chore: post-deployment cleanup and docs update

- Make AI routing rules DB-driven (read from system_configs, removed from config.py)
- Add translation quota tracking to LLM translation (OpenAIProvider)
- Add Alibaba MT ECS RAM role support (STS token, no AccessKey needed)
- Fix admin sidebar link for AI模型配置 page
- Fix Quota.vue API path (quotas → translation-quotas)
- Fix login auto-redirect to dashboard
- Add provider dropdown selects to AI routing config UI
- Clean up stale ai_provider_* system_configs records
- Remove OpencodeGo, Spark providers (code + DB)
- Update deploy config: nginx port 8000, systemd cwd

AGENTS.md

@@ -1,20 +1,31 @@
 # TradeMate (外贸小助手) — Agent Guide
 
+## AI Assistant (Frontend AI Chatbot)
+
+- **Components**: `user-frontend/src/components/AiAssistant.vue` + `admin-frontend/src/components/AiAssistant.vue`
+- **Backend**: `backend/app/api/v1/ai_assistant.py` — `POST /api/v1/ai/chat`, `GET /api/v1/ai/quick-questions`
+- **Action types** (configurable via `ACTION_INSTRUCTIONS`): create_customer, create_product, create_quotation, scan_followups, generate_marketing, discovery_search, navigate, search_users, update_user, update_config, review_certification, process_invoice
+- **Frontend action dispatch**: `AiAssistant.vue` switch/case calls corresponding API from `@/api`
+- **Layout integration**: `<AiAssistant />` in `UserLayout.vue` + `AdminLayout.vue`, floating button bottom-right
+- **Quick questions**: configurable via `ai_assistant_quick_questions` `SystemConfig` key
+- **System prompt**: configurable via `ai_assistant_prompt` `SystemConfig` key
+
 ## Architecture
 
 - **Backend**: `backend/` — FastAPI + SQLAlchemy 1.4 async + asyncpg, single `app.main:app`
 - **Frontends**: `uni-app/` (mobile H5/mini-program), `admin-frontend/` (PC admin), `user-frontend/` (PC workspace)
 - **Config**: `backend/app/config.py` reads from `/.env` (project root) via pydantic BaseSettings
 - **Auth**: JWT (python-jose). Default dep `get_current_user_id` in `backend/app/api/v1/deps.py`
-- **AI Router**: `backend/app/ai/router.py` — singleton `AIRouter`, DB-driven providers. Primary = sensenova, fallbacks = alibaba-mt / opencode_go / nvidia / spark
+- **AI Router**: `backend/app/ai/router.py` — singleton `AIRouter`, DB-driven providers + DB-driven routing rules. Routing reads `ai_routing` from `system_configs` table.
 - **Database**: PostgreSQL via `asyncpg`, pool_size=20
 
 ## AI Providers
 
-- **Active**: Sensenova (商汤), OpencodeGo, NVIDIA, 讯飞 Spark, 阿里机器翻译 (alibaba-mt)
+- **Active**: Sensenova (商汤), NVIDIA, 阿里机器翻译 (alibaba-mt) — 5 providers in DB
-- **Removed (dead code)**: Claude (`claude.py`), DeepL (`deepl.py`), Local (`local.py`) — git rm'd, not yet committed
+- **Removed**: Claude, DeepL, Local, OpencodeGo, 讯飞 Spark — all git rm'd
 - **DB-driven**: `AIProvider` model + `admin_ai.py` API — manage providers at runtime. `router.seed_from_env()` loads from `.env` on startup
-- **Provider type mapping** in `router.py._build_provider()`: sensenova, opencode_go, nvidia, spark, alibaba-mt
+- **ECS RAM role**: 阿里翻译使用 ECS 实例 RAM 角色 `trademate-translate` 获取 STS 临时凭证
+- **Provider type mapping** in `router.py._build_provider()`: sensenova, nvidia, alibaba-mt
 
