后台权限管理支持「自定义」角色：按人存 pages_override，可见页逐页勾选并即时生效

admin_user 加 pages_override(可空 JSON)+ Alembic 迁移;role=="custom" 时可见页由这份逐页
勾选决定,否则跟随角色。create/update 收 pages_override(切回普通角色自动清空),_validate_role
放行 custom 哨兵角色。登录/me 下发有效页时优先用 override → 左侧导航直接按勾选生效。
补 4 个用例覆盖建/改/切回/切入,test_admin_roles 全绿。

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

.env.example

@@ -29,7 +29,7 @@ JG_REQUEST_TIMEOUT_SEC=15
 
 # ===== 无障碍保护存活监控(pull 后置检测;本期不接推送)=====
 HEARTBEAT_MONITOR_ENABLED=true
-HEARTBEAT_TIMEOUT_MINUTES=10
+HEARTBEAT_TIMEOUT_MINUTES=60
 HEARTBEAT_SCAN_INTERVAL_SEC=60
 
 # ===== 短信 (mock 模式) =====
@@ -58,6 +58,13 @@ MT_CPS_DEFAULT_SID=sgbjia
 # 线上国内服务器留空(=直连)。留空且本机直连失败时 /feed、/coupons、/top-sales 会返回空。
 MT_CPS_PROXY=
 
+# ===== 京东联盟 CPS =====
+# 京东联盟/京东宙斯开放平台创建应用后填写。AUTH_KEY 是工具商授权 key,自有应用可留空。
+JD_UNION_APP_KEY=
+JD_UNION_APP_SECRET=
+JD_UNION_SITE_ID=
+JD_UNION_AUTH_KEY=
+
 # ===== Pricebot 上游 (领券/比价业务透传目标) =====
 # 客户端调本服务的 /api/v1/coupon/step 等,我们透传到 pricebot-backend。
 # 本地开发用 localhost:8000。生产部署改成内网地址(如 http://pricebot.internal:8000)。
@@ -117,3 +124,16 @@ PANGLE_REWARD_SECRET=
 # ⚠️ 仅本地联调:true 时开放 POST /api/v1/ad/test-grant,让 debug 客户端看完广告直接发奖,
 # 验证"看广告→金币到账"全链路(未部署公网、穿山甲 S2S 打不到本地时用)。生产必须 false(绕过反作弊)。
 AD_REWARD_TEST_GRANT_ENABLED=false
+
+# ===== 穿山甲 GroMore 数据 API(按天拉收益报表,供后台广告收益报表的「穿山甲后台收益」)=====
+# ⚠️ 与上面发奖回调的 m-key 是【两套不同凭证】:这三样在穿山甲后台「接入中心 → GroMore-API →
+# 聚合数据报告 API」文档页领取。只读拉取 GroMore 天级 revenue(预估)/ api_revenue(收益Api),
+# 不参与发奖。三样齐全才生效;留空 = scripts/sync_pangle_revenue 直接 no-op。
+# 子账号(role_id≠user_id)需主账号在「角色管理」授予「查看全部数据」权限,否则查不到收益(接口 118);
+# role_id 填成 = user_id 即查主账号数据。同步:线上每天 ~10:30 由 timer 跑 python -m scripts.sync_pangle_revenue。
+PANGLE_REPORT_USER_ID=0
+PANGLE_REPORT_ROLE_ID=0
+PANGLE_REPORT_SECURITY_KEY=
+# GroMore AppId(报表 site_id 维度)→ 应用环境;默认取现网两个应用,按需覆盖。
+PANGLE_REPORT_SITE_ID_PROD=5830519
+PANGLE_REPORT_SITE_ID_TEST=5832303

.gitignore

@@ -23,14 +23,17 @@ dist/
 *.db-journal
 *.db-shm
 *.db-wal
-# data/ 整体不入库(运行时数据/上传文件/大二进制)。例外:邀请落地页 dl.html——
-# 它既是生产落地页又是本地测试资产,纳入 git 便于同事一致测试
+# data/ 整体不入库(运行时数据/上传文件/大二进制)。例外:邀请落地页 dl.html 及其
+# 引用的静态插画(coupon-page-bg.png 底图 + sb-brand.png logo)——既是生产落地页资产
+# 又是本地测试资产,纳入 git 便于同事一致测试、且随部署进生产 /media(否则线上 404→落地页毛坯)。
 # (见 docs/邀请功能-实现原理与本地测试.md)。其余(avatars/ / *.apk / app.db 等)仍忽略。
 data/*
 !data/media/
 data/media/*
 !data/media/dl.html
 !data/media/taobao_landing.jpg
+!data/media/coupon-page-bg.png
+!data/media/sb-brand.png
 
 secrets/*
 !secrets/.gitkeep
@@ -45,3 +48,12 @@ secrets/*
 # 运行日志(run.sh 输出, 不入库)
 *.log
 logs/
+
+# Claude Code 自动持久化的权限 allowlist / 个人本地设置(会话专属,不入库)。
+# 需要团队共享的 Claude 配置(commands/ 等)可单独 git add -f,不受此忽略影响。
+.claude/settings.json
+.claude/settings.local.json
+tests/meituan_coupon_bj.tsv
+tests/meituan_coupon_data.tsv
+tests/meituan_coupon_fz.tsv
+tests/meituan_coupon_xm.tsv

CLAUDE.md

@@ -0,0 +1,116 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project overview
+
+Shaguabijia (傻瓜比价) App backend — FastAPI + SQLAlchemy 2.0 + JWT. Covers user auth (Jiguang one-click / SMS), welfare wallet (coins/cash/signin/tasks/savings), WeChat Pay withdrawals, ad-reward callbacks (Pangle/GroMore S2S), Meituan CPS (coupon forwarding / price comparison), and an admin backend.
+
+## Commands
+
+```bash
+# Install
+pip install -e ".[dev]"
+
+# Run app server (port 8770, auto-migrates, auto-reload)
+./run.sh                    # or: uvicorn app.main:app --reload --port 8770
+
+# Run admin server (port 8771, separate process)
+uvicorn app.admin.main:admin_app --reload --port 8771
+
+# Database
+alembic upgrade head        # apply all migrations (idempotent)
+alembic revision --autogenerate -m "description"  # generate new migration
+
+# Tests
+pytest -q                   # all tests
+pytest tests/test_auth.py -q        # single file
+pytest -k "test_sms_login" -q       # single test by name
+
+# Lint
+ruff check .
+ruff check --fix .
+```
+
+## Architecture: two FastAPI apps
+
+This repo runs **two separate FastAPI processes** sharing the same `app/` codebase (models, repositories, integrations, config):
+
+| | App server | Admin server |
+|---|---|---|
+| Entry | `app/main.py` → `app:app` | `app/admin/main.py` → `admin_app` |
+| Port | 8770 | 8771 |
+| Auth | User JWT (`JWT_SECRET_KEY`) | Admin JWT (`ADMIN_JWT_SECRET`, independent) |
+| Audience | Mobile app clients | Internal admin dashboard |
+| Docs | `/docs` (non-prod only) | `/admin/docs` (non-prod only) |
+
+The two apps are intentionally decoupled — `app.main` never imports `app.admin`. Admin has its own auth chain (`app/admin/deps.py`, `app/admin/security.py`), role-based guards (`require_role`), and routers under `app/admin/routers/`.
+
+## Layered request flow
+
+```
+api/v1/  (thin: parse → delegate → respond + HTTP errors)
+  ├── integrations/   (external SDKs: signature, encryption, HTTP calls)
+  └── repositories/   (data access + transactions)
+        └── models/   (SQLAlchemy ORM, DeclarativeBase)
+```
+
+- **`api/v1/`**: Route handlers. Keep these thin — parse request, call repository or integration, return response. Never put business logic or external HTTP here.
+- **`api/deps.py`**: Shared FastAPI dependencies — `get_current_user` (Bearer JWT → User ORM object), `get_db` (request-scoped session).
+- **`integrations/`**: All external service logic — Jiguang REST + RSA decryption, WeChat Pay V3 signing/encryption, Meituan CPS gateway signing, Pangle callback signature verification, SMS sending. This is the layer you change when swapping vendors.
+- **`repositories/`**: Data access. Each file wraps SQLAlchemy queries + transactions for one domain (user, wallet, signin, savings, ad_reward, etc.). Some repositories also call integrations (e.g., `wallet.py` calls `integrations/wxpay.py` for withdrawals).
+- **`models/`**: ORM table definitions (SQLAlchemy 2.0 `Mapped` style, `DeclarativeBase`). Every new model must be imported in `app/models/__init__.py` so Alembic can discover it.
+- **`schemas/`**: Pydantic request/response contracts.
+- **`core/`**: Infrastructure — config (`pydantic-settings`), JWT (`security.py`), in-memory rate limiter (`ratelimit.py`), reward constants (`rewards.py`), logging setup, pricebot router (consistent-hash load balancing), withdraw reconcile worker.
+
+## Internal (server-to-server) endpoints
+
+Endpoints under `app/api/internal/` are for server-to-server communication (pricebot → app-server), NOT for clients. They use a shared secret header `X-Internal-Secret` (compared via `hmac.compare_digest`) instead of user JWT. If `INTERNAL_API_SECRET` is empty, these endpoints return 503.
+
+## Auth system
+
+- **User login**: Jiguang one-click (`integrations/jiguang.py` — REST token verification + RSA decryption with multi-padding retry) or SMS code (mock by default; `SMS_MOCK=true`).
+- **Tokens**: JWT access (2h) + refresh (30d). Both are JWT with `typ` claim (`"access"` vs `"refresh"`) to prevent refresh-as-access. See `core/security.py`.
+- **Admin auth**: Separate JWT secret (`ADMIN_JWT_SECRET`), 12h expiry, no refresh. Username + bcrypt password login. Role-based access via `require_role()` guard in `app/admin/deps.py` (`super_admin` bypasses all role checks).
+- **Rate limiting**: In-memory fixed-window by client IP (`core/ratelimit.py`). Single-worker only; disabled in tests via `RATE_LIMIT_ENABLED=false`.
+
+## Database
+
+- **Dev**: SQLite (`sqlite:///./data/app.db`), `check_same_thread=False`, no connection pool.
+- **Prod**: PostgreSQL — just change `DATABASE_URL` in `.env`. Pool size 10 + max overflow 20, pool_recycle 3600.
+- **Migrations**: Alembic with `render_as_batch` for SQLite compatibility. ~60+ migration files in `alembic/versions/` (filenames are descriptive, not hex prefixes). Migration chain uses `down_revision` within each file.
+- **New models**: Define in `app/models/`, import in `app/models/__init__.py`, then run `alembic revision --autogenerate`.
+
+## Config
+
+All config via `pydantic-settings` in `app/core/config.py`. Single `Settings` class with env vars / `.env` file. Access anywhere via `from app.core.config import settings`. Key patterns:
+- `*_configured` properties gate features gracefully (e.g., `mt_cps_configured`, `wxpay_configured`, `pangle_callback_configured`) — missing credentials → endpoints return empty/503 rather than crashing at startup.
+- Prod validation: `_enforce_prod_secrets` model validator blocks startup if `APP_ENV=prod` with weak JWT secrets.
+
+## Testing
+
+- `tests/conftest.py`: Sets env vars BEFORE imports, creates temp SQLite file, builds all tables with `Base.metadata.create_all()`, tears down with `drop_all()` + unlink.
+- External integrations are monkeypatched in tests (e.g., WeChat Pay, Jiguang, Pangle callbacks) — tests never make real HTTP calls.
+- `TestClient` from FastAPI is used for all tests. Rate limiting is disabled globally in tests.
+
+## Key integration details
+
+- **Jiguang one-click login**: REST call to verify `loginToken`, then RSA decrypt the returned phone number. Multiple padding schemes tried in order (PKCS1v15, OAEP with SHA1/SHA256) because Jiguang's encryption padding varies.
+- **WeChat Pay withdrawals**: V3 API merchant transfer to user WeChat balance. Lazy-loads merchant certificates from `secrets/`. Withdrawal flow: bind WeChat → create withdraw order → auto-reconcile worker polls pending orders.
+- **Pangle ad rewards**: S2S callback verification via SHA256 signature. Multiple `m-key` secrets supported (one per ad placement). Callback is idempotent by `trans_id`. Test grant endpoint (`AD_REWARD_TEST_GRANT_ENABLED`) for local debugging — must be false in prod.
+- **Meituan CPS**: Gateway signature-based API calls. Proxy support (`MT_CPS_PROXY`) for local dev (direct connection causes SSL EOF). Coupon endpoints gracefully return empty when credentials are missing.
+- **Pricebot forwarding**: `/api/v1/coupon/step` and `/api/v1/compare/*` proxy to pricebot-backend. Multi-instance support with consistent-hash routing by `trace_id` (see `core/pricebot_router.py`).
+- **CPS redirect**: `/c/{code}` is a public (no auth) short-link redirect — records a click then 302s to Meituan. Click recording failure never blocks the redirect.
+
+## Money and units
+
+All monetary amounts are in **cents** (`*_cents` fields). Coins/gold have their own unit. Conversion constants are in `core/rewards.py`.
+
+## Scripts
+
+Key operational scripts in `scripts/`:
+- `migrate.sh` — run migrations standalone
+- `create_admin.py` — create admin user
+- `daily_auto_exchange.py` — auto-convert coins to cash (triggered by systemd timer)
+- `reconcile_withdraws.py` — reconcile withdrawal orders with WeChat Pay
+- `sim_pangle_callback.py` — simulate Pangle S2S callback for testing

alembic/versions/1699fc2c069f_add_analytics_event_table.py

@@ -0,0 +1,65 @@
+"""add analytics_event table
+
+Revision ID: 1699fc2c069f
+Revises: bcfcaf07152b
+Create Date: 2026-06-26 16:35:16.133975
+
+"""
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = '1699fc2c069f'
+down_revision: str | Sequence[str] | None = 'bcfcaf07152b'
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table('analytics_event',
+    sa.Column('id', sa.Integer(), autoincrement=True, nullable=False),
+    sa.Column('event', sa.String(length=64), nullable=False),
+    sa.Column('props', sa.JSON(), nullable=True),
+    sa.Column('device_id', sa.String(length=64), nullable=False),
+    sa.Column('user_id', sa.Integer(), nullable=True),
+    sa.Column('session_id', sa.String(length=64), nullable=True),
+    sa.Column('client_ts', sa.BigInteger(), nullable=False),
+    sa.Column('sent_at', sa.BigInteger(), nullable=True),
+    sa.Column('page', sa.String(length=64), nullable=True),
+    sa.Column('client_ip', sa.String(length=64), nullable=True),
+    sa.Column('oem', sa.String(length=32), nullable=True),
+    sa.Column('os', sa.String(length=32), nullable=True),
+    sa.Column('model', sa.String(length=64), nullable=True),
+    sa.Column('app_ver', sa.String(length=32), nullable=True),
+    sa.Column('network', sa.String(length=16), nullable=True),
+    sa.Column('channel', sa.String(length=32), nullable=True),
+    sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.text('(CURRENT_TIMESTAMP)'), nullable=False),
+    sa.PrimaryKeyConstraint('id')
+    )
+    with op.batch_alter_table('analytics_event', schema=None) as batch_op:
+        batch_op.create_index(batch_op.f('ix_analytics_event_created_at'), ['created_at'], unique=False)
+        batch_op.create_index(batch_op.f('ix_analytics_event_device_id'), ['device_id'], unique=False)
+        batch_op.create_index(batch_op.f('ix_analytics_event_event'), ['event'], unique=False)
+        batch_op.create_index(batch_op.f('ix_analytics_event_session_id'), ['session_id'], unique=False)
+        batch_op.create_index(batch_op.f('ix_analytics_event_user_id'), ['user_id'], unique=False)
+
+    # 注:autogenerate 顺带检出 ad_ecpm/cps_*/invite_fingerprint 的历史索引漂移,
+    # 与本次「新增埋点表」无关,已手动移除,避免本迁移误改他人表。
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    with op.batch_alter_table('analytics_event', schema=None) as batch_op:
+        batch_op.drop_index(batch_op.f('ix_analytics_event_user_id'))
+        batch_op.drop_index(batch_op.f('ix_analytics_event_session_id'))
+        batch_op.drop_index(batch_op.f('ix_analytics_event_event'))
+        batch_op.drop_index(batch_op.f('ix_analytics_event_device_id'))
+        batch_op.drop_index(batch_op.f('ix_analytics_event_created_at'))
+
+    op.drop_table('analytics_event')
+    # ### end Alembic commands ###

alembic/versions/761ef181ce7c_merge_jd_cps_order_fields_and_coupon_.py

@@ -0,0 +1,26 @@
+"""merge jd_cps_order_fields and coupon_session_origin_package heads
+
+Revision ID: 761ef181ce7c
+Revises: coupon_session_origin_package, jd_cps_order_fields
+Create Date: 2026-07-01 13:52:16.068808
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '761ef181ce7c'
+down_revision: Union[str, Sequence[str], None] = ('coupon_session_origin_package', 'jd_cps_order_fields')
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    pass
+
+
+def downgrade() -> None:
+    pass

alembic/versions/7db22acee504_merge_invite_cash_analytics_device_heads.py

@@ -0,0 +1,26 @@
+"""merge invite_cash + analytics + device heads
+
+Revision ID: 7db22acee504
+Revises: c2874d2bf705, device_first_protected_at, invite_cash_compare_reward
+Create Date: 2026-06-27 03:06:52.594401
+
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = '7db22acee504'
+down_revision: Union[str, Sequence[str], None] = ('c2874d2bf705', 'device_first_protected_at', 'invite_cash_compare_reward')
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    pass
+
+
+def downgrade() -> None:
+    pass

alembic/versions/ad_pangle_daily_revenue.py

@@ -0,0 +1,63 @@
+"""ad_pangle_daily_revenue: 穿山甲 GroMore 天级收益报表(后台结算口径)
+
+新建表存放从 GroMore 数据 API 按天拉取的收益(revenue 预估 + api_revenue 收益Api),
+粒度 = 日期 × 应用(app_env) × 代码位(our_code_id) × 广告源(adn)。供广告收益报表的
+汇总/趋势级展示「穿山甲后台收益」,与客户端自报 eCPM 折算的预估互为对照。
+
+Revision ID: ad_pangle_daily_revenue
+Revises: 7db22acee504
+Create Date: 2026-06-28
+"""
+
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "ad_pangle_daily_revenue"
+down_revision = "7db22acee504"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    op.create_table(
+        "ad_pangle_daily_revenue",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("report_date", sa.String(length=10), nullable=False),
+        sa.Column("app_env", sa.String(length=16), nullable=False),
+        sa.Column("site_id", sa.String(length=32), nullable=True),
+        sa.Column("our_code_id", sa.String(length=64), nullable=False),
+        sa.Column("adn", sa.String(length=16), nullable=False, server_default=""),
+        sa.Column("revenue_yuan", sa.Float(), nullable=False, server_default="0"),
+        sa.Column("api_revenue_yuan", sa.Float(), nullable=True),
+        sa.Column("ecpm", sa.String(length=32), nullable=True),
+        sa.Column("impressions", sa.Integer(), nullable=False, server_default="0"),
+        sa.Column("currency", sa.String(length=8), nullable=False, server_default="cny"),
+        sa.Column(
+            "synced_at",
+            sa.DateTime(timezone=True),
+            server_default=sa.func.now(),
+            nullable=False,
+        ),
+        sa.UniqueConstraint(
+            "report_date", "app_env", "our_code_id", "adn", name="uq_ad_pangle_daily"
+        ),
+    )
+    op.create_index(
+        "ix_ad_pangle_daily_revenue_report_date",
+        "ad_pangle_daily_revenue",
+        ["report_date"],
+    )
+    op.create_index(
+        "ix_ad_pangle_daily_revenue_our_code_id",
+        "ad_pangle_daily_revenue",
+        ["our_code_id"],
+    )
+
+
+def downgrade() -> None:
+    op.drop_index("ix_ad_pangle_daily_revenue_our_code_id", table_name="ad_pangle_daily_revenue")
+    op.drop_index("ix_ad_pangle_daily_revenue_report_date", table_name="ad_pangle_daily_revenue")
+    op.drop_table("ad_pangle_daily_revenue")

alembic/versions/admin_role_table.py

@@ -0,0 +1,69 @@
+"""admin_role 表(RBAC 角色 → 可见页面)+ 种子内建角色
+
+新增后台 RBAC:角色持有一组页面 key,登录后左侧只展示这些页。super_admin 内建全权(pages 存空、
+effective_pages 特判为全部)、不可编辑删除;operator/finance 播默认页集,可由 super_admin 增删改。
+admin_user.role 弱引用本表 name。全新环境顺序应用即得三条内建角色。
+
+Revision ID: admin_role_table
+Revises: comparison_product_names
+Create Date: 2026-07-04 00:00:00.000000
+"""
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+from alembic import op
+
+revision: str = "admin_role_table"
+down_revision: str | Sequence[str] | None = "comparison_product_names"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+_JSON = sa.JSON().with_variant(postgresql.JSONB(), "postgresql")
+
+# 与 app/admin/permissions.py 的 BUILTIN_ROLES 同口径(迁移内联一份,不耦合 app 常量)。
+# 页集对齐 Prototypes/dashboard/permissions.md 的 ROLES。key 承重(require_role),label 为展示名。
+_BUILTIN = [
+    {"name": "super_admin", "label": "管理员", "pages": [], "is_builtin": True},
+    {"name": "operator", "label": "运营", "pages": [
+        "dashboard", "coupon-data", "ad-revenue-report", "comparison-records",
+        "cps", "device-liveness", "price-reports", "feedbacks",
+    ], "is_builtin": False},
+    {"name": "finance", "label": "财务", "pages": [
+        "dashboard", "ad-revenue-report", "cps", "withdraws",
+    ], "is_builtin": False},
+    {"name": "tech", "label": "技术", "pages": [
+        "dashboard", "device-liveness", "config", "ad-revenue", "event-logs", "audit-logs",
+    ], "is_builtin": False},
+]
+
+
+def upgrade() -> None:
+    op.create_table(
+        "admin_role",
+        sa.Column("id", sa.Integer(), primary_key=True, autoincrement=True),
+        sa.Column("name", sa.String(length=32), nullable=False),
+        sa.Column("label", sa.String(length=32), nullable=False, server_default=""),
+        sa.Column("pages", _JSON, nullable=False),
+        sa.Column("is_builtin", sa.Boolean(), nullable=False, server_default=sa.false()),
+        sa.Column(
+            "created_at", sa.DateTime(timezone=True),
+            server_default=sa.func.now(), nullable=False,
+        ),
+    )
+    op.create_index("ix_admin_role_name", "admin_role", ["name"], unique=True)
+
+    tbl = sa.table(
+        "admin_role",
+        sa.column("name", sa.String),
+        sa.column("label", sa.String),
+        sa.column("pages", _JSON),
+        sa.column("is_builtin", sa.Boolean),
+    )
+    op.bulk_insert(tbl, _BUILTIN)
+
+
+def downgrade() -> None:
+    op.drop_index("ix_admin_role_name", table_name="admin_role")
+    op.drop_table("admin_role")

alembic/versions/admin_user_pages_override.py

@@ -0,0 +1,35 @@
+"""admin_user 加 pages_override 列(「自定义」权限:按人存专属可见页 key 列表)
+
+权限管理页新增「自定义」角色:选它时该成员的可见页不跟随任何共享角色,而由逐页勾选决定,
+存这个人专属的一份页 key 列表。仅 role == "custom" 时有效;普通角色为 None(可见页跟随角色)。
+PG 用 JSONB,SQLite 退化为通用 JSON(同 admin_audit_log.detail / comparison_record.raw_payload)。
+
+Revision ID: admin_user_pages_override
+Revises: admin_user_plain_password
+Create Date: 2026-07-08 00:00:00.000000
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from sqlalchemy.dialects.postgresql import JSONB
+
+from alembic import op
+
+revision: str = "admin_user_pages_override"
+down_revision: str | Sequence[str] | None = "admin_user_plain_password"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+# 与 app/models/admin.py 的 _JSON 一致:PG JSONB / 其它 JSON
+_JSON = sa.JSON().with_variant(JSONB(), "postgresql")
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("admin_user", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("pages_override", _JSON, nullable=True))
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("admin_user", schema=None) as batch_op:
+        batch_op.drop_column("pages_override")

alembic/versions/admin_user_plain_password.py

@@ -0,0 +1,29 @@
+"""admin_user 加 plain_password 列(明文登录密码,仅后台 UI 建/重置的账号留存)
+
+权限管理页需能复看某成员已确定的登录密码转交本人。系统只存哈希无法反推,故对「后台 UI 创建/重置」
+的管理员额外留存一份明文;脚本/起后台建的超管为 None(前端不显示密码)。旧账号无此列值 → 同样不显示。
+
+Revision ID: admin_user_plain_password
+Revises: admin_role_table
+Create Date: 2026-07-04 00:00:00.000000
+"""
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision: str = "admin_user_plain_password"
+down_revision: str | Sequence[str] | None = "admin_role_table"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("admin_user", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("plain_password", sa.String(length=128), nullable=True))
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("admin_user", schema=None) as batch_op:
+        batch_op.drop_column("plain_password")

alembic/versions/c2874d2bf705_add_ad_type_to_ad_feed_reward_and_feed_.py

@@ -0,0 +1,46 @@
+"""add ad_type to ad_feed_reward and feed_scene to ad_ecpm
+
+把信息流广告全面改造成 Draw 信息流(draw):
+- ad_feed_reward_record.ad_type:广告形态 feed(信息流) / draw(Draw 信息流)。可空,旧数据 NULL
+  一律视为 feed(向后兼容);每日上限与因子2(LT)仍按本表全表 unit 累计,不按 ad_type 拆。
+- ad_ecpm_record.feed_scene:点位场景 comparison(比价) / coupon(领券) / welfare(福利),供广告
+  收益报表区分比价/领券 Draw 收益;仅信息流/Draw 上报,激励视频为 NULL。
+
+注:autogenerate 会顺带探测到 analytics_event / cps_* / invite_fingerprint 等无关索引差异(本地库与
+metadata 漂移、analytics 模型尚未并入 __init__),与本次改动无关,已手工剔除——本迁移只 add 两列。
+SQLite 经 env.py 的 render_as_batch 自动走 batch_alter_table 重建表。
+
+Revision ID: c2874d2bf705
+Revises: 1699fc2c069f
+Create Date: 2026-06-26 16:48:52.158609
+
+"""
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "c2874d2bf705"
+down_revision: str | Sequence[str] | None = "1699fc2c069f"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    # 信息流发奖记录:新增广告形态 feed/draw(可空,旧数据 NULL=feed)
+    op.add_column(
+        "ad_feed_reward_record",
+        sa.Column("ad_type", sa.String(length=16), nullable=True),
+    )
+    # eCPM 展示上报:新增点位场景 comparison/coupon/welfare(可空,激励视频/旧数据 NULL)
+    op.add_column(
+        "ad_ecpm_record",
+        sa.Column("feed_scene", sa.String(length=16), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    op.drop_column("ad_ecpm_record", "feed_scene")
+    op.drop_column("ad_feed_reward_record", "ad_type")

alembic/versions/comparison_product_names.py

@@ -0,0 +1,70 @@
+"""comparison_record 加 product_names 列(下单商品名派生串)+ 回填历史行
+
+admin 比价记录页要把原「店/商品」一列拆成「店」+「商品」两列、并支持按商品名搜索。
+items 是 JSON(SQLite 下 json.dumps ensure_ascii 会把中文转义存成 \\uXXXX,无法直接 CAST+LIKE
+命中中文),故把商品名派生成普通文本列,跨库 LIKE 一致、可索引。
+
+写路径(upsert_record / harvest_done)已同步派生;本迁移建列并从已存 items 回填历史行。
+全新环境顺序应用即得空列 + 回填(表本为空,回填 no-op)。
+
+Revision ID: comparison_product_names
+Revises: comparison_record_trace_unique
+Create Date: 2026-07-04 00:00:00.000000
+"""
+
+import json
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+revision: str = "comparison_product_names"
+down_revision: str | Sequence[str] | None = "comparison_record_trace_unique"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def _product_names(items) -> str | None:
+    """items([{name,...}]) → 顿号分隔的商品名串(去重保序)。与 repositories.comparison
+    ._product_names_from_items 同口径,迁移内联一份避免耦合 app 代码。"""
+    if isinstance(items, str):  # SQLite: items 以 JSON 文本存,取回是 str
+        try:
+            items = json.loads(items)
+        except (ValueError, TypeError):
+            return None
+    if not isinstance(items, list) or not items:
+        return None
+    names: list[str] = []
+    for it in items:
+        name = it.get("name") if isinstance(it, dict) else None
+        if not name:
+            continue
+        s = str(name).strip()
+        if s and s not in names:
+            names.append(s)
+    joined = "、".join(names)
+    return joined[:500] or None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("comparison_record", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("product_names", sa.String(length=512), nullable=True))
+
+    # 回填:从已存 items 派生商品名。items 小(菜名列表),一次取回即可;只更新有商品名的行。
+    bind = op.get_bind()
+    rows = bind.execute(
+        sa.text("SELECT id, items FROM comparison_record")
+    ).fetchall()
+    for rid, items in rows:
+        pn = _product_names(items)
+        if pn:
+            bind.execute(
+                sa.text("UPDATE comparison_record SET product_names = :pn WHERE id = :id"),
+                {"pn": pn, "id": rid},
+            )
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("comparison_record", schema=None) as batch_op:
+        batch_op.drop_column("product_names")

alembic/versions/comparison_record_trace_unique.py

@@ -0,0 +1,49 @@
+"""comparison_record: 唯一键 (user_id,trace_id) → trace_id 单列 + user_id 可空
+
+Revision ID: comparison_record_trace_unique
+Revises: feedback_type_reply
+Create Date: 2026-07-03 12:00:00.000000
+
+比价记录改「后端 harvest」：app-server 在帧0(pricebot 出 trace_id)即建行，随 done/finalize
+逐步补全，客户端不再 POST 记录。故本表要两处调整：
+  1. user_id 改**可空**——harvest 建行那刻(软鉴权 / 老客户端匿名)可能还没有 user_id。
+  2. 唯一键 (user_id, trace_id) → **trace_id 单列**——trace_id 由 app-server 签发、全局唯一，
+     一次比价一行；harvest 建行时 user_id 还没有，不能再用复合键去重。
+
+⚠️ 上线前(QA)务必确认 comparison_record 无重复 trace_id：旧复合唯一键**允许**同 trace_id
+   跨不同 user_id（实际上客户端 UUID 从不重复，但约束没拦），若真有重复,建 trace 单列唯一会失败。
+   查:  SELECT trace_id, COUNT(*) c FROM comparison_record GROUP BY trace_id HAVING c > 1;
+
+SQLite 不支持直接 drop/alter 约束/列，用 batch_alter_table（建临时表+拷数据+换名），
+与 coupon_engage_per_package / store_mapping_* 同款；Postgres 直接执行原生 ALTER。
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = 'comparison_record_trace_unique'
+down_revision: Union[str, Sequence[str], None] = 'feedback_type_reply'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table('comparison_record', schema=None) as batch_op:
+        # harvest 帧0 建行时 user_id 可能暂缺 → 放开非空。
+        batch_op.alter_column('user_id', existing_type=sa.Integer(), nullable=True)
+        # 复合唯一 → trace_id 单列唯一。
+        batch_op.drop_constraint('uq_comparison_user_trace', type_='unique')
+        batch_op.create_unique_constraint('uq_comparison_trace', ['trace_id'])
+
+
+def downgrade() -> None:
+    with op.batch_alter_table('comparison_record', schema=None) as batch_op:
+        batch_op.drop_constraint('uq_comparison_trace', type_='unique')
+        batch_op.create_unique_constraint(
+            'uq_comparison_user_trace', ['user_id', 'trace_id']
+        )
+        # 回退非空前提是当时无 null user_id 行(harvest 期可能有孤儿行,回滚需先清理)。
+        batch_op.alter_column('user_id', existing_type=sa.Integer(), nullable=False)

alembic/versions/coupon_session_origin_package.py

@@ -0,0 +1,32 @@
+"""coupon_session 加 origin_package 列(发起来源 App 包名 → admin「发起平台」)
+
+null=App 内(傻瓜比价首页)发起,非空=从美团/淘宝/京东弹券发起。与 trace_url 同理单独成迁移,
+已建表环境靠它补列、全新环境顺序应用,不重复加列。
+
+Revision ID: coupon_session_origin_package
+Revises: coupon_session_trace_url
+Create Date: 2026-06-30 00:00:00.000000
+
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "coupon_session_origin_package"
+down_revision: str | Sequence[str] | None = "coupon_session_trace_url"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("origin_package", sa.String(length=64), nullable=True))
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.drop_column("origin_package")

alembic/versions/coupon_session_table.py

@@ -0,0 +1,81 @@
+"""coupon session table(领券任务全程流水 → admin「领券数据」看板数据源)
+
+一次领券一行(trace_id 唯一):客户端 POST /api/v1/coupon/session 两段上报 —— 发起建行
+(status=started)、收尾(completed/failed/abandoned)按 trace_id 更新同一行。记全程耗时
+elapsed_ms + 各平台耗时 platform_elapsed + 机型/ROM,供 admin 算发起/完成数、耗时分位、机型维度。
+
+Revision ID: coupon_session_table
+Revises: feedback_submit_env
+Create Date: 2026-06-30 00:00:00.000000
+
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from sqlalchemy.dialects import postgresql
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "coupon_session_table"
+down_revision: str | Sequence[str] | None = "feedback_submit_env"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+# PG 用 JSONB,SQLite(本地/测试)退化为通用 JSON(同 model 的 _JSON variant)。
+_JSON = sa.JSON().with_variant(postgresql.JSONB(), "postgresql")
+
+
+def upgrade() -> None:
+    op.create_table(
+        "coupon_session",
+        sa.Column("id", sa.Integer(), autoincrement=True, nullable=False),
+        sa.Column("trace_id", sa.String(length=64), nullable=False),
+        sa.Column("device_id", sa.String(length=64), nullable=False),
+        sa.Column("user_id", sa.Integer(), nullable=True),
+        sa.Column("status", sa.String(length=16), nullable=False),
+        sa.Column("app_env", sa.String(length=16), nullable=True),
+        sa.Column("platforms", _JSON, nullable=True),
+        sa.Column("device_model", sa.String(length=128), nullable=True),
+        sa.Column("rom", sa.String(length=64), nullable=True),
+        sa.Column("started_at", sa.DateTime(timezone=True), nullable=False),
+        sa.Column("started_date", sa.Date(), nullable=False),
+        sa.Column("finished_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("elapsed_ms", sa.Integer(), nullable=True),
+        sa.Column("platform_elapsed", _JSON, nullable=True),
+        sa.Column("claimed_count", sa.Integer(), nullable=True),
+        sa.Column(
+            "created_at",
+            sa.DateTime(timezone=True),
+            server_default=sa.text("(CURRENT_TIMESTAMP)"),
+            nullable=False,
+        ),
+        sa.Column(
+            "updated_at",
+            sa.DateTime(timezone=True),
+            server_default=sa.text("(CURRENT_TIMESTAMP)"),
+            nullable=False,
+        ),
+        sa.PrimaryKeyConstraint("id"),
+        sa.UniqueConstraint("trace_id", name="uq_coupon_session_trace"),
+    )
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.create_index(
+            batch_op.f("ix_coupon_session_user_id"), ["user_id"], unique=False
+        )
+        batch_op.create_index(
+            batch_op.f("ix_coupon_session_app_env"), ["app_env"], unique=False
+        )
+        # admin 主聚合/筛选:按上海自然日 + 环境。
+        batch_op.create_index(
+            "ix_coupon_session_date_env", ["started_date", "app_env"], unique=False
+        )
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.drop_index("ix_coupon_session_date_env")
+        batch_op.drop_index(batch_op.f("ix_coupon_session_app_env"))
+        batch_op.drop_index(batch_op.f("ix_coupon_session_user_id"))
+    op.drop_table("coupon_session")

alembic/versions/coupon_session_trace_url.py

@@ -0,0 +1,32 @@
+"""coupon_session 加 trace_url 列(pricebot done 帧公网调试链接)
+
+建表迁移 coupon_session_table 落地后才追加本列,故单独成一个迁移:已建表的环境(本地/已 upgrade 过)
+靠它补列,全新环境则「建表(无 trace_url)→ 本迁移加列」,两条路一致、不重复加列。
+
+Revision ID: coupon_session_trace_url
+Revises: coupon_session_table
+Create Date: 2026-06-30 00:00:00.000000
+
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "coupon_session_trace_url"
+down_revision: str | Sequence[str] | None = "coupon_session_table"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.add_column(sa.Column("trace_url", sa.String(length=512), nullable=True))
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("coupon_session", schema=None) as batch_op:
+        batch_op.drop_column("trace_url")

alembic/versions/cps_v2_platforms.py

@@ -23,39 +23,64 @@ branch_labels: Union[str, Sequence[str], None] = None
 depends_on: Union[str, Sequence[str], None] = None
 
 
-def upgrade() -> None:
-    # 活动:淘宝淘口令 / 京东链接
-    op.add_column("cps_activity", sa.Column("payload", sa.Text(), nullable=True))
+def _has_column(insp: sa.Inspector, table: str, column: str) -> bool:
+    return any(c["name"] == column for c in insp.get_columns(table))
 
-    # 群:多平台(现有群都是美团,server_default 回填)+ sid 可空
-    op.add_column(
-        "cps_group",
-        sa.Column(
-            "platforms",
-            sa.JSON().with_variant(postgresql.JSONB(), "postgresql"),
-            nullable=False,
-            server_default=sa.text("'[\"meituan\"]'::jsonb"),
-        ),
+
+def upgrade() -> None:
+    bind = op.get_bind()
+    insp = sa.inspect(bind)
+    # SQLite 不认 ::jsonb cast(会报 "unrecognized token: :"),只有 PG 用显式 cast。
+    # 两边列类型都靠上面的 with_variant 区分,这里只管 server_default 文本。
+    platforms_default = (
+        sa.text("'[\"meituan\"]'::jsonb")
+        if bind.dialect.name == "postgresql"
+        else sa.text("'[\"meituan\"]'")
     )
-    op.alter_column("cps_group", "sid", existing_type=sa.String(64), nullable=True)
+
+    # 活动:淘宝淘口令 / 京东链接
+    # 加 add_column 存在性守卫:本迁移历史上在 SQLite 第一次跑时加完 payload 即在
+    # platforms 那步炸掉(::jsonb),DDL 非事务不回滚 → 半应用状态。守卫让其可自愈重跑。
+    if not _has_column(insp, "cps_activity", "payload"):
+        op.add_column("cps_activity", sa.Column("payload", sa.Text(), nullable=True))
+
+    # 群:多平台(现有群都是美团,server_default 回填)+ sid 可空。
+    # SQLite 不支持 ALTER COLUMN,必须用 batch(重建表)才能改 nullable;PG 走原生 ALTER。
+    with op.batch_alter_table("cps_group") as batch_op:
+        if not _has_column(insp, "cps_group", "platforms"):
+            batch_op.add_column(
+                sa.Column(
+                    "platforms",
+                    sa.JSON().with_variant(postgresql.JSONB(), "postgresql"),
+                    nullable=False,
+                    server_default=platforms_default,
+                )
+            )
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=True)
 
     # link:sid 可空 + target 加长
-    op.alter_column("cps_link", "sid", existing_type=sa.String(64), nullable=True)
-    op.alter_column("cps_link", "target_url", existing_type=sa.String(1024), type_=sa.String(2048))
+    with op.batch_alter_table("cps_link") as batch_op:
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=True)
+        batch_op.alter_column("target_url", existing_type=sa.String(1024), type_=sa.String(2048))
 
