save_message(conv_id, role, content):
     if not inserted:
         conn.close()
         return 'deleted'
+    message_id=cur.lastrowid
     if role == 'user':
         row = cur.execute('SELECT title FROM chat_history WHERE id=?', (conv_id,)).fetchone()
         if row and (not row[0] or row[0] == '新对话'):
@@ -239,6 +242,7 @@ def save_message(conv_id, role, content):
     cur.execute('UPDATE chat_history SET utime=? WHERE id=?', (now_str(), conv_id))
     conn.commit()
     conn.close()
+    return message_id
 
 
 def last_user_question(conv_id):

logistics_chat/order_data.py

@@ -14,7 +14,7 @@ from .customer_service import order_login_response
 FIELDS = ('source_order_id', 'tenant_id', 'order_no', 'inbound_code', 'shipment_id',
           'status', 'country_code', 'transport_mode', 'warehouse_code', 'order_role',
           'parent_order_no', 'package_count', 'gross_weight_kg', 'volume_cbm',
-          'history_json', 'tracking_json', 'data_note', 'source_updated_at', 'snapshot_at')
+          'history_json', 'tracking_json', 'exception_reason', 'data_note', 'source_updated_at', 'snapshot_at')
 STATUS_LABELS = {'DRAFT': '草稿', 'SUBMITTED': '已提交', 'APPROVED': '审核通过',
                  'ACCEPTED': '已接单', 'REJECTED': '审核驳回', 'PROCESSING': '履约中',
                  'COMPLETED': '头程订单已完成', 'CANCELLED': '已取消'}
@@ -23,7 +23,7 @@ FIELD_LABELS = {'inbound':'入仓码', 'number':'订单号', 'shipment':'Shipmen
                 'weight':'申报毛重', 'volume':'申报体积', 'role':'订单层级',
                 'parent':'父订单号', 'updated':'来源更新时间'}
 IDENTIFIER = re.compile(r'(?<![A-Za-z0-9_-])[A-Za-z0-9][A-Za-z0-9_-]{5,63}(?![A-Za-z0-9_-])')
-ORDER_DETAIL = r'状态|进度|轨迹|到哪|开船|到港|清关|提柜|还柜|签收|预计|什么时候|何时|多久|重量|毛重|体积|件数|箱数|目的地|目的国|目的仓|仓库|父订单|运费|费用|价格|几件|几箱|公斤|千克|多少钱'
+ORDER_DETAIL = r'状态|进度|轨迹|异常|到哪|开船|到港|清关|提柜|还柜|签收|预计|什么时候|何时|多久|重量|毛重|体积|件数|箱数|目的地|目的国|目的仓|仓库|父订单|运费|费用|价格|几件|几箱|公斤|千克|多少钱'
 
 
 class OrderDataUnavailable(Exception):
@@ -39,13 +39,15 @@ def init_order_table(connection):
         status TEXT NOT NULL, country_code TEXT, transport_mode TEXT, warehouse_code TEXT,
         order_role TEXT, parent_order_no TEXT,
         package_count INTEGER, gross_weight_kg TEXT, volume_cbm TEXT,
-        history_json TEXT NOT NULL DEFAULT '[]', tracking_json TEXT NOT NULL DEFAULT '[]', data_note TEXT,
+        history_json TEXT NOT NULL DEFAULT '[]', tracking_json TEXT NOT NULL DEFAULT '[]', exception_reason TEXT, data_note TEXT,
         source_updated_at TEXT, snapshot_at TEXT,
         UNIQUE(tenant_id, order_no)
     )''')
     columns = {row[1] for row in connection.execute('PRAGMA table_info(order_test_data)')}
     if 'tracking_json' not in columns:
         connection.execute("ALTER TABLE order_test_data ADD COLUMN tracking_json TEXT NOT NULL DEFAULT '[]'")
+    if 'exception_reason' not in columns:
+        connection.execute('ALTER TABLE order_test_data ADD COLUMN exception_reason TEXT')
     connection.execute('CREATE INDEX IF NOT EXISTS idx_order_test_inbound ON order_test_data(inbound_code)')
     connection.execute('CREATE INDEX IF NOT EXISTS idx_order_test_shipment ON order_test_data(shipment_id)')
 
@@ -62,6 +64,10 @@ def import_records(connection, records, replace=False):
         row['source_order_id'] = str(row['source_order_id']) if row['source_order_id'] is not None else None
         row['order_no'] = str(row['order_no']).strip()
         row['status'] = str(row['status'] or 'DRAFT')
+        if row['exception_reason'] is not None:
+            if not isinstance(row['exception_reason'],str) or len(row['exception_reason'])>2000:
+                raise ValueError('exception_reason must be customer-visible text up to 2000 characters')
+            row['exception_reason']=row['exception_reason'].strip() or None
         history = row['history_json'] or []
         if isinstance(history, str): history = json.loads(history)
         if not isinstance(history, list): raise ValueError('history_json must contain a list')
@@ -78,7 +84,8 @@ def import_records(connection, records, replace=False):
             continue
         columns = ','.join(FIELDS)
         placeholders = ','.join('?' for _ in FIELDS)
-        updates = ','.join(key+'=excluded.'+key for key in FIELDS if key not in ('tenant_id', 'order_no'))
+        updates = ','.join(key+'=excluded.'+key for key in FIELDS if key not in ('tenant_id', 'order_no')
+                          and (key!='exception_reason' or key in source))
         connection.execute(f'INSERT INTO order_test_data ({columns}) VALUES ({placeholders}) '
             f'ON CONFLICT(tenant_id,order_no) DO UPDATE SET {updates}', [row[key] for key in FIELDS])
         inserted += 1
@@ -179,92 +186,40 @@ def count_orders(tenant_ids, status=None):
 
 
 def parse_count_query(message):
-    """Recognize bounded total/status questions without ignoring unknown filters."""
-    if is_lookup_help(message):
-        return None
-    noun = r'订单|运单|入仓码|入仓号|\borders?\b|\bshipments?\b'
-    count = r'多少|几(?:个|票|条|张|笔)|总数|数量|总量|统计|数一下|(?:订单|运单)数(?!量)|\bhow\s+many\b|\bcount\b|\bnumber\b|\btotal\b'
-    if not re.search(noun,message,re.I) or not re.search(count,message,re.I):
-        return None
-    if re.search(r'公斤|千克|\bkg\b|重量|毛重|体积|件数|箱数|运费|费用|价格|多少钱',message,re.I):
-        return None
-    remaining = message
-    statuses = []
-    for phrase, status in (('头程订单已完成','COMPLETED'),('已完成','COMPLETED'),('履约中','PROCESSING'),
-                           ('处理中','PROCESSING'),('已取消','CANCELLED'),('审核通过','APPROVED'),
-                           ('审核驳回','REJECTED'),('已接单','ACCEPTED'),('已提交','SUBMITTED'),('草稿','DRAFT')):
-        if phrase in remaining:
-            statuses.append(status)
-            remaining = remaining.replace(phrase,'')
-    remaining = re.sub(count+'|'+noun,'',remaining,flags=re.I)
-    remaining = re.sub(r'当前账号|当前账户|账号下|账户下|名下|可查询|能查到|可以查到|目前|现在|当前|一共|总共|合计|总计|'
-                       r'请问|麻烦|帮我|想知道|想查|想看|请|查询|查一下|查下|查查|看一下|看看|看下|一下|下|全部|所有|共有|我的|我|能|有|的|个|票|条|张|笔|呢|吗|呀|啊|共|'
-                       r'\b(?:what|is|of|do|i|have|are|there|my|in|all|the|currently)\b|[\s，,。.!！?？、：:]','',remaining,flags=re.I)
-    return {'status':statuses[0] if statuses else None, 'unsupported':bool(remaining or len(set(statuses)) > 1)}
+    if is_lookup_help(message):return None
+    from .order_query import parse_query
+    return parse_query(message)
 
 
-def answer_count_query(intent, tenants):
-    if not tenants:
-        return order_login_response()
-    if intent['unsupported']:
-        return {'t':'暂不支持按这个条件统计。可以查询订单总数，或按草稿、已提交、履约中、已完成、已取消等业务状态统计。'}
-    try:
-        counts = count_orders(tenants,intent['status'])
-    except OrderDataUnavailable:
-        return {'t':'订单信息暂时无法读取，请稍后重试。'}
-    if intent['status']:
-        text = f"您当前有 {counts['total']} 票业务状态为“{STATUS_LABELS[intent['status']]}”的订单"
-    else:
-        text = f"您当前共有 {counts['total']} 票订单"
-    return {'t':text + f"，其中 {counts['with_inbound']} 票已登记入仓码。", '_customer_data':True}
+def answer_count_query(intent, tenants,conv_id=None):
+    from .order_listing import answer_summary
+    return answer_summary(intent,tenants,conv_id)
 
 
-def answer_list_query(message, tenants):
-    """Reuse bounded status parsing; list reads require an explicit customer scope."""
-    if is_lookup_help(message) or not re.search(r'订单|运单|\borders?\b|\bshipments?\b',message,re.I):
-        return None
-    if not re.search(r'列出|列表|全部|所有|\blist\b',message,re.I):
-        return None
-    if not tenants:
-        return order_login_response()
-    question=re.sub(r'列出|列表|全部|所有|\blist\b','多少',message,flags=re.I)
-    intent=parse_count_query(question)
-    if intent is None or intent['unsupported']:
-        return {'t':'暂不支持按这个条件列出订单。可查看全部订单或按履约中、已完成等业务状态查询。'}
-    scope=tenant_filter(tenant_ids=tenants)
-    if scope is None:
-        return order_login_response()
-    from .db import connect
-    connection=None
-    try:
-        connection=connect()
-        sql='SELECT order_no,inbound_code,status,warehouse_code FROM order_test_data WHERE 1=1'+scope[0]
-        values=scope[1]
-        if intent['status']:
-            sql+=' AND status=?';values=values+[intent['status']]
-        rows=connection.execute(sql+' ORDER BY id DESC LIMIT 21',values).fetchall()
-        if not rows:return {'t':'当前账号下没有符合条件的订单。','_customer_data':True}
-        lines=['您当前的订单列表'+('（先列出20票）' if len(rows)>20 else '')+'：']
-        for number,inbound,status,warehouse in rows[:20]:
-            lines.append(f'- 入仓码：{inbound or "未登记"}；订单号：{number}；业务状态：{STATUS_LABELS.get(status,status)}；目的仓：{warehouse or "未登记"}')
-        lines.append('发送入仓码可查询该票订单和物流轨迹。')
-        return {'t':'\n'.join(lines),'_customer_data':True}
-    except sqlite3.Error:
-        return {'t':'订单信息暂时无法读取，请稍后重试。'}
-    finally:
-        if connection is not None:connection.close()
+def answer_list_query(message, tenants, conv_id=None):
+    from .order_listing import answer_list
+    return answer_list(message,tenants,conv_id)
 
 
 def extract_reference(message):
     # Explicit postal/monetary fields are not order identifiers, even if invalid.
+    message = re.sub(r'(?:手机号|手机号码|电话号码|联系电话|电话)\s*[:：是为]?\s*(?:\+?86[-\s]?)?1[3-9]\d{9}', ' ', message)
     message = re.sub(r'(?:邮编|zip(?:\s*code)?|post\s*code)\s*[:：为是]?\s*[0-9-]+', ' ', message, flags=re.I)
     message = re.sub(r'(?<![A-Za-z0-9_-])\d+(?:\.\d+)?\s*(?:万)?(?:USD|RMB|CNY|美元|美金|人民币|元)(?![A-Za-z0-9_-])',' ',message,flags=re.I)
     message = re.sub(r'\[(?:图片|文件)\]/uploads/\S+|https?://\S+', '', message)
     message = re.sub(r'(?<![A-Za-z0-9_-])\d{4}-\d{1,2}-\d{1,2}'
                      r'(?:[T ]\d{1,2}:\d{2}(?::\d{2}(?:\.\d+)?)?(?:Z|[+-]\d{2}:?\d{2})?)?'
                      r'(?![A-Za-z0-9_-])', ' ', message)
+    from .pricing import DIMENSION_PATTERNS
+    for pattern in DIMENSION_PATTERNS:
+        def remove_dimensions(match):
+            end=match.start()+len(match[0].rstrip())
+            before=message[match.start()-1:match.start()] if match.start() else ''
+            after=message[end:end+1]
+            return match[0] if re.search(r'[A-Za-z0-9_.-]',before+after) else ' '
+        message=pattern.sub(remove_dimensions,message)
     message = re.sub(r'(?<![A-Za-z0-9_.-])\d+(?:\.\d+)?\s*[*xX×]\s*\d+(?:\.\d+)?\s*[*xX×]\s*\d+(?:\.\d+)?'
-                     r'(?:\s*(?:cm|mm|m|厘米|毫米|米))?(?![A-Za-z0-9_.-])', ' ', message, flags=re.I)
+                     r'(?:\s*(?:inches|inch|in|cm|mm|m|英寸|厘米|毫米|米|["″]))?(?![A-Za-z0-9_.-])', ' ', message, flags=re.I)
     message = re.sub(r'(?<![A-Za-z0-9_-])\d+(?:\.\d+)?\s*'
                      r'(?:kilograms?|kgs?|grams?|g|tons?|lbs?|磅|cbm|cm|mm|公斤|千克|克|吨|立方米|方|m[³3]|箱|件)'
                      r'(?![A-Za-z0-9_])', ' ', message, flags=re.I)
@@ -341,7 +296,7 @@ def is_tracking_request(message):
                r'(?:物流|轨迹|进度)(?:查询|查一下|查下))')
     personal = (r'(?:我的|(?:我)?(?:这|那)(?:票|个|单)|该)(?:货物?|包裹|快递|订单|运单)'
                 r'(?:现在|目前)?(?:到哪(?:里)?了?|什么时候(?:能|会)?到|何时能?到|多久能?到|'
-                r'签收了?(?:吗|没)|状态|进度|轨迹)(?:呢|吗|呀|啊)?')
+                r'(?:已)?(?:清关|开船|到港|出库|提柜)(?:了)?(?:吗|没|了吗|没有)?|签收了?(?:吗|没)|状态|进度|轨迹)(?:呢|吗|呀|啊)?')
     english = (r'(?:where\s+(?:is|are)\s+my|track(?:\s+my)?|check(?:\s+my)?)\s+'
                r'(?:parcels?|packages?|shipments?|orders?)(?:\s+(?:status|progress))?|'
                r'what\s+is\s+my\s+(?:parcel|package|shipment|order)\s+(?:status|progress)')
@@ -384,11 +339,11 @@ def answer_query(message, conv_id):
         if explanation:return {'t':explanation}
     count_intent = parse_count_query(message) if not references else None
     if count_intent is not None:
-        return answer_count_query(count_intent,tenants)
+        return answer_count_query(count_intent,tenants,conv_id)
     if not references:
-        listed=answer_list_query(message,tenants)
+        listed=answer_list_query(message,tenants,conv_id)
         if listed is not None:return listed
-    mutation=bool(re.search(r'改址|修改.*(?:收件|地址|订单)|收件地址.*改|(?:取消|删除).*(?:订单|运单)|(?:订单|运单).*取消',message))
+    mutation=bool(re.search(r'改址|改.*(?:收货|收件).*地址|修改.*(?:收件|地址|订单)|(?:收货|收件)地址.*改|(?:取消|删除).*(?:订单|运单)|(?:订单|运单).*取消',message))
     if references and not tenants:
         return order_login_response()
     if len(references) > 1:
@@ -422,6 +377,8 @@ def answer_query(message, conv_id):
             return order_login_response() if not tenants else {'t':'当前支持查询订单与物流轨迹；改址、取消等操作请联系人工客服办理，尚未修改订单。请提供入仓码。'}
         if detail_question(message) and not is_lookup_help(message):
             return order_login_response() if not tenants else {'t':'请提供入仓码，我先核对该票订单的实际物流记录。'}
+        if re.search(r'哪些.*(?:订单|运单)|(?:订单|运单).*(?:筛选|还没开船|未发货|已发货|未签收)',message) and not is_lookup_help(message):
+            return order_login_response() if not tenants else {'t':'暂不支持按这个物流条件批量筛选订单。可按业务状态查看列表，或提供入仓码逐票核对实际物流节点。'}
         if is_tracking_request(message) or (not is_lookup_help(message) and re.search(r'(?:查|搜|看|找).*(?:订单|运单|物流轨迹|入仓码|入仓号)|(?:入仓码|入仓号).*(?:查|轨迹|状态)|订单状态|物流轨迹|shipment\s*id|'
                      r'(?:列出|搜索|所有|全部|多少|我的).*(?:订单|运单|入仓)|(?:订单|运单).*(?:多少|列表|查询|在哪|到哪|进度)|'
                      r'^\s*(?:查单|查询|查一下|帮我查一下)[。？！?!\s]*$|'