 ## Security
 
@@ -32,7 +43,10 @@
 
 ```bash
 # Backend (from project root — .env is there)
+# Dev: matches Vite proxy (both frontends proxy /api → localhost:8000)
 cd backend && source venv/bin/activate && uvicorn app.main:app --reload --port 8000
+# Production: nginx proxies /api/ → localhost:8002
+cd backend && source venv/bin/activate && uvicorn app.main:app --port 8002
 
 # Mobile H5
 cd uni-app && npm run dev:h5
@@ -66,7 +80,7 @@ alembic revision --autogenerate -m "desc"
 - **Workspace**: `trade.yuzhiran.com/workspace/` — Vue 3 + Element Plus (standalone)
 - **Nginx**: SPA fallbacks for `/app/`, `/admin/`, `/workspace/`
 - **vite config**: each project has its own `base` path and dev port
-- **API**: proxied via nginx `location /api/` to `127.0.0.1:8002`
+- **API**: proxied via nginx `location /api/` to `127.0.0.1:8000`
 
 ## Critical Quirks
 
@@ -80,6 +94,14 @@ alembic revision --autogenerate -m "desc"
 - **AI Router reload**: After modifying AI providers via admin API, call `POST /api/v1/admin/ai/reload` to refresh in-memory providers.
 - **Payment**: Uses unified `pay-api` gateway (`UnifiedPayService`). NOT direct WeChat/Alipay integration. Credentials: `PAY_API_KEY`/`PAY_API_SECRET` from `.env`. HMAC-SHA256 auth. Webhook at `POST /api/v1/payment/webhook` skips CSRF.
 - **Payment gateway config**: `PAY_API_KEY`, `PAY_API_SECRET`, `PAY_API_BASE_URL`, `PAY_WEBHOOK_URL` in `config.py`. `pay_type` param: `"alipay"` (returns `pay_url`) or `"wechat"` (returns `code_url`). `UnifiedPayService` normalizes legacy `native`/`jsapi`/`pc` to `wechat`/`alipay`.
+- **Manual auth on some endpoints**: `keywords` and `competitor-analysis` endpoints use `authorization: str = Header(None)` instead of `Depends(get_current_user_id)`.
+- **MarketingService fallback**: When no AI providers initialized, returns template content instead of crashing.
+- **Onboarding service**: calls `mkt.generate(product_info={"name": ..., ...})`, not keyword args. Check `onboarding.py` for the exact dict shape.
+- **CustomerHealthService**: `get_health_overview` endpoint must use `CustomerHealthService(db)` not `CustomerService(db)`.
+- **CSRF**: Sensitive endpoints (auth/payment/profile) require `X-CSRF-Token` header. Token available via `csrf_token` cookie / `X-CSRF-Token` response header.
+- **AI Router reload**: After modifying AI providers via admin API, call `POST /api/v1/admin/ai/reload` to refresh in-memory providers.
+- **Payment**: Uses unified `pay-api` gateway (`UnifiedPayService`). NOT direct WeChat/Alipay integration. Credentials: `PAY_API_KEY`/`PAY_API_SECRET` from `.env`. HMAC-SHA256 auth. Webhook at `POST /api/v1/payment/webhook` skips CSRF.
+- **Payment gateway config**: `PAY_API_KEY`, `PAY_API_SECRET`, `PAY_API_BASE_URL`, `PAY_WEBHOOK_URL` in `config.py`. `pay_type` param: `"alipay"` (returns `pay_url`) or `"wechat"` (returns `code_url`). `UnifiedPayService` normalizes legacy `native`/`jsapi`/`pc` to `wechat`/`alipay`.
 
 ## Project Conventions
 

PROGRESS.md

@@ -1,7 +1,7 @@
 # TradeMate (外贸小助手) - 项目进度文档
 
-**更新时间**: 2026-05-29 12:00  
+**更新时间**: 2026-06-02 12:00  
-**状态**: ✅ 生产分支活跃开发 — 安全加固完成 + 4 前端项目并行 + 客户挖掘+AI 模型管理
+**状态**: ✅ 生产环境运行中 — AI 路由 DB 驱动 + 翻译配额全链路 + ECS RAM 角色认证
 
 ---
 
@@ -11,9 +11,11 @@
 |------|------|------|
 | 后端 API | http://localhost:8000 | ✅ 运行中 |
 | API 文档 | http://localhost:8000/docs | ✅ 可用 |
-| 前端 H5 (uni-app) | http://localhost:5173 | ✅ 运行中 |
+| 主站 — 营销落地页 | https://trade.yuzhiran.com/ | ✅ 已部署 |
-| 管理后台 (admin-frontend) | http://localhost:5173 | ✅ 运行中 |
+| 主站 — 移动端 H5 | https://trade.yuzhiran.com/app/ | ✅ 已部署 |
-| 用户工作台 (user-frontend) | http://localhost:5174 | ✅ 运行中 |
+| 主站 — 管理后台 | https://trade.yuzhiran.com/admin/ | ✅ 已部署 |
+| 主站 — 用户工作台 | https://trade.yuzhiran.com/workspace/ | ✅ 已部署 |
+| 镜像站 | https://www.yuzhiran.com.cn/ | ✅ 同步部署 |
 | Redis | localhost:6379 | ✅ 运行中 |
 | PostgreSQL | localhost:5432 | ✅ 运行中 |
 
@@ -40,21 +42,30 @@
 |------|------|------|
 | **CORS 白名单** | `app/core/middleware.py` | 限制特定前端域名、方法和请求头 |
 | **速率限制** | `app/core/middleware.py` | 登录 5次/分、注册 3次/时、改密 3次/5分、支付 20次/分、Admin 30次/分 |
-| **CSRF 防护** | `app/core/csrf.py` | 双提交 Cookie 模式中间件；auth/payment/profile 等敏感端点必检；webhook 跳过 |
+| **CSRF 防护** | `app/core/csrf.py` | 双提交 Cookie 模式中间件；payment/profile 等敏感端点必检；login/register/webhook 跳过 |
 | **敏感日志清理** | 全代码审计 | 移除 Token/Key 等敏感信息打印 (T-002) |
 
 ### 3. AI 提供商重构 ✅
 
 | 变更 | 说明 |
 |------|------|
-| **移除** | Claude (`claude.py`)、DeepL (`deepl.py`)、本地模型 (`local.py`) |
+| **移除** | Claude (`claude.py`)、DeepL (`deepl.py`)、本地模型 (`local.py`)、OpencodeGo (`opencode_go.py`)、讯飞 Spark (`spark.py`) |
-| **新增** | 阿里机器翻译 (`alibaba.py`)、NVIDIA (`nvidia.py`) |
+| **当前提供商** | Sensenova (商汤) + NVIDIA + 阿里机器翻译 (ECS RAM 角色) |
-| **主提供商** | Sensenova (商汤, `sensenova.py`) |
+| **DB 驱动配置** | `AIProvider` 模型 + `admin_ai.py` API → 管理后台在线增删改 AI 模型 |
-| **Fallback** | OpencodeGo → NVIDIA → 讯飞 Spark |
+| **路由规则 DB 化** | 从 `config.py` 硬编码改为 `system_configs` 表读取 (key: `ai_routing`)，保存后自动重载 |
-| **DB 驱动配置** | 新增 `AIProvider` 模型 + `admin_ai.py` API → 管理后台在线增删改 AI 模型 |
+| **ECS RAM 角色** | 阿里翻译无需 AccessKey，优先使用 ECS 实例 RAM 角色获取 STS 临时凭证 |
-| **路由规则** | `router.py` 支持 `reload_from_db()` 从数据库实时加载配置 |
 | **Env 自动种子** | 启动时从 `.env` 读取 API Key 自动写入 DB |
 
+### 3.1 当前 AI 提供商
+
+| 名称 | 类型 | 模型 | 用途 |
+|------|------|------|------|
+| Sensenova (商汤) | sensenova | deepseek-v4-flash | 翻译/回复/营销/提取/聊天 |
+| NVIDIA | nvidia | stepfun-ai/step-3.7-flash | 翻译/回复/营销/提取/聊天 |
+| Sensenova (商汤) lite | sensenova | sensenova-6.7-flash-lite | 轻量备用 |
+| NVIDIA qwen | nvidia | qwen/qwen3.5-397b-a17b | 备用 |
+| 阿里翻译 | alibaba-mt | alibaba-mt | 专用翻译 (ecommerce/general) |
+
 ### 4. 客户挖掘 (Discovery) ✅
 
 | 功能 | 文件 | 说明 |
@@ -73,6 +84,26 @@
 | 年费定价 | `payment.py` — 新增 yearly 套餐选项 |
 | 搜索 API 管理 | `admin_search.py` — 管理后台配置搜索提供商 |
 
+### 5.1 支付系统 ✅
+
+| 组件 | 文件 | 说明 |
+|------|------|------|
+| **统一网关** | `services/unified_pay.py` | 对接 宇之然 pay-api，HMAC-SHA256 签名 |
+| **网关抽象** | `services/payment_gateway.py` | `PaymentGateway` 抽象基类 |
+| **支付服务** | `services/payment.py` | 多套餐创建/查询/退款/回调 |
+| **交易记录** | `models/payment_transaction.py` | 订单号/金额/状态/退款流水 |
+| **支付 API** | `api/v1/payment.py` | 下单/查询/退款/Webhook 统一入口 |
+| **管理接口** | `api/v1/admin.py` | 交易列表/统计/后台退款 |
+| **数据迁移** | `alembic/versions/add_payment_transactions_table.py` | `payment_transactions` 表 |
+
+**支付流程**:
+1. 前端 POST `/api/v1/payment/create-order` (`plan` + `pay_type`: `alipay` / `wechat`)
+2. 后端调用 unified pay-api 创建订单，返回 `pay_url`(支付宝) 或 `code_url`(微信扫码)
+3. 用户完成支付后，pay-api 回调 `POST /api/v1/payment/webhook`
+4. 后端更新订单状态 + 激活用户套餐
+
+**凭证**: `PAY_API_KEY` / `PAY_API_SECRET` 从 `.env` 读取（外贸助手密钥），HMAC-SHA256 认证
+
 ### 6. PC 桌面端布局 ✅
 
 | 功能 | 说明 |
@@ -99,6 +130,8 @@
 | 11 | `app/api/v1/auth.py` + `deps.py` | API 500 根因 — 游客 UUID 格式问题 | ✅ 已修复 |
 | 12 | `backend/.env` + `app/main.py` | CORS 配置不当 | ✅ 已修复 |
 | 13 | `uni-app/src/utils/api.js` | 前端直连后端端口 → 跨域 | ✅ 已修复 |
+| 14 | `app/api/v1/auth.py` | 登录/注册 CSRF 鸡生蛋 — 匿名用户无 cookie 导致 403 | ✅ 已修复 |
+| 15 | 4 个前端登录/注册页面 | 后端错误英文直接展示给用户 | ✅ 已修复 |
 
 ### 8. 游客模式 (Guest Mode) ✅
 
@@ -119,7 +152,26 @@
 | AI 模型配置 | 在线增删改 AI 提供商、重载配置、启停控制 |
 | 搜索配置 | 搜索提供商管理 |
 
-### 10. 其他增强
+### 10. 翻译配额全链路 ✅
+
+| 组件 | 说明 |
+|------|------|
+| `TranslationQuota` 模型 | 按月配额管理，支持 ecommerce/general/llm 三个版本 |
+| `TranslationQuotaService` | 自动按月重置、配额检查/扣减/查询/管理 |
+| `AlibabaMTProvider.translate()` | 调用前检查配额，调用后扣减字符 |
+| `OpenAIProvider.translate()` | LLM 翻译也走配额检查 (`llm` 版本) |
+| 后台 `Quota.vue` | 配额管理页 (月限额/启用/重置)，修复了 API 路径 bug |
+
+### 11. 管理后台增强
+
+| 功能 | 说明 |
+|------|------|
+| AI 路由规则 | 配置页新增"AI路由规则"卡片，下拉选择已启用供应商 |
+| 翻译配额页 | 修复路径 bug (`quotas` → `translation-quotas`)，新增 `llm` 配额 |
+| AI 模型配置 | 修复侧边栏链接路径 |
+| 登录跳转 | 登录后自动跳转仪表盘 |
+
+### 12. 其他增强
 
 | 功能 | 说明 |
 |------|------|
@@ -176,13 +228,14 @@
 
 ### 4.2 AI 提供商
 
-| 提供商 | 类型 | 用途 |
+| 提供商 | 类型 | 模型 | 用途 |
-|--------|------|------|
+|--------|------|------|------|
-| Sensenova (商汤) | 主提供商 | 翻译/回复/营销/提取 |
+| Sensenova (商汤) | 主提供商 | deepseek-v4-flash | 翻译/回复/营销/提取/聊天 |
-| OpencodeGo | Fallback | 翻译/回复/营销/提取 |
+| NVIDIA | Fallback | stepfun-ai/step-3.7-flash | 翻译/回复/营销/提取/聊天 |
-| NVIDIA | Fallback | 通用对话 |
+| Sensenova lite | 轻量备用 | sensenova-6.7-flash-lite | 备用 |
-| 讯飞 Spark | Fallback | 通用对话 |
+| NVIDIA qwen | 备用 | qwen/qwen3.5-397b-a17b | 备用 |
-| 阿里机器翻译 | 专用 | 翻译 (translate) |
+| 阿里机器翻译 | 专用翻译 | alibaba-mt | 翻译 (ecommerce/general) |
+| **已删除** | — | — | OpencodeGo / 讯飞 Spark / Claude / DeepL / 本地模型 |
 
 ### 4.3 前端
 
@@ -273,13 +326,39 @@ cd backend && alembic upgrade head
 
 ## 八、部署架构
 
+### 8.1 域名
+
+| 域名 | 用途 |
+|------|------|
+| `trade.yuzhiran.com` | 主站 — 营销落地页 + Web 应用（当前生产域名） |
+| `www.yuzhiran.com.cn` | 之前用作镜像站 AI 教育，外贸项目不再使用 |
+
+### 8.2 目录结构
+
 ```
 trade.yuzhiran.com/
 ├── /              → 静态营销落地页
-├── /app/          → uni-app 构建 (移动端 H5)
+├── /app/          → uni-app 构建 (移动端 H5，当前未部署)
 ├── /admin/        → admin-frontend 构建 (管理后台)
 ├── /workspace/    → user-frontend 构建 (用户工作台)
-└── /api/          → Nginx proxy → 127.0.0.1:8002
+└── /api/          → Nginx proxy → 127.0.0.1:8000 (systemd 管理)
+```
+
+### 8.3 部署流程
+
+```bash
+# 前端构建 & 部署
+cd user-frontend && npm run build && cp -r dist/* /www/wwwroot/trade.yuzhiran.com/workspace/
+cd admin-frontend && npm run build && cp -r dist/* /www/wwwroot/trade.yuzhiran.com/admin/
+
+# 后端重启 (systemd)
+sudo systemctl restart ftrade-backend.service
+
+# 查看启动日志
+sudo journalctl -u ftrade-backend.service -n 20
+
+# 本地开发启动
+cd backend && source venv/bin/activate && uvicorn app.main:app --reload --port 8000
 ```
 
 ---
@@ -288,6 +367,7 @@ trade.yuzhiran.com/
 
 | 日期 | 变更内容 |
 |------|----------|
+| 2026-06-02 | 生产环境部署 + AI 路由 DB 驱动 + 翻译配额扩展至 LLM + ECS RAM 角色认证 + 删除 OpencodeGo/Spark |
 | 2026-05-29 | 安全加固 (T-005): 限流/CSRF/CORS + AI 提供商 DB 管理 + 客户挖掘联系人提取 |
 | 2026-05-28 | 加载反馈 + 搜索历史自动保存 + 超时修复 |
 | 2026-05-26 | 落地页 + 推荐系统 + 用量配额 + 搜索 API 管理 + 年费定价 |

admin-frontend/src/api/index.js

@@ -37,9 +37,9 @@ export function listLogs(params) { return http.get('/admin/logs', { params }) }
 export function listConfig() { return http.get('/admin/config') }
 export function updateConfig(key, value) { return http.put(`/admin/config/${key}`, { value }) }
 
-export function listQuotas() { return http.get('/admin/quotas') }
+export function listQuotas() { return http.get('/admin/translation-quotas') }
-export function updateQuota(version, data) { return http.put(`/admin/quotas/${version}`, data) }
+export function updateQuota(version, data) { return http.put(`/admin/translation-quotas/${version}`, data) }
-export function resetQuota(version) { return http.post(`/admin/quotas/${version}/reset`) }
+export function resetQuota(version) { return http.post(`/admin/translation-quotas/${version}/reset`) }
 
 export function listCertifications(page = 1, size = 50, status = '') {
   return http.get('/admin/certifications', { params: { page, size, status: status || undefined } })
@@ -55,4 +55,7 @@ export function processInvoice(id, action) {
   return http.post(`/admin/invoices/${id}/process`, { action })
 }
 
+export function aiChat(message, history = []) { return http.post('/ai/chat', { message, history }) }
+export function aiQuickQuestions() { return http.get('/ai/quick-questions') }
+
 export default http

admin-frontend/src/components/AiAssistant.vue