     # click:sid 可空 + 事件类型
-    op.alter_column("cps_click", "sid", existing_type=sa.String(64), nullable=True)
-    op.add_column(
-        "cps_click",
-        sa.Column("event_type", sa.String(16), nullable=False, server_default="visit"),
-    )
+    with op.batch_alter_table("cps_click") as batch_op:
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=True)
+        if not _has_column(insp, "cps_click", "event_type"):
+            batch_op.add_column(
+                sa.Column("event_type", sa.String(16), nullable=False, server_default="visit")
+            )
 
 
 def downgrade() -> None:
-    op.drop_column("cps_click", "event_type")
-    op.alter_column("cps_click", "sid", existing_type=sa.String(64), nullable=False)
-    op.alter_column("cps_link", "target_url", existing_type=sa.String(2048), type_=sa.String(1024))
-    op.alter_column("cps_link", "sid", existing_type=sa.String(64), nullable=False)
-    op.drop_column("cps_group", "platforms")
-    op.alter_column("cps_group", "sid", existing_type=sa.String(64), nullable=False)
-    op.drop_column("cps_activity", "payload")
+    with op.batch_alter_table("cps_click") as batch_op:
+        batch_op.drop_column("event_type")
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=False)
+    with op.batch_alter_table("cps_link") as batch_op:
+        batch_op.alter_column("target_url", existing_type=sa.String(2048), type_=sa.String(1024))
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=False)
+    with op.batch_alter_table("cps_group") as batch_op:
+        batch_op.drop_column("platforms")
+        batch_op.alter_column("sid", existing_type=sa.String(64), nullable=False)
+    with op.batch_alter_table("cps_activity") as batch_op:
+        batch_op.drop_column("payload")

alembic/versions/device_first_protected_at.py

@@ -0,0 +1,31 @@
+"""device_liveness 加 first_protected_at(首次开无障碍时刻)
+
+Revision ID: device_first_protected_at
+Revises: bcfcaf07152b
+Create Date: 2026-06-25 00:00:00.000000
+
+admin 设备存活页要展示「首次无障碍开启时间」。touch_heartbeat 在 ever_protected 首次翻 true
+时记一次(后续心跳不覆盖)。仅新增可空列,SQLite 原生支持 add_column、不用 batch;downgrade
+的 drop_column 在 SQLite 走 batch_alter_table 兜底。老设备无此时刻 → 留 NULL(无法准确回填)。
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "device_first_protected_at"
+down_revision: Union[str, Sequence[str], None] = "bcfcaf07152b"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "device_liveness",
+        sa.Column("first_protected_at", sa.DateTime(timezone=True), nullable=True),
+    )
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("device_liveness") as batch_op:
+        batch_op.drop_column("first_protected_at")

alembic/versions/feedback_submit_env.py

@@ -0,0 +1,34 @@
+"""feedback 加提交端环境快照(app_version / device_model / rom_name / android_version)
+
+Revision ID: feedback_submit_env
+Revises: ad_pangle_daily_revenue
+Create Date: 2026-06-29 00:00:00.000000
+
+admin 用户反馈页要展示「提交版本号」「机型OS版本」,需在提交时落库端环境。仅新增可空列,
+SQLite 原生支持 add_column、不用 batch;downgrade 的 drop_column 在 SQLite 走 batch 兜底。
+历史反馈无此快照 → 留 NULL(无法回填);客户端改版带上后的新反馈才有值。
+"""
+from typing import Sequence, Union
+
+import sqlalchemy as sa
+from alembic import op
+
+revision: str = "feedback_submit_env"
+down_revision: Union[str, Sequence[str], None] = "ad_pangle_daily_revenue"
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    op.add_column("feedback", sa.Column("app_version", sa.String(length=32), nullable=True))
+    op.add_column("feedback", sa.Column("device_model", sa.String(length=64), nullable=True))
+    op.add_column("feedback", sa.Column("rom_name", sa.String(length=32), nullable=True))
+    op.add_column("feedback", sa.Column("android_version", sa.String(length=16), nullable=True))
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("feedback") as batch_op:
+        batch_op.drop_column("android_version")
+        batch_op.drop_column("rom_name")
+        batch_op.drop_column("device_model")
+        batch_op.drop_column("app_version")

alembic/versions/feedback_type_reply.py

@@ -0,0 +1,48 @@
+"""feedback 加反馈来源/场景/运营回复(source / scene / admin_reply)
+
+admin「用户反馈」页要展示并筛选「反馈类型」(比价反馈 / 普通反馈),并支持审核时给用户留言:
+  - source: 反馈来源入口(profile=「我的」页 / comparison=比价结果页)。NOT NULL,
+    旧数据 + 普通反馈默认 profile(server_default)。加索引供 admin 按类型筛选。
+  - scene: 比价反馈的问题场景(找错商品/优惠不对…),普通反馈为 NULL。
+  - admin_reply: 运营给用户的回复留言(用户端可见),随「我的反馈」历史下发。
+
+均为新增列(SQLite 原生支持 add_column);downgrade 的 drop_column 在 SQLite 走 batch 兜底。
+
+Revision ID: feedback_type_reply
+Revises: 761ef181ce7c
+Create Date: 2026-07-02 00:00:00.000000
+"""
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "feedback_type_reply"
+down_revision: str | Sequence[str] | None = "761ef181ce7c"
+branch_labels: str | Sequence[str] | None = None
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    op.add_column(
+        "feedback",
+        sa.Column(
+            "source",
+            sa.String(length=16),
+            nullable=False,
+            server_default="profile",
+        ),
+    )
+    op.add_column("feedback", sa.Column("scene", sa.String(length=32), nullable=True))
+    op.add_column("feedback", sa.Column("admin_reply", sa.String(length=256), nullable=True))
+    op.create_index("ix_feedback_source", "feedback", ["source"])
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("feedback") as batch_op:
+        batch_op.drop_index("ix_feedback_source")
+        batch_op.drop_column("admin_reply")
+        batch_op.drop_column("scene")
+        batch_op.drop_column("source")

alembic/versions/invite_cash_compare_reward.py

@@ -0,0 +1,88 @@
+"""invite cash account isolation + compare reward tracking (🅱-1)
+
+Revision ID: invite_cash_compare_reward
+Revises: feedback_review_fields
+Create Date: 2026-06-23 00:00:00.000000
+
+邀请功能 v2 账户隔离 + 比价发奖追踪:
+  - coin_account 加 invite_cash_balance_cents(邀请奖励金独立余额,与金币兑换的 cash 物理隔离)
+  - withdraw_order 加 source(标记提现扣哪个账户,退款退回对应账户;旧单默认 coin_cash)
+  - invite_relation 加比价发奖追踪三列(好友比价多次只发一次)
+  - 新增 invite_cash_transaction 表(邀请奖励金独立流水账本)
+
+加列均 NOT NULL + server_default,存量行自动填默认值,安全。
+"""
+from typing import Sequence, Union
+
+from alembic import op
+import sqlalchemy as sa
+
+
+# revision identifiers, used by Alembic.
+revision: str = 'invite_cash_compare_reward'
+down_revision: Union[str, Sequence[str], None] = 'feedback_review_fields'
+branch_labels: Union[str, Sequence[str], None] = None
+depends_on: Union[str, Sequence[str], None] = None
+
+
+def upgrade() -> None:
+    # 1. 邀请奖励金独立余额(与金币兑换的 cash_balance_cents 物理隔离;红线:两本账不可累加)
+    op.add_column(
+        'coin_account',
+        sa.Column('invite_cash_balance_cents', sa.Integer(), nullable=False, server_default='0'),
+    )
+    # 2. 提现单标记账户来源:coin_cash / invite_cash,退款退回对应账户(旧单默认 coin_cash)
+    op.add_column(
+        'withdraw_order',
+        sa.Column('source', sa.String(length=16), nullable=False, server_default='coin_cash'),
+    )
+    # 3. 邀请关系加比价发奖追踪(好友"下载+登录+比价一次"→ 给邀请人发奖,只发一次)
+    op.add_column(
+        'invite_relation',
+        sa.Column('compare_reward_granted', sa.Boolean(), nullable=False, server_default=sa.false()),
+    )
+    op.add_column(
+        'invite_relation',
+        sa.Column('compare_reward_cents', sa.Integer(), nullable=False, server_default='0'),
+    )
+    op.add_column(
+        'invite_relation',
+        sa.Column('compare_rewarded_at', sa.DateTime(timezone=True), nullable=True),
+    )
+    # 4. 邀请奖励金独立流水表(结构同 cash_transaction;balance_after 记 invite_cash_balance_cents)
+    op.create_table(
+        'invite_cash_transaction',
+        sa.Column('id', sa.Integer(), autoincrement=True, nullable=False),
+        sa.Column('user_id', sa.Integer(), nullable=False),
+        sa.Column('amount_cents', sa.Integer(), nullable=False),
+        sa.Column('balance_after_cents', sa.Integer(), nullable=False),
+        sa.Column('biz_type', sa.String(length=32), nullable=False),
+        sa.Column('ref_id', sa.String(length=64), nullable=True),
+        sa.Column('remark', sa.String(length=128), nullable=True),
+        sa.Column('created_at', sa.DateTime(timezone=True), server_default=sa.func.now(), nullable=False),
+        sa.ForeignKeyConstraint(['user_id'], ['user.id']),
+        sa.PrimaryKeyConstraint('id'),
+    )
+    op.create_index('ix_invite_cash_transaction_user_id', 'invite_cash_transaction', ['user_id'])
+    op.create_index('ix_invite_cash_transaction_created_at', 'invite_cash_transaction', ['created_at'])
+    # 提现退款幂等:一个提现单只退一次(partial unique，对齐 cash_transaction 的 withdraw_refund 去重)
+    op.create_index(
+        'ux_invite_cash_txn_refund_ref',
+        'invite_cash_transaction',
+        ['ref_id'],
+        unique=True,
+        sqlite_where=sa.text("biz_type = 'invite_withdraw_refund' AND ref_id IS NOT NULL"),
+        postgresql_where=sa.text("biz_type = 'invite_withdraw_refund' AND ref_id IS NOT NULL"),
+    )
+
+
+def downgrade() -> None:
+    op.drop_index('ux_invite_cash_txn_refund_ref', table_name='invite_cash_transaction')
+    op.drop_index('ix_invite_cash_transaction_created_at', table_name='invite_cash_transaction')
+    op.drop_index('ix_invite_cash_transaction_user_id', table_name='invite_cash_transaction')
+    op.drop_table('invite_cash_transaction')
+    op.drop_column('invite_relation', 'compare_rewarded_at')
+    op.drop_column('invite_relation', 'compare_reward_cents')
+    op.drop_column('invite_relation', 'compare_reward_granted')
+    op.drop_column('withdraw_order', 'source')
+    op.drop_column('coin_account', 'invite_cash_balance_cents')

alembic/versions/jd_cps_order_fields.py

@@ -0,0 +1,56 @@
+"""add jd cps order fields
+
+Revision ID: jd_cps_order_fields
+Revises: 7db22acee504
+Create Date: 2026-06-28 20:30:00.000000
+"""
+from __future__ import annotations
+
+from alembic import op
+import sqlalchemy as sa
+
+
+revision = "jd_cps_order_fields"
+down_revision = "7db22acee504"
+branch_labels = None
+depends_on = None
+
+
+def upgrade() -> None:
+    with op.batch_alter_table("cps_order") as batch_op:
+        batch_op.add_column(
+            sa.Column("platform", sa.String(length=20), nullable=False, server_default="meituan")
+        )
+        batch_op.add_column(sa.Column("external_order_id", sa.String(length=128), nullable=True))
+        batch_op.add_column(sa.Column("external_row_id", sa.String(length=128), nullable=True))
+        batch_op.add_column(sa.Column("estimated_commission_cents", sa.Integer(), nullable=True))
+        batch_op.add_column(sa.Column("actual_commission_cents", sa.Integer(), nullable=True))
+        batch_op.add_column(sa.Column("jd_valid_code", sa.String(length=16), nullable=True))
+        batch_op.add_column(sa.Column("settle_month", sa.String(length=16), nullable=True))
+        batch_op.add_column(sa.Column("site_id", sa.String(length=128), nullable=True))
+        batch_op.add_column(sa.Column("position_id", sa.String(length=128), nullable=True))
+        batch_op.add_column(sa.Column("pid", sa.String(length=128), nullable=True))
+        batch_op.add_column(sa.Column("sub_union_id", sa.String(length=128), nullable=True))
+        batch_op.create_index("ix_cps_order_platform", ["platform"])
+        batch_op.create_index("ix_cps_order_external_order_id", ["external_order_id"])
+        batch_op.create_index("ix_cps_order_external_row_id", ["external_row_id"])
+        batch_op.create_index("ix_cps_order_jd_valid_code", ["jd_valid_code"])
+
+
+def downgrade() -> None:
+    with op.batch_alter_table("cps_order") as batch_op:
+        batch_op.drop_index("ix_cps_order_jd_valid_code")
+        batch_op.drop_index("ix_cps_order_external_row_id")
+        batch_op.drop_index("ix_cps_order_external_order_id")
+        batch_op.drop_index("ix_cps_order_platform")
+        batch_op.drop_column("sub_union_id")
+        batch_op.drop_column("pid")
+        batch_op.drop_column("position_id")
+        batch_op.drop_column("site_id")
+        batch_op.drop_column("settle_month")
+        batch_op.drop_column("jd_valid_code")
+        batch_op.drop_column("actual_commission_cents")
+        batch_op.drop_column("estimated_commission_cents")
+        batch_op.drop_column("external_row_id")
+        batch_op.drop_column("external_order_id")
+        batch_op.drop_column("platform")

app/admin/main.py

@@ -21,14 +21,18 @@ from app.admin.routers.audit import router as audit_router
 from app.admin.routers.auth import router as auth_router
 from app.admin.routers.comparison import router as comparison_router
 from app.admin.routers.config import router as config_router
+from app.admin.routers.coupon_data import router as coupon_data_router
 from app.admin.routers.cps import router as cps_router
 from app.admin.routers.dashboard import router as dashboard_router
+from app.admin.routers.device_liveness import router as device_liveness_router
 from app.admin.routers.ops_stat_config import router as ops_stat_config_router
+from app.admin.routers.event_logs import router as event_logs_router
 from app.admin.routers.feedback import router as feedback_router
 from app.admin.routers.feedback_qr import router as feedback_qr_router
 from app.admin.routers.onboarding import router as onboarding_router
 from app.admin.routers.ops_marquee_seed import router as ops_marquee_seed_router
 from app.admin.routers.price_report import router as price_report_router
+from app.admin.routers.roles import router as roles_router
 from app.admin.routers.users import router as users_router
 from app.admin.routers.wallet import router as wallet_router
 from app.admin.routers.withdraw import router as withdraw_router
@@ -83,6 +87,7 @@ def health() -> dict[str, str]:
 
 admin_app.include_router(auth_router)
 admin_app.include_router(dashboard_router)
+admin_app.include_router(device_liveness_router)
 admin_app.include_router(ops_stat_config_router)
 admin_app.include_router(ops_marquee_seed_router)
 admin_app.include_router(users_router)
@@ -91,12 +96,15 @@ admin_app.include_router(wallet_router)
 admin_app.include_router(withdraw_router)
 admin_app.include_router(price_report_router)
 admin_app.include_router(feedback_router)
+admin_app.include_router(event_logs_router)
 admin_app.include_router(feedback_qr_router)
 admin_app.include_router(admins_router)
+admin_app.include_router(roles_router)
 admin_app.include_router(audit_router)
 admin_app.include_router(config_router)
 admin_app.include_router(comparison_router)
 admin_app.include_router(cps_router)
+admin_app.include_router(coupon_data_router)
 admin_app.include_router(ad_audit_router)
 admin_app.include_router(ad_config_router)
 admin_app.include_router(ad_revenue_router)

app/admin/permissions.py

@@ -0,0 +1,83 @@
+"""admin 后台「页面权限目录」—— RBAC 的权限侧单一真源。
+
+一个权限 = 一个页面(= 左侧导航项)。角色持有一组 page key,登录后左侧只展示这些页(见前端
+layout.tsx 按 pages 过滤导航)。key 必须与前端路由一级对齐(/dashboard → "dashboard")。
+
+super_admin 为内建全权角色,恒可见全部页(effective_pages 特判)。新增页面 = 这里加一条 +
+前端导航加对应项(key 一致即可被各角色勾选可见)。
+"""
+from __future__ import annotations
+
+SUPER_ADMIN_ROLE = "super_admin"
+# 「自定义」哨兵角色:不是 admin_role 表里的行,而是标记「这个人的可见页由 pages_override 决定」。
+# admin_user.role == CUSTOM_ROLE 时,有效可见页取 admin_user.pages_override(见 auth._admin_out_with_pages)。
+CUSTOM_ROLE = "custom"
+
+# 分组镜像前端导航(app/(main)/layout.tsx 的 NAV_GROUPS);key = 路由一级
+PERMISSION_CATALOG: list[dict] = [
+    {"group": "看板", "pages": [
+        {"key": "dashboard", "label": "数据大盘"},
+        {"key": "coupon-data", "label": "领券数据"},
+        {"key": "ad-revenue-report", "label": "广告收益"},
+        {"key": "comparison-records", "label": "比价记录"},
+        {"key": "cps", "label": "CPS收益"},
+        {"key": "device-liveness", "label": "设备存活"},
+    ]},
+    {"group": "奖励审核", "pages": [
+        {"key": "withdraws", "label": "提现审核"},
+        {"key": "price-reports", "label": "低价审核"},
+        {"key": "feedbacks", "label": "用户反馈"},
+    ]},
+    {"group": "数据配置", "pages": [
+        {"key": "config", "label": "系统配置"},
+        {"key": "ad-revenue", "label": "广告配置"},
+        {"key": "users", "label": "用户管理"},
+    ]},
+    {"group": "其他", "pages": [
+        {"key": "admins", "label": "权限管理"},
+        {"key": "event-logs", "label": "埋点日志"},
+        {"key": "audit-logs", "label": "审计日志"},
+    ]},
+]
+
+# 全部页面 key(super_admin 有效可见 = 此全集;也用于校验角色 pages 合法性)
+ALL_PAGE_KEYS: list[str] = [p["key"] for g in PERMISSION_CATALOG for p in g["pages"]]
+_ALL_SET = set(ALL_PAGE_KEYS)
+
+# 内建角色定义(种子 / ensure 兜底的单一真源):(key, 中文展示名, 默认可见页)。
+# 页集对齐 Prototypes/dashboard/permissions.md 的 ROLES;super_admin 恒全权(pages 空、bypass)。
+# key 承重(require_role / admin_user.role),勿改;label 为 UI 展示名。
+BUILTIN_ROLES: list[dict] = [
+    {"name": SUPER_ADMIN_ROLE, "label": "管理员", "pages": []},
+    {"name": "operator", "label": "运营", "pages": [
+        "dashboard", "coupon-data", "ad-revenue-report", "comparison-records",
+        "cps", "device-liveness", "price-reports", "feedbacks",
+    ]},
+    {"name": "finance", "label": "财务", "pages": [
+        "dashboard", "ad-revenue-report", "cps", "withdraws",
+    ]},
+    {"name": "tech", "label": "技术", "pages": [
+        "dashboard", "device-liveness", "config", "ad-revenue", "event-logs", "audit-logs",
+    ]},
+]
+
+# 内建 key → 中文展示名(展示 / 自愈用)
+BUILTIN_LABELS: dict[str, str] = {r["name"]: r["label"] for r in BUILTIN_ROLES}
+
+
+def sanitize_pages(pages: list[str] | None) -> list[str]:
+    """过滤掉不在目录里的 key(去重、保序),防脏数据/目录收缩后的悬空 key。"""
+    seen: set[str] = set()
+    out: list[str] = []
+    for k in pages or []:
+        if k in _ALL_SET and k not in seen:
+            seen.add(k)
+            out.append(k)
+    return out
+
+
+def effective_pages(role_name: str, role_pages: list[str] | None) -> list[str]:
+    """某角色的有效可见页:super_admin → 全部;其余 → 其 pages 与目录取交(自愈悬空 key)。"""
+    if role_name == SUPER_ADMIN_ROLE:
+        return list(ALL_PAGE_KEYS)
+    return sanitize_pages(role_pages)

app/admin/repositories/ad_audit.py

@@ -4,9 +4,9 @@
 - 看视频:每条 granted = 1 份,第 N 份 = 该用户 granted 的 reward_video **账号累计**顺序号
   (与 ad_reward.grant_ad_reward 里 `_granted_cumulative + 1` 一致;LT 因子不按天重置,
   故复算时要把当日序号叠加上该用户在本日**之前**的累计已发份数)。
-- 信息流:每条按 unit_count 份逐份累加,LT 序号 = 该用户 granted 份数**账号累计**
-  (与 ad_feed_reward._unit_reward_total 的 existing_units 一致;同样不按天重置,
-  复算需叠加本日之前的累计份数)。
+- 信息流:**每条 granted = 1 份**(与 ad_feed_reward.grant_feed_reward 同口径:看满一份即发该条
+  满额,**不按 unit_count 逐份累加**),LT 序号 = 该用户 granted **条数**账号累计
+  (与 ad_feed_reward.granted_unit_total 的 COUNT 一致;不按天重置,复算需叠加本日之前的累计条数)。
 
 非 granted(capped/ecpm_missing)不占用份序号、应发恒 0,据此校验闸口是否确实没发。
 """
@@ -108,14 +108,18 @@ def _reward_video_rows(
     return rows
 
 
-def _feed_prior_granted_units(
+def _feed_prior_granted_count(
     db: Session, *, date: str, user_id: int | None
 ) -> dict[int, int]:
-    """各用户在 date **之前** granted 的信息流份数累计,作为当日复算的 LT 序号起点。"""
+    """各用户在 date **之前** granted 的信息流**条数**累计,作为当日复算的 LT 序号起点。
+
+    与发奖侧 ad_feed_reward.granted_unit_total(COUNT status=granted)对齐:一条广告 = 1 份,
+    LT 按账号累计**条数**递进。**不再用 SUM(unit_count)**——那是「一条按时长折多份」的过时口径,
+    与现行发奖(每条 1 份)漂移,会让 unit_count>1 的记录复算虚高、对账恒「不符」。"""
     stmt = (
         select(
             AdFeedRewardRecord.user_id,
-            func.coalesce(func.sum(AdFeedRewardRecord.unit_count), 0),
+            func.count(),
         )
         .where(
             AdFeedRewardRecord.reward_date < date,
@@ -128,8 +132,33 @@ def _feed_prior_granted_units(
     return {uid: int(n) for uid, n in db.execute(stmt).all()}
 
 
-def _feed_rows(db: Session, *, date: str, user_id: int | None) -> list[dict]:
-    """信息流记录复算。granted 记录逐份累加,LT 序号沿用账号累计份数(含本日之前)。"""
+def _feed_scene_matches(rec: AdFeedRewardRecord, scene: str | None) -> bool:
+    """该信息流记录是否落入请求的展示筛选 scene。
+    - scene=="feed":ad_type in ("feed", NULL)(旧数据 NULL 视为 feed,向后兼容)
+    - scene=="draw":ad_type=="draw"
+    - scene=="feed_all":所有信息流(feed/draw/NULL 都要)——业务已全切 Draw 信息流,收益报表把「Draw 信息流」
+      当作整个信息流口径(含历史误标 feed/NULL),用它避免筛选漏历史。
+    - scene 为 None:不筛(两类都要)。
+    """
+    if scene == "feed":
+        return rec.ad_type in (None, "feed")
+    if scene == "draw":
+        return rec.ad_type == "draw"
+    if scene == "feed_all":
+        return True
+    return True
+
+
+def _feed_rows(
+    db: Session, *, date: str, user_id: int | None, scene: str | None = None
+) -> list[dict]:
+    """信息流记录复算。**每条 granted = 1 份**(与发奖同口径,不按 unit_count 累加),
+    LT 序号沿用账号累计**条数**(含本日之前)。
+
+    **关键:LT 因子账号累计按全表 granted 条数累计(feed+draw 共享同一发奖池/上限),不按 ad_type 拆分**——
+    故无论 scene 怎么筛展示,这里都遍历当日**全部**信息流记录维持 granted_count 累加;scene 只决定
+    哪些行被**留下展示**(由 _feed_scene_matches 判断),不影响累计基线,保证复算序号与正式发奖一致。
+    """
     stmt = (
         select(AdFeedRewardRecord)
         .where(AdFeedRewardRecord.reward_date == date)
@@ -138,46 +167,54 @@ def _feed_rows(db: Session, *, date: str, user_id: int | None) -> list[dict]:
     if user_id is not None:
         stmt = stmt.where(AdFeedRewardRecord.user_id == user_id)
 
-    # 本日之前的累计份数做起点,与 _unit_reward_total 的 existing_units(累计)对齐
-    granted_units: dict[int, int] = _feed_prior_granted_units(db, date=date, user_id=user_id)
+    # 本日之前的累计**条数**做起点,与发奖侧 granted_unit_total(COUNT granted)对齐
+    granted_count: dict[int, int] = _feed_prior_granted_count(db, date=date, user_id=user_id)
     rows: list[dict] = []
     for rec in db.execute(stmt).scalars():
+        keep = _feed_scene_matches(rec, scene)  # 累计照常推进,这里只决定是否展示本行
         if rec.status == "granted":
-            existing = granted_units.get(rec.user_id, 0)
-            units = rec.unit_count
-            expected = sum(
-                rewards.calculate_ad_reward_coin(rec.ecpm_raw, existing + offset)
-                for offset in range(1, units + 1)
-            )
-            granted_units[rec.user_id] = existing + units
-            start = existing + 1 if units > 0 else None
-            end = existing + units if units > 0 else None
+            # 一条广告 = 1 份(与 grant_feed_reward 同口径:看满一份即发该条满额,不按 unit_count 累加)。
+            # nth = 账号累计第几**条**(含本日之前),与发奖侧 granted_unit_total+1 对齐;累计照常推进
+            # (即便 scene 不匹配不展示也要 +1,保证序号与正式发奖一致)。
+            nth = granted_count.get(rec.user_id, 0) + 1
+            granted_count[rec.user_id] = nth
+            if not keep:
+                continue
+            expected = rewards.calculate_ad_reward_coin(rec.ecpm_raw, nth)
             rows.append({
                 "scene": "feed",
+                "ad_type": rec.ad_type or "feed",
+                "feed_scene": rec.feed_scene,
                 "record_id": rec.id,
                 "user_id": rec.user_id,
                 "ad_session_id": rec.ad_session_id,
+                "trace_id": rec.trace_id,
                 "app_env": rec.app_env,
                 "our_code_id": rec.our_code_id,
                 "created_at": rec.created_at,
                 "status": rec.status,
                 "ecpm": rec.ecpm_raw,
                 "ecpm_factor": rewards.ad_ecpm_factor(rewards.parse_ecpm_yuan(rec.ecpm_raw)),
-                "units": units,
-                "lt_index_start": start,
-                "lt_index_end": end,
-                "lt_factor_start": rewards.ad_lt_factor(start) if start else None,
-                "lt_factor_end": rewards.ad_lt_factor(end) if end else None,
+                "units": 1,
+                "lt_index_start": nth,
+                "lt_index_end": nth,
+                "lt_factor_start": rewards.ad_lt_factor(nth),
+                "lt_factor_end": rewards.ad_lt_factor(nth),
                 "expected_coin": expected,
                 "actual_coin": rec.coin,
                 "matched": expected == rec.coin,
             })
         else:
+            if not keep:
+                continue
             rows.append({
                 "scene": "feed",
+                "ad_type": rec.ad_type or "feed",
+                "feed_scene": rec.feed_scene,
                 "record_id": rec.id,
                 "user_id": rec.user_id,
                 "ad_session_id": rec.ad_session_id,
+                "trace_id": rec.trace_id,
                 "app_env": rec.app_env,
                 "our_code_id": rec.our_code_id,
                 "created_at": rec.created_at,
@@ -199,16 +236,19 @@ def _feed_rows(db: Session, *, date: str, user_id: int | None) -> list[dict]:
 def audit_rows(
     db: Session, *, date: str, user_id: int | None, scene: str | None = None
 ) -> list[dict]:
-    """当日逐条发奖复算行(未排序)。scene: None=两类 / "reward_video" / "feed"。
+    """当日逐条发奖复算行(未排序)。scene: None=两类 / "reward_video" / "feed" / "draw"。
 
+    "feed" 与 "draw" 都查 ad_feed_reward_record(同一发奖表),按 ad_type 区分:feed 含历史 NULL,
+    draw 仅 ad_type=="draw"。信息流行额外带 `ad_type`/`feed_scene`,供收益报表区分比价/领券 Draw 收益。
     每行含 `app_env`/`our_code_id`/`expected_coin`/`actual_coin` 等,供金币审计逐条对账,
     也供广告收益报表把「应发/实发」按 用户×类型×应用×代码位 聚合(见 ad_revenue,复用同一复算口径)。
+    **LT 因子账号累计仍按全表 unit 累计(feed+draw 共享),scene 只筛展示,不拆累计。**
     """
     rows: list[dict] = []
     if scene in (None, "reward_video"):
         rows.extend(_reward_video_rows(db, date=date, user_id=user_id))
-    if scene in (None, "feed"):
-        rows.extend(_feed_rows(db, date=date, user_id=user_id))
+    if scene in (None, "feed", "draw", "feed_all"):
+        rows.extend(_feed_rows(db, date=date, user_id=user_id, scene=scene))
     return rows
 
 

app/admin/repositories/ad_revenue.py

@@ -3,34 +3,43 @@
 只读。每行 = 一次广告事件(不再按用户聚合):
 - **激励视频**:一次观看 = 1 条展示(ad_ecpm)+ 1 条发奖(ad_reward),按 ad_session_id 合并成一行,
   直接给出 eCPM / 收益 + 状态 / 应发 / 实发 / 一致;点开看该条金币复算因子。
-- **信息流**:轮播每条展示各一行(impressionId 各自独立);整场发奖(ad_feed_reward,client_event_id)
-  与逐条展示无法对应,单独成「纯发奖」行。
-- 兜底:有展示无发奖(中途关 / 未达发奖)、有发奖无展示(未上报 eCPM)都各自成行。
+- **信息流(比价/领券)**:一次比价 / 一次领券 = 一条整场发奖(ad_feed_reward)一行,给出 eCPM /
+  发奖金币 + 应发 / 实发 / 一致;点开看金币复算因子。⚠️ draw 的逐条展示(ad_ecpm,impressionId 各自
+  独立、与整场发奖无公共键、无法归到「哪一次」)**不再单独占行**(2026-07 按「一次比价/领券放一块」调整)——
+  其展示数 / eCPM / 预估收益仍进全量统计(合计 / 趋势 / 分类大盘 / 穿山甲对照),只是主表不逐条铺开。
+- 兜底:激励视频有展示无发奖(中途关 / 未达发奖)、有发奖无展示(未上报 eCPM)仍各自成行。
 
 展示与收益来自 ad_ecpm_record(收益 = eCPM元 ÷ 1000);应发 / 实发金币复用金币审计逐条复算
 (ad_audit.audit_rows,与正式发奖同一公式口径,不另写公式)。合计与对账在全量上统计,
 不受 limit(只截断 items)影响。
 
-⚠️ 局限:① 历史 Draw 发奖混在 ad_feed_reward_record 无类型标记,金币侧统一记 feed。
-② 跨天 S2S 回调:同一次广告的展示与发奖偶尔落相邻日,各自按 report_date / reward_date 归日。
+每行带 ad_type(reward_video/feed/draw)与 feed_scene(comparison/coupon/welfare),供前端区分
+「比价 Draw 收益」与「领券 Draw 收益」(比价/领券共用同一代码位,只能靠 feed_scene 分)。
+
+⚠️ 局限:① 历史信息流/Draw 发奖 ad_type 为 NULL 的旧记录统一视为 feed(向后兼容);Draw 仅
+ad_type=="draw" 的新记录单独成类。② 跨天 S2S 回调:同一次广告的展示与发奖偶尔落相邻日,各自按
+report_date / reward_date 归日。
 """
 from __future__ import annotations
 
-from datetime import date as _date, datetime, timedelta, timezone
+from datetime import UTC, datetime, timedelta
+from datetime import date as _date
 
 from sqlalchemy import select
 from sqlalchemy.orm import Session
 
 from app.admin.repositories import ad_audit
+from app.admin.repositories import stats as admin_stats
 from app.core import rewards
 from app.models.ad_ecpm import AdEcpmRecord
 from app.models.user import User
+from app.repositories import ad_pangle_revenue
 
 
 def _cn_hour(dt: datetime) -> int:
     """created_at(UTC 口径)→ 北京时间小时(0–23)。naive 当 UTC 处理(sqlite),tz-aware 直接换算(pg)。"""
     if dt.tzinfo is None:
-        dt = dt.replace(tzinfo=timezone.utc)
+        dt = dt.replace(tzinfo=UTC)
     return dt.astimezone(rewards.CN_TZ).hour
 
 
@@ -46,8 +55,10 @@ def _date_range(date_from: str, date_to: str) -> list[str]:
     return out
 
 
-# 审计行的 scene 与报表 ad_type 一一对应
-_SCENE_TO_AD_TYPE = {"reward_video": "reward_video", "feed": "feed"}
+# 报表 ad_type 与审计 scene 取值一致(reward_video / feed / draw):feed 与 draw 同查发奖表
+# ad_feed_reward_record,由 audit 内部按 ad_type 区分(feed 含历史 NULL,draw 仅 ad_type=="draw")。
+_AUDIT_SCENES = {"reward_video", "feed", "draw"}
+
 
 # 发奖复算明细字段(展开下钻看「金币怎么算出来的」)——从 audit 行原样取这些 key。
 _REWARD_DETAIL_KEYS = (
@@ -69,14 +80,21 @@ def ad_revenue_report(
     date_to: str,
     user_id: int | None = None,
     ad_type: str | None = None,
+    feed_scene: str | None = None,
+    app_env: str | None = None,
     granularity: str = "day",
     limit: int = 500,
+    offset: int = 0,
+    sort: str = "time",
 ) -> dict:
     """日期区间(北京时间,闭区间)**逐条广告事件**列表 + 发奖对账。单日时 date_from==date_to。
 
     每个 item = 一次广告事件(展示与发奖按 ad_session_id 合并;信息流展示 / 发奖各自成行)。
-    ad_type: None=全部 / reward_video / feed / draw。granularity=hour 时每行带北京小时(由各自时间算)。
-    limit 只截断 items(事件明细),total 与 total_* / daily 在全量上统计,数字始终可信。
+    ad_type: None=全部 / reward_video / feed / draw。feed_scene: None=全部 /
+    comparison / coupon / welfare,作为全局筛选(同时作用于明细、合计与 daily/hourly 趋势)。
+    granularity=hour 时每行带北京小时(由各自时间算),并额外返回全量 hourly 序列。
+    事件按时间倒序(新→旧)排列;limit/offset 对排序后的全量做分页切片(items 为当前页),
+    total 与 total_* / daily / hourly 在全量上统计,不受分页影响。
     """
     by_hour = granularity == "hour"
 
@@ -84,7 +102,14 @@ def ad_revenue_report(
     #    同时保留全量列表,未被展示合并的成「纯发奖」事件。
     reward_by_session: dict[tuple[int, str], list[dict]] = {}
     all_reward_rows: list[dict] = []
-    audit_scene = _SCENE_TO_AD_TYPE.get(ad_type) if ad_type is not None else None
+    # 报表 ad_type → audit scene:reward_video/feed 直传;**draw(前端「Draw 信息流」)映射成 feed_all**
+    # ——业务已全切 Draw,把「Draw 信息流」当作整个信息流口径(含历史误标 feed/NULL),否则筛选会漏历史。
+    if ad_type == "draw":
+        audit_scene = "feed_all"
+    elif ad_type in _AUDIT_SCENES:
+        audit_scene = ad_type
+    else:
+        audit_scene = None
     if ad_type is None or audit_scene is not None:
         for d in _date_range(date_from, date_to):
             for row in ad_audit.audit_rows(db, date=d, user_id=user_id, scene=audit_scene):
@@ -114,7 +139,10 @@ def ad_revenue_report(
     )
     if user_id is not None:
         stmt = stmt.where(AdEcpmRecord.user_id == user_id)
-    if ad_type is not None:
+    if ad_type == "draw":
+        # draw = 所有信息流展示(业务已全 Draw,含历史误标 feed);展示行只进统计,不占主表行
+        stmt = stmt.where(AdEcpmRecord.ad_type.in_(["draw", "feed"]))
+    elif ad_type is not None:
         stmt = stmt.where(AdEcpmRecord.ad_type == ad_type)
     for rec in db.execute(stmt).scalars():
         rwd = _pop_reward(rec.user_id, rec.ad_session_id)
@@ -123,6 +151,7 @@ def ad_revenue_report(
             "report_date": rec.report_date,
             "user_id": rec.user_id,
             "ad_type": rec.ad_type,
+            "feed_scene": rec.feed_scene,
             "app_env": rec.app_env,
             "our_code_id": rec.our_code_id,
             "created_at": rec.created_at,
@@ -130,10 +159,16 @@ def ad_revenue_report(
             "has_impression": True,
             "impressions": 1,
             "ecpm": rec.ecpm_raw,
-            # 单次展示收益(元)= eCPM元 ÷ 1000(每千次→单次);与发奖同源解析,口径一致。
-            "revenue_yuan": round(rewards.parse_ecpm_yuan(rec.ecpm_raw) / 1000.0, 6),
+            # 单次展示收益(元)= eCPM元 ÷ 1000(每千次→单次)。eCPM 先钳到 AD_ECPM_MAX_FEN(¥500 CPM)
+            # 再折收益,与发奖口径 [rewards.calculate_ad_reward_coin] 一致(2026-06-29 修:原裸 parse_ecpm_yuan
+            # 不钳,伪造/异常天价 eCPM 会把报表预估收益冲到任意大;金币侧已钳、收益侧漏钳)。
+            "revenue_yuan": round(
+                min(rewards.parse_ecpm_yuan(rec.ecpm_raw), rewards.AD_ECPM_MAX_FEN / 100.0) / 1000.0, 6,
+            ),
             "adn": rec.adn,
             "slot_id": rec.slot_id,
+            "sub_rewards": [],
+            "sub_count": 1,
         }
         if rwd is not None:
             ev.update({
@@ -153,35 +188,109 @@ def ad_revenue_report(
             })
         events.append(ev)
 