logistics_chat/order_explanations.py

@@ -6,8 +6,8 @@ import re
 def detail_question(message):
     if re.search(r'报价|运费|海派|海卡|\d+\s*(?:KG|公斤)',message,re.I) and not re.search(r'这票|该订单|我的订单',message):
         return False
-    return bool(re.search(r'上架|没更新|未更新|没有更新|不更新|少了?.*(?:箱|件)|到了?几箱|逐箱|每箱.*签收|'
-                          r'已完成.*(?:签收|送达)|(?:签收|送达).*已完成|催单|催一下|拦截|'
+    return bool(re.search(r'异常原因|为什么异常|什么异常|上架|没更新|未更新|没有更新|不更新|少了?.*(?:箱|件)|到了?几箱|逐箱|每箱.*签收|'
+                          r'已完成.*(?:签收|送达)|(?:签收|送达).*已完成|催单|催一下|拦截|改.*(?:收货|收件).*地址|(?:收货|收件)地址.*改|快递单号|付过款|付款状态|'
                           r'(?:这票|该订单|我的货).*(?:发货|发走|出库|开船)',message))
 
 
@@ -22,8 +22,15 @@ def explain_order(message,row,card):
     if not detail_question(message):return None
     reference=row.get('inbound_code') or row['order_no']
     lines=[f'入仓码 {reference}。' if row.get('inbound_code') else f'订单号 {reference}。']
+    if re.search(r'异常原因|为什么异常|什么异常',message):
+        reason=(row.get('exception_reason') or '').strip()
+        return lines[0]+('已记录的异常原因：'+reason if reason else '当前未登记异常原因，请联系人工核实。')
     if re.search(r'催单|催一下|拦截',message):
         return lines[0]+'催单或拦截需联系人工客服办理；当前仅查询订单，尚未提交催单或拦截申请。'
+    if re.search(r'付过款|付款状态',message):
+        return lines[0]+'当前订单未接入付款状态，不能根据业务状态判断是否付款，请联系人工核对账单。'
+    if '快递单号' in message:
+        return lines[0]+'当前未记录可核实的UPS/FedEx尾程快递单号，入仓码和内部订单号不能直接当作快递单号，请联系人工核对。'
     lines.append(f'登记的业务状态为“{card["status"]}”。')
     actual=card.get('nodes',[]) if card.get('node_kind')=='logistics' else []
     if actual:

logistics_chat/order_listing.py

@@ -0,0 +1,174 @@
+"""Scoped summaries, five-row previews and durable conversational list pages."""
+
+from collections import Counter,deque
+import json
+import re
+import sqlite3
+
+from .db import connect
+from .order_query import parse_query,filters,STAGE_LABELS
+from .order_tracking import current_stage
+
+
+def page_command(message):
+    text=re.sub(r'[\s，,。.!！?？]','',message)
+    match=re.fullmatch(r'(?:请)?(?:订单)?第(\d{1,5})页',text)
+    if match:return int(match.group(1))
+    if re.fullmatch(r'(?:请|继续)?(?:下一页|下页|继续显示|接着显示(?:剩下的)?订单)',text):return 'next'
+    if re.fullmatch(r'(?:请)?(?:上一页|上页)',text):return 'previous'
+    return None
+
+
+def valid_filters(query):
+    from .order_data import STATUS_LABELS
+    return (isinstance(query,dict) and set(query)=={'status','stage','exception'}
+            and (query['status'] is None or query['status'] in STATUS_LABELS)
+            and (query['stage'] is None or query['stage'] in STAGE_LABELS)
+            and isinstance(query['exception'],bool))
+
+
+def load_state(conv_id):
+    from .customer_service import current_customer_id,current_scope_key
+    from .visitor_service import visitor_id
+    if not conv_id:return None
+    connection=connect()
+    try:
+        row=connection.execute('SELECT order_list_state FROM chat_history WHERE id=? AND visitor_id=? '
+            'AND customer_id=? AND customer_scope=?',
+            (conv_id,visitor_id(),current_customer_id(),current_scope_key())).fetchone()
+        state=json.loads(row[0]) if row and row[0] else None
+        if not isinstance(state,dict) or not valid_filters(state.get('query')):return None
+        if type(state.get('page')) is not int or not 1<=state['page']<=10000:return None
+        if state.get('size') not in (5,20) or type(state.get('message_id')) is not int:return None
+        for (text,) in connection.execute("SELECT content FROM messages WHERE conv_id=? AND role='user' AND id>? ORDER BY id",
+                                         (conv_id,state['message_id'])):
+            if page_command(text) is None and parse_query(text) is None and text.strip() not in ('查看全部','查看全部订单'):
+                return None
+        return state
+    except (ValueError,TypeError):return None
+    finally:connection.close()
+
+
+def save_state(conv_id,state,message_id):
+    """Called only after the reply and customer binding have been persisted."""
+    from .customer_service import current_customer_id,current_scope_key
+    from .visitor_service import visitor_id
+    if not conv_id or type(message_id) is not int:return
+    connection=connect()
+    try:
+        state=dict(state)
+        state['message_id']=message_id
+        connection.execute('UPDATE chat_history SET order_list_state=? WHERE id=? AND visitor_id=? '
+            'AND customer_id=? AND customer_scope=? '
+            'AND NOT EXISTS(SELECT 1 FROM messages WHERE conv_id=? AND id>?)',
+            (json.dumps(state,ensure_ascii=False),conv_id,visitor_id(),current_customer_id(),current_scope_key(),conv_id,message_id))
+        connection.commit()
+    finally:connection.close()
+
+
+def read_orders(tenants,query,page=1,size=5):
+    """One scoped snapshot; each order occupies one actual-node bucket."""
+    from .order_data import tenant_filter
+    scope=tenant_filter(tenant_ids=tenants)
+    if scope is None or not valid_filters(query):raise ValueError('Explicit customer scope and filters required')
+    connection=connect()
+    try:
+        connection.row_factory=sqlite3.Row
+        where=' WHERE 1=1'+scope[0];values=scope[1]
+        if query['status']:where+=' AND status=?';values=values+[query['status']]
+        connection.execute('BEGIN')
+        rows=connection.execute('SELECT order_no,inbound_code,status,warehouse_code,tracking_json,exception_reason '
+            'FROM order_test_data'+where+' ORDER BY COALESCE(julianday(source_updated_at),julianday(snapshot_at),0) DESC,id DESC',values)
+        total=with_inbound=0;groups=Counter();selected=[];tail=deque(maxlen=size)
+        for row in rows:
+            item=dict(row)
+            item['exception_reason']=(item['exception_reason'] or '').strip()
+            if query['exception'] and not item['exception_reason']:continue
+            item['stage']=current_stage(item.pop('tracking_json'))
+            if query['stage'] and query['stage']!=item['stage']:continue
+            total+=1;with_inbound+=bool((item['inbound_code'] or '').strip());groups[item['stage']]+=1
+            tail.append(item)
+            if (page-1)*size<total<=page*size:selected.append(item)
+        last=max(1,(total+size-1)//size)
+        if page>last:
+            page=last;selected=list(tail)[-(total-(last-1)*size):] if total else []
+        return {'total':total,'with_inbound':with_inbound,'groups':groups,'rows':selected,'page':page}
+    finally:connection.close()
+
+
+def row_text(row):
+    from .order_data import STATUS_LABELS
+    def clean(value):return re.sub(r'[\r\n\t]+',' ',str(value or '未登记'))
+    text=f'- 入仓码：{clean(row["inbound_code"])}；订单号：{clean(row["order_no"])}；业务状态：{STATUS_LABELS.get(row["status"],row["status"])}；目的仓：{clean(row["warehouse_code"])}；当前节点：{STAGE_LABELS[row["stage"]]}'
+    if row['exception_reason']:text+='；异常原因：'+clean(row['exception_reason'])
+    return text
+
+
+def unsupported(listing=False):
+    return {'t':f'暂不支持按这个条件{"列出订单" if listing else "统计"}。可以查询订单总数、业务状态、明确标记的异常，或按当前实际物流节点统计；请一次指定一个节点。'}
+
+
+def answer_summary(intent,tenants,conv_id=None):
+    from .customer_service import order_login_response
+    from .order_data import STATUS_LABELS
+    if not tenants:return order_login_response()
+    if intent['unsupported']:return unsupported()
+    query=filters(intent)
+    try:
+        data=read_orders(tenants,query)
+        total=data['total']
+        if query['stage']:
+            text=f'您当前有 {total} 票当前节点为“{STAGE_LABELS[query["stage"]]}”的订单'
+        elif query['status']:
+            text=f'您当前有 {total} 票业务状态为“{STATUS_LABELS[query["status"]]}”的订单'
+        elif query['exception']:text=f'您当前有 {total} 票明确标记异常的订单'
+        else:text=f'您当前共有 {total} 票订单'
+        if query['exception'] and (query['stage'] or query['status']):text+='（已明确标记异常）'
+        if query['stage'] and query['status']:text+=f'（业务状态：{STATUS_LABELS[query["status"]]}）'
+        lines=[text+f'，其中 {data["with_inbound"]} 票已登记入仓码。']
+        if intent.get('grouped'):
+            lines.append('按当前实际物流节点分布（每票只计一次）：')
+            lines.extend(f'- {label}：{data["groups"][code]}票' for code,label in STAGE_LABELS.items() if data['groups'][code])
+        if data['rows']:
+            lines.append(f'最近更新的{len(data["rows"])}票：')
+            lines.extend(row_text(row) for row in data['rows'])
+            lines.append('回复“查看全部”展开列表，或发送入仓码查询详情。')
+        state={'query':query,'page':1,'size':5}
+        previous=load_state(conv_id)
+        if previous and previous['query']==query:state={key:previous[key] for key in ('query','page','size')}
+        return {'t':'\n'.join(lines),'_customer_data':True,'_order_list_state':state}
+    except sqlite3.Error:return {'t':'订单信息暂时无法读取，请稍后重试。'}
+
+
+def answer_list(message,tenants,conv_id=None):
+    from .order_data import is_lookup_help,STATUS_LABELS
+    from .customer_service import order_login_response
+    command=page_command(message)
+    expand=message.strip(' ，,。.!！?？') in ('查看全部','查看全部订单')
+    initial=(not is_lookup_help(message) and re.search(r'订单|运单|\borders?\b|\bshipments?\b',message,re.I)
+             and re.search(r'列出|列表|全部|所有|\blist\b',message,re.I))
+    if command is None and not initial and not expand:return None
+    if not tenants:return order_login_response()
+    try:
+        previous=load_state(conv_id)
+        if command is not None or expand:
+            if not previous:return {'t':'请先在当前对话中发送“列出我的订单”或查询订单数量，再翻页查看。'}
+            query=previous['query'];size=20 if expand else previous['size']
+            page=1 if expand else int(command) if isinstance(command,int) else max(1,previous['page']+(1 if command=='next' else -1))
+            if not 1<=page<=10000:return {'t':'请提供1至10000之间的有效页码。'}
+        else:
+            intent=parse_query(message,listing=True)
+            if intent is None or intent['unsupported']:return unsupported(True)
+            query=filters(intent);page=1;size=20
+        data=read_orders(tenants,query,page,size);page=data['page'];total=data['total']
+        state={'query':query,'page':page,'size':size}
+        if not total:return {'t':'当前账号下没有符合条件的订单。','_customer_data':True,'_order_list_state':state}
+        label=STATUS_LABELS[query['status']] if query['status'] else '全部'
+        first=f'先列出{size}票；' if page==1 and total>size else ''
+        extra=('；当前节点：'+STAGE_LABELS[query['stage']] if query['stage'] else '')+('；明确标记异常' if query['exception'] else '')
+        lines=[f'您当前的订单列表（{first}第{page}页；业务状态：{label}{extra}）：',f'共{total}票，当前显示第{(page-1)*size+1}至{(page-1)*size+len(data["rows"])}票。']
+        lines.extend(row_text(row) for row in data['rows'])
+        lines.append('回复“下一页”查看后续订单。' if page*size<total else '当前已是最后一页。')
+        lines.append('发送入仓码可查询该票订单和物流轨迹。')
+        return {'t':'\n'.join(lines),'_customer_data':True,'_order_list_state':state}
+    except sqlite3.Error:return {'t':'订单信息暂时无法读取，请稍后重试。'}