@@ -0,0 +1,363 @@
+<template>
+  <div>
+    <div class="ai-float-btn" @click="visible = true">
+      <span class="ai-float-icon">AI</span>
+    </div>
+
+    <el-dialog
+      v-model="visible"
+      title="TradeMate AI 助手"
+      width="480px"
+      :close-on-click-modal="false"
+      class="ai-dialog"
+      top="5vh"
+      @opened="onOpened"
+    >
+      <div class="ai-messages" ref="msgContainer">
+        <div v-for="(msg, i) in messages" :key="i" class="ai-msg-row" :class="msg.role">
+          <div class="ai-avatar" v-if="msg.role === 'assistant'">
+            <el-icon :size="18" color="#667eea"><Cpu /></el-icon>
+          </div>
+          <div class="ai-msg-body">
+            <div class="ai-msg-bubble">
+              <div class="ai-msg-text">{{ msg.content }}</div>
+
+              <div v-if="msg.actions && msg.actions.length" class="ai-action-card">
+                <div v-for="(action, ai) in msg.actions" :key="ai" class="ai-action-item">
+                  <div class="ai-action-title">{{ action.label }}</div>
+                  <div v-if="action.type === 'navigate'" class="ai-navigate-hint">
+                    <el-tag type="info">跳转至 {{ action.fields?.path }}</el-tag>
+                  </div>
+                  <div v-for="(val, key) in action.fields" :key="key" class="ai-field-row" v-if="action.type !== 'navigate'">
+                    <span class="ai-field-label">{{ fieldLabel(action.type, key) }}</span>
+                    <el-input
+                      v-model="action.fields[key]"
+                      :placeholder="fieldPlaceholder(action.type, key)"
+                      size="small"
+                    />
+                  </div>
+                  <div class="ai-action-btns">
+                    <el-button size="small" @click="cancelAction(i, ai)">取消</el-button>
+                    <el-button size="small" type="primary" @click="confirmAction(i, ai)">确认执行</el-button>
+                  </div>
+                </div>
+              </div>
+            </div>
+          </div>
+        </div>
+
+        <div v-if="loading" class="ai-loading">
+          <el-icon class="is-loading" :size="18"><Loading /></el-icon>
+          <span>思考中...</span>
+        </div>
+
+        <div v-if="showSuggestions" class="ai-suggestions">
+          <div
+            v-for="(s, i) in suggestions"
+            :key="i"
+            class="ai-suggestion"
+            @click="sendQuick(s)"
+          >
+            {{ s }}
+          </div>
+        </div>
+      </div>
+
+      <template #footer>
+        <div class="ai-input-bar">
+          <el-input
+            v-model="inputText"
+            placeholder="输入你的问题..."
+            :disabled="loading"
+            size="default"
+            @keyup.enter="send"
+          />
+          <el-button type="primary" :disabled="!inputText.trim() || loading" @click="send">发送</el-button>
+        </div>
+      </template>
+    </el-dialog>
+  </div>
+</template>
+
+<script setup>
+import { ref, nextTick } from 'vue'
+import { ElMessage } from 'element-plus'
+import { Cpu, Loading } from '@element-plus/icons-vue'
+import { aiChat, aiQuickQuestions, searchUsers, updateUser, updateConfig, reviewCertification, processInvoice } from '@/api'
+
+const visible = ref(false)
+const inputText = ref('')
+const loading = ref(false)
+const suggestions = ref([])
+const msgContainer = ref(null)
+const messages = ref([
+  { role: 'assistant', content: '你好！我是管理后台 AI 助手，可以帮你管理用户、配置、认证审核、发票等。你可以说："查找用户"、"审核认证"等。' },
+])
+const showSuggestions = ref(false)
+
+const fieldLabel = (type, key) => {
+  const labels = {
+    name: '名称', phone: '电话', email: '邮箱', company: '公司',
+    country: '国家', status: '状态', notes: '备注',
+    key: '配置键', value: '配置值',
+    action: '操作', reason: '原因',
+    username: '用户名', role: '角色',
+    version: '版本', quota: '额度',
+    path: '目标页面',
+  }
+  return labels[key] || key
+}
+
+const fieldPlaceholder = (type, key) => {
+  const placeholders = {
+    reason: '审核备注（可选）',
+    action: 'approved / rejected',
+    role: 'admin / user',
+    value: '配置值',
+    quota: '如 100',
+    path: '如 /users, /config, /certifications',
+  }
+  return placeholders[key] || ''
+}
+
+const cancelAction = (msgIdx, actionIdx) => {
+  const msg = messages.value[msgIdx]
+  if (msg.actions) msg.actions.splice(actionIdx, 1)
+}
+
+const confirmAction = async (msgIdx, actionIdx) => {
+  const action = messages.value[msgIdx].actions[actionIdx]
+  const { type, fields } = action
+  loading.value = true
+  try {
+    switch (type) {
+      case 'navigate':
+        if (fields.path) {
+          const router = (await import('@/router')).default
+          router.push(fields.path)
+          ElMessage.success(`正在跳转到 ${fields.path}`)
+          messages.value[msgIdx].actions = []
+        }
+        return
+      case 'search_users':
+        if (!fields.query) { ElMessage.warning('搜索关键词不能为空'); return }
+        await searchUsers(fields.query)
+        break
+      case 'update_user':
+        if (!fields.user_id && !fields.username) { ElMessage.warning('用户标识不能为空'); return }
+        await updateUser(fields.user_id || fields.username, fields)
+        break
+      case 'update_config':
+        if (!fields.key) { ElMessage.warning('配置键不能为空'); return }
+        await updateConfig(fields.key, fields.value)
+        break
+      case 'review_certification':
+        if (!fields.id) { ElMessage.warning('认证ID不能为空'); return }
+        await reviewCertification(fields.id, fields.action || 'approved', fields.reason)
+        break
+      case 'process_invoice':
+        if (!fields.id) { ElMessage.warning('发票ID不能为空'); return }
+        await processInvoice(fields.id, fields.action || 'approve')
+        break
+      default:
+        ElMessage.warning(`未知操作类型: ${type}`)
+        return
+    }
+    ElMessage.success('操作成功')
+    messages.value[msgIdx].actions = []
+  } catch (e) {
+    ElMessage.error(e.message || e.detail || '操作失败')
+  } finally {
+    loading.value = false
+  }
+}
+
+const sendQuick = (text) => {
+  inputText.value = text
+  send()
+}
+
+const fetchSuggestions = async () => {
+  try {
+    const res = await aiQuickQuestions()
+    if (Array.isArray(res)) suggestions.value = res
+  } catch {}
+}
+
+const onOpened = () => {
+  if (suggestions.value.length === 0) fetchSuggestions()
+  showSuggestions.value = messages.value.length <= 1
+  scrollToBottom()
+}
+
+const send = async () => {
+  const msg = inputText.value.trim()
+  if (!msg || loading.value) return
+  inputText.value = ''
+  showSuggestions.value = false
+  messages.value.push({ role: 'user', content: msg })
+  loading.value = true
+  scrollToBottom()
+
+  try {
+    const hist = messages.value.map(m => ({ role: m.role, content: m.content }))
+    const res = await aiChat(msg, hist.slice(0, -1))
+    const newMsg = { role: 'assistant', content: res.reply || '抱歉，我没有理解，请重新描述一下你的问题。' }
+    if (res.actions && res.actions.length) {
+      newMsg.actions = res.actions
+    }
+    messages.value.push(newMsg)
+  } catch {
+    messages.value.push({ role: 'assistant', content: '请求失败，请稍后重试。' })
+  } finally {
+    loading.value = false
+    scrollToBottom()
+  }
+}
+
+const scrollToBottom = async () => {
+  await nextTick()
+  const el = msgContainer.value
+  if (el) el.scrollTop = el.scrollHeight
+}
+</script>
+
+<style scoped>
+.ai-float-btn {
+  position: fixed;
+  right: 30px;
+  bottom: 30px;
+  width: 52px;
+  height: 52px;
+  border-radius: 50%;
+  background: linear-gradient(135deg, #667eea, #764ba2);
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  box-shadow: 0 4px 20px rgba(102, 126, 234, 0.4);
+  z-index: 9999;
+  cursor: pointer;
+  transition: transform 0.2s;
+}
+.ai-float-btn:hover {
+  transform: scale(1.08);
+}
+.ai-float-icon {
+  color: #fff;
+  font-size: 18px;
+  font-weight: bold;
+}
+.ai-dialog {
+  --el-dialog-content-padding: 0;
+}
+.ai-dialog :deep(.el-dialog__body) {
+  padding: 0;
+  height: 480px;
+  overflow: hidden;
+  display: flex;
+  flex-direction: column;
+}
+.ai-messages {
+  flex: 1;
+  padding: 16px;
+  overflow-y: auto;
+  background: #f5f5f5;
+}
+.ai-msg-row {
+  margin-bottom: 14px;
+  display: flex;
+  gap: 8px;
+}
+.ai-msg-row.user {
+  flex-direction: row-reverse;
+}
+.ai-avatar {
+  width: 32px;
+  height: 32px;
+  border-radius: 50%;
+  background: #f0edff;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  flex-shrink: 0;
+}
+.ai-msg-body {
+  max-width: 80%;
+}
+.ai-msg-bubble {
+  padding: 10px 14px;
+  border-radius: 8px;
+  background: #fff;
+  box-shadow: 0 1px 3px rgba(0,0,0,0.06);
+}
+.ai-msg-row.user .ai-msg-bubble {
+  background: #667eea;
+  color: #fff;
+}
+.ai-msg-text {
+  font-size: 13px;
+  line-height: 1.6;
+  white-space: pre-wrap;
+  word-break: break-word;
+}
+.ai-msg-row.user .ai-msg-text {
+  color: #fff;
+}
+.ai-action-card {
+  margin-top: 12px;
+  border-top: 1px solid #eee;
+  padding-top: 10px;
+}
+.ai-action-title {
+  font-size: 14px;
+  font-weight: 600;
+  color: #333;
+  margin-bottom: 10px;
+}
+.ai-field-row {
+  margin-bottom: 8px;
+}
+.ai-field-label {
+  display: block;
+  font-size: 12px;
+  color: #666;
+  margin-bottom: 3px;
+}
+.ai-action-btns {
+  display: flex;
+  gap: 8px;
+  margin-top: 10px;
+}
+.ai-loading {
+  text-align: center;
+  padding: 12px 0;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  gap: 6px;
+  color: #999;
+  font-size: 13px;
+}
+.ai-suggestions {
+  padding: 8px 0;
+}
+.ai-suggestion {
+  background: #f0edff;
+  border-radius: 6px;
+  padding: 8px 12px;
+  margin-bottom: 8px;
+  font-size: 13px;
+  color: #667eea;
+  cursor: pointer;
+  transition: background 0.15s;
+}
+.ai-suggestion:hover {
+  background: #e0dbff;
+}
+.ai-input-bar {
+  display: flex;
+  gap: 8px;
+  padding: 10px 16px;
+  border-top: 1px solid #f0f0f0;
+}
+</style>

admin-frontend/src/layouts/AdminLayout.vue

@@ -51,7 +51,7 @@
             <el-icon><Search /></el-icon>
             <span>搜索配置</span>
           </el-menu-item>
-          <el-menu-item index="/system/ai-providers">
+          <el-menu-item index="/system/search-config/ai-providers">
             <el-icon><Cpu /></el-icon>
             <span>AI 模型配置</span>
           </el-menu-item>
@@ -128,6 +128,7 @@
         </div>
       </el-footer>
     </el-container>
+    <AiAssistant />
   </el-container>
 </template>
 
@@ -135,6 +136,7 @@
 import { ref, computed } from 'vue'
 import { useRoute, useRouter } from 'vue-router'
 import { useAuthStore } from '@/stores/auth'
+import AiAssistant from '@/components/AiAssistant.vue'
 
 const route = useRoute()
 const router = useRouter()

admin-frontend/src/router/index.js

@@ -2,7 +2,7 @@ import { createRouter, createWebHistory } from 'vue-router'
 import AdminLayout from '@/layouts/AdminLayout.vue'
 
 const routes = [
-  { path: '/login', redirect: '/' },
+  { path: '/login', redirect: to => ({ path: '/', query: to.query }) },
   { path: '/', name: 'Landing', component: () => import('@/views/Landing.vue') },
   {
     path: '/dashboard',

admin-frontend/src/views/Config.vue

@@ -7,7 +7,24 @@
           <el-tag size="small" v-if="cfg.description">{{ cfg.description }}</el-tag>
         </div>
       </template>
-      <div v-if="typeof cfg.value === 'object' && !Array.isArray(cfg.value)">