-    # 3) 未被展示合并的发奖行 → 「纯发奖」事件(信息流整场发奖 / 有发奖无展示)。
-    #    收益恒 0(收益只算展示侧,避免与展示行重复计)。
+    # 3) 未被展示合并的发奖行 → 事件:
+    #    - 激励视频(reward_video):逐条成「纯发奖」事件(每次一个 ad_session_id;有发奖无展示等)。
+    #    - 信息流(feed/draw):同一次比价/领券的多条广告共享**整场 ad_session_id**(客户端整场复用),
+    #      按 (user_id, ad_session_id) 聚成**一次比价 / 一次领券**父事件;sub_rewards 为组内逐条明细,
+    #      应发/实发取组内合计;业务已全 Draw → 类型统一 "draw"。session 缺失(极少旧数据)各自单独成组。
+    feed_groups: dict[tuple[int, str], list[dict]] = {}
     for row in all_reward_rows:
         if row["record_id"] in used_reward_ids:
             continue
+        if row["scene"] == "reward_video":
+            events.append({
+                "event_key": f"rwd-{row['record_id']}",
+                "report_date": row["_report_date"],
+                "user_id": row["user_id"],
+                "ad_type": "reward_video",
+                "feed_scene": row.get("feed_scene"),
+                "app_env": row.get("app_env"),
+                "our_code_id": row.get("our_code_id"),
+                "created_at": row["created_at"],
+                "hour": _cn_hour(row["created_at"]) if by_hour else None,
+                "has_impression": False,
+                "impressions": 0,
+                "ecpm": row["ecpm"],
+                "revenue_yuan": 0.0,
+                "adn": None,
+                "slot_id": None,
+                "has_reward": True,
+                "status": row["status"],
+                "expected_coin": int(row["expected_coin"]),
+                "actual_coin": int(row["actual_coin"]),
+                "matched": bool(row["matched"]),
+                "reward_detail": _reward_detail(row),
+                "sub_rewards": [],
+                "sub_count": 1,
+            })
+        else:
+            # 聚合单位 = 一次完整比价/领券流程:优先用 trace_id(比价带 comparisonTraceId、领券带 sessionTraceId,
+            # 整个流程不变;即使中途点广告致浮层关闭重弹、ad_session_id 变了,trace_id 仍不变 → 全流程聚成一行)。
+            # 无 trace_id(历史领券未上报 / 旧数据)回退整场 ad_session_id;再无则 record_id 各自成组、不误并。
+            grp_key = row.get("trace_id") or row.get("ad_session_id") or f"_rid-{row['record_id']}"
+            feed_groups.setdefault((row["user_id"], grp_key), []).append(row)
+
+    # 信息流分组 → 「一次比价 / 一次领券」父事件(收益恒 0:收益只算展示侧,避免与展示行重复计)。
+    for (uid, grp_key), group in feed_groups.items():
+        group.sort(key=lambda r: (r["created_at"], r["record_id"]))
+        rep = group[-1]  # 代表条(最新一条):时间/场景/应用/代码位取它
+        expected_sum = sum(int(g["expected_coin"]) for g in group)
+        actual_sum = sum(int(g["actual_coin"]) for g in group)
+        # 父行 eCPM:组内各条 eCPM(分)均值(展示用,各条不同);无有效值则取代表条
+        ecpm_fens = [rewards.parse_ecpm_fen(g["ecpm"]) for g in group if g.get("ecpm")]
+        avg_ecpm = str(round(sum(ecpm_fens) / len(ecpm_fens))) if ecpm_fens else rep.get("ecpm")
+        # 主表逐行显示用:这次发奖广告的预估收益之和(发奖侧 eCPM 折算,钳顶同展示侧)。只放进
+        # row_revenue_yuan 给主表逐行展示,不进 revenue_yuan/合计/趋势——避免与展示侧 total 重复计。
+        row_revenue = round(sum(
+            min(rewards.parse_ecpm_yuan(g["ecpm"]), rewards.AD_ECPM_MAX_FEN / 100.0) / 1000.0
+            for g in group if g.get("ecpm")
+        ), 6)
         events.append({
-            "event_key": f"rwd-{row['record_id']}",
-            "report_date": row["_report_date"],
-            "user_id": row["user_id"],
-            "ad_type": _SCENE_TO_AD_TYPE.get(row["scene"], row["scene"]),
-            "app_env": row.get("app_env"),
-            "our_code_id": row.get("our_code_id"),
-            "created_at": row["created_at"],
-            "hour": _cn_hour(row["created_at"]) if by_hour else None,
+            "event_key": f"feedgrp-{uid}-{grp_key}",
+            "report_date": rep["_report_date"],
+            "user_id": uid,
+            "ad_type": "draw",  # 业务已全切 Draw 信息流,聚合行统一 draw
+            "feed_scene": rep.get("feed_scene"),
+            "app_env": rep.get("app_env"),
+            "our_code_id": rep.get("our_code_id"),
+            "created_at": rep["created_at"],
+            "hour": _cn_hour(rep["created_at"]) if by_hour else None,
             "has_impression": False,
             "impressions": 0,
-            "ecpm": row["ecpm"],
+            "ecpm": avg_ecpm,
             "revenue_yuan": 0.0,
+            "row_revenue_yuan": row_revenue,
             "adn": None,
             "slot_id": None,
             "has_reward": True,
-            "status": row["status"],
-            "expected_coin": int(row["expected_coin"]),
-            "actual_coin": int(row["actual_coin"]),
-            "matched": bool(row["matched"]),
-            "reward_detail": _reward_detail(row),
+            "status": rep["status"],  # 代表状态(逐条见展开)
+            "expected_coin": expected_sum,
+            "actual_coin": actual_sum,
+            "matched": all(bool(g["matched"]) for g in group),
+            "reward_detail": None,
+            "sub_rewards": [_reward_detail(g) for g in group],
+            "sub_count": len(group),
         })
 
-    events.sort(key=lambda e: (e["report_date"], e["user_id"], e["created_at"]))
+    # 「场景」作为全局筛选(与 user_id/ad_type 一致):同时作用于明细、合计与 daily/hourly 趋势。
+    # feed_scene 仅信息流 / Draw 有值,激励视频与旧数据为 None;选中后只保留该场景事件。
+    if feed_scene is not None:
+        events = [e for e in events if e.get("feed_scene") == feed_scene]
+
+    # app_env 过滤(2026-06-29 新增能力,修隐患:测试应用上报的假 eCPM 如 ¥678 CPM 会污染正式收益合计/平均):
+    # 显式传 "prod"/"test" 只看该环境;不传=全部(维持现状)。**不擅自把默认改成排除 test**——本地 dev 库多为
+    # test 数据、默认排除会使本地报表空,且「正式报表是否含 test」属产品口径。建议前端报表页加 app_env 筛选器
+    # (默认选 prod),或产品确认后再把默认改成排除 test。注:穿山甲后台收益列(total_pangle_*)暂未联动此过滤
+    # (它是独立对照列,且 pangle 的 test 是真实小额、非客户端那种假值)。
+    if app_env is not None:
+        events = [e for e in events if e.get("app_env") == app_env]
+
+    # 排序:time=按时间倒序(新→旧);ecpm=按 eCPM 数值倒序(eCPM 原值是字符串「分」,转数值排;
+    # 纯发奖行用其发奖采用的 eCPM,缺失/非法计 0 排末尾)。
+    if sort == "ecpm":
+        events.sort(key=lambda e: rewards.parse_ecpm_fen(e["ecpm"]), reverse=True)
+    else:
+        events.sort(key=lambda e: (e["report_date"], e["created_at"]), reverse=True)
 
     # 补手机号(admin 展示用,完整不脱敏,与用户 / 钱包 / 比价记录页一致):批量一次查,避免 N+1。
     uids = {e["user_id"] for e in events}
@@ -219,14 +328,118 @@ def ad_revenue_report(
         for d in sorted(daily_map.values(), key=lambda x: x["date"])
     ]
 
+    # 穿山甲后台收益(GroMore 数据 API,T+1 入库 ad_pangle_daily_revenue):汇总 + 按天趋势级展示,
+    # 与上面客户端自报 eCPM 折算的预估并列对照(看 gap)。穿山甲数据**无用户/场景/类型维度**,故仅在
+    # 「全量视图」(未按 user_id / ad_type / feed_scene 过滤)给值;一旦带这些过滤,穿山甲数无法对应口径
+    # → 置 None,前端显示「-」并提示。逐条事件行不动(仍是客户端预估)。
+    pangle_filterable = user_id is None and ad_type is None and feed_scene is None
+    total_pangle_revenue_yuan: float | None = None
+    total_pangle_api_revenue_yuan: float | None = None
+    if pangle_filterable:
+        pangle_aggs = ad_pangle_revenue.aggregate_by_date(db, date_from=date_from, date_to=date_to)
+        if pangle_aggs:
+            by_date = {a["date"]: a for a in pangle_aggs}
+            for d in daily:
+                pa = by_date.get(d["date"])
+                d["pangle_revenue_yuan"] = pa["revenue_yuan"] if pa else None
+                d["pangle_api_revenue_yuan"] = pa["api_revenue_yuan"] if pa else None
+            total_pangle_revenue_yuan = round(sum(a["revenue_yuan"] for a in pangle_aggs), 6)
+            api_vals = [a["api_revenue_yuan"] for a in pangle_aggs if a["api_revenue_yuan"] is not None]
+            total_pangle_api_revenue_yuan = round(sum(api_vals), 6) if api_vals else None
+
+    # 按小时汇总(全量,不受分页 limit/offset 影响):供前端按小时趋势图(单日 granularity=hour 时用)。
+    # 只在 by_hour 下聚合(此时每个 event 带 hour);否则空。前端按天趋势仍用 daily。
+    hourly: list[dict] = []
+    if by_hour:
+        hour_map: dict[int, dict] = {}
+        for e in events:
+            h = e["hour"]
+            if h is None:
+                continue
+            hd = hour_map.get(h)
+            if hd is None:
+                hd = {"hour": h, "impressions": 0, "revenue_yuan": 0.0,
+                      "expected_coin": 0, "actual_coin": 0}
+                hour_map[h] = hd
+            hd["impressions"] += e["impressions"]
+            hd["revenue_yuan"] += e["revenue_yuan"]
+            hd["expected_coin"] += e["expected_coin"]
+            hd["actual_coin"] += e["actual_coin"]
+        hourly = [
+            {**hd, "revenue_yuan": round(hd["revenue_yuan"], 6)}
+            for hd in sorted(hour_map.values(), key=lambda x: x["hour"])
+        ]
+
+    # 分广告类型小计(按 ad_type:展示条数 + 预估收益;eCPM 由前端用 收益÷展示×1000 算)。
+    # 基于全量(已按 feed_scene 过滤)events;前端只取 draw / reward_video 两类展示。
+    type_map: dict[str, dict] = {}
+    for e in events:
+        t = type_map.get(e["ad_type"])
+        if t is None:
+            t = {"impressions": 0, "revenue_yuan": 0.0}
+            type_map[e["ad_type"]] = t
+        t["impressions"] += e["impressions"]
+        t["revenue_yuan"] += e["revenue_yuan"]
+    type_stats = {
+        k: {"impressions": v["impressions"], "revenue_yuan": round(v["revenue_yuan"], 6)}
+        for k, v in type_map.items()
+    }
+
+    # 分场景小计(按 feed_scene:展示条数 + 预估收益),同 type_stats 基于全量 events——
+    # 供数据大盘「领券广告 / 比价广告」卡用。此前大盘是在分页 items 里按 feed_scene 现算,
+    # 2026-07-02 起信息流逐条展示行(唯一带收益 + 场景的行)不再进主表 items,现算恒为 0;
+    # 改为服务端在全量上聚合下发(也顺带不受 limit 分页截断影响)。feed_scene 为空(激励视频 /
+    # 旧数据)不计入任何场景桶。
+    scene_map: dict[str, dict] = {}
+    for e in events:
+        sc = e.get("feed_scene")
+        if not sc:
+            continue
+        s = scene_map.get(sc)
+        if s is None:
+            s = {"impressions": 0, "revenue_yuan": 0.0}
+            scene_map[sc] = s
+        s["impressions"] += e["impressions"]
+        s["revenue_yuan"] += e["revenue_yuan"]
+    scene_stats = {
+        k: {"impressions": v["impressions"], "revenue_yuan": round(v["revenue_yuan"], 6)}
+        for k, v in scene_map.items()
+    }
+
+    # DAU:复用数据大盘活跃用户口径(登录 + 开始比价 + 开始领券,按用户去重),按所选日期区间
+    # 统计(含今日),历史 / 多天区间同样有值。ARPU = 区间预估收益 ÷ 区间活跃用户。全局口径,
+    # 不随 user / ad_type / feed_scene / app_env 筛选变化(活跃用户口径无这些维度)。
+    dau = admin_stats.period_active_dau(
+        db, _date.fromisoformat(date_from), _date.fromisoformat(date_to)
+    )
+
+    # 主表「逐行」= 单次广告行为(2026-07 按「一次比价/领券放一块」聚合):激励视频 = 一次观看一行(展示+发奖
+    # 按 ad_session_id 合并);一次比价 / 一次领券 = 该次整场多条广告按 ad_session_id 聚成一行(展开看逐条)。
+    # 信息流(draw/feed)的逐条展示(ad_ecpm,impressionId 各自独立、与整场发奖无公共键)不再单独占行
+    # ——其展示数 / eCPM / 预估收益已计入上面的全量统计(total_*、daily / hourly、type_stats、穿山甲对照),
+    # 只是主表不逐条铺开;逐条明细在父行展开里看(sub_rewards)。合计 / 趋势 / 分类大盘均基于全量 events,
+    # 不受此过滤影响;total / 分页只作用于主表行。
+    main_rows = [
+        e for e in events
+        if not (e["ad_type"] in ("draw", "feed") and e["has_impression"] and not e["has_reward"])
+    ]
+
     return {
-        "total": len(events),
-        "truncated": len(events) > limit,
+        "total": len(main_rows),
+        "truncated": len(main_rows) > offset + limit,
         "total_impressions": total_impressions,
         "total_revenue_yuan": total_revenue_yuan,
+        # 穿山甲后台收益合计(元):预估 revenue + 收益Api;非全量视图(带 user/类型/场景过滤)或无数据为 None。
+        "total_pangle_revenue_yuan": total_pangle_revenue_yuan,
+        "total_pangle_api_revenue_yuan": total_pangle_api_revenue_yuan,
+        "pangle_revenue_available": total_pangle_revenue_yuan is not None,
         "total_expected_coin": total_expected_coin,
         "total_actual_coin": total_actual_coin,
         "mismatch_count": mismatch_count,
         "daily": daily,
-        "items": events[:limit],
+        "hourly": hourly,
+        "type_stats": type_stats,
+        "scene_stats": scene_stats,
+        "dau": dau,
+        "items": main_rows[offset:offset + limit],
     }

app/admin/repositories/admin_role.py

@@ -0,0 +1,81 @@
+"""admin_role 表 CRUD + 「角色 → 有效可见页」解析。"""
+from __future__ import annotations
+
+from sqlalchemy import select
+from sqlalchemy.orm import Session
+
+from app.admin.permissions import (
+    ALL_PAGE_KEYS,
+    BUILTIN_ROLES,
+    SUPER_ADMIN_ROLE,
+    sanitize_pages,
+)
+from app.models.admin_role import AdminRole
+
+
+def ensure_builtin_roles(db: Session) -> None:
+    """空表时播种内建角色(管理员/运营/财务/技术)。幂等:表非空即 no-op。
+    给 create_all 环境(测试 / 未跑迁移的全新库)兜底,迁移已播种则不重复。
+    仅 super_admin 为 is_builtin(锁定不可改删);运营/财务/技术可编辑页/删除。"""
+    if db.execute(select(AdminRole.id).limit(1)).first() is not None:
+        return
+    db.add_all([
+        AdminRole(
+            name=r["name"], label=r["label"], pages=list(r["pages"]),
+            is_builtin=(r["name"] == SUPER_ADMIN_ROLE),
+        )
+        for r in BUILTIN_ROLES
+    ])
+    db.commit()
+
+
+def list_roles(db: Session) -> list[AdminRole]:
+    ensure_builtin_roles(db)
+    # super_admin(内建)恒排最前,其余按创建序
+    return list(
+        db.execute(
+            select(AdminRole).order_by(AdminRole.is_builtin.desc(), AdminRole.id)
+        ).scalars().all()
+    )
+
+
+def get_role(db: Session, name: str) -> AdminRole | None:
+    return db.execute(
+        select(AdminRole).where(AdminRole.name == name)
+    ).scalar_one_or_none()
+
+
+def create_role(db: Session, *, name: str, label: str, pages: list[str]) -> AdminRole:
+    # 自定义角色:name(key)= label = 用户输入的名称
+    role = AdminRole(name=name, label=label, pages=sanitize_pages(pages), is_builtin=False)
+    db.add(role)
+    db.commit()
+    db.refresh(role)
+    return role
+
+
+def update_role(
+    db: Session, role: AdminRole, *, label: str | None = None, pages: list[str] | None = None
+) -> AdminRole:
+    # 只改展示名 label + 可见页;name(key)不可变(admin_user.role / require_role 承重)
+    if label is not None:
+        role.label = label
+    if pages is not None:
+        role.pages = sanitize_pages(pages)
+    db.commit()
+    db.refresh(role)
+    return role
+
+
+def delete_role(db: Session, role: AdminRole) -> None:
+    db.delete(role)
+    db.commit()
+
+
+def effective_pages_of(db: Session, role_name: str) -> list[str]:
+    """某角色名的有效可见页:super_admin → 全部;其余 → 查表 pages 与目录取交。"""
+    if role_name == SUPER_ADMIN_ROLE:
+        return list(ALL_PAGE_KEYS)
+    ensure_builtin_roles(db)
+    role = get_role(db, role_name)
+    return sanitize_pages(role.pages if role else None)

app/admin/repositories/admin_user.py

@@ -20,12 +20,23 @@ def get_by_username(db: Session, username: str) -> AdminUser | None:
 
 
 def create_admin(
-    db: Session, *, username: str, password: str, role: str = "operator"
+    db: Session,
+    *,
+    username: str,
+    password: str,
+    role: str = "operator",
+    plain_password: str | None = None,
+    pages_override: list[str] | None = None,
 ) -> AdminUser:
+    """建管理员。plain_password 非空则额外留存明文(后台 UI 建的账号传,供权限管理页复看);
+    脚本/起后台建账号不传(留 None → 前端不显示密码)。
+    pages_override:role=="custom" 时传专属可见页 key 列表;普通角色为 None。"""
     admin = AdminUser(
         username=username,
         password_hash=hash_password(password),
         role=role,
+        plain_password=plain_password,
+        pages_override=pages_override,
     )
     db.add(admin)
     db.commit()

app/admin/repositories/coupon_data.py

@@ -0,0 +1,225 @@
+"""admin「领券数据」看板聚合:发起/完成数、耗时均值与分位、按天/小时趋势、逐条明细。
+
+数据源 coupon_session(一次领券一行,客户端 /api/v1/coupon/session 两段上报)。量级不大,全量拉
+区间数据后 Python 聚合(分位 SQLite 无 percentile,统一 Python 算,PG 上也一致)。
+- 发起数 = 区间内全部 session(含 started/completed/failed/abandoned),= 流失统计的基数。
+- 完成数 / 耗时均值 / 分位 = 仅 status==completed 子集(成功跑完才有可比的"领券耗时")。
+- summary/daily/hourly/total 在全量上算,不受分页;items 为排序后当前页。
+"""
+from __future__ import annotations
+
+from datetime import UTC, date as _date, datetime
+
+from sqlalchemy import func, or_, select
+from sqlalchemy.orm import Session
+
+from app.core import rewards
+from app.models.coupon_state import CouponSession
+from app.models.user import User
+
+
+def _cn_hour(dt: datetime) -> int:
+    """started_at(UTC 口径)→ 北京时间小时(0–23)。naive 当 UTC(sqlite),tz-aware 直接换算(pg)。"""
+    if dt.tzinfo is None:
+        dt = dt.replace(tzinfo=UTC)
+    return dt.astimezone(rewards.CN_TZ).hour
+
+
+def _percentile(sorted_vals: list[int], q: float) -> int | None:
+    """线性插值分位(q=0..100,numpy 默认法)。sorted_vals 须已升序;空返回 None。"""
+    if not sorted_vals:
+        return None
+    if len(sorted_vals) == 1:
+        return sorted_vals[0]
+    idx = (len(sorted_vals) - 1) * q / 100.0
+    lo = int(idx)
+    hi = min(lo + 1, len(sorted_vals) - 1)
+    frac = idx - lo
+    return round(sorted_vals[lo] * (1 - frac) + sorted_vals[hi] * frac)
+
+
+def _avg(vals: list[int]) -> int | None:
+    return round(sum(vals) / len(vals)) if vals else None
+
+
+def _session_to_row(r, phone: str | None = None, nickname: str | None = None) -> dict:
+    """CouponSession ORM → 明细行 dict(主表「领券数据」与「用户全部领券」抽屉共用)。"""
+    return {
+        "id": r.id,
+        "trace_id": r.trace_id,
+        "user_id": r.user_id,
+        "user_phone": phone,
+        "user_nickname": nickname,
+        "status": r.status,
+        "platforms": r.platforms,
+        "origin_package": r.origin_package,
+        "elapsed_ms": r.elapsed_ms,
+        "platform_elapsed": r.platform_elapsed,
+        "device_model": r.device_model,
+        "rom": r.rom,
+        "app_env": r.app_env,
+        "started_at": r.started_at,
+        "claimed_count": r.claimed_count,
+        "trace_url": r.trace_url,
+    }
+
+
+def _empty_result() -> dict:
+    return {
+        "summary": {
+            "started_count": 0, "completed_count": 0, "avg_elapsed_ms": None,
+            "p5_ms": None, "p50_ms": None, "p95_ms": None, "p99_ms": None,
+        },
+        "daily": [],
+        "hourly": [],
+        "total": 0,
+        "items": [],
+    }
+
+
+def coupon_data_report(
+    db: Session,
+    *,
+    date_from: str,
+    date_to: str,
+    user: str | None = None,
+    app_env: str | None = None,
+    granularity: str = "day",
+    limit: int = 500,
+    offset: int = 0,
+    sort: str = "time",
+) -> dict:
+    """日期区间(北京自然日 started_date,闭区间)领券数据:汇总卡 + 趋势 + 逐条明细。
+
+    - user:手机号/昵称模糊搜(匹配不到任何用户 → 空结果)。
+    - app_env:prod/dev 精确;None=全部。
+    - sort:time=发起时刻倒序(默认) / elapsed=全程耗时倒序(None 末尾)。
+    """
+    by_hour = granularity == "hour"
+    d_from = _date.fromisoformat(date_from)
+    d_to = _date.fromisoformat(date_to)
+
+    # user 模糊 → 先定位匹配用户 id;匹配不到直接空结果(不全表扫)。
+    user_ids: set[int] | None = None
+    if user:
+        like = f"%{user}%"
+        user_ids = set(db.execute(
+            select(User.id).where(or_(User.phone.like(like), User.nickname.like(like)))
+        ).scalars().all())
+        if not user_ids:
+            return _empty_result()
+
+    stmt = select(CouponSession).where(
+        CouponSession.started_date >= d_from,
+        CouponSession.started_date <= d_to,
+    )
+    if app_env is not None:
+        stmt = stmt.where(CouponSession.app_env == app_env)
+    if user_ids is not None:
+        stmt = stmt.where(CouponSession.user_id.in_(user_ids))
+    rows = list(db.execute(stmt).scalars())
+
+    # ── 汇总卡 ──
+    completed_elapsed = sorted(
+        r.elapsed_ms for r in rows if r.status == "completed" and r.elapsed_ms is not None
+    )
+    summary = {
+        "started_count": len(rows),
+        "completed_count": sum(1 for r in rows if r.status == "completed"),
+        "avg_elapsed_ms": _avg(completed_elapsed),
+        "p5_ms": _percentile(completed_elapsed, 5),
+        "p50_ms": _percentile(completed_elapsed, 50),
+        "p95_ms": _percentile(completed_elapsed, 95),
+        "p99_ms": _percentile(completed_elapsed, 99),
+    }
+
+    # ── 按天趋势(柱=发起/完成数,线=平均耗时)──
+    daily_map: dict[str, dict] = {}
+    for r in rows:
+        d = r.started_date.isoformat()
+        b = daily_map.get(d)
+        if b is None:
+            b = {"date": d, "started_count": 0, "completed_count": 0, "_elapsed": []}
+            daily_map[d] = b
+        b["started_count"] += 1
+        if r.status == "completed":
+            b["completed_count"] += 1
+            if r.elapsed_ms is not None:
+                b["_elapsed"].append(r.elapsed_ms)
+    daily = [
+        {
+            "date": b["date"],
+            "started_count": b["started_count"],
+            "completed_count": b["completed_count"],
+            "avg_elapsed_ms": _avg(b["_elapsed"]),
+        }
+        for b in sorted(daily_map.values(), key=lambda x: x["date"])
+    ]
+
+    # ── 按小时趋势(单日 hour 粒度)──
+    hourly: list[dict] = []
+    if by_hour:
+        hour_map: dict[int, dict] = {}
+        for r in rows:
+            h = _cn_hour(r.started_at)
+            b = hour_map.get(h)
+            if b is None:
+                b = {"hour": h, "started_count": 0, "completed_count": 0, "_elapsed": []}
+                hour_map[h] = b
+            b["started_count"] += 1
+            if r.status == "completed":
+                b["completed_count"] += 1
+                if r.elapsed_ms is not None:
+                    b["_elapsed"].append(r.elapsed_ms)
+        hourly = [
+            {
+                "hour": b["hour"],
+                "started_count": b["started_count"],
+                "completed_count": b["completed_count"],
+                "avg_elapsed_ms": _avg(b["_elapsed"]),
+            }
+            for b in sorted(hour_map.values(), key=lambda x: x["hour"])
+        ]
+
+    # ── 明细:排序 + 分页 + 补用户手机号/昵称(批量,防 N+1)──
+    if sort == "elapsed":
+        rows.sort(key=lambda r: (r.elapsed_ms is None, -(r.elapsed_ms or 0)))
+    else:  # time:发起时刻倒序
+        rows.sort(key=lambda r: r.started_at, reverse=True)
+    page = rows[offset:offset + limit]
+
+    uids = {r.user_id for r in page if r.user_id is not None}
+    user_map: dict[int, tuple[str | None, str | None]] = {}
+    if uids:
+        user_map = {
+            uid: (phone, nickname)
+            for uid, phone, nickname in db.execute(
+                select(User.id, User.phone, User.nickname).where(User.id.in_(uids))
+            ).all()
+        }
+    items = []
+    for r in page:
+        phone, nickname = user_map.get(r.user_id, (None, None)) if r.user_id is not None else (None, None)
+        items.append(_session_to_row(r, phone, nickname))
+
+    return {
+        "summary": summary,
+        "daily": daily,
+        "hourly": hourly,
+        "total": len(rows),
+        "items": items,
+    }
+
+
+def coupon_user_records(db: Session, *, user_id: int, limit: int = 100) -> dict:
+    """某用户全部领券记录(点手机号抽屉用):按发起时刻倒序、不限日期,total=该用户领券总次数。"""
+    rows = list(db.execute(
+        select(CouponSession)
+        .where(CouponSession.user_id == user_id)
+        .order_by(CouponSession.started_at.desc())
+        .limit(limit)
+    ).scalars())
+    total = db.execute(
+        select(func.count()).select_from(CouponSession).where(CouponSession.user_id == user_id)
+    ).scalar_one()
+    return {"items": [_session_to_row(r) for r in rows], "total": int(total)}

app/admin/repositories/cps.py

@@ -7,23 +7,29 @@ from __future__ import annotations
 
 from datetime import datetime, timedelta, timezone
 from decimal import Decimal, InvalidOperation
+from typing import Any
 from uuid import uuid4
 
 from sqlalchemy import desc, func, select
 from sqlalchemy.orm import Session
 
 from app.admin.repositories.queries import _as_utc, offset_paginate
-from app.integrations import meituan
+from app.integrations import jd_union, meituan
 from app.repositories import cps_link as cps_link_repo
 from app.models.cps_activity import CpsActivity
 from app.models.cps_group import CpsGroup
-from app.models.cps_link import CpsClick
+from app.models.cps_link import CpsClick, CpsLink
 from app.models.cps_order import CpsOrder
 from app.models.cps_wx_user import CpsWxUser
 
 # 美团订单状态:取消(4)/风控(5)不计佣金;结算(6)为佣金真正到账
 _INVALID_STATUS = {"4", "5"}
 _SETTLED_STATUS = "6"
+_JD_INVALID_CODES = {
+    "2", "3", "4", "5", "6", "7", "8", "9", "11", "13", "14", "19", "20", "21",
+    "22", "23", "25", "26", "27", "28", "29", "30", "31", "34", "35", "36",
+}
+_JD_UNPAID_CODES = {"15"}
 
 # CPS 点击时序按北京时区分桶(运营看的是北京时间)
 _BJ_TZ = timezone(timedelta(hours=8))
@@ -45,12 +51,58 @@ def _yuan_to_cents(v: object) -> int | None:
 
 def _ts_to_dt(ts: object) -> datetime | None:
     """秒级时间戳 → tz-aware UTC datetime(绝对时刻,前端按北京展示)。"""
-    if not ts:
+    if ts is None:
+        return None
+    if isinstance(ts, datetime):
+        return ts if ts.tzinfo else ts.replace(tzinfo=_BJ_TZ).astimezone(timezone.utc)
+    s = str(ts).strip()
+    if not s or s.lower() == "null":
         return None
     try:
-        return datetime.fromtimestamp(int(ts), tz=timezone.utc)
-    except (ValueError, OSError, TypeError):
+        seconds = float(Decimal(s))
+    except (InvalidOperation, ValueError):
         return None
+    if seconds == 0:
+        return None
+    # 美团文档是秒级时间戳，这里顺手兼容毫秒/微秒，避免上游格式变化导致时间再次落空。
+    if abs(seconds) > 10_000_000_000_000:
+        seconds /= 1_000_000
+    elif abs(seconds) > 10_000_000_000:
+        seconds /= 1_000
+    try:
+        return datetime.fromtimestamp(seconds, tz=timezone.utc)
+    except (OverflowError, OSError, ValueError):
+        return None
+
+
+def _jd_dt_to_utc(value: object) -> datetime | None:
+    """京东时间字符串(北京时间) → UTC aware datetime。"""
+    if value is None:
+        return None
+    s = str(value).strip()
+    if not s:
+        return None
+    for fmt in ("%Y-%m-%d %H:%M:%S", "%Y-%m-%d"):
+        try:
+            dt = datetime.strptime(s, fmt)
+            return dt.replace(tzinfo=_BJ_TZ).astimezone(timezone.utc)
+        except ValueError:
+            continue
+    return None
+
+
+def _text(value: object) -> str | None:
+    if value is None:
+        return None
+    s = str(value).strip()
+    return s or None
+
+
+def _pick(row: dict[str, Any], *keys: str) -> Any:
+    for key in keys:
+        if key in row and row[key] is not None:
+            return row[key]
+    return None
 
 
 # ───────────── 群 ─────────────
@@ -249,6 +301,80 @@ def _map_order_fields(r: dict) -> dict:
     }
 
 
+def _jd_order_key(r: dict[str, Any]) -> str | None:
+    row_id = _text(_pick(r, "id", "rowId", "orderRowId"))
+    if row_id:
+        return f"jd:{row_id}"
+    order_id = _text(_pick(r, "orderId", "parentOrderId"))
+    sku_id = _text(_pick(r, "skuId"))
+    if order_id and sku_id:
+        return f"jd:{order_id}:{sku_id}"
+    if order_id:
+        return f"jd:{order_id}"
+    return None
+
+
+def _map_jd_order_fields(r: dict[str, Any]) -> dict:
+    """京东 order.row.query 单条订单行 → CpsOrder 字段。"""
+    sku_name = _text(_pick(r, "skuName", "goodsName", "productName"))
+    if sku_name and len(sku_name) > 500:
+        sku_name = sku_name[:500]
+    valid_code = _text(_pick(r, "validCode", "valid_code"))
+    actual_fee = _yuan_to_cents(_pick(r, "actualFee", "actual_fee"))
+    estimate_fee = _yuan_to_cents(_pick(r, "estimateFee", "estimate_fee"))
+    commission = actual_fee if actual_fee not in (None, 0) else estimate_fee
+    order_time = _jd_dt_to_utc(_pick(r, "orderTime", "order_time"))
+    return {
+        "platform": "jd",
+        "external_order_id": _text(_pick(r, "orderId", "parentOrderId")),
+        "external_row_id": _text(_pick(r, "id", "rowId", "orderRowId")),
+        "sid": _text(_pick(r, "subUnionId", "sub_union_id")),
+        "act_id": None,
+        "biz_line": None,
+        "trade_type": None,
+        "pay_price_cents": _yuan_to_cents(
+            _pick(r, "actualCosPrice", "estimateCosPrice", "price")
+        ),
+        "commission_cents": commission,
+        "commission_rate": _text(_pick(r, "commissionRate", "commission_rate")),
+        "refund_price_cents": None,
+        "refund_profit_cents": None,
+        "estimated_commission_cents": estimate_fee,
+        "actual_commission_cents": actual_fee,
+        "mt_status": None,
+        "jd_valid_code": valid_code,
+        "invalid_reason": None if _is_jd_valid_code(valid_code) else f"validCode={valid_code}",
+        "product_name": sku_name,
+        "settle_month": _text(_pick(r, "payMonth", "settleMonth", "pay_month")),
+        "site_id": _text(_pick(r, "siteId", "site_id")),
+        "position_id": _text(_pick(r, "positionId", "position_id")),
+        "pid": _text(_pick(r, "pid")),
+        "sub_union_id": _text(_pick(r, "subUnionId", "sub_union_id")),
+        "pay_time": order_time,
+        "mt_update_time": _jd_dt_to_utc(_pick(r, "modifyTime", "updateTime", "modify_time"))
+        or order_time,
+        "raw": r,
+    }
+
+
+def _is_jd_valid_code(valid_code: str | None) -> bool:
+    code = str(valid_code).strip() if valid_code is not None else ""
+    return bool(code and code not in _JD_INVALID_CODES and code not in _JD_UNPAID_CODES)
+
+
+def is_jd_order_valid(order: CpsOrder) -> bool:
+    return _is_jd_valid_code(order.jd_valid_code)
+
+
+def effective_commission_cents(order: CpsOrder) -> int:
+    if order.platform == "jd":
+        if order.actual_commission_cents not in (None, 0):
+            return order.actual_commission_cents or 0
+        if order.estimated_commission_cents is not None:
+            return order.estimated_commission_cents or 0
+    return order.commission_cents or 0
+
+
 def reconcile_orders(
     db: Session, *, start_time: int, end_time: int,
     query_time_type: int = 1, sid: str | None = None, max_pages: int = 200,
@@ -274,6 +400,11 @@ def reconcile_orders(
                 continue
             fetched += 1
             fields = _map_order_fields(r)
+            fields.setdefault("platform", "meituan")
+            fields.setdefault("external_order_id", order_id)
+            fields.setdefault("external_row_id", None)
+            fields.setdefault("estimated_commission_cents", fields.get("commission_cents"))
+            fields.setdefault("actual_commission_cents", None)
             existing = db.execute(
                 select(CpsOrder).where(CpsOrder.order_id == order_id)
             ).scalar_one_or_none()
@@ -291,6 +422,56 @@ def reconcile_orders(
     return {"fetched": fetched, "inserted": inserted, "updated": updated, "pages": pages}
 
 
+def reconcile_jd_orders(
+    db: Session, *, start_time: datetime, end_time: datetime,
+    query_time_type: int = 3, max_pages: int = 100,
+) -> dict:
+    """调京东 order.row.query 拉单 → 按订单行 upsert。
+
+    京东单次查询窗口最多 1 小时,这里按北京自然时间切窗并逐页拉取。
+    """
+    fetched = inserted = updated = pages = 0
+    cur = start_time
+    while cur < end_time:
+        win_end = min(cur + timedelta(hours=1), end_time)
+        page = 1
+        while page <= max_pages:
+            resp = jd_union.query_order_rows(
+                start_time=cur,
+                end_time=win_end,
+                query_time_type=query_time_type,
+                page_index=page,
+                page_size=200,
+            )
+            rows = resp.get("rows") or []
+            has_more = bool(resp.get("has_more"))
+            if not rows:
+                break
+            pages += 1
+            for r in rows:
+                order_id = _jd_order_key(r)
+                if not order_id:
+                    continue
+                fetched += 1
+                fields = _map_jd_order_fields(r)
+                existing = db.execute(
+                    select(CpsOrder).where(CpsOrder.order_id == order_id)
+                ).scalar_one_or_none()
+                if existing is None:
+                    db.add(CpsOrder(order_id=order_id, **fields))
+                    inserted += 1
+                else:
+                    for k, v in fields.items():
+                        setattr(existing, k, v)
+                    updated += 1
+            if not has_more or len(rows) < 200:
+                break
+            page += 1
+        cur = win_end
+    db.commit()
+    return {"fetched": fetched, "inserted": inserted, "updated": updated, "pages": pages}
+
+
 def list_orders(
     db: Session, *, sid: str | None = None, mt_status: str | None = None,
     limit: int = 20, cursor: int | None = None,
@@ -524,3 +705,90 @@ def group_wx_users(db: Session, *, group_id: int, limit: int = 200) -> list[dict
     ]
     result.sort(key=lambda x: x["first_seen"], reverse=True)
     return result[:limit]
+
+
+def group_day_users(
+    db: Session, *, group_id: int, start: datetime, end: datetime, limit: int = 200,
+) -> list[dict]:
+    """该群某天(北京)以用户为单位的领券/点击 + 每人 visit 过的券。
+
+    时间窗为半开区间 [start, end)(end=次日 00:00),避免午夜双计。只统计 openid 非空
+    (可归属到人)的点击 —— 匿名点击(美团/京东 302 多为匿名)不计入。券名 = 该点击 link
+    对应活动名;活动被硬删则兜底 活动#{id}。copy=领券次数、visit=点击次数;coupons 仅
+    取 visit 事件按活动分组、按次数倒序(合计 = visit_count)。排序:领券 desc、再点击 desc。
+    与 group_wx_users 同风格(Python 侧聚合,跨 PG/SQLite 无方言坑)。
+
+    注:每日明细行的 click_pv/copy_pv 计全部点击(含匿名、UV 按 ip,ua);本函数只计 openid
+    用户,故各用户求和 <= 当天行总数,二者口径不同、不必相等。
+    """
+    rows = db.execute(
+        select(CpsClick.openid, CpsClick.event_type, CpsClick.link_id)
+        .where(CpsClick.group_id == group_id)
+        .where(CpsClick.clicked_at >= _as_utc(start))
+        .where(CpsClick.clicked_at < _as_utc(end))
+        .where(CpsClick.openid.is_not(None))
+    ).all()
+    if not rows:
+        return []
+
+    # link_id -> activity_id -> 券名(活动名)
+    link_ids = {r.link_id for r in rows}
+    link_to_act = dict(
+        db.execute(
+            select(CpsLink.id, CpsLink.activity_id).where(CpsLink.id.in_(link_ids))
+        ).all()
+    )
+    act_ids = {aid for aid in link_to_act.values() if aid is not None}
+    act_name = (
+        dict(
+            db.execute(
+                select(CpsActivity.id, CpsActivity.name).where(CpsActivity.id.in_(act_ids))
+            ).all()
+        )
+        if act_ids
+        else {}
+    )
+
+    def _coupon_name(link_id: int) -> str:
+        aid = link_to_act.get(link_id)
+        if aid is None:
+            return f"链接#{link_id}"
+        return act_name.get(aid) or f"活动#{aid}"
+
+    stat: dict[str, dict] = {}
+    for openid, event_type, link_id in rows:
+        s = stat.setdefault(openid, {"copy": 0, "visit": 0, "coupons": {}})
+        if event_type == "copy":
+            s["copy"] += 1
+        else:
+            s["visit"] += 1
+            name = _coupon_name(link_id)
+            s["coupons"][name] = s["coupons"].get(name, 0) + 1
+
+    openids = list(stat.keys())
+    users = {
+        u.openid: u
+        for u in db.execute(
+            select(CpsWxUser).where(CpsWxUser.openid.in_(openids))
+        ).scalars().all()
+    }
+
+    result = [
+        {
+            "openid": openid,
+            "nickname": users[openid].nickname if openid in users else None,
+            "headimgurl": users[openid].headimgurl if openid in users else None,
+            "copy_count": s["copy"],
+            "visit_count": s["visit"],
+            "coupons": [
+                {"name": name, "count": cnt}
+                # 次数倒序;同次数按券名升序兜底,保证 PG 无 ORDER BY 行序下输出稳定
+                for name, cnt in sorted(
+                    s["coupons"].items(), key=lambda kv: (-kv[1], kv[0])
+                )
+            ],
+        }
+        for openid, s in stat.items()
+    ]
+    result.sort(key=lambda x: (x["copy_count"], x["visit_count"]), reverse=True)
+    return result[:limit]

app/admin/repositories/mutations.py

@@ -65,16 +65,19 @@ def review_feedback(
     reward_coins: int | None = None,
     reject_reason: str | None = None,
     review_note: str | None = None,
+    admin_reply: str | None = None,
     commit: bool = True,
 ) -> Feedback:
-    """审核反馈:置 adopted/rejected + 记录奖励/原因/审核人/审核时间。
+    """审核反馈:置 adopted/rejected + 记录奖励/原因/回复留言/审核人/审核时间。
 
     发金币(wallet.grant_coins)由 router 在同一事务里调,确保状态、金币流水、审计一起提交。
+    admin_reply=给用户的回复留言(用户端可见),与 review_note(内部备注)区分。
     """
     feedback.status = status
     feedback.reward_coins = reward_coins
     feedback.reject_reason = reject_reason
     feedback.review_note = review_note
+    feedback.admin_reply = admin_reply
     feedback.reviewed_by_admin_id = reviewed_by_admin_id
     feedback.reviewed_at = datetime.now(CN_TZ).replace(tzinfo=None)
     if commit:

app/admin/repositories/queries.py

@@ -8,19 +8,33 @@ from __future__ import annotations
 from datetime import datetime, timedelta, timezone
 from zoneinfo import ZoneInfo
 
-from sqlalchemy import Select, asc, desc, func, or_, select
+from sqlalchemy import Select, asc, case, desc, func, or_, select
 from sqlalchemy.orm import Session
 