logistics_chat/order_query.py

@@ -0,0 +1,61 @@
+"""Recognize explicit, whitelisted order summary and list filters."""
+
+import re
+
+from .order_tracking import NODE_LABELS
+
+
+NODE_PHRASES = (
+    ('DOMESTIC_RECEIVED',r'已入国内仓|已入仓'),
+    ('LOADING_COMPLETED',r'已完成装柜|已装柜'),
+    ('DEPARTED_WAREHOUSE',r'已离开国内仓库|已离仓|已出库'),
+    ('AT_ORIGIN_PORT',r'重箱已进入起运港或码头|已到起运港|已进起运港'),
+    ('LOADED_ON_VESSEL',r'已装船'),
+    ('DEPARTED_ORIGIN_PORT',r'船舶已离开起运港|已开船|开船'),
+    ('IN_TRANSIT',r'海运途中|运输中|在途|海上'),
+    ('ARRIVED_DESTINATION_PORT',r'船舶已到目的港|已到目的港|已到港|到港'),
+    ('DISCHARGED_FROM_VESSEL',r'已卸船、未清关|卸船未清关|已卸船|卸船'),
+    ('PICKUP_READY',r'已清关、可提柜|已清关|清关完成|可提柜'),
+    ('PICKED_UP',r'已提柜|提柜'),
+    ('EMPTY_RETURNED',r'已还柜|还柜'),
+    ('NO_ACTUAL',r'没有实际轨迹|暂无实际节点|没有轨迹|无轨迹'),
+    ('MIXED',r'多柜进度不一致|多柜进度不同'),
+)
+STAGE_LABELS = {code:labels[1] for code,labels in NODE_LABELS.items()}
+STAGE_LABELS.update(NO_ACTUAL='暂无实际节点',MIXED='多柜进度不同')
+STATUSES = (('头程订单已完成','COMPLETED'),('已完成','COMPLETED'),('履约中','PROCESSING'),
+            ('处理中','PROCESSING'),('已取消','CANCELLED'),('审核通过','APPROVED'),
+            ('审核驳回','REJECTED'),('已接单','ACCEPTED'),('已提交','SUBMITTED'),('草稿','DRAFT'))
+NOUN = r'订单|运单|入仓码|入仓号|\borders?\b|\bshipments?\b'
+COUNT = r'多少|几(?:个|票|条|张|笔)|总数|数量|总量|统计|数一下|(?:订单|运单)数(?!量)|\bhow\s+many\b|\bcount\b|\bnumber\b|\btotal\b'
+
+
+def parse_query(message,listing=False):
+    if not re.search(NOUN,message,re.I):return None
+    if not listing and not (re.search(COUNT,message,re.I) or re.search(r'分布|(?:都|分别).*到哪',message)):return None
+    if re.search(r'公斤|千克|\bkg\b|重量|毛重|体积|件数|箱数|运费|费用|价格|多少钱',message,re.I):return None
+    remaining=message
+    statuses=[];nodes=[]
+    # Unknown/negated logistics conditions must never silently become a total.
+    unsupported=bool(re.search(r'未到港|没到港|还没|未开船|未发货|未签收|已签收|已送达|曾经|曾到|到过|正常订单|无异常',remaining))
+    for code,pattern in NODE_PHRASES:
+        if re.search(pattern,remaining):
+            nodes.append(code);remaining=re.sub(pattern,'',remaining)
+    for phrase,status in STATUSES:
+        if phrase in remaining:
+            statuses.append(status);remaining=remaining.replace(phrase,'')
+    exception=bool(re.search(r'异常',remaining))
+    remaining=re.sub(r'物流异常|异常','',remaining)
+    grouped=not nodes and bool(re.search(r'轨迹|节点|阶段|到哪',remaining))
+    remaining=re.sub(r'物流轨迹|物流节点|当前节点|轨迹|节点|阶段|到哪里|到哪|各自|各个|每个|分别|分布|各|都|了', '',remaining)
+    remaining=re.sub(COUNT+'|'+NOUN,'',remaining,flags=re.I)
+    remaining=re.sub(r'当前账号|当前账户|账号下|账户下|名下|可查询|能查到|可以查到|目前|现在|当前|一共|总共|合计|总计|'
+        r'请问|麻烦|帮我|想知道|想查|想看|请|查询|查一下|查下|查查|看一下|看看|看下|一下|下|列出|列表|显示|全部|所有|共有|我的|我|能|有|的|个|票|条|张|笔|呢|吗|呀|啊|共|'
+        r'\b(?:list|what|is|of|do|i|have|are|there|my|in|all|the|currently)\b|[\s，,。.!！?？、：:]','',remaining,flags=re.I)
+    return {'status':statuses[0] if statuses else None,'stage':nodes[0] if nodes else None,
+            'exception':exception,'grouped':grouped,
+            'unsupported':bool(unsupported or remaining or len(set(statuses))>1 or len(set(nodes))>1)}
+
+
+def filters(intent):
+    return {key:intent.get(key) for key in ('status','stage','exception')}

logistics_chat/order_tracking.py

@@ -33,14 +33,14 @@ def parse_time(value):
         return None
 
 
-def build_tracking(raw):
-    """Latest observation corrects a milestone; forecasts never advance progress."""
+def tracking_facts(raw):
+    """Share the same corrections and actual/estimated split with aggregation."""
     try:
         facts = json.loads(raw) if isinstance(raw, str) else raw
     except (ValueError, TypeError):
-        return None
+        return {},{}
     if not isinstance(facts, list):
-        return None
+        return {},{}
     actual, estimated = {}, {}
     for index, fact in enumerate(facts):
         if not isinstance(fact, dict) or not isinstance(fact.get('code'), str) or fact['code'] not in NODE_LABELS:
@@ -55,8 +55,27 @@ def build_tracking(raw):
             candidate = (observation, index, when, fact)
             if key not in target or candidate[:2] > target[key][:2]:
                 target[key] = candidate
-    if not actual and not estimated:
-        return None
+    return actual,estimated
+
+
+def current_stage(raw):
+    actual,estimated=tracking_facts(raw)
+    progress={}
+    for (shipment,code) in actual:
+        if shipment not in progress or NODE_LABELS[code][0]>NODE_LABELS[progress[shipment]][0]:
+            progress[shipment]=code
+    shipments={shipment for shipment,_ in actual.keys() | estimated.keys() if shipment}
+    if shipments:
+        if not any(shipment in progress for shipment in shipments) and '' in progress:return progress['']
+        progress={shipment:progress.get(shipment,'NO_ACTUAL') for shipment in shipments}
+    codes=set(progress.values())
+    return next(iter(codes)) if len(codes)==1 else 'MIXED' if codes else 'NO_ACTUAL'
+
+
+def build_tracking(raw):
+    """Latest observation corrects a milestone; forecasts never advance progress."""
+    actual,estimated=tracking_facts(raw)
+    if not actual and not estimated:return None
 
     shipments = sorted({key[0] for key in actual.keys() | estimated.keys()} - {''})
     multiple = len(shipments) > 1

logistics_chat/pricing.py

@@ -94,15 +94,36 @@ RATES = {
 
 CBM_CHANNELS = ("卡航", "普卡", "铁路", "海运")
 
-def parse_dims(t):
-    """解析长x宽x高，支持小数、空格及 x/×/* 分隔。
+DIMENSION_NUMBER = r'(?<![\d.+-])(\d+(?:\.\d+)?)(?![\d.])'
+DIMENSION_UNIT = r'\s*(英寸|毫米|厘米|公分|米|(?:inches|inch|in|cm|mm|m)(?=$|[^A-Za-z0-9²³]|[xX])|["″])?\s*'
+DIMENSION_AXIS = DIMENSION_NUMBER + DIMENSION_UNIT
+DIMENSION_PATTERNS = tuple(re.compile(pattern,re.I) for pattern in (
+    DIMENSION_AXIS+r'[*xX×]\s*'+DIMENSION_AXIS+r'[*xX×]\s*'+DIMENSION_AXIS,
+    r'长\s*'+DIMENSION_AXIS+r'宽\s*'+DIMENSION_AXIS+r'高\s*'+DIMENSION_AXIS,
+    DIMENSION_AXIS+r'长\s*'+DIMENSION_AXIS+r'宽\s*'+DIMENSION_AXIS+r'高',
+))
 
-    用"最长连续数字"匹配三组，要求每组前后都不与其它数字相连，
-    这样 "100x80x60 200kg" 里的 60 不会被后面的 200 吞成 6020。
-    """
-    match = re.search(r'(?<![\d.+-])(\d+(?:\.\d+)?)\s*[*xX×]\s*'
-                      r'(\d+(?:\.\d+)?)\s*[*xX×]\s*(\d+(?:\.\d+)?)(?![\d.])', t)
-    return [float(value) for value in match.groups()] if match else None
+
+def dimension_measurement(text):
+    """Keep units attached to the three packaging axes, not product descriptions."""
+    matches=[match for pattern in DIMENSION_PATTERNS if (match:=pattern.search(text))]
+    if not matches:return None
+    match=min(matches,key=lambda item:item.start())
+    parts=match.groups()
+    values=[float(parts[index]) for index in (0,2,4)]
+    aliases={'英寸':'in','inches':'in','inch':'in','"':'in','″':'in',
+             '毫米':'mm','厘米':'cm','公分':'cm','米':'m'}
+    units=[aliases.get(value.lower(),value.lower()) if value else None for value in parts[1::2]]
+    default=next((unit for unit in reversed(units) if unit),'cm')
+    units=[unit or default for unit in units]
+    if len(set(units))==1:return values,units[0]
+    factors={'cm':1,'mm':0.1,'m':100,'in':2.54}
+    return [value*factors[unit] for value,unit in zip(values,units)],'cm'
+
+
+def parse_dims(t):
+    measurement=dimension_measurement(t)
+    return measurement[0] if measurement else None
 
 
 SHIPPING_REGION_PATTERNS = tuple((re.compile(pattern, re.I), region) for region, pattern in (
@@ -142,11 +163,6 @@ def parse_shipping_intent(text):
 
     # 尺寸：在原始文本上解析（空格有助于区分尺寸和后面的重量）
     intent["dims"] = parse_dims(text)
-    if not intent["dims"]:
-        # 支持“长100宽80高60”
-        dl = re.search(r"长(\d+(?:\.\d+)?)宽(\d+(?:\.\d+)?)高(\d+(?:\.\d+)?)", t)
-        if dl:
-            intent["dims"] = [float(x) for x in dl.groups()]
 
     # 重量：带单位的数字。去掉空格后在原始文本上匹配，
     # 但不能把尺寸里的数字（如 120*100*100 的第三个数）当成重量
@@ -177,20 +193,24 @@ def parse_shipping_intent(text):
     return intent
 
 def detect_dim_unit(text):
-    """识别尺寸单位：cm（默认）/ m / mm"""
+    """Recognize explicit dimensions; aggregate CBM does not imply meters."""
+    measurement=dimension_measurement(text)
+    if measurement:return measurement[1]
     low = text.lower()
+    if re.search(r'英寸|\binch(?:es)?(?![A-Za-z])|\d\s*in(?![A-Za-z0-9])|\d\s*["″]',low):
+        return 'in'
     if re.search(r"毫米|\d\s*mm", low):
         return "mm"
-    if re.search(r"米", low) and not re.search(r"厘米", low):
-        return "m"
     if re.search(r"\d\s*cm|\d\s*厘米", low):
         return "cm"
-    if re.search(r"\d\s*m", low):          # 注意：cm 里也含 m，所以放在 cm 之后判断
+    if re.search(r"\d\s*(?:m(?![A-Za-z0-9²³])|米)", low):
         return "m"
     return "cm"
 
 def dims_to_cm(dims, unit):
     """统一换算成厘米，体积重公式按 cm 计算"""
+    if unit == 'in':
+        return [d * 2.54 for d in dims]
     if unit == "m":
         return [d * 100 for d in dims]
     if unit == "mm":

logistics_chat/quote_answers.py

@@ -7,9 +7,29 @@ from .quote_fees import document, amount, material_fee, limit_fees
 
 
 def answer_business(intent,book):
-    from .quote_engine import number
+    from .quote_engine import number,exact_number
 