+      <div v-if="cfg.key === 'ai_routing'">
+        <div v-for="(taskVal, taskKey) in cfg.value" :key="taskKey" class="cfg-nested-group">
+          <div class="cfg-group-title">{{ fieldLabelsMap.ai_routing?.[taskKey] || taskKey }}</div>
+          <div class="cfg-field">
+            <span class="cfg-label">主选</span>
+            <el-select v-model="edits[cfg.key][taskKey].primary" size="small" style="width:300px" filterable>
+              <el-option v-for="p in providers" :key="p.provider_type" :value="p.provider_type" :label="p.name + ' — ' + p.model_name" />
+            </el-select>
+          </div>
+          <div class="cfg-field">
+            <span class="cfg-label">备用</span>
+            <el-select v-model="edits[cfg.key][taskKey].fallback" size="small" style="width:400px" multiple filterable collapse-tags>
+              <el-option v-for="p in providers" :key="p.provider_type" :value="p.provider_type" :label="p.name + ' — ' + p.model_name" />
+            </el-select>
+          </div>
+        </div>
+      </div>
+      <div v-else-if="typeof cfg.value === 'object' && !Array.isArray(cfg.value)">
         <div class="cfg-field" v-for="(v, k) in cfg.value" :key="k">
           <span class="cfg-label">{{ fieldLabel(cfg.key, k) }}</span>
           <el-input v-if="typeof v === 'string'" v-model="edits[cfg.key][k]" size="small" style="width:300px" />
@@ -34,12 +51,33 @@
 import { ref, reactive, onMounted } from 'vue'
 import { ElMessage } from 'element-plus'
 import { listConfig, updateConfig } from '@/api'
+import http from '@/api'
 
 const configs = ref([])
 const edits = reactive({})
-const configLabels = { system_maintenance: '系统维护', feature_flags: '功能开关', translation_providers: '翻译服务', ai_model_config: 'AI模型', cache_ttl: '缓存时间' }
+const providers = ref([])
-const fieldLabelsMap = { system_maintenance: { maintenance_mode: '维护模式', maintenance_message: '维护消息' }, feature_flags: { feature_wechat_login: '微信登录', feature_export: '数据导出' }, translation_providers: { primary: '首选服务', fallback: '备用服务' }, ai_model_config: { default_model: '默认模型', max_tokens: '最大Token' } }
+const configLabels = { system_maintenance: '系统维护', feature_flags: '功能开关', translation_providers: '翻译服务', ai_model_config: 'AI模型', cache_ttl: '缓存时间', ai_routing: 'AI路由规则' }
-function fieldLabel(key, k) { return fieldLabelsMap[key]?.[k] || k }
+const fieldLabelsMap = { system_maintenance: { maintenance_mode: '维护模式', maintenance_message: '维护消息' }, feature_flags: { feature_wechat_login: '微信登录', feature_export: '数据导出' }, translation_providers: { primary: '首选服务', fallback: '备用服务' }, ai_model_config: { default_model: '默认模型', max_tokens: '最大Token' }, ai_routing: { translate: '翻译', reply: '回复建议', marketing: '营销文案', extract: '信息提取', quotation: '报价单', chat: 'AI助手' } }
+const taskFieldLabels = { primary: '主选', fallback: '备用' }
+function fieldLabel(key, k) {
+  if (key === 'ai_routing') return configLabels[key] + ' > ' + (taskFieldLabels[k] || k)
+  return fieldLabelsMap[key]?.[k] || k
+}
+function taskFieldLabel(cfgKey, taskKey, subKey) {
+  if (cfgKey === 'ai_routing') {
+    const taskLabel = fieldLabelsMap.ai_routing?.[taskKey] || taskKey
+    const subLabel = taskFieldLabels[subKey] || subKey
+    return taskLabel + ' > ' + subLabel
+  }
+  return subKey || taskKey
+}
+
+async function loadProviders() {
+  try {
+    const res = await http.get('/admin/ai-providers')
+    providers.value = (res.items || []).filter(p => p.enabled)
+  } catch (_) {}
+}
 
 async function load() {
   try {
@@ -48,13 +86,14 @@ async function load() {
     for (const cfg of configs.value) {
       edits[cfg.key] = JSON.parse(JSON.stringify(cfg.value))
     }
+    await loadProviders()
   } catch (e) { console.error(e) }
 }
 
 async function save(key) {
   try {
     await updateConfig(key, edits[key])
-    ElMessage.success('已保存')
+    ElMessage.success('已保存，AI 路由器已重载')
   } catch (e) { ElMessage.error(e?.detail || '保存失败') }
 }
 
@@ -66,5 +105,7 @@ onMounted(load)
 .cfg-header { display: flex; align-items: center; gap: 12px; font-weight: 600; }
 .cfg-field { display: flex; align-items: center; gap: 12px; padding: 8px 0; }
 .cfg-label { width: 160px; font-size: 13px; color: #666; flex-shrink: 0; }
+.cfg-nested-group { margin-bottom: 12px; padding: 8px 12px; background: #f9fafb; border-radius: 6px; }
+.cfg-group-title { font-weight: 600; font-size: 13px; color: #333; margin-bottom: 4px; padding-bottom: 4px; border-bottom: 1px solid #e5e7eb; }
 .cfg-actions { margin-top: 12px; }
 </style>

admin-frontend/src/views/Landing.vue

@@ -88,7 +88,7 @@ async function submit() {
     const res = await loginApi(form)
     localStorage.setItem('admin_token', res.access_token)
     localStorage.setItem('admin_user', JSON.stringify(res.user || {}))
-    ElMessage.success('登录成功')
+    router.push('/dashboard')
   } catch (e) {
     const map = { 'Invalid credentials': '用户名或密码错误' }
     error.value = map[e?.detail] || e?.detail || '登录失败'

admin-frontend/src/views/Quota.vue

@@ -4,7 +4,7 @@
       <el-card shadow="never">
         <template #header>
           <div class="quota-header">
-            <span class="quota-version">{{ q.version === 'ecommerce' ? '电商版' : '通用版' }}</span>
+            <span class="quota-version">{{ { ecommerce: '电商版', general: '通用版', llm: 'AI模型翻译' }[q.version] || q.version }}</span>
             <el-tag size="small" v-if="q.description">{{ q.description }}</el-tag>
           </div>
         </template>

admin-frontend/vite.config.js

@@ -15,7 +15,7 @@ export default defineConfig({
   server: {
     port: 5173,
     proxy: {
-      '/api': { target: 'http://localhost:8002', changeOrigin: true }
+      '/api': { target: 'http://localhost:8000', changeOrigin: true }
     }
   }
 })

backend/alembic/versions/add_payment_transactions_table.py

@@ -9,7 +9,7 @@ import sqlalchemy as sa
 from sqlalchemy.dialects.postgresql import UUID
 
 revision = "add_payment_transactions"
-down_revision = "add_ai_providers_table"
+down_revision = "add_ai_providers"
 branch_labels = None
 depends_on = None
 

backend/app/ai/providers/__init__.py

@@ -1,8 +1,6 @@
 from .openai import OpenAIProvider
-from .spark import SparkProvider
 from .sensenova import SensenovaProvider
-from .opencode_go import OpencodeGoProvider
 from .nvidia import NvidiaProvider
 from .alibaba import AlibabaMTProvider
 
-__all__ = ["OpenAIProvider", "SparkProvider", "SensenovaProvider", "OpencodeGoProvider", "NvidiaProvider", "AlibabaMTProvider"]
+__all__ = ["OpenAIProvider", "SensenovaProvider", "NvidiaProvider", "AlibabaMTProvider"]

backend/app/ai/providers/alibaba.py

@@ -1,11 +1,13 @@
 from typing import Dict, Any, Optional
 from aliyunsdkcore.client import AcsClient
+from aliyunsdkcore.auth.credentials import StsTokenCredential
 from aliyunsdkalimt.request.v20181012 import TranslateGeneralRequest, TranslateECommerceRequest
 from app.services.translation_quota import TranslationQuotaService
 from app.database import AsyncSessionLocal
 import asyncio
 import json
 import logging
+import os
 
 logger = logging.getLogger(__name__)
 
@@ -16,11 +18,55 @@ ALIBABA_LANG_MAP = {
     "id": "id", "ms": "ms", "tl": "tl", "hi": "hi",
 }
 
+ECS_METADATA_URL = "http://100.100.100.200/latest/meta-data/ram/security-credentials/"
+
+
+def _fetch_ecs_ram_credentials():
+    try:
+        import urllib.request
+        req = urllib.request.Request(ECS_METADATA_URL, method="GET")
+        with urllib.request.urlopen(req, timeout=2) as resp:
+            role_name = resp.read().decode().strip()
+        if not role_name:
+            logger.warning("ECS metadata returned empty role name")
+            return None
+        url = f"{ECS_METADATA_URL}{role_name}"
+        req = urllib.request.Request(url, method="GET")
+        with urllib.request.urlopen(req, timeout=2) as resp:
+            data = json.loads(resp.read().decode())
+        if data.get("Code") == "Success":
+            logger.info(f"Fetched STS token for role {role_name}, expires {data.get('Expiration')}")
+            return (data["AccessKeyId"], data["AccessKeySecret"], data["SecurityToken"])
+        else:
+            logger.warning(f"ECS metadata returned non-success: {data.get('Code')}")
+    except Exception as e:
+        logger.debug(f"ECS metadata fetch failed: {e}")
+    return None
+
+
+def _build_acs_client(access_key_id: str = "", access_key_secret: str = "",
+                      region_id: str = "cn-hangzhou") -> AcsClient:
+    creds = _fetch_ecs_ram_credentials()
+    if creds:
+        ak, sk, token = creds
+        sts_cred = StsTokenCredential(ak, sk, token)
+        client = AcsClient(credential=sts_cred, region_id=region_id)
+        logger.info("Alibaba MT using ECS RAM role (STS token)")
+        return client
+
+    ak = access_key_id or os.getenv("ALIBABA_ACCESS_KEY_ID", "")
+    sk = access_key_secret or os.getenv("ALIBABA_ACCESS_KEY_SECRET", "")
+    if ak and sk:
+        logger.info("Alibaba MT using AccessKey credentials")
+        return AcsClient(ak, sk, region_id)
+
+    raise ValueError("No Alibaba Cloud credentials found (neither ECS RAM role nor AccessKey)")
+
 
 class AlibabaMTProvider:
-    def __init__(self, access_key_id: str, access_key_secret: str,
+    def __init__(self, access_key_id: str = "", access_key_secret: str = "",
                  region_id: str = "cn-hangzhou"):
-        self.client = AcsClient(access_key_id, access_key_secret, region_id)
+        self.client = _build_acs_client(access_key_id, access_key_secret, region_id)
         self._name = "alibaba-mt"
 
     async def translate(self, text: str, source_lang: Optional[str],

backend/app/ai/providers/openai.py

@@ -51,6 +51,20 @@ class OpenAIProvider(AIProvider):
         self._cheap_model = "gpt-4o-mini" if model == "gpt-4o" else model
 
     async def translate(self, text: str, source_lang: Optional[str], target_lang: str, context: Optional[str] = None) -> Dict[str, Any]:
+        from app.services.translation_quota import TranslationQuotaService
+        from app.database import AsyncSessionLocal
+
+        async with AsyncSessionLocal() as db:
+            quota_svc = TranslationQuotaService(db)
+            if not await quota_svc.check_quota("llm"):
+                raise Exception("LLM translation quota exhausted or disabled")
+            result = await self._do_translate(text, source_lang, target_lang, context)
+            if result and result.get("translated_text"):
+                await quota_svc.consume("llm", len(text))
+                await db.commit()
+            return result
+
+    async def _do_translate(self, text: str, source_lang: Optional[str], target_lang: str, context: Optional[str] = None) -> Dict[str, Any]:
         system = SYSTEM_PROMPTS["translate"]
         if context:
             system += f"\nContext: this is about {context}"

backend/app/ai/providers/opencode_go.py

@@ -1,7 +0,0 @@
-from app.ai.providers.openai import OpenAIProvider
-
-
-class OpencodeGoProvider(OpenAIProvider):
-    def __init__(self, api_key: str, model: str = "deepseek-v4-flash", base_url: str = "https://opencode.ai/zen/go/v1"):
-        super().__init__(api_key=api_key, model=model, base_url=base_url)
-        self._name = f"opencode-go-{model}"