+from app.admin.repositories.stats import COMPARE_START_EVENT, COUPON_START_EVENT
 from app.core import rewards
+from app.core.config import settings
 from app.models.ad_feed_reward import AdFeedRewardRecord
 from app.models.ad_reward import AdRewardRecord
 from app.models.admin import AdminAuditLog
+from app.models.analytics_event import AnalyticsEvent
 from app.models.comparison import ComparisonRecord
+from app.models.coupon_state import CouponPromptEngagement
+from app.models.device import DeviceLiveness
 from app.models.feedback import Feedback
 from app.models.onboarding import OnboardingCompletion
 from app.models.price_report import PriceReport
 from app.models.user import User
-from app.models.wallet import CashTransaction, CoinAccount, CoinTransaction, WithdrawOrder
+from app.models.wallet import (
+    CashTransaction,
+    CoinAccount,
+    CoinTransaction,
+    InviteCashTransaction,
+    WithdrawOrder,
+)
+
+# 「最近活跃」计入的行为事件(与大盘 DAU/留存活跃口径一致:开始比价 + 开始领券)
+_ACTIVE_EVENTS = (COMPARE_START_EVENT, COUPON_START_EVENT)
 
 # 折算成可提现现金时,非广告金币来源的排除集(广告单独统计、人工调整不算"赚取")
 _NON_TASK_BIZ_TYPES = ("reward_video", "feed_ad_reward", "admin_grant", "admin_deduct")
@@ -73,6 +87,86 @@ def offset_paginate(
     return items, next_cursor, total
 
 
+def _last_active_parts():
+    """「最近活跃」的两个按 user_id 预聚合派生表(最近开始比价/领券事件、最近领券发起)。
+
+    活跃口径与大盘 DAU/留存一致(2026-07-05 产品定:进入 App≈登录 last_login_at +
+    发起比价 real_compare_start + 发起领券 real_coupon_start/claim_started)。
+    用 LEFT JOIN 预聚合而非相关标量子查询:后者在 PG 上对 users 每行各跑一个 SubPlan
+    (排序键、range 筛选、offset_paginate 的 count 三处叠加),埋点表大了会拖垮列表接口;
+    预聚合借 analytics_event.event 索引只扫两类 start 事件,每次查询聚合一次。
+    """
+    ev_agg = (
+        select(
+            AnalyticsEvent.user_id.label("user_id"),
+            func.max(AnalyticsEvent.created_at).label("last_at"),
+        )
+        .where(
+            AnalyticsEvent.user_id.is_not(None),
+            AnalyticsEvent.event.in_(_ACTIVE_EVENTS),
+        )
+        .group_by(AnalyticsEvent.user_id)
+        .subquery()
+    )
+    eng_agg = (
+        select(
+            CouponPromptEngagement.user_id.label("user_id"),
+            func.max(CouponPromptEngagement.created_at).label("last_at"),
+        )
+        .where(
+            CouponPromptEngagement.user_id.is_not(None),
+            CouponPromptEngagement.engage_type == "claim_started",
+        )
+        .group_by(CouponPromptEngagement.user_id)
+        .subquery()
+    )
+    return ev_agg, eng_agg
+
+
+def _norm_utc(dt: datetime | None) -> datetime | None:
+    """naive 视为 UTC 补 tzinfo(SQLite 读回 naive、PG 读回 aware,混着 max() 会 TypeError)。"""
+    if dt is None:
+        return None
+    return dt if dt.tzinfo is not None else dt.replace(tzinfo=timezone.utc)
+
+
+def _attach_last_active(db: Session, users: list[User]) -> None:
+    """给本页用户瞬态挂 last_active_at(非 DB 列,供 AdminUserListItem from_attributes 读)。
+
+    口径同 [_last_active_expr];按本页 user_id 批量两次 GROUP BY,防 N+1。
+    """
+    uids = [u.id for u in users]
+    if not uids:
+        return
+    ev_map = dict(
+        db.execute(
+            select(AnalyticsEvent.user_id, func.max(AnalyticsEvent.created_at))
+            .where(
+                AnalyticsEvent.user_id.in_(uids),
+                AnalyticsEvent.event.in_(_ACTIVE_EVENTS),
+            )
+            .group_by(AnalyticsEvent.user_id)
+        ).all()
+    )
+    eng_map = dict(
+        db.execute(
+            select(CouponPromptEngagement.user_id, func.max(CouponPromptEngagement.created_at))
+            .where(
+                CouponPromptEngagement.user_id.in_(uids),
+                CouponPromptEngagement.engage_type == "claim_started",
+            )
+            .group_by(CouponPromptEngagement.user_id)
+        ).all()
+    )
+    for u in users:
+        candidates = [
+            _norm_utc(u.last_login_at),
+            _norm_utc(ev_map.get(u.id)),
+            _norm_utc(eng_map.get(u.id)),
+        ]
+        u.last_active_at = max((c for c in candidates if c is not None), default=None)
+
+
 def list_users(
     db: Session,
     *,
@@ -84,16 +178,34 @@ def list_users(
     created_to: datetime | None = None,
     last_login_from: datetime | None = None,
     last_login_to: datetime | None = None,
+    last_active_from: datetime | None = None,
+    last_active_to: datetime | None = None,
     sort_by: str = "id",
     sort_order: str = "desc",
     limit: int = 20,
     cursor: int | None = None,
 ) -> tuple[list[User], int | None, int]:
-    """用户列表(admin 全量)。支持手机号前缀 / 渠道 / 状态 / 昵称模糊 / 注册·最近登录时间范围筛选,
-    按 id·注册时间·最近登录排序。**offset 分页**(cursor=offset):任意列排序下游标语义统一,
+    """用户列表(admin 全量)。支持手机号前缀 / 渠道 / 状态 / 昵称模糊 / 注册·最近登录·最近活跃
+    时间范围筛选,按 id·注册时间·最近登录·最近活跃排序;每页附带计算列 last_active_at
+    (口径见 [_last_active_expr])。**offset 分页**(cursor=offset):任意列排序下游标语义统一,
     代价是翻页期间数据变动可能错位一条——admin 低频场景可接受(同 [list_all_withdraw_orders])。
     日期入参统一转 tz-aware UTC 比较(列为 timestamptz,见 _as_utc)。"""
-    stmt = select(User)
+    # 最近活跃 = max(最近登录, 最近行为事件, 最近领券发起)。PG 用 GREATEST;SQLite 标量 max()
+    # 任一参数 NULL 即返回 NULL,故 LEFT JOIN 未命中侧 coalesce 到 last_login_at 兜底
+    # (注册即登录,该列恒非空)。派生表 1:1(按 user_id 聚合),outerjoin 不会放大行数,
+    # offset_paginate 的 count 不受影响。
+    ev_agg, eng_agg = _last_active_parts()
+    greatest = func.greatest if db.get_bind().dialect.name == "postgresql" else func.max
+    last_active = greatest(
+        User.last_login_at,
+        func.coalesce(ev_agg.c.last_at, User.last_login_at),
+        func.coalesce(eng_agg.c.last_at, User.last_login_at),
+    )
+    stmt = (
+        select(User)
+        .outerjoin(ev_agg, ev_agg.c.user_id == User.id)
+        .outerjoin(eng_agg, eng_agg.c.user_id == User.id)
+    )
     if phone:
         stmt = stmt.where(User.phone.like(f"{phone}%"))  # 前缀匹配
     if register_channel:
@@ -110,20 +222,30 @@ def list_users(
         stmt = stmt.where(User.last_login_at >= _as_utc(last_login_from))
     if last_login_to is not None:
         stmt = stmt.where(User.last_login_at <= _as_utc(last_login_to))
+    if last_active_from is not None:
+        stmt = stmt.where(last_active >= _as_utc(last_active_from))
+    if last_active_to is not None:
+        stmt = stmt.where(last_active <= _as_utc(last_active_to))
 
     sort_cols = {
         "id": User.id,
         "created_at": User.created_at,
         "last_login_at": User.last_login_at,
+        "last_active_at": last_active,
     }
     sort_col = sort_cols.get(sort_by, User.id)
     order_fn = asc if sort_order == "asc" else desc
     id_order = asc(User.id) if sort_order == "asc" else desc(User.id)
-    return offset_paginate(db, stmt, (order_fn(sort_col), id_order), limit=limit, cursor=cursor)
+    items, next_cursor, total = offset_paginate(
+        db, stmt, (order_fn(sort_col), id_order), limit=limit, cursor=cursor
+    )
+    _attach_last_active(db, items)
+    return items, next_cursor, total
 
 
-def _attach_user_info(db: Session, records: list[ComparisonRecord]) -> None:
-    """给每条比价记录瞬态挂 phone/nickname(非 DB 列,供 admin schema from_attributes 读)。"""
+def _attach_user_info(db: Session, records: list[ComparisonRecord | Feedback | PriceReport]) -> None:
+    """给每条记录瞬态挂 phone/nickname(非 DB 列,供 admin schema from_attributes 读)。
+    按 user_id 鸭子类型,比价记录/反馈/上报通用。"""
     uids = {r.user_id for r in records}
     if not uids:
         return
@@ -144,11 +266,14 @@ def list_comparison_records(
     phone: str | None = None,
     status: str | None = None,
     business_type: str | None = None,
+    store: str | None = None,
+    product: str | None = None,
     limit: int = 20,
     cursor: int | None = None,
 ) -> tuple[list[ComparisonRecord], int | None, int]:
     """admin 比价记录列表(debug)。按 user_id 精确 或 phone 前缀定位用户 + status/业务类型筛,
-    offset 分页(创建时间倒序、id 兜底)。join User 取 phone/nickname 瞬态挂记录上。"""
+    store(店名)/product(商品名)子串模糊匹配,offset 分页(创建时间倒序、id 兜底)。
+    join User 取 phone/nickname 瞬态挂记录上。"""
     stmt = select(ComparisonRecord)
     if user_id is not None:
         stmt = stmt.where(ComparisonRecord.user_id == user_id)
@@ -162,6 +287,11 @@ def list_comparison_records(
         stmt = stmt.where(ComparisonRecord.status == status)
     if business_type:
         stmt = stmt.where(ComparisonRecord.business_type == business_type)
+    if store:
+        stmt = stmt.where(ComparisonRecord.store_name.like(f"%{store}%"))
+    if product:
+        # 商品名搜 product_names 派生文本列(非 items JSON:SQLite 下 JSON 中文被转义无法直接 LIKE)。
+        stmt = stmt.where(ComparisonRecord.product_names.like(f"%{product}%"))
     items, next_cursor, total = offset_paginate(
         db, stmt,
         (desc(ComparisonRecord.created_at), desc(ComparisonRecord.id)),
@@ -199,6 +329,156 @@ def list_onboarding_devices(db: Session, *, limit: int = 500) -> list[dict]:
     ]
 
 
+def _heartbeat_seconds_ago(last: datetime | None) -> int | None:
+    """距上次心跳的秒数(兼容 SQLite 取回的 naive datetime,按 UTC 处理)。None = 从没心跳。"""
+    if last is None:
+        return None
+    if last.tzinfo is None:
+        last = last.replace(tzinfo=timezone.utc)
+    return int((datetime.now(timezone.utc) - last).total_seconds())
+
+
+def _device_model_from_id(device_id: str) -> str:
+    """从 device_id(格式 device_<机型>_<hash>)解析机型;不规范则回退原 id。"""
+    parts = device_id.split("_")
+    if len(parts) >= 3 and parts[0] == "device":
+        return "_".join(parts[1:-1]).replace("_", " ")
+    return device_id
+
+
+def _attach_device_derived(devices: list[DeviceLiveness]) -> None:
+    """给每台设备瞬态挂 device_model / online / display_state / offline_seconds(供 schema 读)。
+
+    在线判定:开过无障碍(ever_protected)且距上次心跳 ≤ HEARTBEAT_TIMEOUT_MINUTES。
+    从没开过无障碍(ever_protected=False)= never(不算掉线,避免把装了没用的设备误报掉线)。
+    offline_seconds 仅 offline 时给值(=掉线时长),在线/从未启用为 None。"""
+    timeout_sec = max(1, int(settings.HEARTBEAT_TIMEOUT_MINUTES)) * 60
+    for d in devices:
+        d.device_model = _device_model_from_id(d.device_id)
+        secs = _heartbeat_seconds_ago(d.last_heartbeat_at)
+        if not d.ever_protected:
+            d.online = False
+            d.display_state = "never"
+            d.offline_seconds = None
+        elif secs is not None and secs <= timeout_sec:
+            d.online = True
+            d.display_state = "online"
+            d.offline_seconds = None
+        else:
+            d.online = False
+            d.display_state = "offline"
+            d.offline_seconds = secs
+
+
+def _attach_device_user_info(db: Session, devices: list[DeviceLiveness]) -> None:
+    """给每台设备瞬态挂归属用户 phone/nickname(同 _attach_user_info,供 admin schema 读)。"""
+    uids = {d.user_id for d in devices}
+    if not uids:
+        return
+    rows = db.execute(
+        select(User.id, User.phone, User.nickname).where(User.id.in_(uids))
+    ).all()
+    umap = {uid: (phone, nick) for uid, phone, nick in rows}
+    for d in devices:
+        phone, nick = umap.get(d.user_id, (None, None))
+        d.phone = phone
+        d.nickname = nick
+
+
+def _liveness_cutoff() -> datetime:
+    """掉线判定分界:此刻 - HEARTBEAT_TIMEOUT_MINUTES。心跳早于它 = 掉线(同 list_overdue 口径)。"""
+    timeout_min = max(1, int(settings.HEARTBEAT_TIMEOUT_MINUTES))
+    return datetime.now(timezone.utc) - timedelta(minutes=timeout_min)
+
+
+def list_device_liveness(
+    db: Session,
+    *,
+    status: str | None = None,
+    device_id: str | None = None,
+    phone: str | None = None,
+    user_id: int | None = None,
+    sort_by: str = "status",
+    sort_order: str = "desc",
+    limit: int = 20,
+    cursor: int | None = None,
+) -> tuple[list[DeviceLiveness], int | None, int]:
+    """设备存活列表(admin 全量)。按 在线情况(online/offline/never)/ 设备id(包含)/
+    归属用户(手机号前缀 或 user_id)筛,offset 分页。join user 取 phone/nickname,派生
+    device_model/online/display_state/offline_seconds 挂行上。
+    默认排序 status=掉线置顶(offline → online → never;掉线组内掉得最久在前)。"""
+    cutoff = _liveness_cutoff()
+    stmt = select(DeviceLiveness)
+    # 在线情况派生筛选(口径同 _attach_device_derived:ever_protected + 心跳是否过阈值)
+    if status == "online":
+        stmt = stmt.where(
+            DeviceLiveness.ever_protected.is_(True),
+            DeviceLiveness.last_heartbeat_at.is_not(None),
+            DeviceLiveness.last_heartbeat_at >= cutoff,
+        )
+    elif status == "offline":
+        stmt = stmt.where(
+            DeviceLiveness.ever_protected.is_(True),
+            DeviceLiveness.last_heartbeat_at.is_not(None),
+            DeviceLiveness.last_heartbeat_at < cutoff,
+        )
+    elif status == "never":
+        stmt = stmt.where(DeviceLiveness.ever_protected.is_(False))
+    if device_id and device_id.strip():
+        stmt = stmt.where(DeviceLiveness.device_id.like(f"%{device_id.strip()}%"))
+    if user_id is not None:
+        stmt = stmt.where(DeviceLiveness.user_id == user_id)
+    if phone:
+        stmt = stmt.where(
+            DeviceLiveness.user_id.in_(select(User.id).where(User.phone.like(f"{phone}%")))
+        )
+
+    if sort_by in ("last_heartbeat_at", "created_at"):
+        col = (
+            DeviceLiveness.last_heartbeat_at
+            if sort_by == "last_heartbeat_at"
+            else DeviceLiveness.created_at
+        )
+        order_fn = asc if sort_order == "asc" else desc
+        id_order = asc(DeviceLiveness.id) if sort_order == "asc" else desc(DeviceLiveness.id)
+        sort_clause: tuple = (order_fn(col), id_order)
+    else:
+        # 默认「掉线置顶」:offline(0) → online(1) → never(2);掉线组内按心跳最旧(掉得最久)在前
+        rank = case(
+            (DeviceLiveness.ever_protected.is_(False), 2),
+            (DeviceLiveness.last_heartbeat_at < cutoff, 0),
+            else_=1,
+        )
+        sort_clause = (asc(rank), asc(DeviceLiveness.last_heartbeat_at), desc(DeviceLiveness.id))
+
+    items, next_cursor, total = offset_paginate(db, stmt, sort_clause, limit=limit, cursor=cursor)
+    _attach_device_user_info(db, items)
+    _attach_device_derived(items)
+    return items, next_cursor, total
+
+
+def device_liveness_stats(db: Session) -> dict:
+    """顶部卡片:总设备数 + 在线 / 已掉线 / 未启用(按心跳阈值派生,口径同列表)。"""
+    cutoff = _liveness_cutoff()
+
+    def _count(*conds) -> int:
+        return int(db.execute(select(func.count(DeviceLiveness.id)).where(*conds)).scalar_one())
+
+    total = int(db.execute(select(func.count(DeviceLiveness.id))).scalar_one())
+    never = _count(DeviceLiveness.ever_protected.is_(False))
+    online = _count(
+        DeviceLiveness.ever_protected.is_(True),
+        DeviceLiveness.last_heartbeat_at.is_not(None),
+        DeviceLiveness.last_heartbeat_at >= cutoff,
+    )
+    offline = _count(
+        DeviceLiveness.ever_protected.is_(True),
+        DeviceLiveness.last_heartbeat_at.is_not(None),
+        DeviceLiveness.last_heartbeat_at < cutoff,
+    )
+    return {"total": total, "online": online, "offline": offline, "never": never}
+
+
 def list_all_coin_transactions(
     db: Session,
     *,
@@ -236,6 +516,7 @@ def list_all_withdraw_orders(
     *,
     user_id: int | None = None,
     status: str | None = None,
+    source: str | None = None,
     keyword: str | None = None,
     date_from: datetime | None = None,
     date_to: datetime | None = None,
@@ -255,6 +536,9 @@ def list_all_withdraw_orders(
         stmt = stmt.where(WithdrawOrder.user_id == user_id)
     if status:
         stmt = stmt.where(WithdrawOrder.status == status)
+    # 提现类型:coin_cash(福利页提现)/ invite_cash(邀请提现);None=全部
+    if source:
+        stmt = stmt.where(WithdrawOrder.source == source)
 
     kw = (keyword or "").strip()
     if kw:
@@ -380,6 +664,7 @@ def list_feedbacks(
     db: Session,
     *,
     status: str | None = None,
+    source: str | None = None,
     user_id: int | None = None,
     content: str | None = None,
     created_from: datetime | None = None,
@@ -389,13 +674,15 @@ def list_feedbacks(
     limit: int = 20,
     cursor: int | None = None,
 ) -> tuple[list[Feedback], int | None, int]:
-    """反馈工单列表。支持 状态 / 用户ID / 内容模糊 / 提交时间范围 筛选,按 id·提交时间排序。
-    **offset 分页**(cursor=offset):任意列排序下游标语义统一(同 [list_users]),代价是翻页期间
-    数据变动可能错位一条——admin 低频场景可接受。返回 (items, next_cursor, total),total 供页码分页。
-    created_at 为 timestamptz,日期入参统一转 tz-aware UTC 比较。"""
+    """反馈工单列表。支持 状态 / 反馈类型(source) / 用户ID / 内容模糊 / 提交时间范围 筛选,
+    按 id·提交时间排序。**offset 分页**(cursor=offset):任意列排序下游标语义统一(同 [list_users]),
+    代价是翻页期间数据变动可能错位一条——admin 低频场景可接受。返回 (items, next_cursor, total),
+    total 供页码分页。created_at 为 timestamptz,日期入参统一转 tz-aware UTC 比较。"""
     stmt = select(Feedback)
     if status:
         stmt = stmt.where(Feedback.status == status)
+    if source:
+        stmt = stmt.where(Feedback.source == source)
     if user_id is not None:
         stmt = stmt.where(Feedback.user_id == user_id)
     if content and content.strip():
@@ -409,6 +696,62 @@ def list_feedbacks(
     sort_col = sort_cols.get(sort_by, Feedback.id)
     order_fn = asc if sort_order == "asc" else desc
     id_order = asc(Feedback.id) if sort_order == "asc" else desc(Feedback.id)
+    items, next_cursor, total = offset_paginate(
+        db, stmt, (order_fn(sort_col), id_order), limit=limit, cursor=cursor
+    )
+    _attach_user_info(db, items)  # 列表展示完整手机号(点手机号查该用户全部反馈)
+    return items, next_cursor, total
+
+
+def feedback_summary(db: Session) -> dict:
+    """反馈审核台各状态计数(待审核/已采纳/未采纳/合计)。待审核含历史 new 态(与前端 isPending 一致)。"""
+    rows = db.execute(
+        select(Feedback.status, func.count(Feedback.id)).group_by(Feedback.status)
+    ).all()
+    by_status = {status: int(count) for status, count in rows}
+    return {
+        "pending": by_status.get("pending", 0) + by_status.get("new", 0),
+        "adopted": by_status.get("adopted", 0),
+        "rejected": by_status.get("rejected", 0),
+        "total": sum(by_status.values()),
+    }
+
+
+def list_analytics_events(
+    db: Session,
+    *,
+    event: str | None = None,
+    device_id: str | None = None,
+    user_id: int | None = None,
+    session_id: str | None = None,
+    created_from: datetime | None = None,
+    created_to: datetime | None = None,
+    sort_by: str = "id",
+    sort_order: str = "desc",
+    limit: int = 20,
+    cursor: int | None = None,
+) -> tuple[list[AnalyticsEvent], int | None, int]:
+    """埋点日志列表(admin 全量)。按 事件名 / 设备ID前缀 / 用户ID / 会话ID / 接收时间范围 筛选,
+    按 id·接收时间排序。offset 分页(同 [list_feedbacks])。created_at 为 timestamptz,
+    日期入参统一转 tz-aware UTC 比较。"""
+    stmt = select(AnalyticsEvent)
+    if event:
+        stmt = stmt.where(AnalyticsEvent.event == event)
+    if device_id and device_id.strip():
+        stmt = stmt.where(AnalyticsEvent.device_id.like(f"{device_id.strip()}%"))
+    if user_id is not None:
+        stmt = stmt.where(AnalyticsEvent.user_id == user_id)
+    if session_id and session_id.strip():
+        stmt = stmt.where(AnalyticsEvent.session_id == session_id.strip())
+    if created_from is not None:
+        stmt = stmt.where(AnalyticsEvent.created_at >= _as_utc(created_from))
+    if created_to is not None:
+        stmt = stmt.where(AnalyticsEvent.created_at <= _as_utc(created_to))
+
+    sort_cols = {"id": AnalyticsEvent.id, "created_at": AnalyticsEvent.created_at}
+    sort_col = sort_cols.get(sort_by, AnalyticsEvent.id)
+    order_fn = asc if sort_order == "asc" else desc
+    id_order = asc(AnalyticsEvent.id) if sort_order == "asc" else desc(AnalyticsEvent.id)
     return offset_paginate(db, stmt, (order_fn(sort_col), id_order), limit=limit, cursor=cursor)
 
 
@@ -509,26 +852,19 @@ def withdraw_risk_flags(
     return flags, score
 
 
-def withdraw_ledger_check(db: Session) -> dict:
-    cash_balance_total = int(
-        db.execute(select(func.coalesce(func.sum(CoinAccount.cash_balance_cents), 0))).scalar_one()
-    )
-    cash_txn_total = int(
-        db.execute(select(func.coalesce(func.sum(CashTransaction.amount_cents), 0))).scalar_one()
-    )
+def _check_withdraw_ledger_side(
+    orders: list[WithdrawOrder], txns: list, *, withdraw_biz: str, refund_biz: str
+) -> dict:
+    """对某一本账(普通现金 / 邀请奖励金)做提现单 ↔ 流水的交叉校验。
 
-    orders = list(db.execute(select(WithdrawOrder)).scalars().all())
-    cash_txns = list(
-        db.execute(
-            select(CashTransaction).where(
-                CashTransaction.biz_type.in_(("withdraw", "withdraw_refund"))
-            )
-        ).scalars().all()
-    )
-    withdraw_refs = {txn.ref_id for txn in cash_txns if txn.biz_type == "withdraw"}
+    orders 已按 source 过滤到本账;txns 是本账流水表里 withdraw_biz/refund_biz 两类流水。
+    规则:每单发起应有一条扣款流水(ref_id=out_bill_no);失败/拒绝单应有且仅一条退款流水;
+    非退款终态不应出现退款流水。四个计数全为 0 即本账自洽。
+    """
+    withdraw_refs = {txn.ref_id for txn in txns if txn.biz_type == withdraw_biz}
     refund_counts: dict[str, int] = {}
-    for txn in cash_txns:
-        if txn.biz_type == "withdraw_refund" and txn.ref_id:
+    for txn in txns:
+        if txn.biz_type == refund_biz and txn.ref_id:
             refund_counts[txn.ref_id] = refund_counts.get(txn.ref_id, 0) + 1
 
     missing_withdraw = 0
@@ -543,24 +879,95 @@ def withdraw_ledger_check(db: Session) -> dict:
         if has_refund and order.status not in {"failed", "rejected"}:
             refund_on_non_terminal += 1
 
-    duplicate_refund = sum(1 for count in refund_counts.values() if count > 1)
-    diff = cash_balance_total - cash_txn_total
+    return {
+        "missing_withdraw": missing_withdraw,
+        "missing_refund": missing_refund,
+        "duplicate_refund": sum(1 for count in refund_counts.values() if count > 1),
+        "refund_on_non_terminal": refund_on_non_terminal,
+    }
+
+
+def withdraw_ledger_check(db: Session) -> dict:
+    """现金账本校验:两本物理隔离的账各自对账(产品红线:coin_cash / invite_cash 不串)。
+
+    普通现金:CoinAccount.cash_balance_cents ↔ cash_transaction(withdraw/withdraw_refund);
+    邀请奖励金:CoinAccount.invite_cash_balance_cents ↔ invite_cash_transaction
+              (invite_withdraw/invite_withdraw_refund)。
+    提现单按 source 分流到对应账核对——邀请提现的流水写在 invite_cash_transaction 表,
+    绝不能拿去和普通现金流水比(否则每笔邀请提现单都会被误报「缺扣款/缺退款流水」)。
+    分流口径与 create_withdraw 一致:仅 source==invite_cash 走邀请账,其余(含历史空值)归普通现金。
+    """
+    orders = list(db.execute(select(WithdrawOrder)).scalars().all())
+    coin_orders = [o for o in orders if o.source != "invite_cash"]
+    invite_orders = [o for o in orders if o.source == "invite_cash"]
+
+    # —— 普通现金账(coin_cash) ——
+    cash_balance_total = int(
+        db.execute(select(func.coalesce(func.sum(CoinAccount.cash_balance_cents), 0))).scalar_one()
+    )
+    cash_txn_total = int(
+        db.execute(select(func.coalesce(func.sum(CashTransaction.amount_cents), 0))).scalar_one()
+    )
+    cash_txns = list(
+        db.execute(
+            select(CashTransaction).where(
+                CashTransaction.biz_type.in_(("withdraw", "withdraw_refund"))
+            )
+        ).scalars().all()
+    )
+    coin = _check_withdraw_ledger_side(
+        coin_orders, cash_txns, withdraw_biz="withdraw", refund_biz="withdraw_refund"
+    )
+    cash_diff = cash_balance_total - cash_txn_total
+
+    # —— 邀请奖励金账(invite_cash,独立账户 + 独立流水表) ——
+    invite_balance_total = int(
+        db.execute(
+            select(func.coalesce(func.sum(CoinAccount.invite_cash_balance_cents), 0))
+        ).scalar_one()
+    )
+    invite_txn_total = int(
+        db.execute(
+            select(func.coalesce(func.sum(InviteCashTransaction.amount_cents), 0))
+        ).scalar_one()
+    )
+    invite_txns = list(
+        db.execute(
+            select(InviteCashTransaction).where(
+                InviteCashTransaction.biz_type.in_(("invite_withdraw", "invite_withdraw_refund"))
+            )
+        ).scalars().all()
+    )
+    invite = _check_withdraw_ledger_side(
+        invite_orders, invite_txns,
+        withdraw_biz="invite_withdraw", refund_biz="invite_withdraw_refund",
+    )
+    invite_diff = invite_balance_total - invite_txn_total
+
     ok = (
-        diff == 0
-        and missing_withdraw == 0
-        and missing_refund == 0
-        and duplicate_refund == 0
-        and refund_on_non_terminal == 0
+        cash_diff == 0
+        and invite_diff == 0
+        and all(v == 0 for v in coin.values())
+        and all(v == 0 for v in invite.values())
     )
     return {
         "ok": ok,
+        # 普通现金账(coin_cash:金币兑换的现金)
         "cash_balance_total_cents": cash_balance_total,
         "cash_transaction_total_cents": cash_txn_total,
-        "balance_diff_cents": diff,
-        "missing_withdraw_txn_count": missing_withdraw,
-        "missing_refund_txn_count": missing_refund,
-        "duplicate_refund_txn_count": duplicate_refund,
-        "refund_txn_on_non_terminal_count": refund_on_non_terminal,
+        "balance_diff_cents": cash_diff,
+        "missing_withdraw_txn_count": coin["missing_withdraw"],
+        "missing_refund_txn_count": coin["missing_refund"],
+        "duplicate_refund_txn_count": coin["duplicate_refund"],
+        "refund_txn_on_non_terminal_count": coin["refund_on_non_terminal"],
+        # 邀请奖励金账(invite_cash:与普通现金物理隔离,各自对账)
+        "invite_cash_balance_total_cents": invite_balance_total,
+        "invite_cash_transaction_total_cents": invite_txn_total,
+        "invite_balance_diff_cents": invite_diff,
+        "invite_missing_withdraw_txn_count": invite["missing_withdraw"],
+        "invite_missing_refund_txn_count": invite["missing_refund"],
+        "invite_duplicate_refund_txn_count": invite["duplicate_refund"],
+        "invite_refund_txn_on_non_terminal_count": invite["refund_on_non_terminal"],
     }
 
 
@@ -578,6 +985,7 @@ def get_user_overview(db: Session, user_id: int) -> dict | None:
         "user": user,
         "coin_balance": acc.coin_balance if acc else 0,
         "cash_balance_cents": acc.cash_balance_cents if acc else 0,
+        "invite_cash_balance_cents": acc.invite_cash_balance_cents if acc else 0,
         "total_coin_earned": acc.total_coin_earned if acc else 0,
         "comparison_total": _count(ComparisonRecord, ComparisonRecord.user_id == user_id),
         "comparison_success": _count(
@@ -595,6 +1003,12 @@ def get_user_overview(db: Session, user_id: int) -> dict | None:
     }
 
 
+def _as_utc_naive(value: datetime) -> datetime:
+    """窗口入参 → UTC naive(= _as_utc 去时区),与库里按 naive UTC 存取的 created_at 同口径比较。
+    历史遗留:_window_conds 一直引用本函数却未定义(自定义区间会 NameError),此处补上。"""
+    return _as_utc(value).replace(tzinfo=None)
+
+
 def _window_conds(col, date_from: datetime | None, date_to: datetime | None) -> list:
     """把 [date_from, date_to] 转成对 col(created_at)的过滤条件;都为 None = 全量(注册至今)。"""
     conds = []
@@ -684,6 +1098,13 @@ def user_reward_stats(
     }
 
 
+def _cn_wall_to_utc(dt: datetime) -> datetime:
+    """coin_transaction 存的是北京 wall-clock(naive,见 wallet.grant_coins「存北京 wall-clock」),转成 UTC naive,
+    与广告表(func.now() UTC)统一 —— 让本函数按同一绝对时刻排序、且前端 apiTime(把无时区时间当 UTC 再 +8 展示)
+    口径一致;否则签到会比实际多显示 8 小时(北京时间又被 +8)。"""
+    return dt.replace(tzinfo=rewards.CN_TZ).astimezone(timezone.utc).replace(tzinfo=None)
+
+
 def user_coin_records(
     db: Session,
     user_id: int,
@@ -704,6 +1125,9 @@ def user_coin_records(
     offset = max(cursor or 0, 0)
     fetch = offset + limit + 1
     rows: list[dict] = []
+    # coin_transaction 存北京 wall-clock(其余表存 UTC);签到窗口边界 +8h 对齐北京,过滤/计数才不偏移 8 小时
+    signin_from = date_from + timedelta(hours=8) if date_from is not None else None
+    signin_to = date_to + timedelta(hours=8) if date_to is not None else None
 
     for rec in db.execute(
         select(AdRewardRecord)
@@ -747,7 +1171,7 @@ def user_coin_records(
         .where(
             CoinTransaction.user_id == user_id,
             CoinTransaction.biz_type == "signin",
-            *_window_conds(CoinTransaction.created_at, date_from, date_to),
+            *_window_conds(CoinTransaction.created_at, signin_from, signin_to),
         )
         .order_by(CoinTransaction.created_at.desc())
         .limit(fetch)
@@ -755,7 +1179,8 @@ def user_coin_records(
         rows.append({
             "source": "signin",
             "source_label": "签到",
-            "created_at": rec.created_at,
+            # 北京 wall-clock → UTC,与广告记录统一(前端 apiTime 会 +8 回北京展示,不然签到会多 8 小时)
+            "created_at": _cn_wall_to_utc(rec.created_at),
             "ecpm": None,
             "coin": rec.amount,
         })
@@ -781,27 +1206,72 @@ def user_coin_records(
         + _count(
             CoinTransaction, CoinTransaction.user_id == user_id,
             CoinTransaction.biz_type == "signin",
-            *_window_conds(CoinTransaction.created_at, date_from, date_to),
+            *_window_conds(CoinTransaction.created_at, signin_from, signin_to),
         )
     )
     return rows[offset:offset + limit], (offset + limit if has_more else None), total
 
 
+def _attach_price_report_comparison(db: Session, records: list[PriceReport]) -> None:
+    """给每条上报瞬态挂关联比价记录的 trace + 设备/版本快照:
+    trace_id/trace_url(点 Trace 看完整比价过程)、device_model/rom_name/android_version(机型OS版本列)、
+    app_version(提交版本号列,= 提交时我们 app 的 versionName)。
+    comparison_record_id 为空 / 关联记录查不到 → 全 None。"""
+    cids = {r.comparison_record_id for r in records if r.comparison_record_id is not None}
+    rows = (
+        db.execute(
+            select(
+                ComparisonRecord.id,
+                ComparisonRecord.trace_id,
+                ComparisonRecord.trace_url,
+                ComparisonRecord.device_model,
+                ComparisonRecord.rom_name,
+                ComparisonRecord.android_version,
+                ComparisonRecord.app_version,
+            ).where(ComparisonRecord.id.in_(cids))
+        ).all()
+        if cids
+        else []
+    )
+    cmap = {row.id: row for row in rows}
+    for r in records:
+        row = cmap.get(r.comparison_record_id)
+        r.trace_id = row.trace_id if row else None
+        r.trace_url = row.trace_url if row else None
+        r.device_model = row.device_model if row else None
+        r.rom_name = row.rom_name if row else None
+        r.android_version = row.android_version if row else None
+        r.app_version = row.app_version if row else None
+
+
 def list_price_reports(
     db: Session,
     *,
     status: str | None = None,
     user_id: int | None = None,
+    sort_by: str = "id",
+    sort_order: str = "desc",
     limit: int = 20,
     cursor: int | None = None,
 ) -> tuple[list[PriceReport], int | None, int]:
-    """上报更低价列表(admin 全量,可按状态/用户筛)。offset 分页 + total,id 倒序。"""
+    """上报更低价列表(admin 全量,可按状态/用户筛)。offset 分页 + total,按 id·提交时间排序。
+    join User 取 phone 瞬态挂(列表展示完整手机号);按 comparison_record_id 挂 trace_id/trace_url。"""
     stmt = select(PriceReport)
     if status:
         stmt = stmt.where(PriceReport.status == status)
     if user_id is not None:
         stmt = stmt.where(PriceReport.user_id == user_id)
-    return offset_paginate(db, stmt, (PriceReport.id.desc(),), limit=limit, cursor=cursor)
+
+    sort_cols = {"id": PriceReport.id, "created_at": PriceReport.created_at}
+    sort_col = sort_cols.get(sort_by, PriceReport.id)
+    order_fn = asc if sort_order == "asc" else desc
+    id_order = asc(PriceReport.id) if sort_order == "asc" else desc(PriceReport.id)
+    items, next_cursor, total = offset_paginate(
+        db, stmt, (order_fn(sort_col), id_order), limit=limit, cursor=cursor
+    )
+    _attach_user_info(db, items)
+    _attach_price_report_comparison(db, items)
+    return items, next_cursor, total
 
 
 def price_report_summary(db: Session) -> dict:

app/admin/repositories/stats.py

@@ -5,20 +5,50 @@ user.last_login_at / comparison_record.status / withdraw_order.status)要加索
 """
 from __future__ import annotations
 
-from datetime import datetime, timedelta, timezone
+from collections import Counter
+from datetime import date, datetime, time, timedelta, timezone
+from decimal import Decimal, InvalidOperation
 
-from sqlalchemy import func, select
+from sqlalchemy import case, func, select
 from sqlalchemy.orm import Session
 
+from app.admin.repositories.coupon_data import _percentile
 from app.models.ad_feed_reward import AdFeedRewardRecord
 from app.models.ad_reward import AdRewardRecord
+from app.models.analytics_event import AnalyticsEvent
 from app.models.comparison import ComparisonRecord
+from app.models.coupon_state import (
+    CouponClaimRecord,
+    CouponPromptEngagement,
+    CouponSession,
+)
+from app.models.cps_order import CpsOrder
 from app.models.feedback import Feedback
+from app.models.savings import SavingsRecord
 from app.models.signin import SigninBoostRecord, SigninRecord
 from app.models.user import User
 from app.models.wallet import CoinTransaction, WithdrawOrder
 
 _BEIJING = timezone(timedelta(hours=8))
+COUPON_REWARD_BIZ_TYPES = ("reward_video", "ad_reward", "coupon", "coupon_reward")
+COMPARISON_REWARD_BIZ_TYPES = ("comparison", "compare_reward", "comparison_reward")
+EXCLUDED_REWARD_BIZ_TYPES = ("invite_inviter", "invite_invitee", "admin_grant")
+UNCLASSIFIED_FEED_BIZ_TYPES = ("feed_ad_reward",)
+REGULAR_TASK_EXCLUDED_BIZ_TYPES = (
+    *COUPON_REWARD_BIZ_TYPES,
+    *COMPARISON_REWARD_BIZ_TYPES,
+    *EXCLUDED_REWARD_BIZ_TYPES,
+    *UNCLASSIFIED_FEED_BIZ_TYPES,
+)
+MEITUAN_CPS_INVALID_STATUSES = ("4", "5")
+MEITUAN_CPS_SETTLED_STATUS = "6"
+COMPARE_START_EVENT = "real_compare_start"
+COUPON_START_EVENT = "real_coupon_start"
+JD_CPS_INVALID_CODES = {
+    "2", "3", "4", "5", "6", "7", "8", "9", "11", "13", "14", "19", "20", "21",
+    "22", "23", "25", "26", "27", "28", "29", "30", "31", "34", "35", "36",
+}
+JD_CPS_UNPAID_CODES = {"15"}
 
 
 def _beijing_today_start_utc() -> datetime:
@@ -28,20 +58,161 @@ def _beijing_today_start_utc() -> datetime:
     return start_bj.astimezone(timezone.utc)
 