     text=intent['text'];mode=intent.get('mode')
+    if mode=='weight_basis':
+        return '运费按计费重量计算，不是只看净重。实重按含包装的毛重，与材积重取较大值；美国海派材积重为长×宽×高(cm)÷6000，每箱不足12KG按12KG计费。海卡包税按对应公斤规则，不包税按方方案以实际体积和毛重÷363取大；请按所选方案核算。'
+    if mode=='rounding':
+        weight=(f'若{exact_number(intent["weight"])}KG已是计费重，就按{exact_number(intent["weight"])}KG计算。' if intent.get('weight') else '')
+        return '美国海派按原小数计费，不向上取整；每箱不足12KG仍按12KG计费。'+weight+'计费重先按各箱实重与材积重取大，再按渠道对应的重量门槛匹配档位。海卡等其他方案的进位方式需另行核对。'
+    if mode=='mixing':
+        return '请分别提供各产品的品名、材质、数量、重量，以及是否带电、带磁、液体或粉末，并说明是否放在同一箱。需先核对各品类的承运与申报要求，再确认混装方案，不能仅凭“两种产品”判断可以混装。'
+    if mode=='metal_scope':
+        from .quote_fees import material_charge
+        from .quote_engine import quantity
+        per_kg=material_fee(book)
+        if per_kg is None:return '铁铝附加费只按铁铝货物的计费重收取，当前未匹配到费率，请联系人工核对。'
+        lines=[f'铁铝附加费为{per_kg:g}元/KG，只按铁铝货物的计费重收取；一票只有部分铁铝货物时，不按整票计费重加收。']
+        unit='cbm' if intent.get('tax')=='excluded' and intent.get('cbm') else 'kg'
+        rate={'kind':'truck' if intent.get('kinds')==['truck'] else 'parcel','unit':unit,'service':''}
+        details=material_charge(intent,quantity(intent,rate),unit,book)
+        lines+=details['items']+details['notes']
+        if intent.get('metal_scope')!='partial' and details['items'] and not intent.get('dims'):
+            lines.append('以上按当前提供的重量条件估算，仍需核对各箱尺寸及材积重。')
+        return '\n'.join(dict.fromkeys(lines))
     if mode=='currency':
         return '当前海卡、海派运价中的“元/KG”和“元/CBM”按人民币（RMB/CNY）计；一件代发等资料中标注“$”“USD”“美金”的费用按美元计。不同币种分别列明，未提供结算汇率时不自动换算。'
     if mode=='insurance':
@@ -74,10 +94,17 @@ def answer_business(intent,book):
         return '请提供品名、材质、是否带电及所选渠道，先核对承运条件，再计算对应费用。'
     if mode=='fees':
         lines=[]
+        if re.search(r'含税|含派送|全包|包不包',text):
+            lines.append('报价行标注“包税”或“不包税”的方案按各自税费方式计算；海派的派送还需核对目的仓/邮编，住宅、商业地址及偏远等附加费不能直接当作已全部包含。')
         metal=material_fee(book)
         if metal is not None and re.search(r'铁制|钢制|铝制|铁铝|不锈钢',text):
-            lines.append(f'当前资料铁铝制品附加费为{metal:g}元/KG。')
-            if intent.get('weight'):lines.append(f'先按已提供重量{number(intent["weight"])}KG计算该项：{number(intent["weight"])}×{metal:g}={number(Decimal(intent["weight"])*metal)}元；计费重确定后继续核算。')
+            from .quote_fees import material_charge
+            from .quote_engine import quantity
+            unit='cbm' if intent.get('tax')=='excluded' and intent.get('cbm') else 'kg'
+            charge=quantity(intent,{'kind':'truck' if intent.get('kinds')==['truck'] else 'parcel','unit':unit,'service':''})
+            details=material_charge(intent,charge,unit,book)
+            lines.append(f'铁铝附加费为{metal:g}元/KG，只按铁铝货物的计费重收取。')
+            lines+=details['items']+details['notes']
         if re.search(r'超重|超尺寸|超长',text):
             raw=document(book,('超重收',))
             fee=amount(raw,r'超重收\s*[+＋]?\s*(\d+(?:\.\d+)?)')
@@ -93,6 +120,6 @@ def answer_business(intent,book):
             if remote is not None and extended is not None:
                 lines.append(f'当前美国海派资料：偏远地址加{remote:g}元/KG，超偏远地址加{extended:g}元/KG，遥远地址单询；海派每箱不足12KG按12KG计费。')
             lines.append('报价中的包税/不包税是不同费率口径；材质、报关申报、偏远及超重超尺寸等按条件单列，不代表所有服务费用都已包含。')
-            if metal is not None:lines.append(f'铁铝制品附加费{metal:g}元/KG；多项费用分别列明，同箱超限费用只取最高项。')
+            if metal is not None:lines.append(f'铁铝制品附加费{metal:g}元/KG，只按铁铝货物计费重收取；多项费用分别列明，同箱超限费用只取最高项。')
         return '\n'.join(lines)
     return None

logistics_chat/quote_context.py

@@ -9,7 +9,10 @@ QUANTITY_QUESTION='箱数已更改，请确认新的总重或各箱重量，原
 
 def merge_intent(current,previous):
     result=deepcopy(current)
-    excluded={'mode','text','errors','dim_unit','quantity_note','quantity_uncertain','preserve_total'}
+    excluded={'mode','text','errors','dim_unit','quantity_note','quantity_uncertain','preserve_total','metal_weight_update'}
+    if current.get('metal_weight_update') and previous.get('metal_chargeable_weight') is None:
+        result['metal_chargeable_weight']=None
+        result['errors'].append('请确认铁铝部分的计费重，不能把毛重直接当作计费重。')
     if current.get('warehouse') or current.get('unknown_warehouses'):
         excluded.update(('postal','zone','zones'))
     elif current.get('postal'):
@@ -28,6 +31,11 @@ def merge_intent(current,previous):
         if not current.get('allocations'):excluded.add('allocations')
     if current['dims']:excluded.add('cbm')
     if current['cbm'] is not None:excluded.add('dims')
+    if any(current.get(k) is not None for k in ('weight','dims','cbm','boxes','packages','allocations','transport','kinds','services','tax')) and current.get('metal_chargeable_weight') is None:
+        excluded.add('metal_chargeable_weight')
+    if current.get('material')=='ordinary':excluded.update(('metal_scope','metal_chargeable_weight'))
+    if current.get('metal_scope')=='all' and current.get('metal_chargeable_weight') is None:
+        excluded.add('metal_chargeable_weight')
     for key,value in previous.items():
         if result.get(key) is None and key not in excluded:result[key]=deepcopy(value)
     if current['dims'] is None and 'dims' not in excluded and previous['dims']:

logistics_chat/quote_edits.py

@@ -0,0 +1,14 @@
+"""Discard explicitly rejected quote conditions before parsing their replacements."""
+
+import re
+
+
+def clean_conditions(text):
+    rejected=r'(?:不是|不要|并非|不从)\s*(?:(?:发|寄|运|送)(?:到|往)?|去|到)?\s*'
+    value=(r'(?:[A-Za-z]{2,4}[0-9][A-Za-z0-9]{0,2})(?![A-Za-z0-9])|'
+           r'深圳|华南|东莞|义乌|华东|美西|美中|美东|美国|洛杉矶|纽约|铁制品|钢制品|铝制品|铁铝制品|金属制品')
+    return re.sub(rejected+'(?:'+value+')',' ',text,flags=re.I)
+
+
+def weight_comparison(text):
+    return bool(re.search(r'会不会按|是不是按|是否按|向上进位|小数公斤|计费.*(?:进位|取整)|公斤.*(?:进位|取整)',text))

logistics_chat/quote_engine.py

@@ -116,6 +116,17 @@ def number(value, places=2):
     return format(Decimal(value).quantize(Decimal(10)**-places, rounding=ROUND_HALF_UP), 'f').rstrip('0').rstrip('.') if places else str(value)
 
 
+def exact_number(value):
+    text=format(Decimal(value),'f')
+    return text.rstrip('0').rstrip('.') if '.' in text else text
+
+
+def quantity_number(value,intent):
+    if Decimal(value)==Decimal(intent.get('weight') or '0') or Decimal(value).as_tuple().exponent>=-6:
+        return exact_number(value)
+    return '约'+number(value,6)
+
+
 def render_rates(intent, book, destinations, remote_fees):
     from .quote_fees import charge_details
     from .quote_language import translate_quote
@@ -124,12 +135,14 @@ def render_rates(intent, book, destinations, remote_fees):
     if intent.get('mode')=='time':
         lines.append('美国线路参考时效：')
     elif intent['weight']:
-        lines.append(f"美国 {number(intent['weight'])}KG 基础运费：")
+        lines.append(f"美国 {exact_number(intent['weight'])}KG 基础运费：")
     else:
         lines.append('美国线路报价：')
     if intent.get('quantity_note'):lines.append(intent['quantity_note'])
     if intent.get('allocations'):
         lines.append('按各目的仓分票计算，箱规需逐票核对：')
+        if intent.get('metal_chargeable_weight') is not None:
+            lines.append('请分别提供各目的仓的铁铝货物计费重，整票铁铝基数不能重复套用到每个仓。')
     elif len(intent.get('warehouses') or [])>1:
         lines.append('以下按各目的仓分别比较；实际分票请提供每个仓的重量。')
     if intent.get('air_requested') and not intent.get('air_only'):
@@ -138,6 +151,8 @@ def render_rates(intent, book, destinations, remote_fees):
         part=destination.get('allocation')
         billing=({**intent,'weight':part['weight'],'packages':part.get('packages'),'dims':None,'cbm':None,'boxes':part.get('boxes')}
                  if part else intent)
+        if part and intent.get('metal_chargeable_weight') is not None:
+            billing.update(metal_chargeable_weight=None,metal_scope='partial')
         rates = match_rates(billing, book, destination)
         if intent.get('mode') in ('cheapest','fastest') and rates:
             if intent['mode']=='cheapest':
@@ -181,7 +196,7 @@ def render_rates(intent, book, destinations, remote_fees):
                     total = amount*price
                     details=charge_details(billing,rate,fees,book)
                     total+=details['cny']
-                    rate_text += f"，{number(amount,3)}{'KG' if unit == 'kg' else 'CBM'} = {number(total)}元"
+                    rate_text += f"，{quantity_number(amount,billing)}{'KG' if unit == 'kg' else 'CBM'} = {number(total)}元"
                     if details['items']:rate_text+='（'+ '；'.join(details['items'])+'）'
                     for note in details['notes']:
                         if note not in lines:lines.append(note)

logistics_chat/quote_fees.py

@@ -22,6 +22,35 @@ def material_fee(book):
     return amount(text,r'铁铝制品[^\n]*?[￥¥]\s*(\d+(?:\.\d+)?)\s*/KG')
 
 
+def material_charge(intent,charged,unit,book):
+    result={'cny':Decimal(0),'items':[],'notes':[]}
+    material=intent.get('material')
+    if material not in ('metal','unknown_metal'):return result
+    per_kg=material_fee(book)
+    if per_kg is None:
+        result['notes'].append('铁铝附加费费率尚未明确，当前金额未包含该项，请联系人工确认。')
+        return result
+    if material=='unknown_metal':
+        result['notes'].append('请先确认金属是否属于铁铝制品；铁铝附加费只按铁铝货物的计费重收取，当前金额未包含待核实的材质费用。')
+        return result
+    basis=intent.get('metal_chargeable_weight')
+    if basis is None and intent.get('metal_scope')=='partial':
+        result['notes'].append(f'铁铝附加费{per_kg:g}元/KG只按铁铝货物的计费重收取，不能按整票重量加费；请提供铁铝部分的计费重，当前金额未包含该项。')
+        return result
+    if basis is None and unit!='kg':
+        result['notes'].append(f'铁铝附加费为{per_kg:g}元/KG，当前按方运费未包含该项；需确认铁铝货物计费重量后单列，不能把CBM直接乘公斤附加费。')
+        return result
+    basis=Decimal(basis) if basis is not None else charged
+    if basis is None:return result
+    if unit=='kg' and charged is not None and basis>charged:
+        result['notes'].append('铁铝货物计费重高于当前整票可计算的计费重，请补充各箱实重和尺寸核对；当前金额未包含该项。')
+        return result
+    fee=basis*per_kg
+    result['cny']=fee
+    result['items'].append(f'铁铝附加费{per_kg:g}元/KG×{basis:g}KG={fee:g}元')
+    return result
+
+
 def specimens(intent):
     if intent.get('packages'):
         return [{**p,'dims':p.get('dims') or intent.get('dims'),
@@ -83,13 +112,8 @@ def charge_details(intent,rate,remote,book):
             fee=charged*Decimal(remote['fee']);result['cny']+=fee
             result['items'].append(f'偏远费{remote["fee"]}元/KG×{charged:g}KG={fee:g}元')
         else:result['notes'].append('遥远地址的附加费需单独确认，当前金额未包含该项。')
-    if intent.get('material')=='metal':
-        per_kg=material_fee(book)
-        if per_kg is not None and rate['unit']=='kg':
-            fee=charged*per_kg;result['cny']+=fee
-            result['items'].append(f'铁铝附加费{per_kg:g}元/KG×{charged:g}KG={fee:g}元')
-        elif per_kg is not None:
-            result['notes'].append(f'铁铝附加费为{per_kg:g}元/KG，当前按方运费未包含该项；需确认适用计费重量后单列，不能把CBM直接乘公斤附加费。')
+    metal=material_charge(intent,charged,rate['unit'],book)
+    result['cny']+=metal['cny'];result['items']+=metal['items'];result['notes']+=metal['notes']
     if rate['kind']=='parcel':
         limits=limit_fees(intent,book)
         result['cny']+=limits['cny'];result['items']+=limits['items'];result['notes']+=limits['notes']

logistics_chat/quote_intent.py

@@ -8,7 +8,7 @@ from .pricing import parse_shipping_intent, detect_dim_unit, shipping_region
 US = re.compile(r'美国|美西|美中|美东|洛杉矶|纽约|芝加哥|休斯顿|萨凡纳|'
     r'(?<![A-Za-z])(?:USA|(?-i:US)|United States|America)(?=$|[^A-Za-z]|FBA(?![A-Za-z]))|'
     r'(?:发(?:往|到)?|寄(?:往|到)?|运往|运到|到|往|\bto\s+)\s*us(?=$|[^A-Za-z]|FBA(?![A-Za-z]))',re.I)
-QUOTE_WORDS = re.compile(r'报价|价格|运费|费用|多少钱|计费|包税|海运|空运|海派|海卡|海铁|美森|合德|普船|附加|保险|代发|申报|偏远|超重|超尺寸|币种|\bEXX\b|\b(?:quote|price|cost|freight)\b',re.I)
+QUOTE_WORDS = re.compile(r'报价|价格|运费|费用|收费|多少钱|计费|包税|海运|空运|海派|海卡|海铁|美森|合德|普船|附加|保险|代发|申报|偏远|超重|超尺寸|币种|\bEXX\b|\b(?:quote|price|cost|freight)\b',re.I)
 WEIGHT = re.compile(r'(?<![\d.])(-?\d+(?:\.\d+)?)\s*(kilograms?|kgs?|公斤|千克|grams?|克|g|tons?|吨|lbs?|磅)(?![A-Za-z])',re.I)
 TOPIC_BOUNDARY = re.compile(r'新(?:的)?一(?:票|批)|另外|另一|换个话题|理赔|赔偿|破损|丢失|退件|投诉|查单|轨迹|签收')
 