backend/app/ai/providers/spark.py

@@ -1,90 +0,0 @@
-from typing import Dict, Any, Optional
-import json
-from app.ai.base import AIProvider
-
-
-SYSTEM_PROMPTS = {
-    "translate": "You are a professional translator specialized in foreign trade. "
-                 "Translate business terms accurately. Return ONLY the translated text.",
-    "reply": "You are an experienced foreign trade sales expert. Write professional, "
-             "clear business replies. Return ONLY the reply text.",
-    "marketing": "You are a creative copywriter for international trade. "
-                 "Return ONLY the marketing copy, no explanations.",
-    "extract": "Extract structured data from text. Return ONLY valid JSON.",
-}
-
-
-class SparkProvider(AIProvider):
-    def __init__(self, api_key: str, model: str = "astron-code-latest", base_url: str = None):
-        from app.config import settings
-        try:
-            from openai import AsyncOpenAI
-        except ImportError:
-            raise ImportError("openai>=1.0 is required for SparkProvider")
-        self.client = AsyncOpenAI(
-            api_key=api_key,
-            base_url=base_url or settings.IFLYTEK_API_BASE,
-        )
-        self.model = model
-        self._name = f"spark-{model}"
-
-    async def translate(self, text: str, source_lang: Optional[str], target_lang: str, context: Optional[str] = None) -> Dict[str, Any]:
-        system = SYSTEM_PROMPTS["translate"]
-        if context:
-            system += f"\nContext: {context}"
-        prompt = f"Translate {f'from {source_lang} ' if source_lang and source_lang != 'auto' else ''}to {target_lang}:\n\n{text}"
-        content = await self._call(system, prompt)
-        return {"translated_text": content, "provider": self.name}
-
-    async def reply(self, inquiry: str, context: Optional[Dict[str, Any]] = None, tone: str = "professional", preference_context: Optional[str] = None) -> Dict[str, Any]:
-        system = SYSTEM_PROMPTS["reply"] + f"\nTone: {tone}"
-        if preference_context:
-            system += f"\nUser preference: {preference_context}"
-        ctx = ""
-        if context:
-            ctx = "\n".join(f"{k}: {v}" for k, v in context.items() if v)
-        prompt = f"{ctx}\nCustomer inquiry:\n{inquiry}\n\nWrite a reply:"
-        content = await self._call(system, prompt)
-        return {"reply": content, "provider": self.name}
-
-    async def generate_marketing(self, product_info: Dict[str, Any], target: str, style: str = "professional", language: str = "en", preference_context: Optional[str] = None) -> Dict[str, Any]:
-        system = SYSTEM_PROMPTS["marketing"] + f"\nStyle: {style}\nAudience: {target}\nLanguage: {language}"
-        if preference_context:
-            system += f"\nUser preference: {preference_context}"
-        info = json.dumps(product_info, ensure_ascii=False)
-        prompt = f"Product:\n{info}\n\nGenerate marketing copy:"
-        content = await self._call(system, prompt, max_tokens=1500)
-        return {"content": content, "provider": self.name}
-
-    async def extract_info(self, text: str, schema: Dict[str, Any]) -> Dict[str, Any]:
-        system = SYSTEM_PROMPTS["extract"]
-        prompt = f"Schema:\n{json.dumps(schema, indent=2)}\n\nText:\n{text}\n\nJSON:"
-        content = await self._call(system, prompt, response_format={"type": "json_object"})
-        try:
-            data = json.loads(content)
-            return {"data": data, "confidence": 0.9, "provider": self.name}
-        except json.JSONDecodeError:
-            return {"data": {}, "confidence": 0.0, "provider": self.name}
-
-    async def _call(self, system: str, prompt: str, max_tokens: int = 1000, response_format: Optional[Dict] = None) -> str:
-        kwargs = {
-            "model": self.model,
-            "messages": [
-                {"role": "system", "content": system},
-                {"role": "user", "content": prompt},
-            ],
-            "max_tokens": max_tokens,
-            "temperature": 0.7,
-        }
-        if response_format:
-            kwargs["response_format"] = response_format
-        resp = await self.client.chat.completions.create(**kwargs)
-        return resp.choices[0].message.content
-
-    @property
-    def name(self) -> str:
-        return self._name
-
-    @property
-    def cost_per_1k_tokens(self) -> float:
-        return 0.0

backend/app/ai/router.py

@@ -1,17 +1,26 @@
 from typing import Dict, Any, Optional, List
 from app.ai.base import AIProvider
-from app.ai.providers import SparkProvider, SensenovaProvider, OpencodeGoProvider, NvidiaProvider, AlibabaMTProvider
+from app.ai.providers import SensenovaProvider, NvidiaProvider, AlibabaMTProvider
-from app.config import settings
 from app.ai.trade_corpus import TradeCorpus
+from app.config import settings
 import logging
 
 logger = logging.getLogger(__name__)
 
+DEFAULT_ROUTING: Dict[str, dict] = {
+    "translate": {"primary": "sensenova", "fallback": ["alibaba-mt", "nvidia"]},
+    "reply": {"primary": "sensenova", "fallback": ["nvidia"]},
+    "marketing": {"primary": "sensenova", "fallback": ["nvidia"]},
+    "extract": {"primary": "sensenova", "fallback": ["nvidia"]},
+    "quotation": {"primary": "sensenova", "fallback": ["nvidia"]},
+    "chat": {"primary": "sensenova", "fallback": ["nvidia"]},
+}
+
 
 class AIRouter:
     def __init__(self):
         self.providers: Dict[str, AIProvider] = {}
-        self.routing_rules = settings.AI_ROUTING
+        self.routing_rules = dict(DEFAULT_ROUTING)
         self.corpus = TradeCorpus()
 
     async def reload_from_db(self, db_session) -> int:
@@ -38,8 +47,47 @@ class AIRouter:
         else:
             logger.warning("No enabled AI providers found in DB")
 
+        await self._load_routing_rules(db_session)
         return len(rows)
 
+    async def _load_routing_rules(self, db_session):
+        from app.models.system_config import SystemConfig
+        from sqlalchemy import select
+
+        # Try consolidated key first
+        result = await db_session.execute(
+            select(SystemConfig).where(SystemConfig.key == "ai_routing")
+        )
+        cfg = result.scalar_one_or_none()
+        if cfg and isinstance(cfg.value, dict):
+            self.routing_rules = {**DEFAULT_ROUTING, **cfg.value}
+            logger.info("Loaded routing rules from system_configs (ai_routing)")
+            return
+
+        # Fallback: load individual per-task keys
+        task_keys = {
+            "translate": "ai_provider_translate",
+            "reply": "ai_provider_reply",
+            "marketing": "ai_provider_marketing",
+            "extract": "ai_provider_extract",
+            "quotation": "ai_provider_quotation",
+        }
+        loaded = {}
+        for task, key in task_keys.items():
+            result = await db_session.execute(
+                select(SystemConfig).where(SystemConfig.key == key)
+            )
+            cfg = result.scalar_one_or_none()
+            if cfg and isinstance(cfg.value, dict):
+                loaded[task] = cfg.value
+
+        if loaded:
+            self.routing_rules = {**DEFAULT_ROUTING, **loaded}
+            logger.info(f"Loaded routing rules from system_configs (individual keys): {list(loaded.keys())}")
+        else:
+            self.routing_rules = dict(DEFAULT_ROUTING)
+            logger.info("No routing rules in system_configs, using defaults")
+
     async def seed_from_env(self, db_session) -> int:
         from app.models.ai_provider import AIProvider
 
@@ -53,33 +101,18 @@ class AIRouter:
                 base_url=settings.SENSENOVA_BASE_URL,
                 model_name=settings.SENSENOVA_MODEL, priority=0, enabled=True,
             ))
-        if settings.OPENCODE_GO_API_KEY:
-            seeds.append(AIProvider(
-                name="OpencodeGo", provider_type="opencode_go",
-                api_key=settings.OPENCODE_GO_API_KEY,
-                base_url=settings.OPENCODE_GO_BASE_URL,
-                model_name=settings.OPENCODE_GO_MODEL, priority=1, enabled=True,
-            ))
         if settings.NVIDIA_API_KEY:
             seeds.append(AIProvider(
                 name="NVIDIA", provider_type="nvidia",
                 api_key=settings.NVIDIA_API_KEY,
                 base_url=settings.NVIDIA_BASE_URL,
-                model_name=settings.NVIDIA_MODEL, priority=2, enabled=True,
+                model_name=settings.NVIDIA_MODEL, priority=1, enabled=True,
             ))
-        if settings.IFLYTEK_API_KEY:
-            seeds.append(AIProvider(
-                name="讯飞 Spark", provider_type="spark",
-                api_key=settings.IFLYTEK_API_KEY,
-                base_url=settings.IFLYTEK_API_BASE,
-                model_name=settings.IFLYTEK_MODEL, priority=3, enabled=True,
-            ))
-        if settings.ALIBABA_ACCESS_KEY_ID and settings.ALIBABA_ACCESS_KEY_SECRET:
         seeds.append(AIProvider(
             name="阿里翻译", provider_type="alibaba-mt",
-                api_key=settings.ALIBABA_ACCESS_KEY_ID,
+            api_key=settings.ALIBABA_ACCESS_KEY_ID or "",
-                api_secret=settings.ALIBABA_ACCESS_KEY_SECRET,
+            api_secret=settings.ALIBABA_ACCESS_KEY_SECRET or "",
-                model_name="alibaba-mt", priority=4, enabled=True,
+            model_name="alibaba-mt", priority=3, enabled=True,
         ))
 
         for p in seeds:
@@ -99,12 +132,8 @@ class AIRouter:
             t = p.provider_type
             if t == "sensenova":
                 return SensenovaProvider(api_key=p.api_key, model=p.model_name, base_url=p.base_url)
-            elif t == "opencode_go":
-                return OpencodeGoProvider(api_key=p.api_key, model=p.model_name, base_url=p.base_url)
             elif t == "nvidia":
                 return NvidiaProvider(api_key=p.api_key, model=p.model_name, base_url=p.base_url)
-            elif t == "spark":
-                return SparkProvider(api_key=p.api_key, model=p.model_name, base_url=p.base_url)
             elif t == "alibaba-mt":
                 return AlibabaMTProvider(access_key_id=p.api_key, access_key_secret=p.api_secret or "")
             else:
@@ -117,7 +146,7 @@ class AIRouter:
     def get_providers_for_task(self, task_type: str) -> List[AIProvider]:
         rules = self.routing_rules.get(
             task_type,
-            {"primary": "sensenova", "fallback": ["opencode_go"]},
+            {"primary": "sensenova", "fallback": ["nvidia"]},
         )
         ordered = []
         seen = set()

backend/app/api/v1/admin_ai.py

@@ -42,9 +42,7 @@ class AIProviderUpdate(BaseModel):
 
 PROVIDER_TYPE_LABELS = {
     "sensenova": "Sensenova (商汤)",
-    "opencode_go": "OpencodeGo",
     "nvidia": "NVIDIA",
-    "spark": "讯飞 Spark",
     "alibaba-mt": "阿里翻译",
 }
 

backend/app/api/v1/ai_assistant.py

@@ -20,7 +20,7 @@ router = APIRouter()
 
 
 ACTION_INSTRUCTIONS = """
-当用户想要执行操作时（如添加客户、创建产品、生成报价单等），请执行以下步骤：
+当用户想要执行操作时（如添加客户、创建产品、生成报价单、发送跟进、营销生成等），请执行以下步骤：
 1. 从用户消息中提取所有必要的信息
 2. 在回复末尾附上 JSON 格式的动作块，格式如下：
 
@@ -28,9 +28,20 @@ ACTION_INSTRUCTIONS = """
 [{"type": "create_customer", "label": "添加客户", "fields": {"name": "...", "phone": "...", "email": "...", "company": "...", "country": "...", "notes": "..."}}]
 ```
 