 
-def dashboard_overview(db: Session) -> dict:
+def today_dau(db: Session) -> int:
+    """今日活跃用户数(DAU):登录 + 开始比价 + 开始领券,按用户去重。
+
+    = period_active_dau(今天, 今天);历史 / 多天窗口用 period_active_dau 传区间(广告收益报表复用)。
+    """
+    today_bj = datetime.now(_BEIJING).date()
+    return period_active_dau(db, today_bj, today_bj)
+
+
+def _default_period_end() -> date:
+    """新版大盘不含今日,默认窗口结束日=北京时间昨天。"""
+    return datetime.now(_BEIJING).date() - timedelta(days=1)
+
+
+def _normalize_period(date_from: date | None, date_to: date | None) -> tuple[date, date]:
+    end = date_to or _default_period_end()
+    start = date_from or end
+    if start > end:
+        start, end = end, start
+    return start, end
+
+
+def _period_bounds(date_from: date, date_to: date) -> tuple[datetime, datetime, datetime, datetime]:
+    """返回同一北京自然日窗口的 UTC aware 边界和北京 naive 边界。
+
+    user.created_at / last_login_at 是 UTC aware 口径;比较/金币等历史上有北京 naive
+    写入,所以两套边界同时保留。
+    """
+    start_bj = datetime.combine(date_from, time.min, tzinfo=_BEIJING)
+    end_bj = datetime.combine(date_to + timedelta(days=1), time.min, tzinfo=_BEIJING)
+    start_utc = start_bj.astimezone(timezone.utc)
+    end_utc = end_bj.astimezone(timezone.utc)
+    return (
+        start_utc,
+        end_utc,
+        start_bj.replace(tzinfo=None),
+        end_bj.replace(tzinfo=None),
+    )
+
+
+def _date_range(date_from: date, date_to: date) -> list[date]:
+    days = (date_to - date_from).days
+    return [date_from + timedelta(days=i) for i in range(days + 1)]
+
+
+def _id_set(db: Session, stmt) -> set[int]:
+    return {int(v) for v in db.execute(stmt).scalars().all() if v is not None}
+
+
+def _event_user_ids(
+    db: Session, event_names: tuple[str, ...], start_utc: datetime, end_utc: datetime
+) -> set[int]:
+    return _id_set(
+        db,
+        select(AnalyticsEvent.user_id).where(
+            AnalyticsEvent.user_id.is_not(None),
+            AnalyticsEvent.event.in_(event_names),
+            AnalyticsEvent.created_at >= start_utc,
+            AnalyticsEvent.created_at < end_utc,
+        ),
+    )
+
+
+def _period_active_user_ids(
+    db: Session,
+    *,
+    start_utc: datetime,
+    end_utc: datetime,
+    period_from: date,
+    period_to: date,
+) -> set[int]:
+    """区间去重活跃用户 id 集合:登录(last_login_at)+ 开始比价(real_compare_start)+
+    开始领券(real_coupon_start / claim_started)。
+
+    user / analytics_event 用 UTC aware 边界 [start_utc, end_utc);coupon_prompt_engagement
+    的 engage_date 是北京自然日 date 列,用 [period_from, period_to] 闭区间。今日 / 历史 / 多天通用。
+    """
+    login_user_ids = _id_set(
+        db,
+        select(User.id).where(User.last_login_at >= start_utc, User.last_login_at < end_utc),
+    )
+    compare_start_user_ids = _event_user_ids(db, (COMPARE_START_EVENT,), start_utc, end_utc)
+    coupon_event_user_ids = _event_user_ids(db, (COUPON_START_EVENT,), start_utc, end_utc)
+    coupon_claim_user_ids = _id_set(
+        db,
+        select(CouponPromptEngagement.user_id).where(
+            CouponPromptEngagement.engage_date >= period_from,
+            CouponPromptEngagement.engage_date <= period_to,
+            CouponPromptEngagement.engage_type == "claim_started",
+        ),
+    )
+    return login_user_ids | compare_start_user_ids | coupon_event_user_ids | coupon_claim_user_ids
+
+
+def period_active_dau(db: Session, date_from: date, date_to: date) -> int:
+    """任意北京自然日区间 [date_from, date_to] 的去重活跃用户数。
+
+    与数据大盘 period.users.active 同口径(登录 + 开始比价 + 开始领券);广告收益报表按所选
+    日期区间(含今日)复用,ARPU = 区间预估收益 ÷ 本数。全局口径,不随用户 / 类型 / 场景筛选变化。
+    """
+    period_from, period_to = _normalize_period(date_from, date_to)
+    start_utc, end_utc, _start_local, _end_local = _period_bounds(period_from, period_to)
+    return len(
+        _period_active_user_ids(
+            db,
+            start_utc=start_utc,
+            end_utc=end_utc,
+            period_from=period_from,
+            period_to=period_to,
+        )
+    )
+
+
+def _commission_rate_percent(raw: str | None) -> Decimal | None:
+    """美团 commissionRate 原值: "300"=3%, "10"=0.1%;也兼容 "3%"。"""
+    if raw is None:
+        return None
+    s = str(raw).strip()
+    if not s:
+        return None
+    try:
+        if s.endswith("%"):
+            return Decimal(s[:-1])
+        val = Decimal(s)
+    except (InvalidOperation, ValueError):
+        return None
+    return val / Decimal("100")
+
+
+def _jd_valid_order(order: CpsOrder) -> bool:
+    code = str(order.jd_valid_code).strip() if order.jd_valid_code is not None else ""
+    return bool(code and code not in JD_CPS_INVALID_CODES and code not in JD_CPS_UNPAID_CODES)
+
+
+def dashboard_overview(
+    db: Session, *, date_from: date | None = None, date_to: date | None = None
+) -> dict:
     today_start = _beijing_today_start_utc()
+    period_from, period_to = _normalize_period(date_from, date_to)
+    start_utc, end_utc, start_local, end_local = _period_bounds(period_from, period_to)
 
     def _count(model, *conds) -> int:
         stmt = select(func.count(model.id))
         if conds:
             stmt = stmt.where(*conds)
-        return db.execute(stmt).scalar_one()
+        return int(db.execute(stmt).scalar_one())
 
     def _sum(col, *conds) -> int:
         stmt = select(func.coalesce(func.sum(col), 0))
         if conds:
             stmt = stmt.where(*conds)
-        return db.execute(stmt).scalar_one()
+        return int(db.execute(stmt).scalar_one())
+
+    def _user_id_set(stmt) -> set[int]:
+        return {int(v) for v in db.execute(stmt).scalars().all() if v is not None}
 
     # ===== 用户 =====
     by_status = dict(
@@ -61,6 +232,301 @@ def dashboard_overview(db: Session) -> dict:
     comparison_total = _count(ComparisonRecord)
     comparison_success = _count(ComparisonRecord, ComparisonRecord.status == "success")
     success_rate = round(comparison_success / comparison_total, 4) if comparison_total else 0.0
+    period_comparison_conds = (
+        ComparisonRecord.created_at >= start_local,
+        ComparisonRecord.created_at < end_local,
+    )
+    period_comparison_total = _count(ComparisonRecord, *period_comparison_conds)
+    period_comparison_success = _count(
+        ComparisonRecord,
+        *period_comparison_conds,
+        ComparisonRecord.status == "success",
+    )
+    period_comparison_success_rate = (
+        round(period_comparison_success / period_comparison_total, 4)
+        if period_comparison_total
+        else 0.0
+    )
+    period_saved_positive_count = _count(
+        ComparisonRecord,
+        *period_comparison_conds,
+        ComparisonRecord.status == "success",
+        ComparisonRecord.saved_amount_cents > 0,
+    )
+    period_saved_positive_sum = _sum(
+        ComparisonRecord.saved_amount_cents,
+        *period_comparison_conds,
+        ComparisonRecord.status == "success",
+        ComparisonRecord.saved_amount_cents > 0,
+    )
+    period_avg_saved_cents = (
+        round(period_saved_positive_sum / period_saved_positive_count)
+        if period_saved_positive_count
+        else None
+    )
+    period_avg_duration_ms = db.execute(
+        select(func.avg(ComparisonRecord.total_ms)).where(
+            *period_comparison_conds,
+            ComparisonRecord.total_ms.is_not(None),
+            ComparisonRecord.total_ms > 0,
+        )
+    ).scalar_one()
+    period_avg_duration_ms = (
+        round(float(period_avg_duration_ms))
+        if period_avg_duration_ms is not None
+        else None
+    )
+
+    ordered_exists = (
+        select(SavingsRecord.id)
+        .where(
+            SavingsRecord.user_id == ComparisonRecord.user_id,
+            SavingsRecord.source == "compare",
+            SavingsRecord.shop_name.is_not(None),
+            SavingsRecord.shop_name == ComparisonRecord.store_name,
+        )
+        .exists()
+    )
+    period_ordered_count = _count(
+        ComparisonRecord,
+        *period_comparison_conds,
+        ComparisonRecord.store_name.is_not(None),
+        ordered_exists,
+    )
+
+    # ===== 日期窗口用户 =====
+    period_new_user_ids = _user_id_set(
+        select(User.id).where(User.created_at >= start_utc, User.created_at < end_utc)
+    )
+    period_active_user_ids = _period_active_user_ids(
+        db,
+        start_utc=start_utc,
+        end_utc=end_utc,
+        period_from=period_from,
+        period_to=period_to,
+    )
+    # 留存口径(2026-07-05 产品改):次日留存——窗口内每天 D,取 **D-1 日(前日)新增**用户,
+    # 统计其 D 日活跃(登录/开始比价/开始领券)比例,逐日累加。默认窗口=昨日单天,即
+    # 「前日新增用户的昨日留存」。原口径(窗口内新增∩窗口内活跃)在单日窗口下≈100% 无意义
+    # (注册即登录,当天新增必然当天活跃)。逐日 cohort 在下方 trend 循环内顺带累计。
+    retention_cohort_total = 0
+    retention_retained_total = 0
+    trend_points: list[dict] = []
+    for cur_date in _date_range(period_from, period_to):
+        day_start_utc, day_end_utc, day_start_local, day_end_local = _period_bounds(
+            cur_date, cur_date
+        )
+        daily_comparison_conds = (
+            ComparisonRecord.created_at >= day_start_local,
+            ComparisonRecord.created_at < day_end_local,
+        )
+        daily_active_user_ids = _period_active_user_ids(
+            db,
+            start_utc=day_start_utc,
+            end_utc=day_end_utc,
+            period_from=cur_date,
+            period_to=cur_date,
+        )
+        # 次日留存:cohort = 前一日(D-1)新增用户,留存 = 其中当日(D)活跃者(口径见上)。
+        cohort_ids = _user_id_set(
+            select(User.id).where(
+                User.created_at >= day_start_utc - timedelta(days=1),
+                User.created_at < day_end_utc - timedelta(days=1),
+            )
+        )
+        retention_cohort_total += len(cohort_ids)
+        retention_retained_total += len(cohort_ids & daily_active_user_ids)
+        trend_points.append(
+            {
+                "date": cur_date,
+                "active_users": len(daily_active_user_ids),
+                "new_users": _count(
+                    User,
+                    User.created_at >= day_start_utc,
+                    User.created_at < day_end_utc,
+                ),
+                "comparisons": _count(ComparisonRecord, *daily_comparison_conds),
+            }
+        )
+    period_retention_rate = (
+        round(retention_retained_total / retention_cohort_total, 4)
+        if retention_cohort_total
+        else None
+    )
+
+    period_coin_conds = (
+        CoinTransaction.created_at >= start_local,
+        CoinTransaction.created_at < end_local,
+        CoinTransaction.amount > 0,
+    )
+    period_reward_video_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type.in_(("reward_video", "ad_reward")),
+    )
+    period_feed_ad_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type == "feed_ad_reward",
+    )
+    period_signin_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type == "signin",
+    )
+    period_signin_boost_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type == "signin_boost",
+    )
+    period_task_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type.like("task_%"),
+    )
+    period_coupon_reward_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type.in_(COUPON_REWARD_BIZ_TYPES),
+    )
+    period_comparison_reward_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type.in_(COMPARISON_REWARD_BIZ_TYPES),
+    )
+    period_regular_task_coin_total = _sum(
+        CoinTransaction.amount,
+        *period_coin_conds,
+        CoinTransaction.biz_type.notin_(REGULAR_TASK_EXCLUDED_BIZ_TYPES),
+    )
+    period_cps_orders = list(
+        db.execute(
+            select(CpsOrder).where(
+                CpsOrder.pay_time >= start_utc,
+                CpsOrder.pay_time < end_utc,
+            )
+        ).scalars()
+    )
+    period_meituan_orders = [
+        o for o in period_cps_orders if (o.platform or "meituan") == "meituan"
+    ]
+    period_meituan_valid_orders = [
+        o for o in period_meituan_orders if o.mt_status not in MEITUAN_CPS_INVALID_STATUSES
+    ]
+    period_jd_orders = [o for o in period_cps_orders if o.platform == "jd"]
+    period_jd_valid_orders = [o for o in period_jd_orders if _jd_valid_order(o)]
+    period_jd_invalid_orders = [
+        o for o in period_jd_orders if o.jd_valid_code and not _jd_valid_order(o)
+    ]
+    period_meituan_hit_count = 0
+    period_meituan_miss_count = 0
+    period_meituan_unknown_rate_count = 0
+    for order in period_meituan_valid_orders:
+        rate = _commission_rate_percent(order.commission_rate)
+        if rate is None:
+            period_meituan_unknown_rate_count += 1
+        elif rate < Decimal("1"):
+            period_meituan_miss_count += 1
+        else:
+            period_meituan_hit_count += 1
+    period_meituan_hit_denominator = period_meituan_hit_count + period_meituan_miss_count
+    period_meituan_hit_rate = (
+        round(period_meituan_hit_count / period_meituan_hit_denominator, 4)
+        if period_meituan_hit_denominator
+        else None
+    )
+
+    # ===== 领券核心数据(2026-07-05 产品新增)=====
+    # 数据源:coupon_session(一次领券一行,started_date 北京自然日)+ coupon_claim_record
+    # (一券/点位一天一条终态,claim_date 北京自然日)。点位与 session 不按 trace_id 关联——
+    # record_claims 更新路径不覆盖 trace_id(同设备同券同日重跑归第一次的 trace),按
+    # (device_id, 自然日) 桶关联才可靠;同桶多次发起共享同一份点位终态。
+    period_coupon_sessions = db.execute(
+        select(
+            CouponSession.device_id,
+            CouponSession.started_date,
+            CouponSession.status,
+            CouponSession.elapsed_ms,
+        ).where(
+            CouponSession.started_date >= period_from,
+            CouponSession.started_date <= period_to,
+            # 只统计正式环境,同「领券数据」页默认口径(防 debug 包调试数据串台;
+            # 命中 ix_coupon_session_date_env)。点位表无 app_env 列,但点位指标只经
+            # 下方 prod session 触达的 (device, 日) 桶进入统计,随之收敛到 prod。
+            CouponSession.app_env == "prod",
+        )
+    ).all()
+    coupon_started = len(period_coupon_sessions)
+    coupon_completed_elapsed = sorted(
+        s.elapsed_ms
+        for s in period_coupon_sessions
+        if s.status == "completed" and s.elapsed_ms is not None
+    )
+    # 点位桶:(device, 日) → (点位总数, 成功点位数)。成功口径与「我的」页累计领券一致
+    # (sum_claimed_count,2026-06-15 产品定):success + already_claimed(已领过=持有券)都算成功。
+    point_buckets: dict[tuple[str, date], tuple[int, int]] = {
+        (dev, d): (int(total), int(succ or 0))
+        for dev, d, total, succ in db.execute(
+            select(
+                CouponClaimRecord.device_id,
+                CouponClaimRecord.claim_date,
+                func.count(),
+                func.sum(
+                    case(
+                        (CouponClaimRecord.status.in_(("success", "already_claimed")), 1),
+                        else_=0,
+                    )
+                ),
+            )
+            .where(
+                CouponClaimRecord.claim_date >= period_from,
+                # 上界放宽一天:跨零点场次(23:5x 发起)的点位 claim_date 落在发起日+1,
+                # 桶只经下方 session 触达的键参与计数,放宽不会引入无关数据。
+                CouponClaimRecord.claim_date <= period_to + timedelta(days=1),
+            )
+            .group_by(CouponClaimRecord.device_id, CouponClaimRecord.claim_date)
+        ).all()
+    }
+    # 全部领成功的次数:completed 且其 (device, 日) 桶内点位全部成功(桶为空不算)。
+    coupon_all_success = 0
+    completed_bucket_totals: list[int] = []
+    session_bucket_keys: set[tuple[str, date]] = set()
+    for s in period_coupon_sessions:
+        key = (s.device_id, s.started_date)
+        if key not in point_buckets:
+            # 跨零点回退:发起日桶不存在(点位终态全部落在次日)时取 (device, 发起日+1)。
+            # 仅在发起日桶完全缺失时回退,避免抢占该设备次日 session 自己的桶。
+            next_key = (s.device_id, s.started_date + timedelta(days=1))
+            if next_key in point_buckets:
+                key = next_key
+        bucket = point_buckets.get(key)
+        if bucket is not None:
+            session_bucket_keys.add(key)
+        if s.status != "completed" or bucket is None:
+            continue
+        total, succ = bucket
+        completed_bucket_totals.append(total)
+        if total > 0 and succ == total:
+            coupon_all_success += 1
+    # 每次发起的应领点位数:取「完成过的领券」实际点位数的众数(done 帧会给所有点位终态,
+    # 完成场的点位数=当前配置的全量点位数;数据自校准,配置改点位数无需改代码)。本期无完成场
+    # 时给不出,点位成功率置空。
+    coupon_points_per_session = (
+        Counter(completed_bucket_totals).most_common(1)[0][0]
+        if completed_bucket_totals
+        else None
+    )
+    # 成功点位数:本期 session 触达过的 (device, 日) 桶内成功点位之和(桶级去重,同桶重试不重复计)。
+    coupon_point_success = sum(point_buckets[k][1] for k in session_bucket_keys)
+    # 点位成功率 = 成功点位数 / (发起数 × 应领点位数):中途退出未跑到的点位不产生记录,
+    # 但发起数×点位数把它们计入分母 → 视为失败,符合产品口径;重试会拉低该率(分母按次数计)。
+    coupon_point_success_rate = (
+        round(
+            min(1.0, coupon_point_success / (coupon_started * coupon_points_per_session)), 4
+        )
+        if coupon_started and coupon_points_per_session
+        else None
+    )
 
     return {
         "users": {
@@ -69,7 +535,7 @@ def dashboard_overview(db: Session) -> dict:
             "disabled": by_status.get("disabled", 0),
             "deleted": by_status.get("deleted", 0),
             "new_today": _count(User, User.created_at >= today_start),
-            "dau": _count(User, User.last_login_at >= today_start),
+            "dau": today_dau(db),
         },
         "coins": {
             # 累计发放金币(coin_transaction 里所有 amount>0 之和;负数是兑换/扣减不计)
@@ -119,7 +585,86 @@ def dashboard_overview(db: Session) -> dict:
             "success": comparison_success,
             "success_rate": success_rate,
         },
-        "feedback": {"new": _count(Feedback, Feedback.status.in_(("pending", "new")))},
-        # CPS 收入数据源未接(referral-link 只换链接,转化/佣金未回收)→ 前端显示"待接入"。
-        "cps": {"available": False, "note": "CPS 转化数据未接入(P2)"},
+        "period": {
+            "date_from": period_from,
+            "date_to": period_to,
+            "users": {
+                "new": len(period_new_user_ids),
+                "active": len(period_active_user_ids),
+                "retained_new_users": retention_retained_total,
+                "retention_cohort": retention_cohort_total,
+                "retention_rate": period_retention_rate,
+                "retention_note": (
+                    "次日留存:窗口内每天取前一日新增用户,统计其当日活跃"
+                    "(登录/开始比价/开始领券,按用户去重)比例,逐日累加;"
+                    "默认窗口=昨日,即前日新增用户的昨日留存"
+                ),
+            },
+            "comparison": {
+                "total": period_comparison_total,
+                "success": period_comparison_success,
+                "success_rate": period_comparison_success_rate,
+                "ordered": period_ordered_count,
+                "average_duration_ms": period_avg_duration_ms,
+                "average_saved_cents": period_avg_saved_cents,
+            },
+            "coupon": {
+                "started": coupon_started,
+                "all_success": coupon_all_success,
+                "success_rate": (
+                    round(coupon_all_success / coupon_started, 4) if coupon_started else None
+                ),
+                "point_success": coupon_point_success,
+                "points_per_session": coupon_points_per_session,
+                "point_success_rate": coupon_point_success_rate,
+                "median_elapsed_ms": _percentile(coupon_completed_elapsed, 50),
+            },
+            "coins": {
+                "granted_total": _sum(CoinTransaction.amount, *period_coin_conds),
+                "reward_video_coin_total": period_reward_video_coin_total,
+                "feed_ad_coin_total": period_feed_ad_coin_total,
+                "signin_coin_total": period_signin_coin_total,
+                "signin_boost_coin_total": period_signin_boost_coin_total,
+                "task_coin_total": period_task_coin_total,
+                "coupon_reward_coin_total": period_coupon_reward_coin_total,
+                "comparison_reward_coin_total": period_comparison_reward_coin_total,
+                "regular_task_coin_total": period_regular_task_coin_total,
+            },
+            "cash": {
+                "withdraw_success_cents": _sum(
+                    WithdrawOrder.amount_cents,
+                    WithdrawOrder.status == "success",
+                    WithdrawOrder.created_at >= start_local,
+                    WithdrawOrder.created_at < end_local,
+                ),
+            },
+            "trend": trend_points,
+        },
+        "feedback": {
+            "new": _count(Feedback, Feedback.status.in_(("pending", "new"))),
+        },
+        "cps": {
+            "available": True,
+            "note": "美团/JD CPS 读 cps_order 对账订单;淘宝佣金暂空",
+            "meituan_order_count": len(period_meituan_valid_orders),
+            "meituan_commission_cents": sum(
+                o.commission_cents or 0 for o in period_meituan_valid_orders
+            ),
+            "meituan_hit_count": period_meituan_hit_count,
+            "meituan_miss_count": period_meituan_miss_count,
+            "meituan_unknown_rate_count": period_meituan_unknown_rate_count,
+            "meituan_hit_rate": period_meituan_hit_rate,
+            "jd_order_count": len(period_jd_valid_orders),
+            # 数据大盘京东 CPS 只看实际佣金,不再用预估佣金兜底。
+            "jd_commission_cents": sum(
+                o.actual_commission_cents or 0 for o in period_jd_valid_orders
+            ),
+            "jd_actual_commission_cents": sum(
+                o.actual_commission_cents or 0 for o in period_jd_valid_orders
+            ),
+            "jd_estimated_commission_cents": sum(
+                o.estimated_commission_cents or 0 for o in period_jd_valid_orders
+            ),
+            "jd_invalid_count": len(period_jd_invalid_orders),
+        },
     }

app/admin/routers/ad_audit.py

@@ -26,7 +26,7 @@ def get_ad_coin_audit(
     date: Annotated[str | None, Query(description="北京时间 YYYY-MM-DD,默认今天")] = None,
     user_id: Annotated[int | None, Query(description="只看某用户;不传=全部用户")] = None,
     scene: Annotated[
-        str | None, Query(description="reward_video / feed;不传=两类都要")
+        str | None, Query(description="reward_video / feed / draw;不传=全部")
     ] = None,
     limit: Annotated[int, Query(ge=1, le=500)] = 100,
     only_mismatch: Annotated[

app/admin/routers/ad_revenue.py

@@ -11,7 +11,13 @@ from fastapi import APIRouter, Depends, HTTPException, Query
 
 from app.admin.deps import AdminDb, get_current_admin
 from app.admin.repositories import ad_revenue
-from app.admin.schemas.ad_revenue import AdRevenueDaily, AdRevenueReportOut, AdRevenueRow
+from app.admin.schemas.ad_revenue import (
+    AdRevenueDaily,
+    AdRevenueHourly,
+    AdRevenueReportOut,
+    AdRevenueRow,
+    AdRevenueTypeStat,
+)
 from app.core.rewards import cn_today
 
 router = APIRouter(
@@ -43,10 +49,28 @@ def get_ad_revenue_report(
         str | None,
         Query(description="reward_video / feed / draw;不传=全部类型"),
     ] = None,
+    feed_scene: Annotated[
+        str | None,
+        Query(
+            description="comparison(比价) / coupon(领券) / welfare(福利);不传=全部场景。"
+            "全局筛选,同时影响明细 / 合计 / 趋势"
+        ),
+    ] = None,
+    app_env: Annotated[
+        str | None,
+        Query(
+            description="prod(正式应用) / test(测试应用);不传=全部。"
+            "建议正式收益报表选 prod,避免测试应用的假 eCPM 污染收益合计/平均"
+        ),
+    ] = None,
     granularity: Annotated[
         str, Query(description="day=按天 / hour=按小时(北京时间);区间>1 天建议用 day")
     ] = "day",
-    limit: Annotated[int, Query(ge=1, le=1000)] = 500,
+    limit: Annotated[int, Query(ge=1, le=1000, description="每页条数(分页大小)")] = 500,
+    offset: Annotated[int, Query(ge=0, description="分页偏移(已跳过的条数)=(页码-1)×每页条数")] = 0,
+    sort: Annotated[
+        str, Query(description="排序:time=时间倒序(默认) / ecpm=按 eCPM 数值倒序")
+    ] = "time",
 ) -> AdRevenueReportOut:
     today = cn_today()
     d_from = _parse_day(date_from, field="date_from", default=today)
@@ -58,16 +82,24 @@ def get_ad_revenue_report(
 
     result = ad_revenue.ad_revenue_report(
         db, date_from=d_from.isoformat(), date_to=d_to.isoformat(),
-        user_id=user_id, ad_type=ad_type, granularity=granularity, limit=limit,
+        user_id=user_id, ad_type=ad_type, feed_scene=feed_scene, app_env=app_env,
+        granularity=granularity, limit=limit, offset=offset, sort=sort,
     )
     return AdRevenueReportOut(
         date_from=d_from.isoformat(),
         date_to=d_to.isoformat(),
         daily=[AdRevenueDaily(**d) for d in result["daily"]],
+        hourly=[AdRevenueHourly(**h) for h in result["hourly"]],
+        type_stats={k: AdRevenueTypeStat(**v) for k, v in result["type_stats"].items()},
+        scene_stats={k: AdRevenueTypeStat(**v) for k, v in result["scene_stats"].items()},
+        dau=result["dau"],
         total=result["total"],
         truncated=result["truncated"],
         total_impressions=result["total_impressions"],
         total_revenue_yuan=result["total_revenue_yuan"],
+        total_pangle_revenue_yuan=result["total_pangle_revenue_yuan"],
+        total_pangle_api_revenue_yuan=result["total_pangle_api_revenue_yuan"],
+        pangle_revenue_available=result["pangle_revenue_available"],
         total_expected_coin=result["total_expected_coin"],
         total_actual_coin=result["total_actual_coin"],
         mismatch_count=result["mismatch_count"],

app/admin/routers/admins.py

@@ -5,6 +5,8 @@ from fastapi import APIRouter, Depends, HTTPException, Request
 
 from app.admin.audit import write_audit
 from app.admin.deps import AdminDb, CurrentAdmin, get_client_ip, require_role
+from app.admin.permissions import CUSTOM_ROLE, SUPER_ADMIN_ROLE, sanitize_pages
+from app.admin.repositories import admin_role as role_repo
 from app.admin.repositories import admin_user as admin_repo
 from app.admin.schemas.admin import AdminCreateRequest, AdminUpdateRequest
 from app.admin.schemas.auth import AdminOut
@@ -23,9 +25,31 @@ def _active_super_count(db: AdminDb) -> int:
     )
 
 
-@router.get("", response_model=list[AdminOut], summary="管理员列表")
+def _validate_role(db: AdminDb, role: str) -> None:
+    """角色必须是 super_admin / custom(自定义) / admin_role 表里已存在的角色,否则 400。"""
+    if role in (SUPER_ADMIN_ROLE, CUSTOM_ROLE):
+        return
+    role_repo.ensure_builtin_roles(db)  # 空表(测试/全新库)兜底播种,再校验
+    if role_repo.get_role(db, role) is None:
+        raise HTTPException(status_code=400, detail=f"角色不存在: {role}")
+
+
+def _clean_override(role: str, pages_override: list[str] | None) -> list[str] | None:
+    """按最终角色算出该存的 pages_override:custom → 勾选集(过滤非法 key,可空列表);
+    非 custom → None(切回普通角色即清空自定义页)。"""
+    if role == CUSTOM_ROLE:
+        return sanitize_pages(pages_override)
+    return None
+
+
+@router.get("", response_model=list[AdminOut], summary="管理员列表(含明文密码,super_admin 专属)")
 def list_admins(db: AdminDb) -> list[AdminOut]:
-    return [AdminOut.model_validate(a) for a in admin_repo.list_admins(db)]
+    out: list[AdminOut] = []
+    for a in admin_repo.list_admins(db):
+        item = AdminOut.model_validate(a)
+        item.password = a.plain_password  # 明文:仅本 super_admin 专属路由下发,供权限管理页复看
+        out.append(item)
+    return out
 
 
 @router.post("", response_model=AdminOut, summary="创建管理员")
@@ -34,12 +58,17 @@ def create_admin(
 ) -> AdminOut:
     if admin_repo.get_by_username(db, body.username) is not None:
         raise HTTPException(status_code=409, detail="用户名已存在")
+    _validate_role(db, body.role)
+    override = _clean_override(body.role, body.pages_override)
     new = admin_repo.create_admin(
-        db, username=body.username, password=body.password, role=body.role
+        db, username=body.username, password=body.password, role=body.role,
+        plain_password=body.password,  # UI 建的账号留存明文,供权限管理页复看
+        pages_override=override,
     )
     write_audit(
         db, admin, action="admin.create", target_type="admin", target_id=new.id,
-        detail={"username": new.username, "role": new.role}, ip=get_client_ip(request), commit=True,
+        detail={"username": new.username, "role": new.role, "pages_override": override},
+        ip=get_client_ip(request), commit=True,
     )
     return AdminOut.model_validate(new)
 
@@ -67,8 +96,12 @@ def update_admin(
     if demotes_super and _active_super_count(db) <= 1:
         raise HTTPException(status_code=400, detail="不能降级/禁用最后一个超级管理员")
 
+    if body.role is not None:
+        _validate_role(db, body.role)
+
     changes: dict = {}
-    if body.role is not None and body.role != target.role:
+    role_changed = body.role is not None and body.role != target.role
+    if role_changed:
         changes["role"] = {"before": target.role, "after": body.role}
         target.role = body.role
     if body.status is not None and body.status != target.status:
@@ -77,6 +110,18 @@ def update_admin(
     if body.password is not None:
         changes["password"] = "reset"
         target.password_hash = hash_password(body.password)
+        target.plain_password = body.password  # 同步留存明文,权限管理页复看保持一致
+
+    # 自定义可见页:按「最终角色」(target.role,已应用完角色变更)决定该存什么。
+    # - 切到/维持 custom 且传了 pages_override → 用勾选集;切到 custom 没传 → 清成空列表。
+    # - 切回普通角色 → 清空 override(_clean_override 返回 None)。
+    # - 角色没变、只传 pages_override(编辑现有 custom 用户的勾选)→ 也更新。
+    if role_changed or body.pages_override is not None:
+        new_override = _clean_override(target.role, body.pages_override)
+        if new_override != target.pages_override:
+            changes["pages_override"] = {"before": target.pages_override, "after": new_override}
+            target.pages_override = new_override
+
     if not changes:
         raise HTTPException(status_code=400, detail="无任何变更字段")
     db.commit()
@@ -86,3 +131,27 @@ def update_admin(
         detail=changes, ip=get_client_ip(request), commit=True,
     )
     return AdminOut.model_validate(target)
+
+
+@router.delete("/{admin_id}", summary="删除管理员(带审计)")
+def delete_admin(admin_id: int, request: Request, admin: CurrentAdmin, db: AdminDb) -> dict:
+    target = admin_repo.get_by_id(db, admin_id)
+    if target is None:
+        raise HTTPException(status_code=404, detail="管理员不存在")
+    if admin_id == admin.id:
+        raise HTTPException(status_code=400, detail="不能删除自己")
+    # 防自锁:删掉某个 active super_admin 前,确认后仍至少剩 1 个,否则进「零可用超管」死局。
+    if (
+        target.role == "super_admin"
+        and target.status == "active"
+        and _active_super_count(db) <= 1
+    ):
+        raise HTTPException(status_code=400, detail="不能删除最后一个超级管理员")
+    username = target.username
+    db.delete(target)
+    db.commit()
+    write_audit(
+        db, admin, action="admin.delete", target_type="admin", target_id=admin_id,
+        detail={"username": username, "role": target.role}, ip=get_client_ip(request), commit=True,
+    )
+    return {"deleted": True}

app/admin/routers/auth.py

@@ -6,6 +6,8 @@ import logging
 from fastapi import APIRouter, Depends, HTTPException
 
 from app.admin.deps import AdminDb, CurrentAdmin
+from app.admin.permissions import CUSTOM_ROLE, sanitize_pages
+from app.admin.repositories import admin_role as role_repo
 from app.admin.repositories import admin_user as admin_repo
 from app.admin.schemas.auth import AdminLoginRequest, AdminLoginResponse, AdminOut
 from app.admin.security import create_admin_token
@@ -17,6 +19,17 @@ logger = logging.getLogger("shagua.admin.auth")
 router = APIRouter(prefix="/admin/api/auth", tags=["admin-auth"])
 
 
+def _admin_out_with_pages(admin, db: AdminDb) -> AdminOut:  # noqa: ANN001
+    """AdminOut + 有效可见页(前端左侧导航按此过滤)。
+    role=="custom" → 用这个人的 pages_override(按人自定义,过滤悬空 key);其余走角色解析。"""
+    out = AdminOut.model_validate(admin)
+    if admin.role == CUSTOM_ROLE:
+        out.pages = sanitize_pages(admin.pages_override)
+    else:
+        out.pages = role_repo.effective_pages_of(db, admin.role)
+    return out
+
+
 @router.post(
     "/login",
     response_model=AdminLoginResponse,
@@ -39,10 +52,10 @@ def login(req: AdminLoginRequest, db: AdminDb) -> AdminLoginResponse:
     return AdminLoginResponse(
         access_token=token,
         expires_in=expires_in,
-        admin=AdminOut.model_validate(admin),
+        admin=_admin_out_with_pages(admin, db),
     )
 
 
-@router.get("/me", response_model=AdminOut, summary="当前管理员")
-def me(admin: CurrentAdmin) -> AdminOut:
-    return AdminOut.model_validate(admin)
+@router.get("/me", response_model=AdminOut, summary="当前管理员(含有效可见页)")
+def me(admin: CurrentAdmin, db: AdminDb) -> AdminOut:
+    return _admin_out_with_pages(admin, db)

app/admin/routers/comparison.py

@@ -32,12 +32,15 @@ def list_comparison_records(
     phone: Annotated[str | None, Query(description="手机号前缀")] = None,
     status: Annotated[str | None, Query(pattern="^(success|failed|cancelled)$")] = None,
     business_type: Annotated[str | None, Query()] = None,
+    store: Annotated[str | None, Query(description="店名子串模糊匹配")] = None,
+    product: Annotated[str | None, Query(description="商品名子串模糊匹配")] = None,
     limit: Annotated[int, Query(ge=1, le=100)] = 20,
     cursor: Annotated[int | None, Query()] = None,
 ) -> CursorPage[AdminComparisonListItem]:
     items, next_cursor, total = queries.list_comparison_records(
         db, user_id=user_id, phone=phone, status=status,
-        business_type=business_type, limit=limit, cursor=cursor,
+        business_type=business_type, store=store, product=product,
+        limit=limit, cursor=cursor,
     )
     return CursorPage(
         items=[AdminComparisonListItem.model_validate(r) for r in items],

app/admin/routers/config.py

@@ -55,9 +55,14 @@ def _item(db, key: str) -> ConfigItemOut:
     raise HTTPException(status_code=404, detail="未知配置项")
 
 
-@router.get("", response_model=list[ConfigItemOut], summary="所有可配项 + 当前值")
+@router.get("", response_model=list[ConfigItemOut], summary="所有可配项 + 当前值(不含 hidden)")
 def list_config(db: AdminDb) -> list[ConfigItemOut]:
-    return [ConfigItemOut(**item) for item in app_config.list_all(db)]
+    # hidden 项(已下线/由专用页管理,如福利页任务·里程碑·看广告调参、首页轮播数据源)不在本页渲染。
+    return [
+        ConfigItemOut(**item)
+        for item in app_config.list_all(db)
+        if not CONFIG_DEFS[item["key"]].get("hidden")
+    ]
 
 
 @router.patch("/{key}", response_model=ConfigItemOut, summary="改某项配置(带审计)")

app/admin/routers/coupon_data.py

@@ -0,0 +1,106 @@
+"""admin「领券数据」看板:发起/完成数 + 领券耗时(均值 + P5/P50/P95/P99)+ 按天趋势 + 逐条明细。
+
+任意已登录 admin 可看(只读)。聚合逻辑在 app/admin/repositories/coupon_data.py。
+数据源 coupon_session(客户端 /api/v1/coupon/session 两段上报)。
+"""
+from __future__ import annotations
+
+from datetime import date as _date
+from typing import Annotated
+
+from fastapi import APIRouter, Depends, HTTPException, Query
+
+from app.admin.deps import AdminDb, get_current_admin
+from app.admin.repositories import coupon_data
+from app.admin.schemas.coupon_data import (
+    CouponDataDaily,
+    CouponDataHourly,
+    CouponDataOut,
+    CouponDataRow,
+    CouponDataSummary,
+    CouponUserRecordsOut,
+)
+from app.core.rewards import cn_today
+
+router = APIRouter(
+    prefix="/admin/api/coupon-data",
+    tags=["admin-coupon-data"],
+    dependencies=[Depends(get_current_admin)],
+)
+
+# 区间最大跨度(天);超出拒绝,避免一次拉过多天拖垮接口(对齐广告收益报表)。
+_MAX_RANGE_DAYS = 92
+
+
+def _parse_day(value: str | None, *, field: str, default: _date) -> _date:
+    if value is None:
+        return default
+    try:
+        return _date.fromisoformat(value)
+    except ValueError as e:
+        raise HTTPException(status_code=422, detail=f"{field} 需为 YYYY-MM-DD") from e
+
+
+@router.get(
+    "",
+    response_model=CouponDataOut,
+    summary="领券数据看板(发起/完成数 + 耗时分位 + 按天趋势 + 逐条明细)",
+)
+def get_coupon_data(
+    db: AdminDb,
+    date_from: Annotated[str | None, Query(description="起始日 北京 YYYY-MM-DD,默认今天")] = None,
+    date_to: Annotated[str | None, Query(description="结束日 北京 YYYY-MM-DD,闭区间,默认=date_from")] = None,
+    user: Annotated[str | None, Query(description="用户手机号/昵称模糊搜;不传=全部")] = None,
+    app_env: Annotated[str, Query(description="prod(默认) / dev / all(全部环境)")] = "prod",
+    granularity: Annotated[
+        str, Query(description="day=按天 / hour=按小时(北京);区间>1 天建议 day")
+    ] = "day",
+    limit: Annotated[int, Query(ge=1, le=1000, description="每页条数(分页大小)")] = 500,
+    offset: Annotated[int, Query(ge=0, description="分页偏移(已跳过条数)=(页码-1)×每页条数")] = 0,
+    sort: Annotated[
+        str, Query(description="排序:time=发起时间倒序(默认) / elapsed=耗时倒序")
+    ] = "time",
+) -> CouponDataOut:
+    today = cn_today()
+    d_from = _parse_day(date_from, field="date_from", default=today)
+    d_to = _parse_day(date_to, field="date_to", default=d_from)
+    if d_to < d_from:
+        raise HTTPException(status_code=422, detail="date_to 不能早于 date_from")
+    if (d_to - d_from).days + 1 > _MAX_RANGE_DAYS:
+        raise HTTPException(status_code=422, detail=f"区间最长 {_MAX_RANGE_DAYS} 天")
+
+    # 报表默认只看 prod(对齐广告报表防串台口径);app_env=all 时不过滤、看全部环境。
+    env = None if app_env == "all" else app_env
+    result = coupon_data.coupon_data_report(
+        db, date_from=d_from.isoformat(), date_to=d_to.isoformat(),
+        user=user, app_env=env, granularity=granularity,
+        limit=limit, offset=offset, sort=sort,
+    )
+    return CouponDataOut(
+        date_from=d_from.isoformat(),
+        date_to=d_to.isoformat(),
+        summary=CouponDataSummary(**result["summary"]),
+        daily=[CouponDataDaily(**d) for d in result["daily"]],
+        hourly=[CouponDataHourly(**h) for h in result["hourly"]],
+        total=result["total"],
+        items=[CouponDataRow(**r) for r in result["items"]],
+    )
+
+
+@router.get(
+    "/user-records",
+    response_model=CouponUserRecordsOut,
+    summary="某用户全部领券记录(点手机号抽屉:领券次数 + 记录列表)",
+)
+def get_user_coupon_records(
+    db: AdminDb,
+    user_id: Annotated[int, Query(description="用户 id")],
+    limit: Annotated[int, Query(ge=1, le=500, description="最多返回条数")] = 100,
+    sort_by: Annotated[str, Query(description="兼容 UserRecordsDrawer 参数;固定按发起时间倒序")] = "created_at",
+    sort_order: Annotated[str, Query(description="兼容参数,忽略")] = "desc",
+) -> CouponUserRecordsOut:
+    result = coupon_data.coupon_user_records(db, user_id=user_id, limit=limit)
+    return CouponUserRecordsOut(
+        items=[CouponDataRow(**r) for r in result["items"]],
+        total=result["total"],
+    )

app/admin/routers/cps.py

@@ -1,13 +1,13 @@
-"""admin CPS 分发与对账:群/活动管理 + 生成落地页短链 + 美团订单对账 + 统计。
+"""admin CPS 分发与对账:群/活动管理 + 生成落地页短链 + 联盟订单对账 + 统计。
 