@@ -31,7 +31,7 @@ def chinese_number(text):
 
 
 def normalize(text):
-    result=re.sub(r'([零〇一二两三四五六七八九十百千万点]+)(?=\s*(?:公斤|千克|克|吨|磅|箱|件|立方米|立方|方|kg|cbm))',
+    result=re.sub(r'([零〇一二两三四五六七八九十百千万点]+)(?=\s*(?:公斤|千克|克|吨|磅|(?:个)?箱|件|立方米|立方|方|kg|cbm))',
         lambda m:chinese_number(m.group(1)),text,flags=re.I)
     return re.sub(r'负(?=\d)','-',result)
 
@@ -48,14 +48,22 @@ def kg(value,unit):
 def request_mode(text):
     # Business topics outrank generic verbs such as “怎么算”.
     if re.search(r'保险|保价|投保|insured|insurance',text,re.I):return 'insurance'
+    from .quote_edits import weight_comparison
+    if weight_comparison(text):return 'rounding'
+    if re.search(r'毛重|净重|计费重.*(?:什么|区别)|材积重.*(?:什么|区别)',text):return 'weight_basis'
+    if re.search(r'混装|混箱|不同产品.*一箱',text):return 'mixing'
+    if (re.search(r'(?:铁|钢|铝|金属).*(?:附加|加费|加钱|整票)|整票.*(?:铁|钢|铝|金属)',text)
+            and not re.search(r'最终总价|一共多少钱|总共多少钱|合计多少钱|最终价格|最终费用|到门总价',text)):
+        return 'metal_scope'
     if re.search(r'纯电池|带电|内置电池|液体|粉末|食品|医疗用品|仿牌',text):return 'acceptance'
     from .shipping_faq import preparation_topic
     if preparation_topic(text):return 'preparation'
-    if re.search(r'最便宜|便宜的|最省钱|最划算|最低价|cheapest|least expensive',text,re.I):return 'cheapest'
+    if re.search(r'最便宜|便宜的|便宜一?点|优惠|折扣|打折|最省钱|最划算|最低价|cheapest|least expensive',text,re.I):return 'cheapest'
+    if re.search(r'最终总价|一共多少钱|总共多少钱|合计多少钱|最终价格|最终费用|到门总价',text):return 'total'
     if re.search(r'最快|快的|fastest',text,re.I):return 'fastest'
     if re.search(r'一件代发|散箱入库|整柜卸货|卸货费|仓租|仓储|贴标费',text):return 'fulfillment'
     if re.search(r'FDA|申报费|报关|续页|品名',text,re.I):return 'customs'
-    if re.search(r'附加费|还.*(?:费|加钱)|(?:额外|其他|其它).*(?:费用|收费)|包含.*费|包税.*(?:包含|包括)|超重费|超尺寸|住宅|私人|商业地址',text):return 'fees'
+    if re.search(r'附加费|含税|含派送|全包|包不包|还.*(?:费|加钱)|(?:额外|其他|其它).*(?:费用|收费)|包含.*费|包税.*(?:包含|包括)|超重费|超尺寸|住宅|私人|商业地址',text):return 'fees'
     if re.search(r'电池|带电|液体|粉末|食品|医疗|能(?:不能|发|走)|可以.*(?:发|走)',text):return 'acceptance'
     if re.search(r'人民币|美元|美金|币种|currency|\b(?:RMB|CNY|USD)\b',text,re.I) and not WEIGHT.search(text):return 'currency'
     if re.search(r'怎么(?:算|计算)|计算(?:过程|出来)|怎么算出来|calculation',text,re.I):return 'calculation'
@@ -65,27 +73,32 @@ def request_mode(text):
 
 def parse_intent(message,warehouse_codes):
     text=normalize(message)
+    from .quote_edits import clean_conditions
+    text=clean_conditions(text)
     text=re.sub(r'(?:而?不是|并非|不要按|不是按)\s*(?:每[箱件]|单[箱件])?\s*'+WEIGHT.pattern,' ',text,flags=re.I)
-    parsed=parse_shipping_intent(text)
+    from .quote_material import material_intent
+    materials=material_intent(text,WEIGHT,kg)
+    quantity_text=materials['quantity_text']
+    parsed=parse_shipping_intent(quantity_text)
     country='US' if US.search(text) else ('other' if shipping_region(text) is not None else None)
     candidates=re.findall(r'(?<![A-Za-z0-9])[A-Za-z][A-Za-z0-9]{2,7}(?![A-Za-z0-9])',text.upper())
     warehouses=list(dict.fromkeys(t for t in candidates if t in warehouse_codes))
     unknown=[t for t in candidates if t not in warehouse_codes and re.fullmatch(r'[A-Z]{2,4}[0-9][A-Z0-9]{0,2}',t)
              and not t.startswith(('USD','RMB','CNY'))]
-    errors=[]
+    errors=materials['errors'][:]
     from .quote_allocations import allocations
-    assigned=allocations(text,warehouses,WEIGHT,kg)
-    weights=[m for m in WEIGHT.finditer(text) if not text[:m.start()].rstrip().endswith(('x','X','×','*'))]
+    assigned=allocations(quantity_text,warehouses,WEIGHT,kg)
+    weights=[m for m in WEIGHT.finditer(quantity_text) if not quantity_text[:m.start()].rstrip().endswith(('x','X','×','*'))]
     weight=str(kg(*weights[0].groups())) if weights else None
-    boxes=re.search(r'(\d+)\s*(?:箱|件|cartons?|boxes?)',text,re.I)
+    boxes=re.search(r'(\d+)\s*(?:个)?(?:箱(?:子)?|件|cartons?|boxes?)',quantity_text,re.I)
     count=int(boxes.group(1)) if boxes else None
     packages=[]
     package_re=re.compile(r'(第)?(\d+)\s*(?:箱|件)\s*(?:[，,]?\s*每(?:箱|件))?\s*'+WEIGHT.pattern.replace(r'(?<![\d.])',''),re.I)
-    matches=list(package_re.finditer(text))
+    matches=list(package_re.finditer(quantity_text))
     package_dimensions={}
     for index,match in enumerate(matches):
         ordinal,n,value,unit=match.groups()
-        segment=text[match.end():matches[index+1].start() if index+1<len(matches) else len(text)]
+        segment=quantity_text[match.end():matches[index+1].start() if index+1<len(matches) else len(quantity_text)]
         dims=parse_shipping_intent(segment)['dims']
         packages.append({'count':1 if ordinal else int(n),'weight':str(kg(value,unit)),
                          'dims':dims,'dim_unit':detect_dim_unit(segment)})
@@ -93,7 +106,7 @@ def parse_intent(message,warehouse_codes):
     for part in assigned or []:
         for piece in part.get('packages') or []:
             piece.update(package_dimensions.get(piece.pop('position'),{}))
-    each=re.search(r'每(?:箱|件)\s*'+WEIGHT.pattern.replace(r'(?<![\d.])',''),text,re.I)
+    each=re.search(r'每(?:箱|件)\s*'+WEIGHT.pattern.replace(r'(?<![\d.])',''),quantity_text,re.I)
     if each and count and not packages:
         packages=[{'count':count,'weight':str(kg(*each.groups())),'dims':parsed['dims'],'dim_unit':detect_dim_unit(text)}]
     if packages:
@@ -103,16 +116,16 @@ def parse_intent(message,warehouse_codes):
             errors.append('箱数与分箱明细不一致，请确认总箱数及每箱重量。')
         count=package_count
         weight=str(sum(Decimal(p['weight'])*p['count'] for p in packages))
-        total=re.search(r'(?:总重|合计重量|一共重)\s*'+WEIGHT.pattern.replace(r'(?<![\d.])',''),text,re.I)
+        total=re.search(r'(?:总重|合计重量|一共重)\s*'+WEIGHT.pattern.replace(r'(?<![\d.])',''),quantity_text,re.I)
         if total and kg(*total.groups())!=Decimal(weight):errors.append('总重与分箱重量合计不一致，请确认。')
         # Per-carton dimensions must not spill onto another carton without dimensions.
         if len(packages)>1:parsed['dims']=None
     elif assigned:
         weight=str(sum(Decimal(a['weight']) for a in assigned))
         if set(a['warehouse'] for a in assigned)!=set(warehouses):errors.append('请分别提供每个目的仓的重量，避免遗漏分票。')
-        total=re.search(r'(?:总重|合计重量)\s*'+WEIGHT.pattern,text,re.I)
+        total=re.search(r'(?:总重|合计重量)\s*'+WEIGHT.pattern,quantity_text,re.I)
         if total and kg(*total.groups())!=Decimal(weight):errors.append('总重与各目的仓分配重量不一致，请确认。')
-    elif len(weights)>1:
+    elif len(weights)>1 and request_mode(text)!='rounding':
         errors.append('请明确总重，或分别提供每箱重量，避免把不同重量混在一起。')
     if assigned and packages and Decimal(weight)!=sum(Decimal(a['weight']) for a in assigned):
         errors.append('分仓重量与分箱重量不一致，请分别确认每仓的箱数和重量。')
@@ -149,8 +162,9 @@ def parse_intent(message,warehouse_codes):
         'unknown_warehouses':unknown or None,'postal':postal,'zone':zones[0] if zones else None,'zones':zones or None,
         'tax':'excluded' if re.search(r'不包税|自税|不含税|tax excluded',text,re.I) else ('included' if re.search(r'包税|tax included',text,re.I) else None),
         'errors':errors,'mode':request_mode(text),'english':True if english else (False if re.search(r'[\u4e00-\u9fff]',message) else None),
-        'material':('ordinary' if re.search(r'普通货|普货|塑料|木制|非铁铝',text) else
-                    ('metal' if re.search(r'铁制|钢制|铝制|铁铝|不锈钢',text) else None)),
+        'material':materials['material'],'metal_scope':materials['metal_scope'],
+        'metal_chargeable_weight':materials['metal_chargeable_weight'],
+        'metal_weight_update':materials['metal_weight_update'],
         'address_type':('non_fba' if re.search(r'私人|住宅|商业地址|非亚马逊',text) else ('fba' if warehouses else None)),
         'preserve_total':True if re.search(r'总重不变|重量不变',text) else None,'text':text}
 
@@ -166,4 +180,4 @@ def is_followup(message,intent):
     if intent.get('mode')!='quote':return not (intent['country'] and intent['weight'])
     if re.fullmatch(r'(?:邮编|zip|postcode)?\s*\d{5}(?:-\d{4})?(?:呢)?',text,re.I):return True
     if re.fullmatch(r'(?:义乌|华东|深圳|华南|美国|美西|美中|美东|东[12]区|西[12]区|中[12]区|不包税|包税)(?:交仓|仓)?(?:呢)?',text):return True
-    return bool(len(text)<=60 and not (intent['country'] and intent['weight']) and any(intent.get(k) is not None for k in ('services','kinds','transport','weight','dims','cbm','tax','origin')))
+    return bool(len(text)<=60 and not (intent['country'] and intent['weight']) and any(intent.get(k) is not None for k in ('services','kinds','transport','weight','dims','cbm','tax','origin','zone','material','metal_scope','metal_chargeable_weight')))

logistics_chat/quote_material.py

@@ -0,0 +1,33 @@
+"""Separate metal chargeable-weight declarations from the shipment's gross weight."""
+
+import re
+from decimal import Decimal
+
+
+def material_intent(text,weight_pattern,to_kg):
+    declarations=re.sub(r'铁铝(?:附加费|加费|费用)','',text)
+    metal=bool(re.search(r'铁制|钢制|铝制|铁铝|不锈钢|(?:铁|钢|铝)(?:件|货|扣|罐|盒|板|管|架|材)',declarations))
+    generic='金属' in text
+    ordinary=bool(re.search(r'普通货|普货|非铁铝',text))
+    partial=bool(re.search(r'只有|部分|其中|有些|只是|一部分',text) or
+                 metal and re.search(r'塑料|木制|布料',text))
+    scope='partial' if partial and (metal or generic) else ('all' if re.search(r'全是|全部是|都是|整票是|货物是|换成|改成',text) and metal else None)
+    material=('ordinary' if '非铁铝' in text else 'metal' if metal else 'unknown_metal' if generic
+              else 'ordinary' if ordinary or re.search(r'塑料|木制',text) else None)
+    name=r'(?:铁铝|铁制品|钢制品|铝制品|铁件|钢件|铝件|金属)(?:部分|货物)?(?:的)?'
+    patterns=(name+r'\s*计费重(?:量)?\s*(?:为|是|[:：])?\s*'+weight_pattern.pattern,
+              name+r'\s*按\s*'+weight_pattern.pattern+r'\s*计费',
+              r'(?:铁铝|铁制品|钢制品|铝制品|铁件|钢件|铝件|金属)部分(?:的计费重(?:量)?)?\s*'
+              r'(?:还是|仍然是|仍是|改为|改成|换成)\s*'+weight_pattern.pattern)
+    charged=None;cleaned=text;errors=[];oral=False
+    for index,pattern in enumerate(patterns):
+        matches=list(re.finditer(pattern,text,re.I))
+        for match in matches:
+            value=to_kg(*match.groups())
+            if charged is not None and Decimal(charged)!=value:errors.append('铁铝货物的计费重不一致，请确认。')
+            charged=str(value);scope='partial'
+            oral=oral or (index==2 and '计费重' not in match[0])
+            if not 0<value<=1000000:errors.append('请提供大于0的有效铁铝货物计费重。')
+            cleaned=cleaned[:match.start()]+' '*(match.end()-match.start())+cleaned[match.end():]
+    return {'material':material,'metal_scope':scope,'metal_chargeable_weight':charged,
+            'quantity_text':cleaned,'errors':errors,'metal_weight_update':True if oral else None}

logistics_chat/quote_scope.py

@@ -0,0 +1,25 @@
+"""Keep unanswered US shipping questions outside legacy regional model knowledge."""
+
+import re
+
+from .quote_intent import US
+from .pricing import shipping_region
+
+
+def answer_us_fallback(message,conv_id):
+    if re.search(r'理赔|赔偿|索赔|破损|丢失|退件|投诉',message):return None
+    if re.search(r'换个话题|另外问',message):return None
+    from .quote_edits import clean_conditions
+    current=clean_conditions(message)
+    active=bool(US.search(current))
+    if not active and shipping_region(current) is not None:return None
+    if not active and conv_id:
+        from .quote_service import previous_quote_intents
+        active=any(intent['country']=='US' for intent in previous_quote_intents(conv_id,set()))
+    if not active:return None
+    from .quote_data import load_book
+    if not load_book():return {'t':'当前尚未包含可用的美国报价表，请提供目的地仓码或邮编及货物信息，联系人工确认。'}
+    english=not re.search(r'[\u4e00-\u9fff]',message) and bool(re.search(r'[A-Za-z]{3,}',message))
+    if english:
+        return {'t':'Please specify the US route or destination warehouse/ZIP code and the point you want to confirm. Unlisted charges or service conditions require confirmation by customer support.'}
+    return {'t':'请说明您想确认的美国渠道、目的仓码或邮编，以及具体的费用或发货要求；未明确的服务条件需要联系人工核对。'}

logistics_chat/quote_service.py

@@ -9,29 +9,39 @@ from .quote_data import load_book, postcode_fee
 from .quote_engine import US, QUOTE_WORDS, parse_intent, is_followup, variants, render_rates
 