-支持的 action type：
+支持的 action type 及字段说明：
-- create_customer：添加客户，fields 支持 name(必填), phone, email, company, country, notes
+- create_customer：添加客户，fields 支持 name(必填), phone, email, company, country, website, notes
-- create_product：添加产品（开发中）
+- create_product：添加产品，fields 支持 name(必填), name_en, description, description_en, category, price, price_unit(默认USD), moq, keywords(逗号分隔)
+- create_quotation：生成报价单，fields 支持 customer_name(必填), product_info(必填), quantity(必填), price, terms
+- scan_followups：扫描待跟进客户，fields 不需要（空对象）
+- send_followup：发送跟进消息，fields 支持 customer_name(必填), message(必填)
+- generate_marketing：生成营销素材，fields 支持 product_name(必填), target_market, tone(如professional/casual), language
+- discovery_search：搜索潜在客户，fields 支持 keywords(必填), country, industry
+- navigate：跳转到指定页面，fields 支持 path(必填, 如 /customers /products /quotations /marketing /discovery /followup /translate /team /analytics)
+- search_users：搜索用户，fields 支持 query(必填)
+- update_user：修改用户信息，fields 支持 user_id(必填), username, phone, email, role, status
+- update_config：更新系统配置，fields 支持 key(必填), value(必填)
+- review_certification：审核认证，fields 支持 id(必填), action(approved/rejected), reason
+- process_invoice：处理发票，fields 支持 id(必填), action(approve/reject)
 