 平台:meituan(actId+sid 转链 + query_order 对账) / taobao(整段淘口令) / jd(链接)。
-淘宝/京东无 API → 只统计点击(咱落地页 PV/UV + 淘宝复制),对账字段显示 "-"。
+淘宝暂未接 API → 只统计点击(咱落地页 PV/UV + 淘宝复制),对账字段显示 "-"。
 群/活动管理 = operator;订单对账(涉佣金) = finance;只读列表/统计 = 登录即可。
 """
 from __future__ import annotations
 
 import time
-from datetime import datetime, timedelta, timezone
+from datetime import date as _date, datetime, time as _dt_time, timedelta, timezone
 from typing import Annotated
 
 from fastapi import APIRouter, Depends, File, HTTPException, Query, Request, UploadFile
@@ -34,6 +34,7 @@ from app.admin.schemas.cps import (
 from app.core import media
 from app.core.config import settings
 from app.integrations import meituan
+from app.integrations.jd_union import JdUnionError
 from app.integrations.meituan import MeituanCpsError
 from app.models.admin import AdminUser
 from app.models.cps_activity import CpsActivity
@@ -340,24 +341,108 @@ def generate_referral_links(
 
 
 # ───────────── 订单对账 ─────────────
-@router.post("/orders/reconcile", response_model=CpsReconcileResult, summary="拉取美团订单对账")
+_BEIJING = timezone(timedelta(hours=8))
+
+
+def _parse_day(value: str | None, *, field: str) -> _date | None:
+    if value is None:
+        return None
+    try:
+        return _date.fromisoformat(value)
+    except ValueError as e:
+        raise HTTPException(status_code=422, detail=f"{field} 需为 YYYY-MM-DD") from e
+
+
+def _reconcile_range_to_ts(
+    date_from: _date | None, date_to: _date | None, days: int
+) -> tuple[int, int]:
+    if date_from is None and date_to is None:
+        now = int(time.time())
+        return now - days * 86400, now
+
+    start_day = date_from or date_to
+    end_day = date_to or date_from
+    if start_day is None or end_day is None:
+        raise HTTPException(status_code=422, detail="日期参数不完整")
+    if start_day > end_day:
+        start_day, end_day = end_day, start_day
+    if (end_day - start_day).days + 1 > 90:
+        raise HTTPException(status_code=422, detail="美团订单查询最长 90 天")
+
+    start_dt = datetime.combine(start_day, _dt_time.min, tzinfo=_BEIJING)
+    end_dt = datetime.combine(end_day + timedelta(days=1), _dt_time.min, tzinfo=_BEIJING)
+    return int(start_dt.timestamp()), int(end_dt.timestamp())
+
+
+def _reconcile_range_to_bj_dt(
+    date_from: _date | None, date_to: _date | None, days: int
+) -> tuple[datetime, datetime]:
+    start_ts, end_ts = _reconcile_range_to_ts(date_from, date_to, days)
+    return (
+        datetime.fromtimestamp(start_ts, tz=_BEIJING),
+        datetime.fromtimestamp(end_ts, tz=_BEIJING),
+    )
+
+
+def _merge_reconcile_result(total: dict, current: dict) -> None:
+    for key in ("fetched", "inserted", "updated", "pages"):
+        total[key] = int(total.get(key, 0)) + int(current.get(key, 0))
+
+
+@router.post("/orders/reconcile", response_model=CpsReconcileResult, summary="拉取联盟订单对账")
 def reconcile_orders(
     request: Request,
     admin: Annotated[AdminUser, Depends(require_role("finance"))],
     db: AdminDb,
-    days: Annotated[int, Query(ge=1, le=90)] = 7,
+    date_from: Annotated[str | None, Query(description="起始日 YYYY-MM-DD")] = None,
+    date_to: Annotated[str | None, Query(description="结束日 YYYY-MM-DD")] = None,
+    days: Annotated[int, Query(ge=1, le=90, description="未传日期时默认拉近 N 天")] = 7,
     sid: Annotated[str | None, Query(max_length=64)] = None,
+    query_time_type: Annotated[int, Query(ge=1, le=3)] = 1,
+    platform: Annotated[str, Query(pattern="^(all|meituan|jd)$")] = "all",
 ) -> CpsReconcileResult:
-    now = int(time.time())
+    parsed_from = _parse_day(date_from, field="date_from")
+    parsed_to = _parse_day(date_to, field="date_to")
+    result = {"fetched": 0, "inserted": 0, "updated": 0, "pages": 0}
     try:
-        result = cps_repo.reconcile_orders(
-            db, start_time=now - days * 86400, end_time=now, sid=sid,
-        )
+        if platform in {"all", "meituan"}:
+            start_ts, end_ts = _reconcile_range_to_ts(parsed_from, parsed_to, days)
+            mt_result = cps_repo.reconcile_orders(
+                db,
+                start_time=start_ts,
+                end_time=end_ts,
+                query_time_type=query_time_type if query_time_type in (1, 2) else 2,
+                sid=sid,
+            )
+            _merge_reconcile_result(result, mt_result)
+        if platform in {"all", "jd"}:
+            if sid:
+                raise HTTPException(status_code=422, detail="京东订单刷新不支持 sid 筛选")
+            start_dt, end_dt = _reconcile_range_to_bj_dt(parsed_from, parsed_to, days)
+            jd_result = cps_repo.reconcile_jd_orders(
+                db,
+                start_time=start_dt,
+                end_time=end_dt,
+                query_time_type=query_time_type,
+            )
+            _merge_reconcile_result(result, jd_result)
     except MeituanCpsError as e:
         raise HTTPException(status_code=502, detail=f"美团拉单失败: {e}") from e
+    except JdUnionError as e:
+        raise HTTPException(status_code=502, detail=f"京东拉单失败: {e}") from e
     write_audit(
         db, admin, action="cps.order.reconcile", target_type="cps_order", target_id=None,
-        detail={"days": days, "sid": sid, **result}, ip=get_client_ip(request), commit=True,
+        detail={
+            "platform": platform,
+            "date_from": date_from,
+            "date_to": date_to,
+            "days": days,
+            "sid": sid,
+            "query_time_type": query_time_type,
+            **result,
+        },
+        ip=get_client_ip(request),
+        commit=True,
     )
     return CpsReconcileResult(**result)
 
@@ -468,7 +553,7 @@ def group_daily(
     while cur <= last:
         cp = click_points[idx] if idx < len(click_points) else None
         row = {
-            "date": cur.strftime("%m-%d"),
+            "date": cur.strftime("%Y-%m-%d"),
             "click_pv": cp["click_pv"] if cp else 0,
             "click_uv": cp["click_uv"] if cp else 0,
             "copy_pv": cp["copy_pv"] if cp else 0,
@@ -496,3 +581,28 @@ def group_wx_users(group_id: int, db: AdminDb) -> dict:
     if group is None:
         raise HTTPException(status_code=404, detail="群不存在")
     return {"users": cps_repo.group_wx_users(db, group_id=group_id)}
+
+
+@router.get("/groups/{group_id}/day-users", summary="某天该群按用户的领券/点击 + 每人点过的券")
+def group_day_users(
+    group_id: int,
+    db: AdminDb,
+    date: Annotated[str, Query(description="北京日期 YYYY-MM-DD")],
+) -> dict:
+    group = cps_repo.get_group(db, group_id)
+    if group is None:
+        raise HTTPException(status_code=404, detail="群不存在")
+    bj = timezone(timedelta(hours=8))
+    try:
+        day0 = datetime.strptime(date, "%Y-%m-%d").replace(tzinfo=bj)
+    except ValueError as e:
+        raise HTTPException(status_code=400, detail="date 格式应为 YYYY-MM-DD") from e
+    start = day0.replace(hour=0, minute=0, second=0, microsecond=0)
+    end = start + timedelta(days=1)
+    users = cps_repo.group_day_users(db, group_id=group_id, start=start, end=end)
+    return {
+        "group_id": group.id,
+        "group_name": group.name,
+        "date": date,
+        "users": users,
+    }

app/admin/routers/dashboard.py

@@ -1,7 +1,9 @@
 """admin 数据大盘(只读聚合)。"""
 from __future__ import annotations
 
-from fastapi import APIRouter, Depends
+from datetime import date
+
+from fastapi import APIRouter, Depends, Query
 
 from app.admin.deps import AdminDb, get_current_admin
 from app.admin.repositories import stats
@@ -15,5 +17,11 @@ router = APIRouter(
 
 
 @router.get("/overview", response_model=DashboardOverview, summary="大盘核心指标")
-def overview(db: AdminDb) -> DashboardOverview:
-    return DashboardOverview.model_validate(stats.dashboard_overview(db))
+def overview(
+    db: AdminDb,
+    date_from: date | None = Query(None, description="北京时间自然日起始日 YYYY-MM-DD"),
+    date_to: date | None = Query(None, description="北京时间自然日结束日 YYYY-MM-DD"),
+) -> DashboardOverview:
+    return DashboardOverview.model_validate(
+        stats.dashboard_overview(db, date_from=date_from, date_to=date_to)
+    )

app/admin/routers/device_liveness.py

@@ -0,0 +1,63 @@
+"""admin 设备存活监控(只读):列设备心跳/在线掉线 + 顶部统计卡片。
+
+数据源 device_liveness 表(心跳 last_heartbeat_at + liveness_state + kill_alert_pending,
+见 app/models/device.py)。在线/掉线、掉线时长由 repo 按 HEARTBEAT_TIMEOUT_MINUTES 阈值派生。
+纯读:无写、无审计。任意登录管理员可看(同大盘/设备管理,无角色门)。
+"""
+from __future__ import annotations
+
+from typing import Annotated
+
+from fastapi import APIRouter, Depends, Query
+
+from app.admin.deps import AdminDb, get_current_admin
+from app.admin.repositories import queries
+from app.admin.schemas.common import CursorPage
+from app.admin.schemas.device import DeviceLivenessItem, DeviceLivenessStats
+
+router = APIRouter(
+    prefix="/admin/api/device-liveness",
+    tags=["admin-device-liveness"],
+    dependencies=[Depends(get_current_admin)],
+)
+
+
+@router.get("/stats", response_model=DeviceLivenessStats, summary="设备存活统计(顶部卡片)")
+def device_stats(db: AdminDb) -> DeviceLivenessStats:
+    return DeviceLivenessStats(**queries.device_liveness_stats(db))
+
+
+@router.get(
+    "",
+    response_model=CursorPage[DeviceLivenessItem],
+    summary="设备存活列表(在线情况/设备id/归属用户 筛选 + 排序 + 分页,默认掉线置顶)",
+)
+def list_devices(
+    db: AdminDb,
+    status: Annotated[str | None, Query(pattern="^(online|offline|never)$")] = None,
+    device_id: Annotated[str | None, Query(max_length=128)] = None,
+    phone: Annotated[str | None, Query(max_length=20)] = None,
+    user_id: Annotated[int | None, Query()] = None,
+    sort_by: Annotated[
+        str, Query(pattern="^(status|last_heartbeat_at|created_at)$")
+    ] = "status",
+    sort_order: Annotated[str, Query(pattern="^(asc|desc)$")] = "desc",
+    limit: Annotated[int, Query(ge=1, le=100)] = 20,
+    cursor: Annotated[int | None, Query()] = None,
+) -> CursorPage[DeviceLivenessItem]:
+    items, next_cursor, total = queries.list_device_liveness(
+        db,
+        status=status,
+        device_id=device_id,
+        phone=phone,
+        user_id=user_id,
+        sort_by=sort_by,
+        sort_order=sort_order,
+        limit=limit,
+        cursor=cursor,
+    )
+    return CursorPage(
+        items=[DeviceLivenessItem.model_validate(d) for d in items],
+        next_cursor=next_cursor,
+        total=total,
+    )

app/admin/routers/event_logs.py

@@ -0,0 +1,52 @@
+"""admin 埋点日志:列表 + 按事件 / 设备 / 用户 / 会话 / 时间筛选(只读,同库直接查 analytics_event)。"""
+from __future__ import annotations
+
+from datetime import datetime
+from typing import Annotated
+
+from fastapi import APIRouter, Depends, Query
+
+from app.admin.deps import AdminDb, get_current_admin
+from app.admin.repositories import queries
+from app.admin.schemas.analytics import AnalyticsEventOut
+from app.admin.schemas.common import CursorPage
+
+router = APIRouter(
+    prefix="/admin/api/event-logs",
+    tags=["admin-event-logs"],
+    dependencies=[Depends(get_current_admin)],
+)
+
+
+@router.get("", response_model=CursorPage[AnalyticsEventOut], summary="埋点日志列表")
+def list_event_logs(
+    db: AdminDb,
+    event: Annotated[str | None, Query(max_length=64)] = None,
+    device_id: Annotated[str | None, Query(max_length=64)] = None,
+    user_id: Annotated[int | None, Query()] = None,
+    session_id: Annotated[str | None, Query(max_length=64)] = None,
+    created_from: Annotated[datetime | None, Query()] = None,
+    created_to: Annotated[datetime | None, Query()] = None,
+    sort_by: Annotated[str, Query(pattern="^(id|created_at)$")] = "id",
+    sort_order: Annotated[str, Query(pattern="^(asc|desc)$")] = "desc",
+    limit: Annotated[int, Query(ge=1, le=100)] = 20,
+    cursor: Annotated[int | None, Query()] = None,
+) -> CursorPage[AnalyticsEventOut]:
+    items, next_cursor, total = queries.list_analytics_events(
+        db,
+        event=event,
+        device_id=device_id,
+        user_id=user_id,
+        session_id=session_id,
+        created_from=created_from,
+        created_to=created_to,
+        sort_by=sort_by,
+        sort_order=sort_order,
+        limit=limit,
+        cursor=cursor,
+    )
+    return CursorPage(
+        items=[AnalyticsEventOut.model_validate(e) for e in items],
+        next_cursor=next_cursor,
+        total=total,
+    )

app/admin/routers/feedback.py

@@ -10,7 +10,12 @@ from app.admin.audit import write_audit
 from app.admin.deps import AdminDb, get_client_ip, get_current_admin, require_role
 from app.admin.repositories import mutations, queries
 from app.admin.schemas.common import CursorPage, OkResponse
-from app.admin.schemas.feedback import FeedbackApproveRequest, FeedbackOut, FeedbackRejectRequest
+from app.admin.schemas.feedback import (
+    FeedbackApproveRequest,
+    FeedbackOut,
+    FeedbackRejectRequest,
+    FeedbackSummary,
+)
 from app.models.admin import AdminUser
 from app.models.feedback import Feedback
 from app.repositories import wallet as wallet_repo
@@ -31,6 +36,8 @@ def _ensure_pending(fb: Feedback) -> None:
 def list_feedbacks(
     db: AdminDb,
     status: Annotated[str | None, Query()] = None,
+    # 反馈类型筛选:profile(普通反馈)/ comparison(比价反馈);None=全部
+    source: Annotated[str | None, Query(pattern="^(profile|comparison)$")] = None,
     user_id: Annotated[int | None, Query()] = None,
     content: Annotated[str | None, Query(max_length=100)] = None,
     created_from: Annotated[datetime | None, Query()] = None,
@@ -43,6 +50,7 @@ def list_feedbacks(
     items, next_cursor, total = queries.list_feedbacks(
         db,
         status=status,
+        source=source,
         user_id=user_id,
         content=content,
         created_from=created_from,
@@ -59,6 +67,11 @@ def list_feedbacks(
     )
 
 
+@router.get("/summary", response_model=FeedbackSummary, summary="反馈审核统计(各状态计数)")
+def feedback_summary(db: AdminDb) -> FeedbackSummary:
+    return FeedbackSummary.model_validate(queries.feedback_summary(db))
+
+
 @router.post("/{feedback_id}/handle", response_model=OkResponse, summary="标记反馈已处理")
 def handle_feedback(
     feedback_id: int,
@@ -91,6 +104,7 @@ def approve_feedback(
         status="adopted",
         reward_coins=payload.reward_coins,
         review_note=payload.note,
+        admin_reply=payload.reply,
         reviewed_by_admin_id=admin.id,
         commit=False,
     )
@@ -113,6 +127,7 @@ def approve_feedback(
             "after": "adopted",
             "reward_coins": payload.reward_coins,
             "note": payload.note,
+            "reply": payload.reply,
         },
         ip=get_client_ip(request),
         commit=False,
@@ -142,6 +157,7 @@ def reject_feedback(
         status="rejected",
         reject_reason=payload.reason,
         review_note=payload.note,
+        admin_reply=payload.reply,
         reviewed_by_admin_id=admin.id,
         commit=False,
     )
@@ -156,6 +172,7 @@ def reject_feedback(
             "after": "rejected",
             "reason": payload.reason,
             "note": payload.note,
+            "reply": payload.reply,
         },
         ip=get_client_ip(request),
         commit=False,

app/admin/routers/ops_marquee_seed.py

@@ -8,6 +8,7 @@ from __future__ import annotations
 from typing import Annotated
 
 from fastapi import APIRouter, Depends, HTTPException, Query, Request
+from pydantic import BaseModel
 
 from app.admin.audit import write_audit
 from app.admin.deps import AdminDb, get_client_ip, get_current_admin, require_role
@@ -22,7 +23,11 @@ from app.admin.schemas.ops_marquee_seed import (
 )
 from app.models.admin import AdminUser
 from app.models.ops_marquee_seed import OpsMarqueeSeed
-from app.repositories import ops_marquee
+from app.repositories import app_config, ops_marquee
+
+
+class MarqueeModeUpdate(BaseModel):
+    mode: str  # mixed / real / seed
 
 router = APIRouter(
     prefix="/admin/api/marquee-seeds",
@@ -52,9 +57,42 @@ def list_seeds(db: AdminDb) -> list[OpsMarqueeSeedOut]:
 def preview_feed(
     db: AdminDb,
     limit: Annotated[int, Query(ge=1, le=30)] = 8,
+    mode: Annotated[str | None, Query(description="mixed/real/seed;不传=当前持久化模式")] = None,
 ) -> OpsSavingsFeedPreviewOut:
-    """返回客户端实际会看到的 feed(真实记录会插队、种子随机抽取/金额随机/名字合成),供运营对效果。"""
-    return OpsSavingsFeedPreviewOut(items=ops_marquee.get_feed(db, limit=limit))
+    """返回客户端实际会看到的 feed(真实记录会插队、种子随机抽取/金额随机/名字合成),供运营对效果。
+
+    mode 显式指定则预览该模式(**不改持久化配置**,供前端切换开关时实时预览);不传则用当前持久化模式。
+    """
+    if mode is not None and mode not in ops_marquee.FEED_MODES:
+        raise HTTPException(status_code=400, detail="mode 需为 mixed / real / seed")
+    return OpsSavingsFeedPreviewOut(items=ops_marquee.get_feed(db, limit=limit, mode=mode))
+
+
+# 注:/mode 两个端点须在 /{seed_id} 之前注册,否则 PATCH /mode 会被 /{seed_id} 抢先按 id 解析。
+@router.get("/mode", summary="首页轮播数据源模式(mixed/real/seed)")
+def get_feed_mode(db: AdminDb) -> dict:
+    """当前轮播取数模式:mixed=真实优先+种子补位(默认)/ real=只真实 / seed=只种子·合成。"""
+    return {"mode": ops_marquee.get_feed_mode(db)}
+
+
+@router.patch("/mode", summary="改首页轮播数据源模式(带审计)")
+def set_feed_mode(
+    body: MarqueeModeUpdate,
+    request: Request,
+    admin: Annotated[AdminUser, Depends(require_role("operator"))],
+    db: AdminDb,
+) -> dict:
+    if body.mode not in ops_marquee.FEED_MODES:
+        raise HTTPException(status_code=400, detail="mode 需为 mixed / real / seed")
+    before = ops_marquee.get_feed_mode(db)
+    app_config.set_value(db, "marquee_feed_mode", body.mode, admin_id=admin.id, commit=False)
+    write_audit(
+        db, admin, action="ops_marquee_seed.set_mode", target_type="config",
+        target_id="marquee_feed_mode", detail={"before": before, "after": body.mode},
+        ip=get_client_ip(request), commit=False,
+    )
+    db.commit()
+    return {"mode": body.mode}
 
 
 @router.post("", response_model=OpsMarqueeSeedOut, summary="新增轮播种子(带审计)")

app/admin/routers/price_report.py

@@ -37,11 +37,14 @@ def list_price_reports(
     db: AdminDb,
     status: Annotated[str | None, Query()] = None,
     user_id: Annotated[int | None, Query()] = None,
+    sort_by: Annotated[str, Query(pattern="^(id|created_at)$")] = "id",
+    sort_order: Annotated[str, Query(pattern="^(asc|desc)$")] = "desc",
     limit: Annotated[int, Query(ge=1, le=100)] = 20,
     cursor: Annotated[int | None, Query()] = None,
 ) -> CursorPage[PriceReportOut]:
     items, next_cursor, total = queries.list_price_reports(
-        db, status=status, user_id=user_id, limit=limit, cursor=cursor,
+        db, status=status, user_id=user_id,
+        sort_by=sort_by, sort_order=sort_order, limit=limit, cursor=cursor,
     )
     return CursorPage(
         items=[PriceReportOut.model_validate(r) for r in items],

app/admin/routers/roles.py

@@ -0,0 +1,129 @@
+"""admin RBAC 角色管理(仅 super_admin):列角色 / 权限目录 / 增删改角色。均写审计。
+
+角色 = 一组「可见页面」(见 app/admin/permissions.py)。角色登录后台后左侧只展示其 pages 对应导航项。
+只有 super_admin(内建全权角色)能进本组端点(dependencies=require_role() 无参 = 仅 super_admin)。
+"""
+from __future__ import annotations
+
+from fastapi import APIRouter, Depends, HTTPException, Request
+from sqlalchemy import func, select
+
+from app.admin.audit import write_audit
+from app.admin.deps import AdminDb, CurrentAdmin, get_client_ip, require_role
+from app.admin.permissions import (
+    PERMISSION_CATALOG,
+    SUPER_ADMIN_ROLE,
+    effective_pages,
+    sanitize_pages,
+)
+from app.admin.repositories import admin_role as role_repo
+from app.admin.schemas.role import (
+    PermissionGroup,
+    RoleCreateRequest,
+    RoleOut,
+    RoleUpdateRequest,
+)
+from app.models.admin import AdminUser
+from app.models.admin_role import AdminRole
+
+router = APIRouter(
+    prefix="/admin/api/roles",
+    tags=["admin-roles"],
+    dependencies=[Depends(require_role())],  # 无参 = 仅 super_admin
+)
+
+
+def _usage_counts(db: AdminDb) -> dict[str, int]:
+    rows = db.execute(
+        select(AdminUser.role, func.count(AdminUser.id)).group_by(AdminUser.role)
+    ).all()
+    return {r: c for r, c in rows}
+
+
+def _to_out(role: AdminRole, usage: dict[str, int]) -> RoleOut:
+    return RoleOut(
+        id=role.id,
+        name=role.name,
+        label=role.label or role.name,
+        pages=effective_pages(role.name, role.pages),
+        is_builtin=role.is_builtin,
+        in_use=usage.get(role.name, 0),
+    )
+
+
+@router.get("", response_model=list[RoleOut], summary="角色列表(含可见页 + 使用数)")
+def list_roles(db: AdminDb) -> list[RoleOut]:
+    usage = _usage_counts(db)
+    return [_to_out(r, usage) for r in role_repo.list_roles(db)]
+
+
+@router.get("/catalog", response_model=list[PermissionGroup], summary="页面权限目录(分组)")
+def get_catalog() -> list[PermissionGroup]:
+    return [PermissionGroup(**g) for g in PERMISSION_CATALOG]
+
+
+@router.post("", response_model=RoleOut, summary="新增角色(带审计)")
+def create_role(
+    body: RoleCreateRequest, request: Request, admin: CurrentAdmin, db: AdminDb
+) -> RoleOut:
+    name = body.name.strip()
+    if not name:
+        raise HTTPException(status_code=400, detail="角色名称不能为空")
+    if name == SUPER_ADMIN_ROLE:
+        raise HTTPException(status_code=400, detail="super_admin 为内建角色,不能新建")
+    if role_repo.get_role(db, name) is not None:
+        raise HTTPException(status_code=409, detail="角色名称已存在")
+    # 自定义角色:name(key)= label = 输入名称
+    role = role_repo.create_role(db, name=name, label=name, pages=body.pages)
+    write_audit(
+        db, admin, action="role.create", target_type="role", target_id=str(role.id),
+        detail={"name": name, "pages": role.pages}, ip=get_client_ip(request), commit=True,
+    )
+    return _to_out(role, _usage_counts(db))
+
+
+@router.patch("/{role_id}", response_model=RoleOut, summary="改角色(展示名/可见页,带审计)")
+def update_role(
+    role_id: int, body: RoleUpdateRequest, request: Request, admin: CurrentAdmin, db: AdminDb
+) -> RoleOut:
+    role = db.get(AdminRole, role_id)
+    if role is None:
+        raise HTTPException(status_code=404, detail="角色不存在")
+    if role.is_builtin:
+        raise HTTPException(status_code=400, detail="内建角色「管理员」不可编辑")
+
+    new_label = body.label.strip() if body.label is not None else None
+    changes: dict = {}
+    if new_label and new_label != role.label:
+        changes["label"] = {"before": role.label, "after": new_label}
+    if body.pages is not None:
+        changes["pages"] = {"before": role.pages, "after": sanitize_pages(body.pages)}
+    if not changes:
+        raise HTTPException(status_code=400, detail="无任何变更字段")
+
+    # 只改展示名 + 可见页;key(name)不可变,故无需级联 admin_user.role
+    role_repo.update_role(db, role, label=new_label, pages=body.pages)
+    write_audit(
+        db, admin, action="role.update", target_type="role", target_id=str(role_id),
+        detail=changes, ip=get_client_ip(request), commit=True,
+    )
+    return _to_out(role, _usage_counts(db))
+
+
+@router.delete("/{role_id}", summary="删角色(带审计;内建/在用不可删)")
+def delete_role(role_id: int, request: Request, admin: CurrentAdmin, db: AdminDb) -> dict:
+    role = db.get(AdminRole, role_id)
+    if role is None:
+        raise HTTPException(status_code=404, detail="角色不存在")
+    if role.is_builtin:
+        raise HTTPException(status_code=400, detail="内建角色不可删除")
+    used = _usage_counts(db).get(role.name, 0)
+    if used > 0:
+        raise HTTPException(status_code=400, detail=f"该角色仍有 {used} 名管理员在用,请先改派再删")
+    name = role.name
+    role_repo.delete_role(db, role)
+    write_audit(
+        db, admin, action="role.delete", target_type="role", target_id=str(role_id),
+        detail={"name": name}, ip=get_client_ip(request), commit=True,
+    )
+    return {"deleted": True}

app/admin/routers/users.py

@@ -42,7 +42,12 @@ def list_users(
     created_to: Annotated[datetime | None, Query()] = None,
     last_login_from: Annotated[datetime | None, Query()] = None,
     last_login_to: Annotated[datetime | None, Query()] = None,
-    sort_by: Annotated[str, Query(pattern="^(id|created_at|last_login_at)$")] = "id",
+    # 最近活跃(登录/发起比价/发起领券取最大,见 queries._last_active_expr)筛选与排序
+    last_active_from: Annotated[datetime | None, Query()] = None,
+    last_active_to: Annotated[datetime | None, Query()] = None,
+    sort_by: Annotated[
+        str, Query(pattern="^(id|created_at|last_login_at|last_active_at)$")
+    ] = "id",
     sort_order: Annotated[str, Query(pattern="^(asc|desc)$")] = "desc",
     limit: Annotated[int, Query(ge=1, le=100)] = 20,
     cursor: Annotated[int | None, Query()] = None,
@@ -51,6 +56,7 @@ def list_users(
         db, phone=phone, register_channel=register_channel, status=status,
         nickname=nickname, created_from=created_from, created_to=created_to,
         last_login_from=last_login_from, last_login_to=last_login_to,
+        last_active_from=last_active_from, last_active_to=last_active_to,
         sort_by=sort_by, sort_order=sort_order, limit=limit, cursor=cursor,
     )
     return CursorPage(
@@ -210,22 +216,28 @@ def grant_user_cash(
     db: AdminDb,
 ) -> OkResponse:
     """给指定用户增/减或设值现金(分)。delta:正=发放、负=扣减;set:直接设为目标值。
+    account=coin_cash(金币兑现金)/ invite_cash(邀请奖励金):两本账物理隔离、各调各的。
     主要用于让无现金用户直接测试提现。"""
     user = user_repo.get_user_by_id(db, user_id)
     if user is None:
         raise HTTPException(status_code=404, detail="用户不存在")
-    # set=设为目标值:读当前余额算差值,仍复用 grant_cash 写一笔流水(沿用原子/审计/扣负保护)
+    # 按目标账户选「余额字段 + 变动入口」(grant_invite_cash 与 grant_cash 同构)
+    is_invite = body.account == "invite_cash"
+    balance_attr = "invite_cash_balance_cents" if is_invite else "cash_balance_cents"
+    grant_fn = wallet_repo.grant_invite_cash if is_invite else wallet_repo.grant_cash
+    acct_label = "邀请奖励金" if is_invite else "现金"
+    before: int | None = None
+    # set=设为目标值:读当前余额算差值,仍复用 grant 写一笔流水(沿用原子/审计/扣负保护)
     if body.mode == "set":
         if body.amount_cents < 0:
-            raise HTTPException(status_code=400, detail="目标现金值不能为负")
+            raise HTTPException(status_code=400, detail=f"目标{acct_label}值不能为负")
         # lock=True:锁账户行,防连点/并发各读同一 before 算同一 delta 双写,余额错位
-        before = wallet_repo.get_or_create_account(
-            db, user_id, commit=False, lock=True
-        ).cash_balance_cents
+        acc_locked = wallet_repo.get_or_create_account(db, user_id, commit=False, lock=True)
+        before = getattr(acc_locked, balance_attr)
         delta = body.amount_cents - before
         if delta == 0:
             raise HTTPException(
-                status_code=400, detail=f"当前现金已为 {body.amount_cents} 分,无需调整"
+                status_code=400, detail=f"当前{acct_label}已为 {body.amount_cents} 分,无需调整"
             )
     else:
         if body.amount_cents == 0:
@@ -234,18 +246,20 @@ def grant_user_cash(
         # 负数扣减时不允许扣成负余额(运营误操作保护);lock=True 防并发扣穿
         if delta < 0:
             acc_now = wallet_repo.get_or_create_account(db, user_id, commit=False, lock=True)
-            if acc_now.cash_balance_cents + delta < 0:
+            if getattr(acc_now, balance_attr) + delta < 0:
                 raise HTTPException(
-                    status_code=400, detail=f"扣减后现金为负(当前余额 {acc_now.cash_balance_cents} 分)"
+                    status_code=400,
+                    detail=f"扣减后{acct_label}为负(当前余额 {getattr(acc_now, balance_attr)} 分)",
                 )
     biz_type = "admin_grant" if delta > 0 else "admin_deduct"
-    # grant_cash 只 flush 不 commit;审计同 commit=False;最后一起 commit → 原子(改钱+留痕)
-    acc, _ = wallet_repo.grant_cash(
+    # grant 只 flush 不 commit;审计同 commit=False;最后一起 commit → 原子(改钱+留痕)
+    acc, _ = grant_fn(
         db, user_id, delta, biz_type=biz_type, remark=f"admin:{body.reason}"[:128],
     )
     detail = {
+        "account": body.account,
         "amount_cents": delta,
-        "balance_after_cents": acc.cash_balance_cents,
+        "balance_after_cents": getattr(acc, balance_attr),
         "reason": body.reason,
     }
     if body.mode == "set":

app/admin/routers/withdraw.py

@@ -50,6 +50,8 @@ def list_withdraws(
     db: AdminDb,
     user_id: Annotated[int | None, Query()] = None,
     status: Annotated[str | None, Query()] = None,
+    # 提现类型筛选:coin_cash(福利页提现)/ invite_cash(邀请提现);None=全部
+    source: Annotated[str | None, Query(pattern="^(coin_cash|invite_cash)$")] = None,
     keyword: Annotated[str | None, Query(max_length=100)] = None,
     date_from: Annotated[datetime | None, Query()] = None,
     date_to: Annotated[datetime | None, Query()] = None,
@@ -69,6 +71,7 @@ def list_withdraws(
         db,
         user_id=user_id,
         status=status,
+        source=source,
         keyword=keyword,
         date_from=date_from,
         date_to=date_to,

app/admin/schemas/ad_config.py

@@ -9,8 +9,8 @@ class AdConfigOut(BaseModel):
 
     app_id: str
     reward_code_id: str
-    compare_feed_code_id: str
-    coupon_feed_code_id: str
+    compare_draw_code_id: str
+    coupon_draw_code_id: str
     reward_mkey: str
     reward_enabled: bool
     compare_ad_enabled: bool
@@ -23,8 +23,8 @@ class AdConfigUpdate(BaseModel):
 
     app_id: str | None = None
     reward_code_id: str | None = None
-    compare_feed_code_id: str | None = None
-    coupon_feed_code_id: str | None = None
+    compare_draw_code_id: str | None = None
+    coupon_draw_code_id: str | None = None
     reward_mkey: str | None = None
     reward_enabled: bool | None = None
     compare_ad_enabled: bool | None = None

app/admin/schemas/ad_revenue.py

@@ -40,15 +40,38 @@ class AdRevenueRecord(BaseModel):
 
 
 class AdRevenueDaily(BaseModel):
-    """按日期汇总的一天(供前端按天趋势图;全量,不受 limit 影响)。"""
+    """按日期汇总的一天(供前端按天趋势图;全量,不受分页影响)。"""
 
     date: str = Field(..., description="北京时间 YYYY-MM-DD")
     impressions: int = Field(..., description="当天展示条数合计")
-    revenue_yuan: float = Field(..., description="当天预估收益合计(元)")
+    revenue_yuan: float = Field(..., description="当天客户端预估收益合计(元;eCPM 折算)")
+    pangle_revenue_yuan: float | None = Field(
+        None, description="当天穿山甲后台预估收益(元;GroMore revenue);非全量视图/无数据为空"
+    )
+    pangle_api_revenue_yuan: float | None = Field(
+        None, description="当天穿山甲收益Api(元;GroMore api_revenue,更接近结算);未配/当天/无数据为空"
+    )
     expected_coin: int = Field(..., description="当天应发金币合计")
     actual_coin: int = Field(..., description="当天实发金币合计")
 
 
+class AdRevenueHourly(BaseModel):
+    """按北京小时(0–23)汇总的一小时(供前端按小时趋势图;全量,不受分页影响,单日 granularity=hour 时非空)。"""
+
+    hour: int = Field(..., description="北京时间小时 0–23")
+    impressions: int = Field(..., description="该小时展示条数合计")
+    revenue_yuan: float = Field(..., description="该小时预估收益合计(元)")
+    expected_coin: int = Field(..., description="该小时应发金币合计")
+    actual_coin: int = Field(..., description="该小时实发金币合计")
+
+
+class AdRevenueTypeStat(BaseModel):
+    """按广告类型(ad_type)的小计:展示条数 + 预估收益(eCPM 由前端用 收益÷展示×1000 算)。"""
+
+    impressions: int = Field(..., description="该类型展示条数合计")
+    revenue_yuan: float = Field(..., description="该类型预估收益合计(元)")
+
+
 class AdRevenueRow(BaseModel):
     """一次广告事件(逐条一行):激励视频展示与发奖按 ad_session_id 合并;信息流展示 / 发奖各自成行。"""
 
@@ -56,7 +79,12 @@ class AdRevenueRow(BaseModel):
     report_date: str = Field(..., description="该事件所属日期(北京时间 YYYY-MM-DD)")
     user_id: int
     user_phone: str | None = Field(None, description="用户手机号(admin 展示用,完整;用户已删 / 查不到为空)")
-    ad_type: str = Field(..., description="reward_video(激励视频) / feed(信息流) / draw(历史 Draw 信息流)")
+    ad_type: str = Field(..., description="reward_video(激励视频) / feed(信息流) / draw(Draw 信息流);历史 NULL 视为 feed")
+    feed_scene: str | None = Field(
+        None,
+        description="点位场景:comparison(比价) / coupon(领券) / welfare(福利);供区分比价/领券 Draw 收益;"
+        "激励视频与旧数据为空",
+    )
     app_env: str | None = Field(None, description="我们的应用:prod(傻瓜比价正式) / test(测试应用);旧数据为空")
     our_code_id: str | None = Field(None, description="我们后台配置的代码位 ID(104xxx);旧数据为空")
     hour: int | None = Field(None, description="北京时间小时 0–23(granularity=hour 时有值;按天为 null)")
@@ -66,6 +94,11 @@ class AdRevenueRow(BaseModel):
     impressions: int = Field(..., description="本行展示条数:有展示=1 / 纯发奖=0(供日汇总、趋势图复用)")
     ecpm: str | None = Field(None, description="eCPM 原始值(分/千次);展示行取展示值,纯发奖行取发奖采用值")
     revenue_yuan: float = Field(..., description="本次展示预估收益(元)= eCPM元 ÷ 1000;纯发奖行=0")
+    row_revenue_yuan: float | None = Field(
+        None,
+        description="主表逐行展示用的预估收益(元):一次比价/领券聚合行=该次发奖广告 eCPM 折算之和;"
+        "其它行为空(前端回退取 revenue_yuan)。不进合计/趋势,避免与展示侧重复计",
+    )
     adn: str | None = Field(None, description="实际填充 ADN 子渠道(pangle/gdt…);纯发奖行为空")
     slot_id: str | None = Field(None, description="底层 mediation rit(非我们配置的广告位 ID);纯发奖行为空")
     # ── 发奖侧 ──
@@ -78,6 +111,15 @@ class AdRevenueRow(BaseModel):
         None,
         description="发奖复算明细(eCPM/因子1/份数/LT/因子2/应发/实发/一致);点行展开下钻用,纯展示为空",
     )
+    sub_rewards: list[AdRevenueRecord] = Field(
+        default_factory=list,
+        description="一次比价/领券聚合行的组内逐条发奖明细(同一整场 ad_session_id 的多条广告);"
+        "点行展开渲染多行。激励视频/纯展示行为空(单条看 reward_detail)",
+    )
+    sub_count: int = Field(
+        1,
+        description="本行聚合的发奖条数:一次比价/领券=该次广告条数(≥1);激励视频/纯展示=1",
+    )
 
 
 class AdRevenueReportOut(BaseModel):
@@ -86,10 +128,43 @@ class AdRevenueReportOut(BaseModel):
     date_from: str = Field(..., description="报表起始日期(北京时间 YYYY-MM-DD)")
     date_to: str = Field(..., description="报表结束日期(北京时间 YYYY-MM-DD,闭区间;单日时与 date_from 相同)")
     daily: list[AdRevenueDaily] = Field(..., description="按日期汇总序列(全量,供按天趋势图)")
-    total: int = Field(..., description="广告事件总数(全量,不受 limit 影响)")
-    truncated: bool = Field(..., description="明细是否被 limit 截断")
+    hourly: list[AdRevenueHourly] = Field(
+        default_factory=list,
+        description="按小时汇总序列(全量,供按小时趋势图;按天查询时为空)",
+    )
+    type_stats: dict[str, AdRevenueTypeStat] = Field(
+        default_factory=dict,
+        description="按广告类型(ad_type)小计 {ad_type: {impressions, revenue_yuan}};前端取 draw / reward_video 做分类大盘",
+    )
+    scene_stats: dict[str, AdRevenueTypeStat] = Field(
+        default_factory=dict,
+        description="按信息流场景(feed_scene)小计 {comparison/coupon/welfare: {impressions, revenue_yuan}};"
+        "全量统计(不受分页截断),供数据大盘「领券广告 / 比价广告」卡;feed_scene 为空的事件不计入",
+    )
+    dau: int | None = Field(
+        None,
+        description="所选日期区间的去重活跃用户数(口径同数据大盘 period.users.active:登录 + 开始比价 + "
+        "开始领券)。按 date_from~date_to 区间统计,含今日、近 7 天、近 30 天等任意区间;全局口径,"
+        "不随 user_id / ad_type / feed_scene / app_env 筛选变化",
+    )
+    total: int = Field(..., description="广告事件总数(全量,不受分页影响;= 当前筛选下的分页总条数)")
+    truncated: bool = Field(..., description="当前页之后是否还有更多事件(len(events) > offset + limit)")
     total_impressions: int = Field(..., description="全量展示条数合计")
-    total_revenue_yuan: float = Field(..., description="全量收益合计(元)")
+    total_revenue_yuan: float = Field(..., description="全量客户端预估收益合计(元;eCPM 折算)")
+    total_pangle_revenue_yuan: float | None = Field(
+        None,
+        description="全量穿山甲后台预估收益合计(元;GroMore revenue)。穿山甲无用户/类型/场景维度,"
+        "仅「全量视图」(未按 user_id/ad_type/feed_scene 过滤)时有值,否则为 null",
+    )
+    total_pangle_api_revenue_yuan: float | None = Field(
+        None,
+        description="全量穿山甲收益Api合计(元;GroMore api_revenue,各 ADN 回传、更接近结算);"
+        "未配 Reporting / 查当天 / 非全量视图 时为 null",
+    )
+    pangle_revenue_available: bool = Field(
+        False,
+        description="本次结果是否带穿山甲后台收益(=全量视图且已同步到数据)。false 时前端「穿山甲收益」显示「-」",
+    )
     total_expected_coin: int = Field(..., description="全量应发金币合计")
     total_actual_coin: int = Field(..., description="全量实发金币合计")
     mismatch_count: int = Field(..., description="应发≠实发的发奖条数(=0 说明全部按公式发放)")

app/admin/schemas/admin.py

@@ -6,21 +6,25 @@ from typing import Literal
 
 from pydantic import BaseModel, ConfigDict, Field
 
-_Role = Literal["super_admin", "finance", "operator"]
+# 角色不再硬编码枚举:改为任意角色名(自定义角色由 admin_role 表管理),存在性在路由层校验。
 
 
 class AdminCreateRequest(BaseModel):
     username: str = Field(..., min_length=3, max_length=64)
     password: str = Field(..., min_length=8, max_length=72)  # bcrypt ≤72 字节
-    role: _Role = "operator"
+    role: str = Field("operator", min_length=1, max_length=32)
+    # 仅当 role == "custom":这个人专属可见页 key 列表(逐页勾选结果)。其余角色不传/忽略。
+    pages_override: list[str] | None = None
 
 
 class AdminUpdateRequest(BaseModel):
-    """改角色 / 启用禁用 / 重置密码,字段都可选(只改传了的)。"""
+    """改角色 / 启用禁用 / 重置密码 / 自定义可见页,字段都可选(只改传了的)。"""
 