 
+def previous_quote_intents(conv_id,codes):
+    """Read the current quote back to its topic boundary, without a turn cutoff."""
+    earlier_intents=[]
+    if conv_id:
+        connection=connect()
+        try:
+            recent=connection.execute("SELECT content FROM messages WHERE conv_id=? AND role='user' ORDER BY id DESC",(conv_id,))
+            for (previous,) in recent:
+                if re.search(r'理赔|赔偿|破损|丢失|退件|投诉|查单|轨迹|签收|换个话题|另外问',previous):break
+                earlier=parse_intent(previous,codes)
+                if earlier['country']=='other':break
+                if not (QUOTE_WORDS.search(previous) or US.search(previous) or is_followup(previous,earlier) or
+                        any(earlier.get(key) for key in ('weight','dims','cbm','warehouse','postal'))):break
+                earlier_intents.append(earlier)
+                if not is_followup(previous,earlier):break
+        finally:
+            connection.close()
+    return earlier_intents
+
+
 def intent_with_context(message, conv_id, book):
     from .quote_context import merge_intent
 
     codes = {row['code'] for row in book['warehouses']}
     codes.update(value.upper() for rate in book['rates'] for value in rate['destinations'])
     intent = parse_intent(message,codes)
-    earlier_intents=[]
-    if conv_id and is_followup(message,intent):
-        connection=connect()
-        try:
-            recent=connection.execute("SELECT content FROM messages WHERE conv_id=? AND role='user' ORDER BY id DESC LIMIT 10",(conv_id,)).fetchall()
-        finally:
-            connection.close()
-        for (previous,) in recent:
-            if re.search(r'理赔|赔偿|破损|丢失|退件|投诉|查单|轨迹|签收',previous):break
-            earlier=parse_intent(previous,codes)
-            if earlier['country']=='other':break
-            if not (QUOTE_WORDS.search(previous) or US.search(previous) or is_followup(previous,earlier) or
-                    any(earlier.get(key) for key in ('weight','dims','cbm','warehouse','postal'))):break
-            earlier_intents.append(earlier)
-            if not is_followup(previous,earlier):break
+    earlier_intents=previous_quote_intents(conv_id,codes) if is_followup(message,intent) else []
+    if intent.get('metal_weight_update') and not earlier_intents:
+        intent['metal_chargeable_weight']=None
+        intent['errors'].append('请确认铁铝部分的计费重，不能把毛重直接当作计费重。')
     if earlier_intents:
         state=earlier_intents[-1]
+        if state.get('metal_weight_update'):state['metal_chargeable_weight']=None
         for update in list(reversed(earlier_intents[:-1]))+[intent]:
             state=merge_intent(update,state)
         intent=state
@@ -73,19 +83,18 @@ def answer_quote(message, conv_id):
         dimensions += [piece['dims'] for piece in intent.get('packages') or [] if piece.get('dims')]
         if any(not 0 < value <= 100000 for dims in dimensions for value in dims):
             return {'t':'请提供大于 0 的有效长宽高。'}
+        if intent['air_only']:
+            return {'t':'当前美国报价资料包含美森、EXX、合德及普船的海卡、海铁和海派。发我目的仓码或邮编、重量，可以列出这些线路的价格；空运费率当前资料没有列出，空运承运条件也需单独确认。'}
         from .quote_answers import answer_business
         preparation=None
         if intent.get('mode')=='preparation':
             from .shipping_faq import answer_preparation
-            if intent['air_only']:return {'t':'当前美国资料未列明空运的交仓和发货要求，请确认具体空运渠道后联系人工核对。'}
             preparation=answer_preparation(intent,book)
             if preparation_topic(message)!='schedule' or not re.search(r'\d+\s*(?:KG|公斤|千克|克|吨|方|CBM)',intent['text'],re.I):
                 return {'t':preparation}
             intent['mode']='cheapest' if re.search(r'最便宜|最划算|最低价',message) else 'quote'
         business=answer_business(intent,book)
         if business is not None:return {'t':business}
-        if intent['air_only']:
-            return {'t':'当前美国报价资料包含美森、EXX、合德及普船的海卡、海铁和海派。发我目的仓码或邮编、重量，可以列出这些线路的价格；空运费率当前资料没有列出。'}
         destinations = variants(intent,book)
         fees = []
         for destination in destinations:
@@ -94,6 +103,10 @@ def answer_quote(message, conv_id):
                 return {'t':f'当前海派资料不接收邮编 {postal} 的地址，请提供其他目的仓码或联系人工确认可用线路。'}
             fees.append(postcode_fee(book['id'],postal) if postal else None)
         reply=render_rates(intent,book,destinations,fees)
+        if intent.get('mode')=='total':
+            reply+='\n以上为已提供条件下可计算的运费，各方案按所列包税/不包税方式分别核算；目的仓、箱规及未确定附加费需核实后才能确认最终应付总价。'
+        if intent.get('mode')=='cheapest' and re.search(r'便宜一?点|折扣|优惠|打折',message):
+            reply+='\n以上按当前价表比较，折扣或合作价格需人工确认，尚未修改单价。'
         return {'t':reply+('\n'+preparation if preparation else '')}
     except (sqlite3.Error, ValueError, KeyError, TypeError, InvalidOperation):
         return {'t':'报价数据暂时无法读取，请稍后重试。'}

logistics_chat/shipping_faq.py

@@ -7,9 +7,10 @@ def preparation_topic(text):
     if re.search(r'理赔|赔偿|索赔|货损|丢件|破损',text):return None
     if re.search(r'截仓|截单|船期|排仓|哪天开船|(?:什么时候|几号).*开船|能.*开船|'
                  r'(?:今天|明天|这周|下周).*(?:发货|发走|开船)|赶得上|来得及|(?:保证|必须|月底).*到',text):return 'schedule'
-    if re.search(r'上门.*(?:提货|取货|收货|揽收)|提货服务|交仓地址|仓库地址|送哪个仓|仓库在哪|收货时间',text):return 'delivery'
-    if re.search(r'报关资料|清关资料|报关.*(?:准备|材料)|申报比例|低申报|怎么申报|如何申报|认证资料',text):return 'documents'
+    if re.search(r'上门.*(?:提货|取货|收货|揽收)|提货服务|交仓地址|仓库地址|仓.*(?:地址|电话|联系人|几点)|送哪个仓|仓库在哪|收货时间|\bpick\s*up\b|\bpickup\b',text,re.I):return 'delivery'
+    if re.search(r'报关资料|清关资料|报关.*(?:准备|材料)|申报比例|低申报|怎么申报|如何申报|认证资料|熏蒸',text):return 'documents'
     if re.search(r'查验.*(?:费|承担)|谁.*承担.*查验',text):return 'inspection'
+    if re.search(r'(?:东|西|中)[12](?:区)?.*(?:区别|分区|哪区)|分区.*(?:什么意思|怎么分)',text):return 'zones'
     if '一件代发' not in text and re.search(r'贴标|FBA标|标签|唛头|产地标|外箱.*贴|贴.*条码|made\s+in\s+china',text,re.I):return 'labels'
     if re.search(r'哪些渠道|哪些线路|什么渠道|能发哪些仓|支持.*仓|(?:能发|能走|可以发).*美国',text):return 'coverage'
     return None
@@ -23,10 +24,14 @@ def answer_preparation(intent,book):
     if topic=='coverage':
         services=list(dict.fromkeys(rate['service'] for rate in book['rates']))
         return '美国可查询的线路：'+ '、'.join(services)+'。请提供目的仓码或邮编、重量及交货仓，我再匹配具体可用线路。'
+    if topic=='zones':
+        return '东1、东2等是报价资料中的不同目的地区域，需要按目的仓码或邮编匹配；同一仓码存在多个分区记录时会分别列出。请提供目的仓码或邮编，我按现有记录核对。'
     if topic=='schedule':
         return '请提供计划交仓日期、交货仓和目的仓码。具体截仓、装柜和开船安排需联系人工核对；当前参考时效不能作为月底或指定日期必达的承诺。'
     if topic=='delivery':
         raw=document(book,('不提供上门提货服务',))
+        if intent.get('english'):
+            return ('The current US route sheet requires delivery to the designated warehouse and does not offer pickup. ' if raw else '')+'Please confirm the route and Yiwu or South China origin with customer support before delivery; warehouse address, contact and receiving hours need confirmation.'
         lines=[]
         if raw:lines.append('当前美国渠道资料要求客户自行送货至对应仓库，不提供上门提货服务。')
         lines.append('请确认所选渠道及义乌或华南交仓，仓库地址、联系人和收货时间请联系人工核对后再送仓。')
@@ -39,6 +44,9 @@ def answer_preparation(intent,book):
         if origin:lines.append('产品和包装还需有“Made in China”产地标识。')
         return '\n'.join(lines) or '请提供渠道和目的仓码，由人工核对外箱标签及代贴要求。'
     if topic=='documents':
+        if '熏蒸' in intent['text']:
+            raw=document(book,('熏蒸','备货阶段'))
+            return (raw.strip()+'\n' if raw else '')+'请确认是木制成品还是木质包装，并提供具体品名及渠道，核对是否需要熏蒸及证明材料后再安排发货。'
         if re.search(r'申报比例|低申报|怎么申报|如何申报',intent['text']):
             return '请提供品名、材质、用途、实际交易金额及销售链接，由业务核对申报资料。申报价需结合实际交易信息确认，不能按统一比例直接代算。'
         raw=document(book,('单独报关需提供资料',))