 如果用户没有提供足够信息，请先询问缺少的字段，不要生成 action。
 如果用户明确表示要执行操作但缺少信息，生成 action 但标注缺失的字段。

backend/app/config.py

@@ -32,13 +32,7 @@ class Settings(BaseSettings):
     SENSENOVA_BASE_URL: str = "https://token.sensenova.cn/v1"
     SENSENOVA_MODEL: str = "deepseek-v4-flash"
 
-    IFLYTEK_API_KEY: Optional[str] = None
-    IFLYTEK_API_BASE: str = "https://maas-api.cn-huabei-1.xf-yun.com/v2"
-    IFLYTEK_MODEL: str = "astron-code-latest"
 
-    OPENCODE_GO_API_KEY: Optional[str] = None
-    OPENCODE_GO_BASE_URL: str = "https://opencode.ai/zen/go/v1"
-    OPENCODE_GO_MODEL: str = "minimax-m2.7"
 
     NVIDIA_API_KEY: Optional[str] = None
     NVIDIA_BASE_URL: str = "https://integrate.api.nvidia.com/v1"
@@ -74,15 +68,6 @@ class Settings(BaseSettings):
     SENTRY_DSN: Optional[str] = None
     DEBUG: bool = True
 
-    AI_ROUTING: dict = {
-        "translate": {"primary": "sensenova", "fallback": ["alibaba-mt", "opencode_go"]},
-        "reply": {"primary": "sensenova", "fallback": ["opencode_go"]},
-        "marketing": {"primary": "sensenova", "fallback": ["opencode_go"]},
-        "extract": {"primary": "sensenova", "fallback": ["opencode_go"]},
-        "quotation": {"primary": "sensenova", "fallback": ["opencode_go"]},
-        "chat": {"primary": "sensenova", "fallback": ["opencode_go", "nvidia"]},
-    }
-
     FREE_DAILY_TRANSLATE_CHARS: int = 5000
     FREE_DAILY_REPLIES: int = 20
     FREE_DAILY_MARKETING: int = 5

backend/app/services/admin.py

@@ -288,11 +288,14 @@ class AdminService:
 
     async def _seed_default_configs(self):
         defaults = [
-            SystemConfig(key="ai_provider_translate", value={"primary": "sensenova", "fallback": ["alibaba-mt", "opencode_go"]}, description="翻译任务 AI 模型选择"),
+            SystemConfig(key="ai_routing", value={
-            SystemConfig(key="ai_provider_reply", value={"primary": "sensenova", "fallback": ["opencode_go"]}, description="回复建议 AI 模型选择"),
+                "translate": {"primary": "sensenova", "fallback": ["alibaba-mt", "nvidia"]},
-            SystemConfig(key="ai_provider_marketing", value={"primary": "sensenova", "fallback": ["opencode_go"]}, description="营销文案 AI 模型选择"),
+                "reply": {"primary": "sensenova", "fallback": ["nvidia"]},
-            SystemConfig(key="ai_provider_extract", value={"primary": "sensenova", "fallback": ["opencode_go"]}, description="信息提取 AI 模型选择"),
+                "marketing": {"primary": "sensenova", "fallback": ["nvidia"]},
-            SystemConfig(key="ai_provider_quotation", value={"primary": "sensenova", "fallback": ["opencode_go"]}, description="报价单 AI 模型选择"),
+                "extract": {"primary": "sensenova", "fallback": ["nvidia"]},
+                "quotation": {"primary": "sensenova", "fallback": ["nvidia"]},
+                "chat": {"primary": "sensenova", "fallback": ["nvidia"]},
+            }, description="AI 路由规则：各任务的主选/备用供应商"),
             SystemConfig(key="feature_guest_mode", value={"enabled": True}, description="游客模式开关"),
             SystemConfig(key="feature_wechat_login", value={"enabled": False}, description="微信登录开关"),
             SystemConfig(key="feature_registration", value={"enabled": True}, description="新用户注册开关"),
@@ -305,6 +308,19 @@ class AdminService:
             self.db.add(cfg)
         await self.db.flush()
 
+    async def _migrate_routing_configs(self):
+        from sqlalchemy import delete
+        # Remove stale individual routing keys (replaced by consolidated ai_routing)
+        stale_prefixes = ["ai_provider_translate", "ai_provider_reply", "ai_provider_marketing",
+                          "ai_provider_extract", "ai_provider_quotation"]
+        for key in stale_prefixes:
+            await self.db.execute(
+                delete(SystemConfig).where(SystemConfig.key == key)
+            )
+        if stale_prefixes:
+            await self.db.flush()
+            logger.info("Cleaned up stale ai_provider_* routing configs")
+
     async def list_config(self) -> List[Dict[str, Any]]:
         result = await self.db.execute(
             select(func.count(SystemConfig.id))
@@ -312,6 +328,28 @@ class AdminService:
         if result.scalar() == 0:
             await self._seed_default_configs()
 
+        await self._migrate_routing_configs()
+
+        # Ensure consolidated ai_routing exists
+        result = await self.db.execute(
+            select(SystemConfig).where(SystemConfig.key == "ai_routing")
+        )
+        if not result.scalar_one_or_none():
+            self.db.add(SystemConfig(
+                key="ai_routing",
+                value={
+                    "translate": {"primary": "sensenova", "fallback": ["alibaba-mt", "nvidia"]},
+                    "reply": {"primary": "sensenova", "fallback": ["nvidia"]},
+                    "marketing": {"primary": "sensenova", "fallback": ["nvidia"]},
+                    "extract": {"primary": "sensenova", "fallback": ["nvidia"]},
+                    "quotation": {"primary": "sensenova", "fallback": ["nvidia"]},
+                    "chat": {"primary": "sensenova", "fallback": ["nvidia"]},
+                },
+                description="AI 路由规则：各任务的主选/备用供应商",
+            ))
+            await self.db.flush()
+            logger.info("Seeded ai_routing config")
+
         result = await self.db.execute(
             select(SystemConfig).order_by(SystemConfig.key)
         )
@@ -336,6 +374,12 @@ class AdminService:
         config.value = value
         config.updated_at = datetime.utcnow()
         await self.db.flush()
+
+        if key == "ai_routing":
+            from app.ai.router import get_ai_router
+            await get_ai_router().reload_from_db(self.db)
+            logger.info("AI router reloaded after ai_routing config update")
+
         return {
             "key": config.key,
             "value": config.value,

backend/app/services/translation_quota.py

@@ -12,6 +12,10 @@ class TranslationQuotaService:
     def __init__(self, db: AsyncSession):
         self.db = db
 
+    def _default_desc(self, version: str) -> str:
+        labels = {"ecommerce": "阿里云翻译电商版", "general": "阿里云翻译通用版", "llm": "AI模型翻译"}
+        return labels.get(version, f"阿里云翻译{version}版")
+
     async def _get_or_create(self, version: str) -> TranslationQuota:
         result = await self.db.execute(
             select(TranslationQuota).where(TranslationQuota.version == version)
@@ -25,7 +29,7 @@ class TranslationQuotaService:
                 used_chars=0,
                 current_month=now.strftime("%Y-%m"),
                 enabled=True,
-                description=f"阿里云翻译{version}版",
+                description=self._default_desc(version),
             )
             self.db.add(quota)
             await self.db.flush()
@@ -57,8 +61,7 @@ class TranslationQuotaService:
         return remaining
 
     async def get_all_quotas(self) -> list:
-        default_versions = ["ecommerce", "general"]
+        for v in ("ecommerce", "general", "llm"):
-        for v in default_versions:
             await self._get_or_create(v)
 
         result = await self.db.execute(select(TranslationQuota).order_by(TranslationQuota.version))

backend/tests/test_config.py

@@ -14,11 +14,12 @@ class TestConfig:
         assert settings.REFRESH_TOKEN_EXPIRE_DAYS == 30
 
     def test_ai_routing_config(self):
-        assert "translate" in settings.AI_ROUTING
+        from app.ai.router import DEFAULT_ROUTING
-        assert "reply" in settings.AI_ROUTING
+        assert "translate" in DEFAULT_ROUTING
-        assert "marketing" in settings.AI_ROUTING
+        assert "reply" in DEFAULT_ROUTING
-        assert "extract" in settings.AI_ROUTING
+        assert "marketing" in DEFAULT_ROUTING
-        assert "primary" in settings.AI_ROUTING["translate"]
+        assert "extract" in DEFAULT_ROUTING
+        assert "primary" in DEFAULT_ROUTING["translate"]
 
     def test_free_tier_limits(self):
         assert settings.FREE_DAILY_TRANSLATE_CHARS == 5000

user-frontend/src/api/index.js

@@ -116,4 +116,7 @@ export function sendWhatsApp(data) { return http.post('/whatsapp/send', data) }
 
 export function getUsageStats() { return http.get('/usage/stats') }
 
+export function aiChat(message, history = []) { return http.post('/ai/chat', { message, history }) }
+export function aiQuickQuestions() { return http.get('/ai/quick-questions') }
+
 export default http