-    role: _Role | None = None
+    role: str | None = Field(None, min_length=1, max_length=32)
     status: Literal["active", "disabled"] | None = None
     password: str | None = Field(None, min_length=8, max_length=72)
+    # 改成/更新「自定义」可见页;role 切回普通角色时后端会清空 override(见路由)。
+    pages_override: list[str] | None = None
 
 
 class AdminAuditLogOut(BaseModel):

app/admin/schemas/analytics.py

@@ -0,0 +1,33 @@
+"""admin 埋点日志列表响应。"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from pydantic import BaseModel, ConfigDict
+
+
+class AnalyticsEventOut(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: int
+    event: str
+    # Who
+    device_id: str
+    user_id: int | None
+    # When
+    session_id: str | None
+    client_ts: int
+    sent_at: int | None
+    created_at: datetime  # 服务端接收时间(server_at)
+    # Where
+    page: str | None
+    client_ip: str | None
+    # How
+    oem: str | None
+    os: str | None
+    model: str | None
+    app_ver: str | None
+    network: str | None
+    channel: str | None
+    # What 专属
+    props: dict | None

app/admin/schemas/auth.py

@@ -20,6 +20,15 @@ class AdminOut(BaseModel):
     status: str
     created_at: datetime
     last_login_at: datetime | None = None
+    # 该管理员当前角色的有效可见页(= 左侧导航项 key);仅登录 / /me 填充,列表接口默认空。
+    # 前端据此过滤左侧导航(super_admin = 全部页;role=="custom" = pages_override)。见 permissions.py。
+    pages: list[str] = []
+    # 「自定义」可见页原始勾选集(role=="custom" 时非空);列表接口下发,供权限管理页编辑回填勾选。
+    # 普通角色为 None。from_attributes 自动从 ORM 列取。
+    pages_override: list[str] | None = None
+    # 明文登录密码:仅「管理员账号列表」(super_admin 专属路由)填充,供权限管理页编辑时复看;
+    # 无留存(脚本建的超管 / 旧账号)为 None → 前端不显示。登录 / /me 不下发(保持 None)。
+    password: str | None = None
 
 
 class AdminLoginResponse(BaseModel):

app/admin/schemas/comparison.py

@@ -23,6 +23,7 @@ class AdminComparisonListItem(BaseModel):
     status: str
     information: str | None = None
     store_name: str | None = None
+    product_names: str | None = None   # 下单商品名派生串(顿号分隔;「商品」列展示 + 商品搜索)
     source_platform_name: str | None = None
     best_platform_name: str | None = None
     source_price_cents: int | None = None

app/admin/schemas/coupon_data.py

@@ -0,0 +1,83 @@
+"""admin「领券数据」看板 schemas:汇总卡 + 按天/小时趋势 + 逐条领券明细。
+
+数据源 coupon_session(一次领券一行)。耗时单位 ms(前端按需折秒);均值/分位只统计 completed。
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from pydantic import BaseModel, Field
+
+
+class CouponDataSummary(BaseModel):
+    """汇总卡:发起/完成数 + 耗时均值与分位(P5/P50/P95/P99,基于 completed 的 elapsed_ms)。"""
+
+    started_count: int = Field(..., description="发起数(区间内所有领券 session)")
+    completed_count: int = Field(..., description="完成数(status=completed)")
+    avg_elapsed_ms: int | None = Field(None, description="平均耗时(ms,仅 completed;无数据为空)")
+    p5_ms: int | None = Field(None, description="耗时 5 分位(ms)")
+    p50_ms: int | None = Field(None, description="耗时 50 分位(ms,中位数)")
+    p95_ms: int | None = Field(None, description="耗时 95 分位(ms)")
+    p99_ms: int | None = Field(None, description="耗时 99 分位(ms)")
+
+
+class CouponDataDaily(BaseModel):
+    """按天趋势(全量,不受分页影响):柱=发起/完成数,线=平均耗时。"""
+
+    date: str = Field(..., description="北京时间 YYYY-MM-DD")
+    started_count: int
+    completed_count: int
+    avg_elapsed_ms: int | None = Field(None, description="当天平均耗时(ms,仅 completed)")
+
+
+class CouponDataHourly(BaseModel):
+    """按北京小时(0–23)趋势(单日 granularity=hour 时非空)。"""
+
+    hour: int = Field(..., description="北京时间小时 0–23")
+    started_count: int
+    completed_count: int
+    avg_elapsed_ms: int | None = None
+
+
+class CouponDataRow(BaseModel):
+    """一条领券明细(一次领券任务)。"""
+
+    id: int = Field(..., description="coupon_session 主键(抽屉 rowKey 用)")
+    trace_id: str
+    user_id: int | None = None
+    user_phone: str | None = Field(None, description="手机号(admin 展示;匿名领券/查不到为空)")
+    user_nickname: str | None = Field(None, description="昵称")
+    status: str = Field(..., description="started / completed / failed / abandoned")
+    platforms: list[str] | None = Field(None, description="发起勾选平台")
+    origin_package: str | None = Field(None, description="发起来源 App 包名;null=App 内(傻瓜比价首页)发起")
+    elapsed_ms: int | None = Field(None, description="全程耗时(ms)")
+    platform_elapsed: dict[str, int] | None = Field(
+        None, description="各平台耗时 {meituan-waimai/taobao-shanguang/jd-waimai: ms}"
+    )
+    device_model: str | None = None
+    rom: str | None = None
+    app_env: str | None = None
+    started_at: datetime = Field(..., description="发起时刻(明细「时间」列)")
+    claimed_count: int | None = None
+    trace_url: str | None = Field(None, description="pricebot 公网 trace 链接(仅 completed 有);admin 渲染可点链接,无则显示可复制 trace_id")
+
+
+class CouponDataOut(BaseModel):
+    """领券数据看板响应:汇总卡 + 趋势 + 明细分页。"""
+
+    date_from: str
+    date_to: str
+    summary: CouponDataSummary
+    daily: list[CouponDataDaily] = Field(default_factory=list, description="按天趋势(全量)")
+    hourly: list[CouponDataHourly] = Field(
+        default_factory=list, description="按小时趋势(单日 hour 粒度时非空)"
+    )
+    total: int = Field(..., description="明细总条数(全量,不受分页)")
+    items: list[CouponDataRow] = Field(..., description="逐条领券明细(当前页)")
+
+
+class CouponUserRecordsOut(BaseModel):
+    """某用户全部领券记录(点手机号抽屉用):total=该用户领券总次数,items=记录列表(UserRecordsDrawer 渲染)。"""
+
+    items: list[CouponDataRow]
+    total: int

app/admin/schemas/cps.py

@@ -1,7 +1,7 @@
 """admin CPS 分发与对账 schemas。金额统一「分」(cents),前端 yuan() 展示。
 
-平台:meituan(actId+sid 转链对账) / taobao(淘口令,只统计点击) / jd(链接,只统计点击)。
-对账类字段对淘宝/京东为 None → 前端显示 "-"(无法对账)。
+平台:meituan(actId+sid 转链对账) / taobao(淘口令,只统计点击) / jd(链接 + 订单 API 对账)。
+对账类字段对淘宝为 None → 前端显示 "-"(暂未对账)。
 """
 from __future__ import annotations
 
@@ -116,15 +116,22 @@ class CpsOrderOut(BaseModel):
     model_config = ConfigDict(from_attributes=True)
 
     id: int
+    platform: str = "meituan"
     order_id: str
+    external_order_id: str | None = None
+    external_row_id: str | None = None
     sid: str | None = None
     act_id: str | None = None
     pay_price_cents: int | None = None
     commission_cents: int | None = None
+    estimated_commission_cents: int | None = None
+    actual_commission_cents: int | None = None
     commission_rate: str | None = None
     mt_status: str | None = None
+    jd_valid_code: str | None = None
     invalid_reason: str | None = None
     product_name: str | None = None
+    settle_month: str | None = None
     pay_time: datetime | None = None
 
 

app/admin/schemas/dashboard.py

@@ -1,6 +1,8 @@
 """admin 大盘 schemas(对应 stats.dashboard_overview 的嵌套结构)。"""
 from __future__ import annotations
 
+from datetime import date
+
 from pydantic import BaseModel
 
 
@@ -38,6 +40,78 @@ class DashboardComparison(BaseModel):
     success_rate: float
 
 
+class DashboardPeriodUsers(BaseModel):
+    new: int
+    active: int
+    # 次日留存(2026-07-05 起):retained_new_users = 窗口内逐日「前一日新增且当日活跃」用户数之和,
+    # retention_cohort = 对应的前一日新增基数之和,retention_rate = 两者之比。
+    retained_new_users: int
+    retention_cohort: int = 0
+    retention_rate: float | None = None
+    retention_note: str
+
+
+class DashboardPeriodComparison(BaseModel):
+    total: int
+    success: int
+    success_rate: float
+    ordered: int
+    average_duration_ms: int | None = None
+    average_saved_cents: int | None = None
+
+
+class DashboardPeriodCoupon(BaseModel):
+    """领券核心数据(2026-07-05 产品新增)。点位=一张券(coupon_claim_record 一天一条终态);
+    成功口径 success+already_claimed(与「我的」页累计领券一致)。"""
+
+    started: int = 0
+    # 全部领成功的次数:completed 且当日该设备全部点位成功
+    all_success: int = 0
+    success_rate: float | None = None
+    # 本期 session 触达的点位中成功的条数(同设备同日去重)
+    point_success: int = 0
+    # 每次发起的应领点位数(本期完成场实际点位数的众数;无完成场为空)
+    points_per_session: int | None = None
+    # 点位成功率 = point_success / (started × points_per_session);未跑到的点位计入分母视为失败
+    point_success_rate: float | None = None
+    # 耗时中位数(仅 completed 的 elapsed_ms,同「领券数据」页口径)
+    median_elapsed_ms: int | None = None
+
+
+class DashboardPeriodCoins(BaseModel):
+    granted_total: int
+    reward_video_coin_total: int = 0
+    feed_ad_coin_total: int = 0
+    signin_coin_total: int = 0
+    signin_boost_coin_total: int = 0
+    task_coin_total: int = 0
+    coupon_reward_coin_total: int = 0
+    comparison_reward_coin_total: int = 0
+    regular_task_coin_total: int = 0
+
+
+class DashboardPeriodCash(BaseModel):
+    withdraw_success_cents: int
+
+
+class DashboardTrendPoint(BaseModel):
+    date: date
+    active_users: int
+    new_users: int
+    comparisons: int
+
+
+class DashboardPeriod(BaseModel):
+    date_from: date
+    date_to: date
+    users: DashboardPeriodUsers
+    comparison: DashboardPeriodComparison
+    coupon: DashboardPeriodCoupon = DashboardPeriodCoupon()
+    coins: DashboardPeriodCoins
+    cash: DashboardPeriodCash
+    trend: list[DashboardTrendPoint] = []
+
+
 class DashboardFeedback(BaseModel):
     new: int
 
@@ -45,6 +119,17 @@ class DashboardFeedback(BaseModel):
 class DashboardCps(BaseModel):
     available: bool
     note: str
+    meituan_order_count: int = 0
+    meituan_commission_cents: int = 0
+    meituan_hit_count: int = 0
+    meituan_miss_count: int = 0
+    meituan_unknown_rate_count: int = 0
+    meituan_hit_rate: float | None = None
+    jd_order_count: int = 0
+    jd_commission_cents: int = 0
+    jd_actual_commission_cents: int = 0
+    jd_estimated_commission_cents: int = 0
+    jd_invalid_count: int = 0
 
 
 class DashboardOverview(BaseModel):
@@ -52,5 +137,6 @@ class DashboardOverview(BaseModel):
     coins: DashboardCoins
     cash: DashboardCash
     comparison: DashboardComparison
+    period: DashboardPeriod
     feedback: DashboardFeedback
     cps: DashboardCps

app/admin/schemas/device.py

@@ -0,0 +1,54 @@
+"""admin 设备存活监控 schemas(只读)。
+
+数据源 device_liveness 表(见 app/models/device.py)。online / display_state /
+offline_seconds 为后端按 HEARTBEAT_TIMEOUT_MINUTES 阈值派生的瞬态字段(非 DB 列),
+在 repo 里算好挂到行上,供 from_attributes 读出。
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from pydantic import BaseModel, ConfigDict
+
+
+class DeviceLivenessItem(BaseModel):
+    model_config = ConfigDict(from_attributes=True)
+
+    id: int
+    user_id: int
+    # join user 取(瞬态;理论上 user 恒存在,留空仅防脏数据)
+    phone: str | None = None
+    nickname: str | None = None
+
+    device_id: str
+    device_model: str | None = None      # 由 device_id 解析(device_<机型>_<hash>);非 DB 列
+    platform: str
+    app_version: str | None = None
+    registration_id: str | None = None  # 非空 = 拿到极光 token、可推送
+
+    ever_protected: bool                 # 是否开过无障碍(=该设备对功能有意义)
+    first_protected_at: datetime | None = None   # 首次开无障碍时刻(老设备为 null)
+    last_heartbeat_at: datetime | None = None
+    last_report_protection_on: bool      # 心跳里上报的无障碍开关(恒 true,仅观测)
+    liveness_state: str                  # unknown / alive / silent(当前未用) / notified
+    notified_at: datetime | None = None
+    kill_alert_pending: bool             # 掉线召回待客户端 ack(与 state 解耦)
+    created_at: datetime
+    updated_at: datetime
+
+    # ===== 后端派生(非 DB 列)=====
+    online: bool = False                 # ever_protected 且距上次心跳 ≤ 阈值
+    # 显示态:online=在线 / offline=已掉线 / never=从未启用(没开过无障碍)
+    display_state: str = "never"
+    # 掉线时长(秒):仅 offline 时有值;在线 / 从未启用为 None
+    offline_seconds: int | None = None
+
+
+class DeviceLivenessStats(BaseModel):
+    """顶部卡片:总设备数 + 在线 / 已掉线 / 未启用(按心跳阈值派生的 operator 口径,
+    不暴露内部状态机 unknown/alive/silent/notified)。"""
+
+    total: int
+    online: int
+    offline: int
+    never: int

app/admin/schemas/feedback.py

@@ -15,14 +15,28 @@ class FeedbackOut(BaseModel):
     user_id: int
     content: str
     contact: str
+    # 反馈来源入口:profile(「我的」页)/ comparison(比价结果页)——admin 展示/筛选「反馈类型」
+    source: str = "profile"
+    # 比价反馈的问题场景(找错商品/优惠不对…);普通反馈为 None
+    scene: str | None = None
     images: list[str] | None = None
     status: str
     reject_reason: str | None = None
     reward_coins: int | None = None
     review_note: str | None = None
+    # 运营给用户的回复留言(用户端可见,采纳/未采纳都可填)
+    admin_reply: str | None = None
     reviewed_by_admin_id: int | None = None
     reviewed_at: datetime | None = None
     created_at: datetime
+    # 提交端环境快照(feedback 表列):提交版本号 / 机型OS版本;改版前的历史反馈为 None
+    app_version: str | None = None
+    device_model: str | None = None
+    rom_name: str | None = None
+    android_version: str | None = None
+    # 联表瞬态字段(queries._attach_user_info 挂):列表展示完整手机号,点手机号查该用户全部反馈
+    phone: str | None = None
+    nickname: str | None = None
 
 
 class FeedbackApproveRequest(BaseModel):
@@ -31,9 +45,20 @@ class FeedbackApproveRequest(BaseModel):
         le=FEEDBACK_REWARD_MAX_COINS,
         description="采纳后发放金币数",
     )
-    note: str | None = Field(default=None, max_length=256, description="采纳要点/审核备注")
+    note: str | None = Field(default=None, max_length=256, description="采纳要点/审核备注(内部)")
+    reply: str | None = Field(default=None, max_length=256, description="给用户的回复留言,用户端可见")
 
 
 class FeedbackRejectRequest(BaseModel):
     reason: str = Field(min_length=1, max_length=256, description="未采纳原因,用户端可见")
     note: str | None = Field(default=None, max_length=256, description="运营内部审核备注")
+    reply: str | None = Field(default=None, max_length=256, description="给用户的回复留言,用户端可见")
+
+
+class FeedbackSummary(BaseModel):
+    """审核台顶部各状态计数(pending 含历史 new 态)。"""
+
+    pending: int
+    adopted: int
+    rejected: int
+    total: int

app/admin/schemas/price_report.py

@@ -31,6 +31,17 @@ class PriceReportOut(BaseModel):
     reward_coins: int | None = None
     reviewed_at: datetime | None = None
     created_at: datetime
+    # 联表瞬态字段:phone/nickname 由 _attach_user_info 挂(展示完整手机号、点击查该用户全部上报);
+    # 其余由 _attach_price_report_comparison 从关联比价记录挂:trace_*(Trace 列点跳调试链接)、
+    # device_model/rom_name/android_version(机型OS版本列)、app_version(提交版本号列)
+    phone: str | None = None
+    nickname: str | None = None
+    trace_id: str | None = None
+    trace_url: str | None = None
+    device_model: str | None = None
+    rom_name: str | None = None
+    android_version: str | None = None
+    app_version: str | None = None
 
 
 class PriceReportRejectRequest(BaseModel):

app/admin/schemas/role.py

@@ -0,0 +1,35 @@
+"""admin RBAC 角色 + 权限目录 schemas。"""
+from __future__ import annotations
+
+from pydantic import BaseModel, Field
+
+
+class PermissionPage(BaseModel):
+    key: str
+    label: str
+
+
+class PermissionGroup(BaseModel):
+    group: str
+    pages: list[PermissionPage]
+
+
+class RoleOut(BaseModel):
+    id: int
+    name: str             # 角色 key(承重,不可变)
+    label: str            # 展示名(UI 显示)
+    pages: list[str]      # 有效可见页 key(super_admin = 全部页)
+    is_builtin: bool      # 内建角色(super_admin):不可编辑/删除
+    in_use: int           # 使用该角色的管理员数(删除前端提示 / 拦截用)
+
+
+class RoleCreateRequest(BaseModel):
+    # 新增自定义角色:name 即用户输入的名称(同时作 key 与展示名)
+    name: str = Field(..., min_length=1, max_length=32)
+    pages: list[str] = Field(default_factory=list)
+
+
+class RoleUpdateRequest(BaseModel):
+    # 只改展示名 + 可见页;key 不可变
+    label: str | None = Field(None, min_length=1, max_length=32)
+    pages: list[str] | None = None

app/admin/schemas/user.py

@@ -17,8 +17,12 @@ class AdminUserListItem(BaseModel):
     status: str
     debug_trace_enabled: bool = False
     wechat_openid: str | None = None
+    wechat_nickname: str | None = None
     created_at: datetime
     last_login_at: datetime
+    # 最近活跃 = max(最近登录, 最近发起比价, 最近发起领券);列表页由 queries._attach_last_active
+    # 瞬态挂上。其他复用本 schema 的入口(用户 360 等)没挂该属性 → None(前端显示 '-')。
+    last_active_at: datetime | None = None
 
 
 class AdminUserOverview(BaseModel):
@@ -29,6 +33,7 @@ class AdminUserOverview(BaseModel):
     user: AdminUserListItem
     coin_balance: int
     cash_balance_cents: int
+    invite_cash_balance_cents: int  # 邀请奖励金余额(与 cash_balance_cents 物理隔离)
     total_coin_earned: int
     comparison_total: int
     comparison_success: int
@@ -88,6 +93,11 @@ class GrantCoinsRequest(BaseModel):
 
 
 class GrantCashRequest(BaseModel):
+    # 目标账户:金币兑换的现金(cash_balance_cents)与邀请奖励金(invite_cash_balance_cents)物理隔离,
+    # 各调各的、不可串。默认 coin_cash 兼容旧调用。
+    account: Literal["coin_cash", "invite_cash"] = Field(
+        "coin_cash", description="目标账户:coin_cash=金币兑现金账户 / invite_cash=邀请奖励金账户"
+    )
     mode: Literal["delta", "set"] = Field(
         "delta", description="delta=增减(amount_cents 为变动量) / set=设为(amount_cents 为目标值,须≥0)"
     )

app/admin/schemas/wallet.py

@@ -41,6 +41,8 @@ class WithdrawOrderOut(BaseModel):
     user_id: int
     out_bill_no: str
     amount_cents: int
+    # 提现类型:coin_cash(福利页提现,金币兑换的现金)/ invite_cash(邀请提现,邀请奖励金)
+    source: str = "coin_cash"
     user_name: str | None = None  # 提现实名(审核核对 + 打款用)
     status: str
     wechat_state: str | None = None
@@ -128,6 +130,7 @@ class WithdrawBulkResult(BaseModel):
 
 class WithdrawLedgerCheckOut(BaseModel):
     ok: bool
+    # 普通现金账(coin_cash:金币兑换的现金)
     cash_balance_total_cents: int
     cash_transaction_total_cents: int
     balance_diff_cents: int
@@ -135,6 +138,14 @@ class WithdrawLedgerCheckOut(BaseModel):
     missing_refund_txn_count: int
     duplicate_refund_txn_count: int
     refund_txn_on_non_terminal_count: int
+    # 邀请奖励金账(invite_cash:与普通现金物理隔离,各自对账)。默认 0 向后兼容。
+    invite_cash_balance_total_cents: int = 0
+    invite_cash_transaction_total_cents: int = 0
+    invite_balance_diff_cents: int = 0
+    invite_missing_withdraw_txn_count: int = 0
+    invite_missing_refund_txn_count: int = 0
+    invite_duplicate_refund_txn_count: int = 0
+    invite_refund_txn_on_non_terminal_count: int = 0
 
 
 class WxpayHealthCheckOut(BaseModel):

app/api/deps.py

@@ -54,5 +54,29 @@ def get_current_user(
     return user
 
 
+def get_current_user_optional(
+    credentials: Annotated[HTTPAuthorizationCredentials | None, Depends(_bearer)],
+    db: Annotated[Session, Depends(get_db)],
+) -> User | None:
+    """软鉴权:有合法 Bearer 就返回 user,否则(无 header / 无效 token / 用户禁用)一律返回
+    None,**不 raise**。
+
+    比价 step / finalize 灰度期用:新客户端带 JWT → 拿到 user_id 绑定 harvest 行;
+    老客户端(不带 JWT)→ None,harvest 行 user_id 暂空,由其后续 /compare/record 上报补齐。
+    等新版覆盖率够高,再把这几条端点从软鉴权收紧成硬 get_current_user。
+    """
+    if credentials is None or credentials.scheme.lower() != "bearer":
+        return None
+    try:
+        payload = decode_token(credentials.credentials, expected_type="access")
+        user = db.get(User, int(payload["sub"]))
+    except (TokenError, KeyError, ValueError, TypeError):
+        return None
+    if user is None or user.status != "active":
+        return None
+    return user
+
+
 CurrentUser = Annotated[User, Depends(get_current_user)]
+OptionalUser = Annotated[User | None, Depends(get_current_user_optional)]
 DbSession = Annotated[Session, Depends(get_db)]

app/api/internal/launch_confirm.py

@@ -11,6 +11,7 @@ from __future__ import annotations
 
 import hmac
 import logging
+from datetime import datetime, timedelta, timezone
 from typing import Annotated
 
 from fastapi import APIRouter, Header, HTTPException, status
@@ -21,6 +22,7 @@ from app.repositories import launch_confirm_sample as repo
 from app.schemas.launch_confirm_sample import (
     LaunchConfirmSampleIn,
     LaunchConfirmSampleOut,
+    LaunchConfirmSampleRow,
 )
 
 logger = logging.getLogger("shagua.internal.launch_confirm")
@@ -61,3 +63,34 @@ def report_launch_confirm_sample(
         payload.exec_success, payload.trace_id,
     )
     return LaunchConfirmSampleOut(id=sid)
+
+
+@router.get(
+    "/launch-confirm-samples",
+    response_model=list[LaunchConfirmSampleRow],
+    summary="启动确认窗兜底样本列表(沉淀脚本 distill 读; server→server, 不走 JWT)",
+)
+def list_launch_confirm_samples(
+    db: DbSession,
+    x_internal_secret: Annotated[str | None, Header()] = None,
+    exec_success: bool | None = None,
+    host_package: str | None = None,
+    since_days: int | None = None,
+    limit: int = 1000,
+) -> list[LaunchConfirmSampleRow]:
+    """读样本供 pricebot 的 distill_launch_confirm.py 聚合沉淀回 PROFILES。
+
+    与上报端点同一把共享密钥;exec_success/host_package/since_days 均可选,limit 默认 1000。
+    """
+    _check_secret(x_internal_secret)
+    since = None
+    if since_days and since_days > 0:
+        since = datetime.now(timezone.utc) - timedelta(days=since_days)
+    rows = repo.list_samples(
+        db,
+        exec_success=exec_success,
+        host_package=host_package,
+        since=since,
+        limit=limit,
+    )
+    return [LaunchConfirmSampleRow.model_validate(r) for r in rows]

app/api/v1/ad.py

@@ -9,8 +9,8 @@
 """
 from __future__ import annotations
 
-import logging
 import json
+import logging
 import uuid
 
 from fastapi import APIRouter, Depends, HTTPException, Request, status
@@ -18,8 +18,8 @@ from fastapi import APIRouter, Depends, HTTPException, Request, status
 from app.api.deps import CurrentUser, DbSession
 from app.core import rewards
 from app.core.config import settings
-from app.integrations import pangle
 from app.core.ratelimit import rate_limit
+from app.integrations import pangle
 from app.repositories import ad_ecpm as crud_ecpm
 from app.repositories import ad_feed_reward as crud_feed
 from app.repositories import ad_reward as crud_ad
@@ -120,6 +120,21 @@ def pangle_callback(request: Request, db: DbSession) -> PangleCallbackOut:
     ad_session_id = extra.get("ad_session_id")
     ecpm = params.get("ecpm")
 
+    # 环境隔离:激励视频 mediaExtra 里带「这次观看属于哪个后端环境」(srv_env=dev/prod,客户端按
+    # BuildConfig.DEBUG 决定),穿山甲 S2S 回调原样带回。回调 URL 在穿山甲后台只配一个(指向生产),
+    # 所以测试包(dev)看广告的 S2S 也会打到生产——若不拦,生产会把奖发给「本库里同 user_id 的另一个
+    # 真人」(user_id 是跨库不隔离的裸数字 → 跨库串号)。这里只处理「属于本服环境」的回调:环境不符
+    # 直接受理但不发币、不写任何记录,保证各环境后台广告收益页只含本环境用户。is_verify=true 让穿山甲
+    # 不再重试(测试用户的币由 localhost 的 test-grant 单独发,不依赖这条 S2S)。
+    # 兼容:旧客户端不带 srv_env(取不到)→ 视为本环境,照常处理,不误伤存量正式用户。
+    callback_env = extra.get("srv_env")
+    if callback_env and callback_env != settings.APP_ENV:
+        logger.info(
+            "pangle callback foreign env skip: callback_env=%s self_env=%s user_id=%d trans_id=%s",
+            callback_env, settings.APP_ENV, user_id, trans_id,
+        )
+        return PangleCallbackOut(is_verify=True, reason=REASON_OK)
+
     existing = crud_ad.find_by_trans(db, trans_id)
     if existing is not None:
         logger.info(
@@ -270,12 +285,13 @@ def ecpm_report(payload: EcpmReportIn, user: CurrentUser, db: DbSession) -> Ecpm
         ad_type=payload.ad_type, ecpm_raw=payload.ecpm,
         ad_session_id=payload.ad_session_id,
         adn=payload.adn, slot_id=payload.slot_id,
+        feed_scene=payload.feed_scene,
         app_env=payload.app_env, our_code_id=payload.our_code_id,
     )
     logger.info(
-        "ad ecpm report user_id=%d type=%s session=%s ecpm=%s adn=%s slot=%s app=%s code=%s",
-        user.id, payload.ad_type, payload.ad_session_id, payload.ecpm, payload.adn, payload.slot_id,
-        payload.app_env, payload.our_code_id,
+        "ad ecpm report user_id=%d type=%s scene=%s session=%s ecpm=%s adn=%s slot=%s app=%s code=%s",
+        user.id, payload.ad_type, payload.feed_scene, payload.ad_session_id, payload.ecpm,
+        payload.adn, payload.slot_id, payload.app_env, payload.our_code_id,
     )
     return EcpmReportOut(ok=True)
 
@@ -386,6 +402,7 @@ def feed_reward(payload: FeedRewardIn, user: CurrentUser, db: DbSession) -> Feed
         client_event_id=payload.client_event_id,
         ecpm=payload.ecpm,
         duration_seconds=payload.duration_seconds,
+        ad_type=payload.ad_type,
         ad_session_id=payload.ad_session_id,
         adn=payload.adn,
         slot_id=payload.slot_id,
@@ -394,6 +411,7 @@ def feed_reward(payload: FeedRewardIn, user: CurrentUser, db: DbSession) -> Feed
         app_env=payload.app_env,
         our_code_id=payload.our_code_id,
         aborted=payload.aborted,
+        display_coin=payload.display_coin,
     )
     logger.info(
         "feed ad reward user_id=%d event=%s status=%s units=%d coin=%d",

app/api/v1/analytics.py

@@ -0,0 +1,31 @@
+"""客户端埋点上报接口。
+
+POST /api/v1/analytics/events — 批量接收新手引导(及后续)埋点,append 落 analytics_event 表。
+**不强制登录**(未登录态也要采集行为):user_id 由客户端在 body 里可选带上,不靠 Bearer。
+服务端补 client_ip(X-Forwarded-For)与 created_at(接收时间 = server_at)。
+"""
+from __future__ import annotations
+
+from fastapi import APIRouter, Request
+
+from app.api.deps import DbSession
+from app.repositories import analytics as analytics_repo
+from app.schemas.analytics import AnalyticsBatchIn, AnalyticsIngestOut
+
+router = APIRouter(prefix="/api/v1/analytics", tags=["analytics"])
+
+
+def _client_ip(request: Request) -> str:
+    """取客户端 IP:生产经 nginx 反代优先 X-Forwarded-For 第一段,否则直连 IP(同 admin get_client_ip)。"""
+    xff = request.headers.get("x-forwarded-for")
+    if xff:
+        return xff.split(",")[0].strip()
+    return request.client.host if request.client else ""
+
+
+@router.post("/events", response_model=AnalyticsIngestOut, summary="批量上报埋点事件")
+def ingest_events(
+    batch: AnalyticsBatchIn, request: Request, db: DbSession
+) -> AnalyticsIngestOut:
+    n = analytics_repo.record_batch(db, batch, client_ip=_client_ip(request))
+    return AnalyticsIngestOut(received=n)

app/api/v1/auth.py

@@ -12,11 +12,11 @@ from __future__ import annotations
 
 import logging
 
-from fastapi import APIRouter, Depends, HTTPException, status
+from fastapi import APIRouter, HTTPException, Request, status
 
 from app.api.deps import CurrentUser, DbSession
 from app.core import test_account
-from app.core.ratelimit import rate_limit
+from app.core.ratelimit import enforce_rate_limit
 from app.core.security import TokenError, decode_token, issue_token_pair
 from app.integrations.jiguang import JiguangError, mask_phone, verify_and_get_phone
 from app.integrations.sms import SmsError, send_code, verify_code
@@ -38,6 +38,12 @@ logger = logging.getLogger("shagua.auth")
 
 router = APIRouter(prefix="/api/v1/auth", tags=["auth"])
 
+# 手机号登录防刷:同一设备(device_id) + 同一 IP 每小时最多的登录尝试次数(成功/失败都计)。
+SMS_LOGIN_MAX_PER_HOUR = 5
+# 发码防刷:同一设备(device_id) + 同一 IP 每小时最多的发码次数。
+# 堵「换手机号绕开单号 60s 冷却」的洞 —— 冷却是单号维度,一机换号能绕开。
+SMS_SEND_MAX_PER_HOUR_PER_DEVICE = 5
+
 
 def _login_response(
     user, *, onboarding_completed: bool, force_onboarding: bool = False
@@ -84,15 +90,27 @@ def jverify_login(req: JverifyLoginRequest, db: DbSession) -> TokenWithUser:
     "/sms/send",
     response_model=SmsSendResponse,
     summary="发送短信验证码",
-    dependencies=[Depends(rate_limit(10, 60, "sms-send"))],  # 同 IP 每分钟≤10 次(防一 IP 刷不同号)
 )
-def sms_send(req: SmsSendRequest) -> SmsSendResponse:
+def sms_send(req: SmsSendRequest, request: Request) -> SmsSendResponse:
     # 测试账号:不真发短信(号码非真实手机号,真发会失败/浪费),直接放行让客户端进入填码界面。
     # 真正的"免验证码"在 /sms/login 跳过校验;每日上限也在 login 处算(send 不耗额度)。
+    # (也不受下面设备发码限流约束:QA 联调要反复发码。)
     if test_account.is_test_account(req.phone):
         logger.info("test_account sms_send short-circuit (不真发)")
         return SmsSendResponse(sent=True, mock=True, cooldown_sec=0)
 
+    # 防刷:同一设备(device_id) + 同一 IP 每小时最多 SMS_SEND_MAX_PER_HOUR_PER_DEVICE 次发码。
+    # 补「换手机号绕开单号 60s 冷却」的洞(冷却是单号维度,一机换号能绕);设备维度按机器封顶,
+    # 挡短信轰炸/烧钱。放在真发(send_code)之前 → 超限直接拦下、不真发短信。
+    enforce_rate_limit(
+        request,
+        scope="sms-send-device",
+        subject=req.device_id,
+        limit=SMS_SEND_MAX_PER_HOUR_PER_DEVICE,
+        window_sec=3600,
+        detail="操作过于频繁,请稍后再试",
+    )
+
     try:
         cooldown = send_code(req.phone)
     except SmsError as e:
@@ -107,11 +125,11 @@ def sms_send(req: SmsSendRequest) -> SmsSendResponse:
     "/sms/login",
     response_model=TokenWithUser,
     summary="手机号+验证码登录",
-    dependencies=[Depends(rate_limit(20, 60, "sms-login"))],  # 防撞库爆破(另有单码失败次数上限)
 )
-def sms_login(req: SmsLoginRequest, db: DbSession) -> TokenWithUser:
+def sms_login(req: SmsLoginRequest, request: Request, db: DbSession) -> TokenWithUser:
     # 测试账号:免验证码登录 + 每日上限 + 每次都走新手引导(详见 app/core/test_account.py)。
     # 放在最前面:命中即不校验验证码;先扣当日额度,超限直接拒,挡住有人猜到号后脚本刷。
+    # 测试账号走自己的每日额度、不受下面 (设备+IP) 每小时限流约束(QA 需在一小时内反复登录联调)。
     if test_account.is_test_account(req.phone):
         if not test_account.try_consume_quota():
             raise HTTPException(status_code=429, detail="测试账号今日使用次数已达上限,请明天再试")
@@ -123,6 +141,19 @@ def sms_login(req: SmsLoginRequest, db: DbSession) -> TokenWithUser:
         # 不读/不写 onboarding_completion 表。
         return _login_response(user, onboarding_completed=False, force_onboarding=True)
 
+    # 防刷:同一设备(device_id) + 同一 IP 每小时最多 SMS_LOGIN_MAX_PER_HOUR 次登录尝试。放在验证码校验
+    # **之前** → 输错验证码的失败尝试也计数,才挡得住撞库/爆破(另有单码失败 SMS_MAX_VERIFY_ATTEMPTS 次即作废兜底)。
+    # ⚠️ 按设备而非手机号 → 一台机器换不同手机号刷登录也受限(防一机狂登多号);device_id 空(老客户端)时
+    #    退化为该 IP 下所有空设备聚一桶,仍受限。
+    enforce_rate_limit(
+        request,
+        scope="sms-login-device",
+        subject=req.device_id,
+        limit=SMS_LOGIN_MAX_PER_HOUR,
+        window_sec=3600,
+        detail="登录尝试过于频繁,请稍后再试",
+    )
+
     if not verify_code(req.phone, req.code):
         raise HTTPException(status_code=400, detail="invalid sms code")
 

app/api/v1/compare.py

@@ -1,48 +1,114 @@
-"""外卖比价业务透传端点。
+"""外卖比价业务透传端点 + 后端 harvest 落库。
 
-把客户端的 POST /api/v1/intent/recognize 和 /api/v1/price/step 请求原样转发给
-pricebot-backend 的 /api/intent/recognize 和 /api/price/step。MVP 阶段**不鉴权**
-(同 coupon/step,见 docs/待办与技术债.md P1:device_id 透传区分设备,待补 JWT +
-device_id↔user_id 绑定后才能做用户级画像)。
+把客户端 POST /api/v1/intent/* 和 /api/v1/price/step 转发给 pricebot,同时**由 app-server
+在透传里直接落库比价记录**(不再靠客户端 POST /compare/record):
+  - 帧0(客户端首帧不带 trace_id)→ app-server 用 uuid 签发 trace_id、注入转发 body、回给
+    客户端;并按 trace_id 建 running 行(harvest_running)。
+  - 最终 done 帧 → 更新成 success/failed + 结果(harvest_done)。发奖不在这里:#113 已把
+    邀请奖口径从「比价」移到「实际下单」(order.py)。
+  - /trace/finalize(用户终止/Phase1 未识别,无 done)→ 更新成 cancelled/failed
+    (harvest_abort,**不降级 success**)。
+trace_url 从 pricebot 响应**顶层 trace_url** 取(pricebot 每帧都带,见其 goal_engine.process_step)。
 