tests/test_customer_gap_fixes.py

@@ -0,0 +1,254 @@
+"""Oral-language, US-region isolation and scoped-page regressions using synthetic facts."""
+
+from decimal import Decimal
+from pathlib import Path
+import re
+import tempfile
+import unittest
+from unittest.mock import patch
+
+from logistics_chat import create_app,db
+from logistics_chat.order_data import import_records
+from logistics_chat.pricing import detect_dim_unit,dims_to_cm
+from logistics_chat.quote_data import import_snapshot
+from logistics_chat.quote_intent import parse_intent
+from logistics_chat.quote_engine import quantity
+from test_quotes import fixture_snapshot,frames
+
+
+class CustomerGapTests(unittest.TestCase):
+    def setUp(self):
+        folder=tempfile.TemporaryDirectory();self.addCleanup(folder.cleanup)
+        root=Path(folder.name)
+        self.app=create_app({'TESTING':True,'DATABASE':root/'data.db','UPLOAD_DIR':root/'uploads',
+            'BACKUP_DIR':root/'backups','LOG_DIR':root/'logs','DEEPSEEK_API_KEY':'',
+            'CUSTOMER_DEMO_TENANT_ID':'qa-a','RATE_LIMIT_PER_MIN':1000})
+        self.client=self.app.test_client();self.client.post('/api/visitor/session',json={'key':'1'*64})
+        book=fixture_snapshot()
+        book['warehouses'].append({'code':'YYY2','zone':'西1区','state':'AA','postal':'90001',
+            'address':'','source_sheet':'合成分区','source_cell':'A8'})
+        texts=['偏远地址加收4元/kg；超偏远地址加收6元/kg','铁铝制品另外加附加费￥5/KG',
+            '正常货物 0KG<单件重量<22KG 0CM<最长单边长度<120CM 0CM<第二边<75CM 最长边<260CM',
+            '超重收+70/件','客户自行交货，不提供上门提货服务',
+            '禁止运输的货物：纯电池、液体','需熏蒸的货物由客户在备货阶段自行完成']
+        book['documents']=[{'sheet':'合成渠道','cell':f'A{i+1}','text':text} for i,text in enumerate(texts)]
+        self.order={'tenant_id':'qa-a','order_no':'FOCUS20260001','inbound_code':'QAFOCUS20260001',
+            'country_code':'US','status':'COMPLETED','package_count':2,
+            'tracking_json':[{'code':'IN_TRANSIT','actual_at':'2026-10-01 10:00'}]}
+        with self.app.app_context():
+            c=db.connect();import_snapshot(c,book,activate=True)
+            import_records(c,[self.order,{**self.order,'tenant_id':'qa-b','order_no':'OTHER20260099','inbound_code':'QAOTHER20260099'}])
+            c.commit();c.close()
+
+    def conversation(self):return self.client.post('/api/new_chat').json['id']
+
+    def events(self,q,cid=None,client=None):
+        with patch('logistics_chat.ai_client.post_completion') as model:
+            result=frames((client or self.client).post('/stream_chat',json={'msg':q,'conv_id':cid}))
+            model.assert_not_called()
+        self.assertEqual(result[-1],{'d':1},result)
+        return result
+
+    def ask(self,q,cid=None):return self.events(q,cid)[0]['t']
+
+    def login(self):
+        self.assertTrue(self.client.post('/api/customer/login',json={'username':'admin','password':'123456'}).json['ok'])
+
+    def test_us_natural_questions_never_enter_old_regional_model_faqs(self):
+        cases=('你们这个报价是按毛重还是净重？','运费能便宜一点吗？','这个价格含税含派送吗？',
+               '一共多少钱，给我最终总价','这个物流服务范围具体怎么执行？')
+        for q in cases:
+            with self.subTest(q=q):
+                cid=self.conversation();self.ask('美国美西30KG，义乌普船海派',cid)
+                text=self.ask(q,cid)
+                self.assertNotIn('21KG',text);self.assertNotIn('250',text);self.assertNotIn('7元/KG',text)
+        text=self.ask('我想发点货到美国，怎么收费？')
+        self.assertIn('元/KG',text);self.assertIn('多少公斤',text)
+
+    def test_final_total_respects_tax_labels_and_keeps_unconfirmed_fees_separate(self):
+        cid=self.conversation();self.ask('美国美西2箱每箱15KG，40x30x20cm，义乌普船海派',cid)
+        text=self.ask('一共多少钱，给我最终总价',cid)
+        self.assertIn('30KG = 60元',text);self.assertIn('· 包税',text)
+        self.assertNotIn('价格不含税',text);self.assertIn('最终应付总价',text)
+        self.assertIn('折扣或合作价格需人工确认',self.ask('运费能便宜一点吗',cid))
+
+    def test_negative_warehouse_and_origin_edits_do_not_keep_rejected_values(self):
+        cid=self.conversation();self.ask('美国30KG发ZZZ1，深圳普船海派',cid)
+        text=self.ask('不是发ZZZ1，是YYY2',cid)
+        self.assertNotIn('ZZZ1',text);self.assertIn('YYY2',text)
+        text=self.ask('不要深圳，还是义乌交仓',cid)
+        self.assertIn('· 义乌',text);self.assertNotIn('· 华南',text)
+        text=self.ask('50KG呢',cid)
+        self.assertIn('YYY2',text);self.assertNotIn('ZZZ1',text);self.assertIn('美国 50KG',text)
+
+    def test_inches_convert_before_charge_weight_and_acceptance_limits(self):
+        intent=parse_intent('美国30KG，60x40x30英寸，1箱普船海派',set())
+        self.assertEqual(intent['dim_unit'],'in')
+        self.assertEqual(dims_to_cm([60,40,30],'in'),[152.4,101.6,76.2])
+        self.assertGreater(quantity(intent,{'kind':'parcel','unit':'kg'}),Decimal('190'))
+        text=self.ask('美国美西30KG，60x40x30英寸，1箱普船海派，义乌')
+        self.assertIn('人工确认是否接收',text)
+        for text in ('60x40x30 inches','60x40x30in','60x40x30"'):
+            with self.subTest(text=text):
+                self.assertEqual(detect_dim_unit(text),'in')
+                self.assertIn('人工确认是否接收',self.ask('美国美西30KG，'+text+'，1箱普船海派，义乌'))
+        self.assertEqual(detect_dim_unit('60x40x30，合计1立方米'),'cm')
+        self.assertEqual(detect_dim_unit('60x40x30，合计1m3'),'cm')
+        self.assertEqual(detect_dim_unit('1.2x0.8x0.6m呢'),'m')
+
+    def test_spoken_carton_and_dimensions_are_used_without_reasking(self):
+        intent=parse_intent('美国30KG一个箱子，60长40宽30高，怎么算？',set())
+        self.assertEqual(intent['boxes'],1);self.assertEqual(intent['dims'],[60,40,30])
+        text=self.ask('美国美西30KG一个箱子，60长40宽30高，义乌普船海派')
+        self.assertIn('最高项70元/件',text)
+        self.assertNotIn('请补充箱数',text)
+
+    def test_box_units_are_local_and_each_axis_can_carry_its_own_unit(self):
+        cases=(('12英寸显示器，包装60x40x30cm',[60,40,30],'cm',Decimal('30')),
+               ('80cm×70cm×60cm',[80,70,60],'cm',Decimal('56')),
+               ('80cmx70cmx60cm',[80,70,60],'cm',Decimal('56')),
+               ('800mm*700mm*600mm',[800,700,600],'mm',Decimal('56')),
+               ('长80厘米宽70厘米高60厘米',[80,70,60],'cm',Decimal('56')),
+               ('0.8m×70cm×600mm',[80,70,60],'cm',Decimal('56')),
+               ('12英寸显示器，包装60x40x30',[60,40,30],'cm',Decimal('30')))
+        for wording,dims,unit,charge in cases:
+            with self.subTest(wording=wording):
+                intent=parse_intent('美国美西30KG，'+wording+'，1箱义乌普船海派',set())
+                self.assertEqual(intent['dims'],dims);self.assertEqual(intent['dim_unit'],unit)
+                self.assertEqual(quantity(intent,{'kind':'parcel','unit':'kg'}),charge)
+                text=self.ask('美国美西30KG，'+wording+'，1箱义乌普船海派')
+                self.assertIn(f'{charge}KG =',text)
+                self.assertNotIn('请补充箱数',text)
+
+    def test_oral_partial_weight_edits_never_overwrite_the_whole_shipment(self):
+        cid=self.conversation()
+        self.ask('美国美西30KG，部分铁制品，铁铝部分计费重10KG，义乌普船海派',cid)
+        for q,amount in (('铁铝部分还是10KG',110),('铁铝部分改成8KG',100),('铁铝部分仍是8KG',100)):
+            text=self.ask(q,cid)
+            self.assertIn('美国 30KG',text);self.assertIn(f'30KG = {amount}元',text)
+        self.assertIn('大于0',self.ask('铁铝部分改为-1KG',cid))
+        fresh=self.conversation()
+        self.assertIn('请确认铁铝部分的计费重',self.ask('美国美西30KG，铁铝部分还是10KG，义乌普船海派',fresh))
+        self.assertNotIn('×10KG',self.ask('一共多少钱，给我最终总价',fresh))
+
+    def test_long_quote_chain_retains_conditions_but_topic_changes_stop_reuse(self):
+        cid=self.conversation();self.ask('美国美西30KG，义乌普船海派',cid)
+        for _ in range(15):self.assertIn('每箱不足12KG',self.ask('这个报价按毛重还是净重？',cid))
+        self.assertIn('美国 50KG',self.ask('50KG呢',cid))
+        self.assertIn('美国渠道',self.ask('这个物流服务范围具体怎么执行？',cid))
+        self.assertIn('请先',self.ask('下一页',cid))
+        from logistics_chat.quote_service import answer_quote
+        from logistics_chat.quote_scope import answer_us_fallback
+        with self.app.test_request_context('/'):
+            self.assertIsNone(answer_quote('新的一票50KG发德国报价',cid))
+            db.save_message(cid,'user','换个话题，我要查订单')
+            self.assertIsNone(answer_us_fallback('50KG呢',cid))
+
+    def test_comparing_two_weights_is_not_treated_as_two_cartons(self):
+        cid=self.conversation();self.ask('美国30KG海派报价',cid)
+        text=self.ask('单箱12.5KG会不会按13KG收费？',cid)
+        self.assertIn('按原小数计费，不向上取整',text)
+        self.assertIn('12.5KG',text)
+        self.assertNotIn('请明确总重',text)
+
+    def test_confirmed_fractional_weight_and_partial_metal_basis_are_used(self):
+        self.assertIn('12.5KG = 25元',self.ask('美国美西1箱12.5KG，40x30x20cm，义乌普船海派'))
+        self.assertIn('12.3456KG = 24.69元',self.ask('美国美西1箱12.3456KG，40x30x20cm，义乌普船海派'))
+        cid=self.conversation()
+        text=self.ask('美国美西30KG，部分铁制品，义乌普船海派',cid)
+        self.assertIn('30KG = 60元',text);self.assertIn('不能按整票重量加费',text)
+        self.assertNotIn('30KG = 210元',text)
+        text=self.ask('铁铝部分计费重10KG',cid)
+        self.assertIn('美国 30KG',text);self.assertIn('30KG = 110元',text)
+        self.assertIn('×10KG=50元',text)
+        self.assertIn('30KG = 110元',self.ask('一共多少钱，给我最终总价',cid))
+        text=self.ask('改成40KG',cid)
+        self.assertIn('当前金额未包含该项',text);self.assertNotIn('×10KG',text)
+        text=self.ask('全部换成铁制品40KG',cid)
+        self.assertIn('40KG = 280元',text)
+
+    def test_unknown_metal_and_cbm_quotes_do_not_spread_partial_fee_over_the_whole_load(self):
+        text=self.ask('美国美西30KG，只是金属扣子，整票都按铁铝加费吗，义乌普船海派')
+        self.assertIn('只按铁铝货物的计费重',text);self.assertIn('确认金属是否属于铁铝',text)
+        text=self.ask('美国30KG发ZZZ1，1方不包税海卡，部分铁制品，铁铝部分计费重10KG，义乌')
+        self.assertIn('1CBM = 150元',text);self.assertIn('×10KG=50元',text)
+        self.assertIn('大于0',self.ask('美国30KG，部分铁制品，铁铝部分计费重-1KG，义乌海派'))
+        text=self.ask('美国20KG发ZZZ1，10KG发YYY2，部分铁制品，铁铝部分计费重10KG，义乌普船海派')
+        self.assertIn('不能重复套用到每个仓',text)
+        self.assertIn('20KG = 100元',text);self.assertIn('12KG = 24元',text)
+        self.assertNotIn('铁铝附加费5元/KG×10KG',text)
+
+    def test_booking_phone_invoice_and_file_capabilities_do_not_invent_workflows(self):
+        cases=(('就按刚才方案帮我下单','尚未创建订单'),('手机号能查吗','暂不支持手机号'),
+               ('你们可以开发票吗，开票加几个点','开票税点'),('付款账号发我一下','收款账号'),
+               ('可以月结吗','月结资格'),('帮我导出订单Excel','暂不提供订单Excel导出'),
+               ('Excel能直接读里面箱规报价吗','不会自动读取Excel'),
+               ('我换了个手机，之前聊天记录还在吗','不会自动同步恢复'))
+        for q,expected in cases:
+            with self.subTest(q=q):self.assertIn(expected,self.ask(q))
+        with patch('logistics_chat.order_data.lookup_order') as lookup:
+            self.assertIn('暂不支持手机号',self.ask('手机号19900000000能查吗'))
+            lookup.assert_not_called()
+        with self.app.app_context():
+            c=db.connect();self.assertEqual(c.execute('SELECT COUNT(*) FROM order_test_data').fetchone()[0],2);c.close()
+
+    def test_address_and_pickup_questions_are_not_prices_and_english_stays_english(self):
+        cid=self.conversation();self.ask('美国30KG海派报价',cid)
+        self.assertIn('地址、联系人和收货时间',self.ask('义乌仓具体地址是什么？几点可以送？',cid))
+        text=self.ask('Can you pick up my goods in Shenzhen for US sea freight?')
+        self.assertIn('does not offer pickup',text);self.assertNotIn('CNY/kg',text)
+        self.assertFalse(re.search(r'[\u4e00-\u9fff]',text))
+
+    def test_air_battery_question_does_not_apply_sea_battery_requirements(self):
+        text=self.ask('想走美国空运，包税带电的，有吗？')
+        self.assertIn('空运费率当前资料没有列出',text);self.assertNotIn('MSDS',text)
+        self.assertIn('备货阶段',self.ask('美国木制品需要熏蒸吗'))
+
+    def seed_pages(self):
+        with self.app.app_context():
+            c=db.connect();import_records(c,[{'tenant_id':'qa-a','order_no':f'PAGE2026{i:06d}',
+                'inbound_code':f'QAPAGE2026{i:06d}','status':'PROCESSING'} for i in range(1,26)])
+            c.commit();c.close()
+
+    def test_pages_use_fresh_authorized_count_and_preserve_filter_and_private_history(self):
+        self.seed_pages()
+        self.assertTrue(self.events('下一页')[0]['login_required'])
+        self.login();cid=self.conversation()
+        text=self.ask('列出我的全部订单',cid)
+        self.assertIn('共26票',text);self.assertEqual(text.count('- 入仓码：'),20)
+        self.assertNotIn('QAOTHER',text)
+        text=self.ask('下一页',cid)
+        self.assertIn('第2页',text);self.assertEqual(text.count('- 入仓码：'),6);self.assertIn('共26票',text)
+        self.assertIn('最后一页',self.ask('下一页',cid))
+        self.assertIn('第1页',self.ask('上一页',cid))
+        self.ask('列出履约中的订单',cid)
+        text=self.ask('下一页',cid)
+        self.assertIn('共25票',text);self.assertEqual(text.count('- 入仓码：'),5);self.assertNotIn('QAFOCUS',text)
+        self.assertIn('第1页',self.ask('第1页',cid))
+        self.client.post('/api/customer/logout')
+        self.assertEqual(self.client.get('/api/get_messages',query_string={'conv_id':cid}).json,[])
+
+    def test_page_commands_cannot_reuse_another_topic_or_unbounded_offset(self):
+        self.login();cid=self.conversation();self.ask('列出我的订单',cid)
+        self.assertIn('有效页码',self.ask('第0页',cid))
+        self.ask('列出我的订单',cid)
+        self.ask('美国30KG海派报价',cid)
+        self.assertIn('请先',self.ask('下一页',cid))
+
+    def test_count_interruptions_do_not_erase_the_active_list_page(self):
+        self.seed_pages();self.login();cid=self.conversation()
+        self.ask('列出我的全部订单',cid)
+        self.ask('我有多少个运单',cid)
+        text=self.ask('下一页',cid)
+        self.assertIn('第2页',text);self.assertEqual(text.count('- 入仓码：'),6)
+
+    def test_clearance_followup_modification_and_unknown_payment_do_not_call_model(self):
+        self.assertTrue(self.events('我的订单已清关了吗')[0]['login_required'])
+        self.login();cid=self.conversation();self.events('QAFOCUS20260001',cid)
+        event=self.events('我的订单已清关了吗？',cid)[0]
+        self.assertEqual(event['card']['no'],'FOCUS20260001')
+        self.assertEqual(event['card']['tracking_status'],'海运途中')
+        self.assertIn('尚未修改订单',self.ask('帮我把收货地址改一下',cid))
+        self.assertIn('未接入付款状态',self.ask('这票订单付过款了吗',cid))
+        self.assertIn('不能直接当作快递单号',self.ask('这票对应的UPS快递单号是多少',cid))
+        self.assertIn('暂不支持按这个物流条件',self.ask('哪些订单还没开船',cid))

tests/test_order_counts.py

@@ -78,14 +78,15 @@ class OrderCountTests(unittest.TestCase):
                              'What is the number of my orders?'):
                 with self.subTest(question=question):
                     events = self.events(question,cid)
-                    self.assertEqual(events[0],{'t':'您当前共有 2 票订单，其中 1 票已登记入仓码。'})
+                    self.assertTrue(events[0]['t'].startswith('您当前共有 2 票订单，其中 1 票已登记入仓码。'))
+                    self.assertEqual(events[0]['t'].count('- 入仓码：'),2)
                     self.assertEqual(events[-1],{'d':1})
             model.assert_not_called()
         with self.app.app_context():
             connection = db.connect()
             order_data.import_records(connection,[{'tenant_id':'tenant-a','order_no':'COUNT20260004','inbound_code':'INB20260004'}])
             connection.commit(); connection.close()
-        self.assertEqual(self.events('我有多少个运单')[0]['t'],'您当前共有 3 票订单，其中 2 票已登记入仓码。')
+        self.assertTrue(self.events('我有多少个运单')[0]['t'].startswith('您当前共有 3 票订单，其中 2 票已登记入仓码。'))
 
     def test_business_status_filters_are_exact_and_unknown_conditions_never_return_the_total(self):
         self.login()
@@ -152,7 +153,7 @@ class OrderCountTests(unittest.TestCase):
             with self.app.app_context():
                 with self.assertRaises(order_data.OrderDataUnavailable):
                     order_data.count_orders(('tenant-a',))
-        with patch('logistics_chat.order_data.count_orders',side_effect=order_data.OrderDataUnavailable):
+        with patch('logistics_chat.order_listing.read_orders',side_effect=sqlite3.OperationalError):
             events = self.events('我有多少个运单')
         self.assertIn('暂时无法读取',events[0]['t'])
 