user-frontend/src/components/AiAssistant.vue

@@ -0,0 +1,382 @@
+<template>
+  <div>
+    <div class="ai-float-btn" @click="visible = true">
+      <span class="ai-float-icon">AI</span>
+    </div>
+
+    <el-dialog
+      v-model="visible"
+      title="TradeMate AI 助手"
+      width="480px"
+      :close-on-click-modal="false"
+      class="ai-dialog"
+      top="5vh"
+      @opened="onOpened"
+    >
+      <div class="ai-messages" ref="msgContainer">
+        <div v-for="(msg, i) in messages" :key="i" class="ai-msg-row" :class="msg.role">
+          <div class="ai-avatar" v-if="msg.role === 'assistant'">
+            <el-icon :size="18" color="#667eea"><Cpu /></el-icon>
+          </div>
+          <div class="ai-msg-body">
+            <div class="ai-msg-bubble">
+              <div class="ai-msg-text">{{ msg.content }}</div>
+
+              <div v-if="msg.actions && msg.actions.length" class="ai-action-card">
+                <div v-for="(action, ai) in msg.actions" :key="ai" class="ai-action-item">
+                  <div class="ai-action-title">{{ action.label }}</div>
+                  <div v-for="(val, key) in action.fields" :key="key" class="ai-field-row">
+                    <span class="ai-field-label">{{ fieldLabel(action.type, key) }}</span>
+                    <el-input
+                      v-model="action.fields[key]"
+                      :placeholder="fieldPlaceholder(action.type, key)"
+                      size="small"
+                    />
+                  </div>
+                  <div class="ai-action-btns">
+                    <el-button size="small" @click="cancelAction(i, ai)">取消</el-button>
+                    <el-button size="small" type="primary" @click="confirmAction(i, ai)">确认执行</el-button>
+                  </div>
+                </div>
+              </div>
+            </div>
+          </div>
+        </div>
+
+        <div v-if="loading" class="ai-loading">
+          <el-icon class="is-loading" :size="18"><Loading /></el-icon>
+          <span>思考中...</span>
+        </div>
+
+        <div v-if="showSuggestions" class="ai-suggestions">
+          <div
+            v-for="(s, i) in suggestions"
+            :key="i"
+            class="ai-suggestion"
+            @click="sendQuick(s)"
+          >
+            {{ s }}
+          </div>
+        </div>
+      </div>
+
+      <template #footer>
+        <div class="ai-input-bar">
+          <el-input
+            v-model="inputText"
+            placeholder="输入你的问题..."
+            :disabled="loading"
+            size="default"
+            @keyup.enter="send"
+          />
+          <el-button type="primary" :disabled="!inputText.trim() || loading" @click="send">发送</el-button>
+        </div>
+      </template>
+    </el-dialog>
+  </div>
+</template>
+
+<script setup>
+import { ref, nextTick } from 'vue'
+import { ElMessage } from 'element-plus'
+import { Cpu, Loading } from '@element-plus/icons-vue'
+import {
+  aiChat, aiQuickQuestions,
+  createCustomer, createProduct, createQuotation,
+  scanFollowups, generateMarketing, discoverySearch,
+} from '@/api'
+
+const visible = ref(false)
+const inputText = ref('')
+const loading = ref(false)
+const suggestions = ref([])
+const msgContainer = ref(null)
+const messages = ref([
+  { role: 'assistant', content: '你好！我是 TradeMate AI 助手，可以帮你操作外贸工具的各项功能。你可以说："帮我添加一个客户"、"创建一个产品"、"生成报价单"等。' },
+])
+const showSuggestions = ref(false)
+
+const fieldLabel = (type, key) => {
+  const labels = {
+    name: '名称', phone: '电话', email: '邮箱', company: '公司',
+    country: '国家', website: '网站', notes: '备注',
+    name_en: '英文名称', description: '描述', description_en: '英文描述',
+    category: '分类', price: '价格', price_unit: '货币', moq: 'MOQ',
+    keywords: '关键词',
+    customer_name: '客户名称', product_info: '产品信息', quantity: '数量',
+    terms: '交易条款',
+    message: '消息内容',
+    product_name: '产品名称', target_market: '目标市场', tone: '语气',
+    language: '语言',
+    industry: '行业',
+  }
+  return labels[key] || key
+}
+
+const fieldPlaceholder = (type, key) => {
+  const placeholders = {
+    phone: '如 +86-13800138000',
+    email: '如 contact@example.com',
+    price: '如 12.50',
+    price_unit: '默认 USD',
+    keywords: '逗号分隔',
+    quantity: '如 1000 pcs',
+    target_market: '如 美国、欧洲',
+  }
+  return placeholders[key] || ''
+}
+
+const cancelAction = (msgIdx, actionIdx) => {
+  const msg = messages.value[msgIdx]
+  if (msg.actions) msg.actions.splice(actionIdx, 1)
+}
+
+const confirmAction = async (msgIdx, actionIdx) => {
+  const action = messages.value[msgIdx].actions[actionIdx]
+  const { type, fields } = action
+  loading.value = true
+  try {
+    switch (type) {
+      case 'create_customer':
+        if (!fields.name) { ElMessage.warning('客户名称不能为空'); return }
+        await createCustomer(fields)
+        break
+      case 'create_product':
+        if (!fields.name) { ElMessage.warning('产品名称不能为空'); return }
+        if (fields.keywords && typeof fields.keywords === 'string') {
+          fields.keywords = fields.keywords.split(/[,，]/).map(s => s.trim()).filter(Boolean)
+        }
+        await createProduct(fields)
+        break
+      case 'create_quotation':
+        if (!fields.customer_name) { ElMessage.warning('客户名称不能为空'); return }
+        if (!fields.product_info) { ElMessage.warning('产品信息不能为空'); return }
+        if (!fields.quantity) { ElMessage.warning('数量不能为空'); return }
+        await createQuotation({
+          customer_name: fields.customer_name,
+          items: [{ description: fields.product_info, quantity: fields.quantity, price: fields.price || '0' }],
+          terms: fields.terms || '',
+        })
+        break
+      case 'scan_followups':
+        await scanFollowups()
+        break
+      case 'generate_marketing':
+        if (!fields.product_name) { ElMessage.warning('产品名称不能为空'); return }
+        await generateMarketing({
+          product_info: { name: fields.product_name },
+          target_market: fields.target_market,
+          tone: fields.tone,
+          language: fields.language,
+        })
+        break
+      case 'discovery_search':
+        if (!fields.keywords) { ElMessage.warning('关键词不能为空'); return }
+        await discoverySearch({
+          keywords: fields.keywords.split(/[,，]/).map(s => s.trim()).filter(Boolean),
+          country: fields.country,
+          industry: fields.industry,
+        })
+        break
+      default:
+        ElMessage.warning(`未知操作类型: ${type}`)
+        return
+    }
+    ElMessage.success('操作成功')
+    messages.value[msgIdx].actions = []
+  } catch (e) {
+    ElMessage.error(e.message || e.detail || '操作失败')
+  } finally {
+    loading.value = false
+  }
+}
+
+const sendQuick = (text) => {
+  inputText.value = text
+  send()
+}
+
+const fetchSuggestions = async () => {
+  try {
+    const res = await aiQuickQuestions()
+    if (Array.isArray(res)) suggestions.value = res
+  } catch {}
+}
+
+const onOpened = () => {
+  if (suggestions.value.length === 0) fetchSuggestions()
+  showSuggestions.value = messages.value.length <= 1
+  scrollToBottom()
+}
+
+const send = async () => {
+  const msg = inputText.value.trim()
+  if (!msg || loading.value) return
+  inputText.value = ''
+  showSuggestions.value = false
+  messages.value.push({ role: 'user', content: msg })
+  loading.value = true
+  scrollToBottom()
+
+  try {
+    const hist = messages.value.map(m => ({ role: m.role, content: m.content }))
+    const res = await aiChat(msg, hist.slice(0, -1))
+    const newMsg = { role: 'assistant', content: res.reply || '抱歉，我没有理解，请重新描述一下你的问题。' }
+    if (res.actions && res.actions.length) {
+      newMsg.actions = res.actions
+    }
+    messages.value.push(newMsg)
+  } catch {
+    messages.value.push({ role: 'assistant', content: '请求失败，请稍后重试。' })
+  } finally {
+    loading.value = false
+    scrollToBottom()
+  }
+}
+
+const scrollToBottom = async () => {
+  await nextTick()
+  const el = msgContainer.value
+  if (el) el.scrollTop = el.scrollHeight
+}
+</script>
+
+<style scoped>
+.ai-float-btn {
+  position: fixed;
+  right: 30px;
+  bottom: 30px;
+  width: 52px;
+  height: 52px;
+  border-radius: 50%;
+  background: linear-gradient(135deg, #667eea, #764ba2);
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  box-shadow: 0 4px 20px rgba(102, 126, 234, 0.4);
+  z-index: 9999;
+  cursor: pointer;
+  transition: transform 0.2s;
+}
+.ai-float-btn:hover {
+  transform: scale(1.08);
+}
+.ai-float-icon {
+  color: #fff;
+  font-size: 18px;
+  font-weight: bold;
+}
+.ai-dialog {
+  --el-dialog-content-padding: 0;
+}
+.ai-dialog :deep(.el-dialog__body) {
+  padding: 0;
+  height: 480px;
+  overflow: hidden;
+  display: flex;
+  flex-direction: column;
+}
+.ai-messages {
+  flex: 1;
+  padding: 16px;
+  overflow-y: auto;
+  background: #f5f5f5;
+}
+.ai-msg-row {
+  margin-bottom: 14px;
+  display: flex;
+  gap: 8px;
+}
+.ai-msg-row.user {
+  flex-direction: row-reverse;
+}
+.ai-avatar {
+  width: 32px;
+  height: 32px;
+  border-radius: 50%;
+  background: #f0edff;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  flex-shrink: 0;
+}
+.ai-msg-body {
+  max-width: 80%;
+}
+.ai-msg-bubble {
+  padding: 10px 14px;
+  border-radius: 8px;
+  background: #fff;
+  box-shadow: 0 1px 3px rgba(0,0,0,0.06);
+}
+.ai-msg-row.user .ai-msg-bubble {
+  background: #667eea;
+  color: #fff;
+}
+.ai-msg-text {
+  font-size: 13px;
+  line-height: 1.6;
+  white-space: pre-wrap;
+  word-break: break-word;
+}
+.ai-msg-row.user .ai-msg-text {
+  color: #fff;
+}
+.ai-action-card {
+  margin-top: 12px;
+  border-top: 1px solid #eee;
+  padding-top: 10px;
+}
+.ai-action-title {
+  font-size: 14px;
+  font-weight: 600;
+  color: #333;
+  margin-bottom: 10px;
+}
+.ai-field-row {
+  margin-bottom: 8px;
+}
+.ai-field-label {
+  display: block;
+  font-size: 12px;
+  color: #666;
+  margin-bottom: 3px;
+}
+.ai-action-btns {
+  display: flex;
+  gap: 8px;
+  margin-top: 10px;
+}
+.ai-loading {
+  text-align: center;
+  padding: 12px 0;
+  display: flex;
+  align-items: center;
+  justify-content: center;
+  gap: 6px;
+  color: #999;
+  font-size: 13px;
+}
+.ai-suggestions {
+  padding: 8px 0;
+}
+.ai-suggestion {
+  background: #f0edff;
+  border-radius: 6px;
+  padding: 8px 12px;
+  margin-bottom: 8px;
+  font-size: 13px;
+  color: #667eea;
+  cursor: pointer;
+  transition: background 0.15s;
+}
+.ai-suggestion:hover {
+  background: #e0dbff;
+}
+.ai-input-bar {
+  display: flex;
+  gap: 8px;
+  padding: 10px 16px;
+  border-top: 1px solid #f0f0f0;
+}
+</style>

user-frontend/src/layouts/UserLayout.vue

@@ -102,6 +102,7 @@
         </div>
       </footer>
     </div>
+    <AiAssistant />
   </div>
 </template>
 
@@ -110,6 +111,7 @@ import { ref, computed, onMounted } from 'vue'
 import { useRoute, useRouter } from 'vue-router'
 import { useAuthStore } from '@/stores/auth'
 import { getUnreadCount } from '@/api'
+import AiAssistant from '@/components/AiAssistant.vue'
 
 const route = useRoute()
 const router = useRouter()

user-frontend/src/views/WorkspaceLanding.vue

@@ -193,7 +193,8 @@ async function register() {
     regForm.password = ''
   } catch (e) {
     const map = { 'Phone already registered': '该手机号已被注册', 'Invalid credentials': '手机号或密码错误' }
-    regError.value = map[e?.detail] || e?.detail || '注册失败'
+    const detail = typeof e === 'string' ? e : e?.detail
+    regError.value = map[detail] || detail || '注册失败，请检查网络连接'
   } finally {
     regLoading.value = false
   }

user-frontend/vite.config.js

@@ -15,7 +15,7 @@ export default defineConfig({
   server: {
     port: 5174,
     proxy: {
-      '/api': { target: 'http://localhost:8002', changeOrigin: true }
+      '/api': { target: 'http://localhost:8000', changeOrigin: true }
     }
   }
 })