@@ -164,7 +165,7 @@ class OrderCountTests(unittest.TestCase):
                 (' \t\r\n','COUNT20260002','tenant-a'))
             order_data.import_records(connection,[{'tenant_id':'tenant-a','order_no':'COUNT20260005','inbound_code':'INB20260001'}])
             connection.commit(); connection.close()
-        self.assertEqual(self.events('我有多少个运单')[0]['t'],'您当前共有 3 票订单，其中 2 票已登记入仓码。')
+        self.assertTrue(self.events('我有多少个运单')[0]['t'].startswith('您当前共有 3 票订单，其中 2 票已登记入仓码。'))
         with self.app.app_context():
             connection = db.connect()
             connection.execute("DELETE FROM order_test_data WHERE tenant_id='tenant-a'")

tests/test_order_statistics.py

@@ -0,0 +1,166 @@
+"""Statistics use explicit exception reasons and actual, order-level progress."""
+
+import json
+from pathlib import Path
+import sqlite3
+import tempfile
+import unittest
+from unittest.mock import patch
+
+from logistics_chat import create_app,db,order_data
+from logistics_chat.order_tracking import current_stage
+
+
+def fact(code,shipment='',forecast=False):
+    return {'code':code,'shipment_key':shipment,
+            'estimated_at' if forecast else 'actual_at':'2026-10-01 10:00'}
+
+
+class OrderStatisticsTests(unittest.TestCase):
+    def setUp(self):
+        folder=tempfile.TemporaryDirectory();self.addCleanup(folder.cleanup);root=Path(folder.name)
+        self.app=create_app({'TESTING':True,'DATABASE':root/'data.db','UPLOAD_DIR':root/'uploads',
+            'BACKUP_DIR':root/'backups','LOG_DIR':root/'logs','DEEPSEEK_API_KEY':'',
+            'CUSTOMER_DEMO_TENANT_ID':'a','RATE_LIMIT_PER_MIN':1000})
+        self.client=self.app.test_client();self.client.post('/api/visitor/session',json={'key':'1'*64})
+        rows=[{'tenant_id':'a','order_no':f'STAT2026{i:04}','inbound_code':f'INSTAT2026{i:04}',
+            'status':'PROCESSING','source_updated_at':f'2026-10-{i+1:02} 10:00',
+            'tracking_json':[fact('IN_TRANSIT'),fact('ARRIVED_DESTINATION_PORT'),fact('EMPTY_RETURNED',forecast=True)],
+            'exception_reason':f'待核实资料{i}' if i<6 else None} for i in range(10)]
+        rows += [{'tenant_id':'a','order_no':'CLEAR20260001','status':'PROCESSING',
+                  'tracking_json':[fact('ARRIVED_DESTINATION_PORT'),fact('PICKUP_READY')],'exception_reason':'等待客户资料'},
+                 {'tenant_id':'a','order_no':'CLEAR20260002','status':'PROCESSING','tracking_json':[fact('PICKUP_READY')]},
+                 {'tenant_id':'a','order_no':'MIXED20260001','tracking_json':[fact('ARRIVED_DESTINATION_PORT','A'),fact('PICKUP_READY','B')]},
+                 {'tenant_id':'a','order_no':'FORECAST20260001','tracking_json':[fact('ARRIVED_DESTINATION_PORT',forecast=True)]},
+                 {'tenant_id':'a','order_no':'BUSINESS20260001','status':'COMPLETED','tracking_json':'[]'},
+                 {'tenant_id':'a','order_no':'REJECTED20260001','status':'REJECTED','exception_reason':'  \t\n'},
+                 {'tenant_id':'b','order_no':'SECRET20260001','status':'PROCESSING','tracking_json':[fact('ARRIVED_DESTINATION_PORT')],'exception_reason':'不能泄露'}]
+        with self.app.app_context():
+            c=db.connect();order_data.import_records(c,rows);c.commit();c.close()
+
+    def login(self):
+        self.assertTrue(self.client.post('/api/customer/login',json={'username':'admin','password':'123456'}).json['ok'])
+
+    def ask(self,q,cid=None):
+        with patch('logistics_chat.ai_client.post_completion') as model:
+            response=self.client.post('/stream_chat',json={'msg':q,'conv_id':cid})
+            events=[json.loads(line[6:]) for line in response.get_data(as_text=True).split('\n\n') if line.startswith('data: ')]
+            model.assert_not_called()
+        self.assertEqual(events[-1],{'d':1},events)
+        self.assertNotIn('_order_list_state',str(events));self.assertNotIn('_customer_data',str(events))
+        self.assertNotIn('SECRET',str(events));self.assertNotIn('不能泄露',str(events))
+        return events[0]
+
+    def test_guests_never_read_aggregate_or_exception_facts(self):
+        with patch('logistics_chat.order_listing.read_orders') as read:
+            for q in ('我有多少异常订单','到港的有多少运单','各个轨迹节点分别有多少订单','查看全部','下一页'):
+                self.assertTrue(self.ask(q)['login_required'])
+            read.assert_not_called()
+
+    def test_summaries_show_at_most_five_recent_rows_and_keep_intersecting_filters(self):
+        self.login()
+        total=self.ask('我有多少个订单')['t']
+        self.assertIn('共有 16 票',total);self.assertEqual(total.count('- 入仓码：'),5)
+        abnormal=self.ask('我有多少个异常订单')['t']
+        self.assertIn('7 票明确标记异常',abnormal);self.assertEqual(abnormal.count('- 入仓码：'),5)
+        self.assertNotIn('REJECTED20260001',abnormal)
+        arrived=self.ask('现在到港的有多少运单')['t']
+        self.assertIn('10 票当前节点为“船舶已到目的港”',arrived)
+        self.assertEqual(arrived.count('- 入仓码：'),5)
+        self.assertIn('STAT20260009',arrived);self.assertNotIn('STAT20260000',arrived)
+        self.assertNotIn('CLEAR20260001',arrived)
+        text=self.ask('已清关的异常订单有多少')['t']
+        self.assertIn('1 票当前节点为“已清关、可提柜”',text);self.assertIn('等待客户资料',text)
+
+    def test_distribution_counts_each_order_once_and_never_uses_forecasts_or_business_completion(self):
+        self.login()
+        text=self.ask('各个轨迹节点分别有多少订单')['t']
+        for token in ('船舶已到目的港：10票','已清关、可提柜：2票','多柜进度不同：1票','暂无实际节点：3票'):
+            self.assertIn(token,text)
+        self.assertNotIn('已还柜：',text);self.assertNotIn('已签收：',text)
+
+    def test_expand_and_pages_keep_filters_and_refresh_counts(self):
+        self.login();cid=self.client.post('/api/new_chat').json['id']
+        self.ask('我有多少异常订单',cid)
+        text=self.ask('下一页',cid)['t']
+        self.assertIn('第2页',text);self.assertEqual(text.count('- 入仓码：'),2)
+        text=self.ask('查看全部',cid)['t']
+        self.assertIn('共7票',text);self.assertEqual(text.count('- 入仓码：'),7)
+        self.assertIn('明确标记异常',text)
+        with self.app.app_context():
+            c=db.connect();c.execute('UPDATE order_test_data SET exception_reason=NULL WHERE tenant_id=?',('a',));c.commit();c.close()
+        self.assertIn('没有符合条件',self.ask('下一页',cid)['t'])
+
+    def test_pages_do_not_cross_conversations_or_login_boundaries(self):
+        self.login();cid=self.client.post('/api/new_chat').json['id']
+        self.ask('到港的订单有多少',cid)
+        self.assertIn('第2页',self.ask('下一页',cid)['t'])
+        self.ask('当前订单总数',cid)
+        self.assertIn('共16票',self.ask('查看全部',cid)['t'])
+        self.client.post('/api/customer/logout')
+        self.assertTrue(self.ask('下一页',cid)['login_required'])
+        self.assertEqual(self.client.get('/api/get_messages',query_string={'conv_id':cid}).json,[])
+        self.login();self.assertIn('共16票',self.ask('查看全部',cid)['t'])
+        other=self.client.post('/api/new_chat').json['id']
+        self.assertIn('请先',self.ask('下一页',other)['t'])
+
+    def test_unknown_or_ambiguous_filters_do_not_degenerate_to_totals(self):
+        self.login()
+        for q in ('今天有多少异常订单','未到港的订单有多少','到港和已清关分别有多少订单','已签收的有多少运单'):
+            with self.subTest(q=q):self.assertIn('暂不支持按这个条件统计',self.ask(q)['t'])
+
+    def test_exception_reasons_are_editable_and_old_reimports_preserve_them(self):
+        self.login()
+        with self.app.app_context():
+            c=db.connect();order_data.import_records(c,[{'tenant_id':'a','order_no':'CLEAR20260001','status':'PROCESSING'}],replace=True);c.commit();c.close()
+        self.assertIn('等待客户资料',self.ask('CLEAR20260001的异常原因是什么')['t'])
+        cid=self.client.post('/api/new_chat').json['id'];self.ask('CLEAR20260001',cid)
+        self.assertIn('等待客户资料',self.ask('为什么异常',cid)['t'])
+        with self.app.app_context():
+            c=db.connect();order_data.import_records(c,[{'tenant_id':'a','order_no':'CLEAR20260001','exception_reason':''}],replace=True);c.commit();c.close()
+        self.assertIn('未登记异常原因',self.ask('CLEAR20260001的异常原因是什么')['t'])
+
+    def test_missing_actuals_and_bad_tracking_do_not_count_as_arrived(self):
+        self.assertEqual(current_stage([fact('ARRIVED_DESTINATION_PORT','A'),fact('PICKUP_READY','B',True)]),'MIXED')
+        self.assertEqual(current_stage([fact('ARRIVED_DESTINATION_PORT',forecast=True)]),'NO_ACTUAL')
+        self.assertEqual(current_stage('{broken'),'NO_ACTUAL')
+        self.assertEqual(current_stage([fact('UNKNOWN')]),'NO_ACTUAL')
+        self.assertEqual(current_stage([fact('DOMESTIC_RECEIVED'),fact('ARRIVED_DESTINATION_PORT','A',True)]),'DOMESTIC_RECEIVED')
+
+    def test_node_names_do_not_collide_with_business_status_or_drop_negations(self):
+        from logistics_chat.order_query import parse_query
+        intent=parse_query('已完成装柜的订单有多少')
+        self.assertFalse(intent['unsupported']);self.assertIsNone(intent['status'])
+        self.assertEqual(intent['stage'],'LOADING_COMPLETED')
+        self.assertTrue(parse_query('未清关的订单有多少')['unsupported'])
+        self.assertTrue(parse_query('没有异常的订单有多少')['unsupported'])
+
+    def test_state_survives_app_restart_and_rejects_late_reply_updates(self):
+        self.login();cid=self.client.post('/api/new_chat').json['id']
+        self.ask('我有多少异常订单',cid)
+        restarted=create_app(dict(self.app.config,INITIALIZE_RUNTIME=False))
+        client=restarted.test_client()
+        # Copy opaque cookies without exporting them.
+        for cookie in self.client._cookies.values():client.set_cookie(cookie.key,cookie.value)
+        response=client.post('/stream_chat',json={'msg':'下一页','conv_id':cid}).get_data(as_text=True)
+        self.assertIn('第2页',response)
+        from logistics_chat.order_listing import save_state
+        cookies='; '.join(cookie.key+'='+cookie.value for cookie in self.client._cookies.values())
+        with self.app.test_request_context('/',headers={'Cookie':cookies}):
+            self.app.preprocess_request()
+            from logistics_chat.customer_service import current_customer_id
+            self.assertIsNotNone(current_customer_id())
+            c=db.connect();prior=c.execute('SELECT order_list_state FROM chat_history WHERE id=?',(cid,)).fetchone()[0]
+            state=json.loads(prior);c.close()
+            db.save_message(cid,'user','换个话题，我要报价')
+            save_state(cid,{'query':state['query'],'page':1,'size':5},state['message_id'])
+            c=db.connect();self.assertEqual(c.execute('SELECT order_list_state FROM chat_history WHERE id=?',(cid,)).fetchone()[0],prior);c.close()
+
+    def test_schema_migration_preserves_old_orders(self):
+        c=sqlite3.connect(':memory:')
+        try:
+            c.execute('CREATE TABLE order_test_data (id INTEGER PRIMARY KEY,tenant_id TEXT,order_no TEXT,status TEXT,inbound_code TEXT,shipment_id TEXT)')
+            c.execute("INSERT INTO order_test_data VALUES(1,'a','OLD20260001','PROCESSING',NULL,NULL)")
+            order_data.init_order_table(c);order_data.init_order_table(c)
+            self.assertEqual(c.execute('SELECT order_no,exception_reason,tracking_json FROM order_test_data').fetchone(),('OLD20260001',None,'[]'))
+        finally:c.close()