-- Phase 1 /intent/recognize:从源平台(淘宝闪购/美团/京东外卖)购物车页识别店名+菜品+
-  价格,一次性。
-- Phase 2 /price/step:多轮循环,在目标平台复现订单读到手价,直到 done。
+鉴权:软鉴权(OptionalUser)——新客户端带 JWT → 绑 user_id;老客户端不带 → user_id 暂空,
+由其后续 /compare/record 上报补(灰度期两条写路径按 trace_id reconcile,success 不被降级)。
 
-真正的目标驱动比价逻辑在 pricebot-backend(GoalEngine,另一个 repo),本接口只是透传壳。
-电商(ecom)那两个端点 food MVP 暂不需要,以后放开 scene 时再加 /ecom/intent/recognize
-和 /ecom/step 两行即可。
-
-pricebot 协议文档:
-  pricebot-backend/docs/main/02_api_protocol.md
+真正的目标驱动比价逻辑在 pricebot-backend(GoalEngine,另一个 repo),本接口是透传壳 + 落库。
+pricebot 协议文档: pricebot-backend/docs/main/02_api_protocol.md
 """
 from __future__ import annotations
 
 import json
 import logging
+import time
+import uuid
 from typing import Any
 
 import httpx
 from fastapi import APIRouter, HTTPException, Request, status
+from fastapi.concurrency import run_in_threadpool
 
+from app.api.deps import OptionalUser
 from app.core.config import settings
+from app.core.logging import trace_id_ctx
+from app.core.pricebot_client import get_pricebot_client
 from app.core.pricebot_router import pick_pricebot
+from app.db.session import SessionLocal
+from app.repositories import comparison as crud_compare
 
 logger = logging.getLogger("shagua.compare")
 
 router = APIRouter(prefix="/api/v1", tags=["compare"])
 
 
-async def _passthrough(request: Request, upstream_path: str) -> dict[str, Any]:
-    """把请求体原样转发给 pricebot-backend 的 {upstream_path}。
+# ============================================================
+# harvest 落库(阻塞 SQLAlchemy → run_in_threadpool,独立 SessionLocal,不阻塞事件循环;
+# 任何写库异常都吞掉、绝不连累比价透传返回 —— 同 coupon.py 现有 best-effort 写法)。
+# ============================================================
 
-    跟 coupon_step 同款透传壳:不鉴权、不做 schema 校验,仅读 device_id/trace_id/step
-    打日志。比价单帧是大上下文 / 逐帧 LLM,超时用 PRICEBOT_COMPARE_TIMEOUT_SEC(60s,
-    比领券的 30s 长)。
+
+def _harvest_running_blocking(
+    trace_id: str, user_id: int | None, business_type: str,
+    device_id: str | None, device_info: dict | None, trace_url: str | None,
+) -> None:
+    with SessionLocal() as db:
+        crud_compare.harvest_running(
+            db, trace_id=trace_id, user_id=user_id, business_type=business_type,
+            device_id=device_id, device_info=device_info, trace_url=trace_url,
+        )
+    logger.info(
+        "harvest running row (user=%s)", user_id,
+        extra={"phase": "harvest_running", "status": "running", "user_id": user_id},
+    )
+
+
+def _harvest_done_blocking(
+    trace_id: str, user_id: int | None, done_params: dict, business_type: str,
+    device_id: str | None, device_info: dict | None, trace_url: str | None,
+) -> None:
+    with SessionLocal() as db:
+        rec, newly_success = crud_compare.harvest_done(
+            db, trace_id=trace_id, user_id=user_id, done_params=done_params,
+            business_type=business_type, device_id=device_id,
+            device_info=device_info, trace_url=trace_url,
+        )
+        logger.info(
+            "harvest done → %s saved=%s newly=%s", rec.status,
+            rec.saved_amount_cents, newly_success,
+            extra={"phase": "harvest_done", "status": rec.status,
+                   "saved_cents": rec.saved_amount_cents,
+                   "best_platform": rec.best_platform_id, "newly_success": newly_success,
+                   "user_id": user_id},
+        )
+        # 不在此处发邀请奖:#113 已把发奖口径从「比价」移到「实际下单」(order.py),harvest
+        # 只记录比价、不发奖。否则比价先于下单 + try_reward 幂等闸会让奖落在「比价」这步,
+        # 架空 #113 的「下单才发奖」防刷意图(newly_success 仅留作日志观测)。
+
+
+def _harvest_abort_blocking(
+    trace_id: str, status_hint: str, reason: str | None, trace_url: str | None,
+) -> None:
+    with SessionLocal() as db:
+        rec = crud_compare.harvest_abort(
+            db, trace_id=trace_id, status=status_hint, reason=reason, trace_url=trace_url,
+        )
+    logger.info(
+        "harvest abort → %s", (rec.status if rec else "no-row"),
+        extra={"phase": "harvest_abort",
+               "status": (rec.status if rec else None), "reason": reason},
+    )
+
+
+async def _forward(
+    request: Request, upstream_path: str, user, *, harvest_first_frame: bool = True,
+) -> tuple[dict[str, Any], str, dict[str, Any]]:
+    """透传给 pricebot 的 {upstream_path},并 mint trace_id + 首帧建 running 行。
+    返回 (resp_json, trace_id, meta)。
+
+    - 客户端首帧不带 trace_id → app-server 用 uuid 签发、写进转发 body、回填进响应顶层
+      `trace_id`(客户端据此拿到后续帧都带上)。带了(老客户端 / 后续帧)→ 原样用、走原始 bytes 快路。
+    - 仅**本次 mint(=首帧)**建 running 行(idempotent);后续帧不再写库。
     """
-    # 读原始字节,避免"反序列化→再序列化"的双重 JSON(省 ~一半透传 CPU,让 app-server
-    # 单 worker 也扛得住高并发)。只 json.loads 一次拿 trace_id 做亲和 + 打日志,转发时
-    # 直接发原始 bytes(content=raw),不重新 dumps。
     raw = await request.body()
     try:
         meta = json.loads(raw)
@@ -51,26 +117,41 @@ async def _passthrough(request: Request, upstream_path: str) -> dict[str, Any]:
     if not isinstance(meta, dict):
         meta = {}
 
-    # 按 trace_id 一致性 hash 选 pricebot 实例(同一比价的所有帧落同一进程,内存维护 state)
-    base = pick_pricebot(meta.get("trace_id"))
+    trace_id = meta.get("trace_id")
+    minted = False
+    if not trace_id:
+        trace_id = str(uuid.uuid4())
+        meta["trace_id"] = trace_id
+        raw = json.dumps(meta).encode()  # 仅首帧重新序列化(注入 trace_id);后续帧走原始 bytes
+        minted = True
+
+    # 请求级 trace_id 贯穿:此后本请求(含 run_in_threadpool 里的 harvest)每行日志自动带 trace_id
+    trace_id_ctx.set(trace_id)
+
+    base = pick_pricebot(trace_id)
     url = f"{base.rstrip('/')}{upstream_path}"
     timeout = settings.PRICEBOT_COMPARE_TIMEOUT_SEC
+    step = meta.get("step")
 
     logger.info(
-        "compare %s device_id=%s trace_id=%s step=%s",
-        upstream_path,
-        meta.get("device_id"),
-        meta.get("trace_id"),
-        meta.get("step"),
+        "→ pricebot %s step=%s%s", upstream_path, step, " [mint]" if minted else "",
+        extra={"phase": "forward", "endpoint": upstream_path, "step": step,
+               "device_id": meta.get("device_id"), "minted": minted,
+               "query": meta.get("query"), "user_id": (user.id if user else None)},
     )
 
+    t0 = time.monotonic()
     try:
-        async with httpx.AsyncClient(timeout=timeout) as client:
-            resp = await client.post(
-                url, content=raw, headers={"Content-Type": "application/json"}
-            )
+        client = get_pricebot_client()
+        resp = await client.post(
+            url, content=raw, headers={"Content-Type": "application/json"}, timeout=timeout
+        )
     except httpx.RequestError as e:
-        logger.error("[pricebot] request failed: %s", e)
+        logger.error(
+            "← pricebot %s 不可达: %s", upstream_path, e,
+            extra={"phase": "resp", "endpoint": upstream_path, "step": step,
+                   "error": "upstream_unreachable"},
+        )
         raise HTTPException(
             status_code=status.HTTP_502_BAD_GATEWAY,
             detail=f"pricebot upstream unreachable: {e}",
@@ -78,49 +159,112 @@ async def _passthrough(request: Request, upstream_path: str) -> dict[str, Any]:
 
     if resp.status_code >= 500:
         logger.error(
-            "[pricebot] 5xx status=%d body=%s",
-            resp.status_code,
-            resp.text[:500],
+            "← pricebot %s 5xx=%d", upstream_path, resp.status_code,
+            extra={"phase": "resp", "endpoint": upstream_path, "step": step,
+                   "http_status": resp.status_code, "error": "upstream_5xx"},
         )
         raise HTTPException(
             status_code=status.HTTP_502_BAD_GATEWAY,
             detail=f"pricebot upstream returned {resp.status_code}",
         )
 
-    return resp.json()
+    resp_json = resp.json()
+    cost_ms = int((time.monotonic() - t0) * 1000)
+    if not isinstance(resp_json, dict):
+        return {}, trace_id, meta
+    # 回传 app-server 签发的 trace_id(新客户端首帧据此拿到,后续帧都带上它)
+    resp_json.setdefault("trace_id", trace_id)
+    _action = resp_json.get("action") or {}
+    logger.info(
+        "← pricebot %s cmd=%s cont=%s %dms", upstream_path,
+        _action.get("command"), resp_json.get("continue"), cost_ms,
+        extra={"phase": "resp", "endpoint": upstream_path, "step": step,
+               "command": _action.get("command"), "continue": resp_json.get("continue"),
+               "cost_ms": cost_ms},
+    )
+
+    # 首帧建 running 行(仅本次 mint;老客户端自带 trace_id → 不 mint → 由 done / 其 POST 建行)
+    if minted and harvest_first_frame:
+        try:
+            await run_in_threadpool(
+                _harvest_running_blocking, trace_id,
+                (user.id if user else None), "food",
+                meta.get("device_id"), meta.get("device_info"),
+                resp_json.get("trace_url"),
+            )
+        except Exception as e:  # noqa: BLE001
+            logger.warning("harvest_running failed trace=%s: %s", trace_id, e)
+
+    return resp_json, trace_id, meta
 
 
-@router.post("/intent/recognize", summary="外卖比价 Phase 1 意图识别 (透传到 pricebot)")
-async def intent_recognize(request: Request) -> dict[str, Any]:
-    return await _passthrough(request, "/api/intent/recognize")
+@router.post("/intent/recognize", summary="外卖比价 Phase 1 意图识别 (透传 + 建 running 行)")
+async def intent_recognize(request: Request, user: OptionalUser) -> dict[str, Any]:
+    resp, _, _ = await _forward(request, "/api/intent/recognize", user)
+    return resp
 
 
-@router.post("/intent/step", summary="外卖比价 Phase 1 多帧意图识别 (透传到 pricebot, 仅淘宝源)")
-async def intent_step(request: Request) -> dict[str, Any]:
-    # 多帧版意图识别(展开+滚动采集→提取): 循环调用直到 done(done 帧顶层带
-    # result+calibration)。目前仅淘宝源走这条, 其它源走上面单次 /intent/recognize。
-    return await _passthrough(request, "/api/intent/step")
+@router.post("/intent/step", summary="外卖比价 Phase 1 多帧意图识别 (透传, 仅淘宝源)")
+async def intent_step(request: Request, user: OptionalUser) -> dict[str, Any]:
+    resp, _, _ = await _forward(request, "/api/intent/step", user)
+    return resp
 
 
-@router.post(
-    "/intent/precoupon/step",
-    summary="外卖比价 Phase 0 意图识别前先用券 (透传到 pricebot, 仅美团源)",
-)
-async def intent_precoupon_step(request: Request) -> dict[str, Any]:
-    # 美团源平台『意图识别前先用券』多帧循环: 客户端在调 /intent/recognize 之前先循环
-    # 调本端点到 done(continue=false)。订单页底部有『点击使用X红包』就自动选最大免费券
-    # 用上, 已用券/无券则首帧秒过。与 /intent/step 同属 intent 域, 复用同一透传壳。
-    return await _passthrough(request, "/api/intent/precoupon/step")
+@router.post("/intent/precoupon/step", summary="外卖比价 Phase 0 识别前先用券 (透传, 仅美团源)")
+async def intent_precoupon_step(request: Request, user: OptionalUser) -> dict[str, Any]:
+    resp, _, _ = await _forward(request, "/api/intent/precoupon/step", user)
+    return resp
 
 
-@router.post("/price/step", summary="外卖比价 Phase 2 步进 (透传到 pricebot)")
-async def price_step(request: Request) -> dict[str, Any]:
-    return await _passthrough(request, "/api/price/step")
+@router.post("/price/step", summary="外卖比价 Phase 2 步进 (透传 + done 落库)")
+async def price_step(request: Request, user: OptionalUser) -> dict[str, Any]:
+    resp, trace_id, meta = await _forward(request, "/api/price/step", user)
+    # 最终 done 帧(command=done 且 continue=false)→ harvest 更新成终态。
+    # (单平台中途 done 被 pricebot 改写成 wait+continue=true,不会命中这里,同 coupon 语义。)
+    action = resp.get("action") or {}
+    if action.get("command") == "done" and not resp.get("continue", True):
+        done_params = action.get("params") or {}
+        try:
+            await run_in_threadpool(
+                _harvest_done_blocking, trace_id, (user.id if user else None),
+                done_params, "food",
+                meta.get("device_id"), meta.get("device_info"),
+                resp.get("trace_url") or done_params.get("trace_url"),
+            )
+        except Exception as e:  # noqa: BLE001
+            logger.warning("harvest_done failed trace=%s: %s", trace_id, e)
+    return resp
 
 
-@router.post("/trace/finalize", summary="比价 trace 收尾上云 (透传到 pricebot, 终止/未识别拿 trace_url)")
-async def trace_finalize(request: Request) -> dict[str, Any]:
-    # 用户终止 / Phase1 未识别没走到 done 帧, pricebot 没上云也没回传 trace_url。客户端收尾时
-    # 打这个, _passthrough 按 trace_id 一致性 hash 落到处理这条 trace 的同一 pricebot 进程
-    # (dir_cache 在那, 才能算对 trace 目录), 由后者打包上云返回 {trace_url}。
-    return await _passthrough(request, "/api/trace/finalize")
+@router.post("/trace/epilogue", summary="比价结果页尾声帧 (透传到 pricebot, 用户视角截图入 trace)")
+async def trace_epilogue(request: Request, user: OptionalUser) -> dict[str, Any]:
+    # App 收到 done、渲染完结果页后,把自己页面的截图(base64)传给 pricebot 存进 trace
+    # 目录并触发重传 —— trace 里补上"用户实际看到的汇总页"(步骤帧只有目标 App 画面)。
+    # body ~几百 KB(截图 base64), 纯透传壳(harvest_first_frame=False:不建行/不落库,
+    # 该 trace 的记录已由 done/finalize 落终态)。#112 原走 _passthrough,合并到 harvest
+    # 分支后统一走 _forward(_passthrough 已并入它)。
+    resp, _, _ = await _forward(
+        request, "/api/trace/epilogue", user, harvest_first_frame=False,
+    )
+    return resp
+
+
+@router.post("/trace/finalize", summary="比价 trace 收尾上云 (透传 + 夭折落库)")
+async def trace_finalize(request: Request, user: OptionalUser) -> dict[str, Any]:
+    # 用户终止 / Phase1 未识别没到 done 帧: pricebot 打包半截上云返回 {trace_url};
+    # app-server 顺手把该 trace 的 running 行更新成夭折终态(**不降级 success**)。
+    # 客户端 finalize body 带 status(cancelled/failed)+ reason;老客户端只带 trace_id →
+    # 默认 cancelled,其后续 /compare/record 上报再补精确态。
+    resp, trace_id, meta = await _forward(
+        request, "/api/trace/finalize", user, harvest_first_frame=False,
+    )
+    try:
+        await run_in_threadpool(
+            _harvest_abort_blocking, trace_id,
+            (meta.get("status") or "cancelled"),
+            (meta.get("reason") or meta.get("information")),
+            (resp.get("trace_url") if isinstance(resp, dict) else None),
+        )
+    except Exception as e:  # noqa: BLE001
+        logger.warning("harvest_abort failed trace=%s: %s", trace_id, e)
+    return resp

app/api/v1/compare_record.py

@@ -1,15 +1,13 @@
-"""比价记录 endpoint(「我的比价记录」数据源)。
+"""比价记录 endpoint(「我的比价记录」+ admin 数据源)。
 
 路由前缀 `/api/v1/compare`:
-  POST /record         上报一次比价结果(幂等:同 user+trace_id 覆盖)
+  POST /record         (灰度期兼容)老客户端上报,按 **trace_id** 幂等 + 不降级 success
   GET  /records        比价记录列表(游标分页)
   GET  /records/{id}   单条详情(含 raw_payload 全量)
 
-**均需鉴权**(CurrentUser)——与同文件无关的不鉴权透传 `compare.py` 分开:那个是
-转发壳(MVP 不鉴权),这里是按用户维度落库的业务接口,必须有 user_id。
-
-注:本轮只做 server 端,客户端(android 仓)在 done 帧后调 POST /record 上报的改动
-另起一轮(见 app-server docs/待办与技术债.md P1)。
+**均需鉴权**(CurrentUser)。⚠️ 写路径现以 `compare.py` 透传壳的**后端 harvest** 为主
+(帧0 建 running 行 → done/finalize 落终态,新客户端不再 POST);本 POST /record 仅灰度期
+给老客户端用,与 harvest 按 trace_id reconcile。新版覆盖够高后可下线本 POST(阶段3)。
 """
 from __future__ import annotations
 
@@ -21,15 +19,15 @@ from app.api.deps import CurrentUser, DbSession
 from app.db.session import SessionLocal
 from app.models.comparison import ComparisonRecord
 from app.repositories import comparison as crud_compare
-from app.services.pricebot_llm_calls import fetch_llm_calls
 from app.schemas.compare_record import (
     CompareStatsOut,
     ComparisonRecordCreatedOut,
     ComparisonRecordDetailOut,
     ComparisonRecordIn,
-    ComparisonRecordPage,
     ComparisonRecordOut,
+    ComparisonRecordPage,
 )
+from app.services.pricebot_llm_calls import fetch_llm_calls
 
 logger = logging.getLogger("shagua.compare_record")
 
@@ -52,6 +50,8 @@ def report_record(
     # 任务做,不阻塞上报响应(顺带给 pricebot 落盘留足余量)。upsert 已 commit,后台用
     # 独立 session 按 record id 回填 llm_calls + 派生 llm_call_count/retry_count。
     background_tasks.add_task(_backfill_llm_calls, rec.id, rec.trace_id)
+    # 注:邀请发奖已从"比价成功"挪到"实际下单"(见 api/v1/order.py report_order)——
+    # 冰拍板:被邀请人完成比价并实际下单才算邀请成功,仅完成比价不再发奖。
     logger.info(
         "compare record user=%s trace=%s biz=%s status=%s saved=%s (llm_calls backfill queued)",
         user.id,

app/api/v1/coupon.py

@@ -20,6 +20,7 @@ from fastapi.concurrency import run_in_threadpool
 
 from app.api.deps import CurrentUser, DbSession
 from app.core.config import settings
+from app.core.pricebot_client import get_pricebot_client
 from app.core.pricebot_router import pick_pricebot
 from app.db.session import SessionLocal
 from app.repositories import coupon_state as coupon_repo
@@ -28,6 +29,7 @@ from app.schemas.coupon_state import (
     CouponPromptDismissIn,
     CouponPromptShouldShowOut,
     CouponPromptShownIn,
+    CouponSessionIn,
     CouponStatsOut,
 )
 
@@ -141,10 +143,10 @@ async def coupon_step(
     )
 
     try:
-        async with httpx.AsyncClient(timeout=timeout) as client:
-            resp = await client.post(
-                url, content=raw, headers={"Content-Type": "application/json"}
-            )
+        client = get_pricebot_client()
+        resp = await client.post(
+            url, content=raw, headers={"Content-Type": "application/json"}, timeout=timeout
+        )
     except httpx.RequestError as e:
         logger.error("[pricebot] request failed: %s", e)
         raise HTTPException(
@@ -193,6 +195,34 @@ async def coupon_step(
     return resp_json
 
 
+@router.post("/session", summary="领券任务流水上报(admin 领券数据看板数据源)")
+def coupon_session(payload: CouponSessionIn, db: DbSession) -> dict[str, bool]:
+    """客户端两段上报一次领券流水(发起 started / 收尾 completed-failed-abandoned),按 trace_id upsert
+    到 coupon_session。不鉴权(同领券循环 MVP,按 device_id/trace_id);供 admin「领券数据」看板算
+    发起/完成数、耗时分位、机型维度。写库失败不应连累客户端(本就 fire-and-forget),吞掉返回 ok。"""
+    try:
+        coupon_repo.upsert_coupon_session(
+            db,
+            trace_id=payload.trace_id,
+            device_id=payload.device_id,
+            status=payload.status,
+            started_at_ms=payload.started_at_ms,
+            user_id=payload.user_id,
+            platforms=payload.platforms,
+            origin_package=payload.origin_package,
+            device_model=payload.device_model,
+            rom=payload.rom,
+            app_env=payload.app_env,
+            elapsed_ms=payload.elapsed_ms,
+            platform_elapsed=payload.platform_elapsed,
+            claimed_count=payload.claimed_count,
+            trace_url=payload.trace_url,
+        )
+    except Exception as e:  # noqa: BLE001
+        logger.warning("coupon session write failed: %s", e)
+    return {"ok": True}
+
+
 @router.post("/prompt/shown", summary="领券引导窗弹出即上报(按 App 记 shown)")
 def coupon_prompt_shown(payload: CouponPromptShownIn, db: DbSession) -> dict[str, bool]:
     """客户端弹出引导窗那刻调 → 记一条今日 engagement(shown),今天**这个 App** 不再自动弹。

app/api/v1/cps_redirect.py

@@ -94,16 +94,35 @@ def cps_landing(code: str, request: Request, db: Session = Depends(get_db)):
         return HTMLResponse(
             _taobao_landing_html(link.target_url, image_url, code, openid, has_uinfo)
         )
-    # 美团短链 / 京东链接:直接 302 跳
+    # 美团短链 / 京东链接:微信内 + 已拿 openid + 还没授权过头像 → 先出「领券前授权」插页,
+    # 点按钮(交互手势,避免微信快照页)走 snsapi_userinfo 拿昵称头像,授权回来经
+    # /c/{code}?authed=1 自动 302 跳券。其余(非微信 / 未配授权 / 无 openid / 已授权)直接 302。
+    # 产品决定:每次未授权都提示(不做冷却、不种"已提示"cookie)。取消授权由插页前端自动
+    # 转跳目标券页(无单独"暂不授权"出口),拒绝也能拿到券(对齐「绝不阻断领券」)。
+    if (
+        settings.wx_oauth_active
+        and _is_wechat(request)
+        and openid is not None
+        and request.cookies.get(_WX_UINFO_COOKIE) != "1"
+    ):
+        return HTMLResponse(_coupon_auth_landing_html(link.target_url, code))
     return RedirectResponse(link.target_url, status_code=302)
 
 
 @router.get("/wx/oauth/cb", include_in_schema=False)
-def wx_oauth_cb(code: str, state: str, db: Session = Depends(get_db)):
+def wx_oauth_cb(
+    code: str | None = None, state: str | None = None, db: Session = Depends(get_db)
+):
     """微信网页授权回调。state='base:{原code}'(静默拿 openid) 或 'uinfo:{原code}'(补昵称头像)。
     换 openid(+userinfo)→ upsert 用户 → 种 cookie → 302 回落地页。任何失败兜底回落地页,
     绝不阻断用户领券。"""
-    kind, _, orig_code = state.partition(":")
+    kind, _, orig_code = (state or "").partition(":")
+    # 授权弹窗点「取消」时,部分微信版本会带空 code 回调本端点(而非退回落地页)。无 code 换
+    # 不到 openid,直接兜底跳回落地页(美团/京东落地页前端会再转跳目标券页),不报 422。
+    if not code:
+        return RedirectResponse(
+            f"/c/{orig_code}" if orig_code else _FALLBACK_URL, status_code=302
+        )
     try:
         token = wx_oauth.exchange_code(code)  # {openid, access_token, scope, ...}
         openid = token["openid"]
@@ -139,6 +158,12 @@ def cps_copy(code: str, request: Request, db: Session = Depends(get_db)) -> dict
     return {"ok": True}
 
 
+def _js_str(s: str) -> str:
+    """把字符串安全嵌进 <script> 的 JS 字面量:json.dumps 不转义 '<',值里含 '</script>'
+    会逃逸标签 → 把 '<' 转成 \\u003c(JS 仍解析为 '<')。auth/target/淘口令等都走它。"""
+    return json.dumps(s).replace("<", "\\u003c")
+
+
 def _taobao_landing_html(
     token: str, image_url: str, code: str, openid: str | None, has_uinfo: bool
 ) -> str:
@@ -156,8 +181,8 @@ def _taobao_landing_html(
     return (
         _TAOBAO_HTML
         .replace("__IMAGE_URL__", safe_img)
-        .replace("__UINFO_URL__", json.dumps(uinfo_url))
-        .replace("__TOKEN_JS__", json.dumps(token))
+        .replace("__UINFO_URL__", _js_str(uinfo_url))
+        .replace("__TOKEN_JS__", _js_str(token))
     )
 
 
@@ -210,3 +235,66 @@ if(location.search.indexOf('authed=1')>=0){doCopy();history.replaceState(null,''
 </script>
 </body>
 </html>"""
+
+
+def _coupon_auth_landing_html(target_url: str, code: str) -> str:
+    """美团/京东「领券前先授权头像」插页(单按钮)。
+
+    唯一按钮「立即领取优惠券」→ snsapi_userinfo 授权(用户点击=交互手势,避免微信快照页)。
+    - 点「允许」→ 回调拿昵称头像 + 种 wx_uinfo cookie → 经 /c/{code}?authed=1 自动 302 跳券。
+    - 点「取消」→ 微信退回本插页,前端 pageshow/visibilitychange 检测到"刚点过领取"即直接
+      转跳 target(不再设单独的"暂不授权"出口),保证拒绝也能拿到券(对齐「绝不阻断领券」)。
+    AUTH/TARGET 经 json.dumps 安全嵌入 JS 字符串(URL 里的 & 不会被 HTML 转义,直跳更稳)。
+    """
+    redirect_uri = f"{settings.CPS_REDIRECT_BASE.rstrip('/')}/wx/oauth/cb"
+    auth_url = wx_oauth.build_authorize_url(redirect_uri, "snsapi_userinfo", f"uinfo:{code}")
+    return (
+        _COUPON_AUTH_HTML
+        .replace("__AUTH_URL__", _js_str(auth_url))
+        .replace("__TARGET_URL__", _js_str(target_url))
+    )
+
+
+_COUPON_AUTH_HTML = """<!DOCTYPE html>
+<html lang="zh">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width,initial-scale=1,maximum-scale=1,user-scalable=no">
+<title>领取优惠券</title>
+<style>
+*{margin:0;padding:0;box-sizing:border-box;-webkit-tap-highlight-color:transparent}
+body{font-family:-apple-system,"PingFang SC",sans-serif;background:#fff0ef;color:#333;min-height:100vh;display:flex;align-items:center;justify-content:center;padding:24px}
+.card{width:100%;max-width:340px;background:#fff;border-radius:18px;padding:34px 24px;text-align:center;box-shadow:0 8px 24px rgba(255,59,59,.12)}
+.title{font-size:22px;font-weight:800;color:#ff3b3b;margin-bottom:10px}
+.sub{font-size:14px;color:#999;line-height:1.6;margin-bottom:28px}
+.btn{display:block;width:100%;background:linear-gradient(90deg,#ff5b5b,#ff3b3b);color:#fff;font-size:18px;font-weight:800;font-family:inherit;text-align:center;padding:15px;border:none;border-radius:28px;cursor:pointer;box-shadow:0 6px 16px rgba(255,59,59,.4)}
+.btn:active{transform:scale(.98)}
+</style>
+</head>
+<body>
+<div class="card">
+<div class="title">🎁 优惠券待领取</div>
+<div class="sub">点击下方按钮，领取你的专属优惠券</div>
+<button class="btn" onclick="claim()">立即领取优惠券</button>
+</div>
+<script>
+var AUTH=__AUTH_URL__,TARGET=__TARGET_URL__;
+function claim(){
+  try{sessionStorage.setItem('mt_auth_ts',String(Date.now()))}catch(e){}
+  location.href=AUTH;
+}
+// 从授权页返回(点了「取消」/未完成授权):若刚点过领取(3 分钟内)→ 直接进目标券页,
+// 保证拒绝也能拿到券(绝不阻断领券)。pageshow 覆盖 BFCache 与整页重载,visibilitychange
+// 兜底"页面保活只切可见性"的情况;靠 sessionStorage 时间戳防陈旧/误触发。
+function maybeForward(){
+  var ts=0;try{ts=parseInt(sessionStorage.getItem('mt_auth_ts')||'0',10)}catch(e){}
+  if(ts&&Date.now()-ts<180000){
+    try{sessionStorage.removeItem('mt_auth_ts')}catch(e){}
+    location.href=TARGET;
+  }
+}
+window.addEventListener('pageshow',maybeForward);
+document.addEventListener('visibilitychange',function(){if(document.visibilityState==='visible')maybeForward()});
+</script>
+</body>
+</html>"""

app/api/v1/feedback.py

@@ -31,7 +31,18 @@ router = APIRouter(prefix="/api/v1/feedback", tags=["feedback"])
 _MAX_IMAGES = 6
 _CONTENT_MAX = 200
 _CONTACT_MAX = 128
+_SCENE_MAX = 32
 _VALID_RECORD_STATUS = {"pending", "adopted", "rejected"}
+_VALID_SOURCES = {"profile", "comparison"}
+
+
+def _resolve_source(source: str, scene: str) -> str:
+    """反馈来源:客户端显式传的 source 优先(profile / comparison);未传或非法时,
+    按 scene 有无派生——比价结果页反馈才带 scene,故 scene 非空即视作 comparison。"""
+    src = source.strip().lower()
+    if src in _VALID_SOURCES:
+        return src
+    return "comparison" if scene.strip() else "profile"
 
 
 def _app_status(db_status: str) -> str:
@@ -46,10 +57,12 @@ def _record_out(fb) -> FeedbackRecordOut:
     return FeedbackRecordOut(
         id=fb.id,
         content=fb.content,
+        scene=getattr(fb, "scene", None),
         images=fb.images or [],
         status=_app_status(fb.status),
         reject_reason=getattr(fb, "reject_reason", None),
         reward_coins=getattr(fb, "reward_coins", None),
+        admin_reply=getattr(fb, "admin_reply", None),
         created_at=fb.created_at,
     )
 
@@ -61,6 +74,16 @@ async def submit_feedback(
     content: str = Form(...),
     # 原型改版后客户端不再采集联系方式;保留字段以兼容旧端 + 后续可能复用,默认空串。
     contact: str = Form(default=""),
+    # 反馈来源入口:profile(「我的」页)/ comparison(比价结果页)。新端显式传;旧端不带 →
+    # 空串 → 按 scene 有无派生。admin 据此展示/筛选「反馈类型」。
+    source: str = Form(default=""),
+    # 比价结果页反馈的「问题场景」(找错商品/优惠不对…);普通反馈不带 → 空串 → 存 NULL。
+    scene: str = Form(default=""),
+    # 提交端环境快照(admin 反馈页展示「提交版本号」「机型OS版本」);旧端不带 → 空串 → 存 NULL。
+    app_version: str = Form(default=""),
+    device_model: str = Form(default=""),
+    rom_name: str = Form(default=""),
+    android_version: str = Form(default=""),
     images: list[UploadFile] = File(default=[]),
 ) -> FeedbackOut:
     content = content.strip()
@@ -71,6 +94,8 @@ async def submit_feedback(
         raise HTTPException(status_code=400, detail="反馈内容过长")
     if len(contact) > _CONTACT_MAX:
         raise HTTPException(status_code=400, detail="联系方式过长")
+    scene = scene.strip()[:_SCENE_MAX]
+    resolved_source = _resolve_source(source, scene)
 
     files = [f for f in (images or []) if f is not None and f.filename]
     if len(files) > _MAX_IMAGES:
@@ -86,8 +111,13 @@ async def submit_feedback(
 
     fb = feedback_repo.create_feedback(
         db, user_id=user.id, content=content, contact=contact, images=urls,
+        source=resolved_source, scene=scene or None,
+        app_version=app_version.strip(), device_model=device_model.strip(),
+        rom_name=rom_name.strip(), android_version=android_version.strip(),
+    )
+    logger.info(
+        "feedback id=%d user_id=%d source=%s images=%d", fb.id, user.id, resolved_source, len(urls)
     )
-    logger.info("feedback id=%d user_id=%d images=%d", fb.id, user.id, len(urls))
     return FeedbackOut.model_validate(fb)
 
 

app/api/v1/invite.py

@@ -38,7 +38,7 @@ logger = logging.getLogger("shagua.invite")
 router = APIRouter(prefix="/api/v1/invite", tags=["invite"])
 
 _BIND_MESSAGES = {
-    "success": "邀请绑定成功,金币已到账",
+    "success": "邀请绑定成功",
     "already_bound": "你已绑定过邀请人",
     "invalid_code": "邀请码无效",
     "self_invite": "不能填写自己的邀请码",
@@ -84,6 +84,8 @@ def _parse_device_model(ua: str) -> str:
 def my_invite(user: CurrentUser, db: DbSession) -> InviteInfoOut:
     code = invite_repo.ensure_code(db, user)
     invited, coins = invite_repo.get_stats(db, user.id)
+    reward_balance, reward_withdrawn = invite_repo.get_reward_stats(db, user.id)
+    days_left, is_fresh_round, countdown_text = invite_repo.compute_invite_countdown(user.created_at)
     sep = "&" if "?" in settings.INVITE_LANDING_URL else "?"
     share_url = f"{settings.INVITE_LANDING_URL}{sep}ref={code}"
     return InviteInfoOut(
@@ -91,6 +93,11 @@ def my_invite(user: CurrentUser, db: DbSession) -> InviteInfoOut:
         share_url=share_url,
         invited_count=invited,
         coins_earned=coins,
+        reward_balance_cents=reward_balance,
+        reward_withdrawn_cents=reward_withdrawn,
+        countdown_days_left=days_left,
+        countdown_is_fresh_round=is_fresh_round,
+        countdown_text=countdown_text,
     )
 
 
@@ -154,7 +161,7 @@ def landing_track(
     return LandingTrackOut(status="ok")
 
 
-@router.post("/bind", response_model=BindInviteOut, summary="绑定邀请人(注册即生效,双方发金币)")
+@router.post("/bind", response_model=BindInviteOut, summary="绑定邀请人(注册即生效,绑定不发奖)")
 def bind_invite(
     req: BindInviteIn, user: CurrentUser, db: DbSession, request: Request
 ) -> BindInviteOut:

app/api/v1/meituan.py

@@ -25,9 +25,19 @@ from app.schemas.meituan import (
     ReferralLinkResponse,
     TopSalesRequest,
 )
+from app.utils.meituan_city import get_meituan_city
 
 logger = logging.getLogger("shagua.meituan")
 
+
+def _resolve_city_id(latitude: float, longitude: float) -> str:
+    """经纬度 → 美团城市 ID;解析失败返 ""（调用方应降级返空）。"""
+    try:
+        return get_meituan_city(latitude, longitude).get("city_id", "")
+    except Exception:
+        logger.exception("get_meituan_city 失败")
+        return ""
+
 router = APIRouter(prefix="/api/v1/meituan", tags=["meituan-cps"])
 
 
@@ -175,13 +185,19 @@ def feed(req: FeedRequest, db: Session = Depends(get_db)) -> FeedResponse:
         status = "degraded" if (not cards and wm_fail and dd_fail) else ("ok" if cards else "empty")
         return FeedResponse(items=cards, has_next=wm_hn or dd_hn, page=req.page, status=status)
 
-    # 智能推荐(rec):走【离线库】筛佣金率 ≥ 3%,分页返回(SQL 侧去重+排序+分页,秒级、不打美团)。
+    # 智能推荐(rec):走【离线库】筛佣金率 ≥ 3%,按城市过滤,分页返回(SQL 侧去重+排序+分页,秒级、不打美团)。
     # 实测库里佣金≥3% 去重后仅 ~578 条(几乎全是外卖;到店团购佣金普遍 <3%):实时按"同城热销榜单"
     # 拉既撞限流、又填不满(该榜单中位佣金 ~0.8%,筛完每页剩 0-1 条),故从库出。佣金阈值逻辑不变。
     if tab == "rec":
+        city_id = _resolve_city_id(lat, lon)
+        if not city_id:
+            return FeedResponse(items=[], has_next=False, page=req.page, status="degraded")
         PAGE = 20
         try:
-            base = select(MeituanCoupon).where(MeituanCoupon.commission_percent >= 3.0)
+            base = select(MeituanCoupon).where(
+                MeituanCoupon.commission_percent >= 3.0,
+                MeituanCoupon.city_id == city_id,
+            )
             deduped = base.distinct(MeituanCoupon.dedup_key).order_by(
                 MeituanCoupon.dedup_key,
                 MeituanCoupon.commission_percent.desc(),
@@ -211,6 +227,9 @@ def feed(req: FeedRequest, db: Session = Depends(get_db)) -> FeedResponse:
                 card.distance_text = None
                 card.distance_meters = None
                 cards.append(card)
+        if not cards and req.page == 1:
+            # 命中城市却 0 券:该城确无 ≥3% 券,或 ETL 灌的 city_id 与 city_dict 口径不一致。
+            logger.info("[feed] rec city_id=%s 命中 0 券(该城确无券?或 ETL/city_dict 的 city_id 口径不一致)", city_id)
         return FeedResponse(items=cards, has_next=has_next, page=req.page,
                             status="ok" if cards else "empty")
 
@@ -254,14 +273,24 @@ def referral_link(req: ReferralLinkRequest) -> ReferralLinkResponse:
 
 
 @router.post("/top-sales", response_model=CouponListResponse,
-             summary="销量最高(从离线库 meituan_coupon 按销量降序 + 跨源去重,不实时打美团)")
+             summary="销量最高(从离线库 meituan_coupon 按销量降序 + 跨源去重,按城市过滤,不实时打美团)")
 def top_sales(req: TopSalesRequest, db: Session = Depends(get_db)) -> CouponListResponse:
+    # 按设备经纬度定位城市,只查同城券;老客户端不带坐标 → 降级返空(不 422、不误返全城)。
+    if req.latitude is None or req.longitude is None:
+        return CouponListResponse(items=[], has_next=False, search_id=None, status="degraded")
+    city_id = _resolve_city_id(req.latitude, req.longitude)
+    if not city_id:
+        return CouponListResponse(items=[], has_next=False, search_id=None, status="degraded")
+
     # 去重 + 排序 + 分页全在 SQL 做,每页只取并解析当前页 ~20 条。
     # (之前实现每翻一页都全表拉取 + 全量 from_raw 解析,翻页慢 → 客户端滑动卡顿/翻不动。)
     # 库为空(prod 刚部署 / ETL 未跑完)时返空 + status=empty,不崩;库查询异常降级 degraded。
     try:
         # 1) DISTINCT ON (dedup_key):每个去重键(品牌|名|价)只留销量最高那条(同销量再按佣金)
-        base = select(MeituanCoupon).where(MeituanCoupon.sale_volume_num.isnot(None))
+        base = select(MeituanCoupon).where(
+            MeituanCoupon.sale_volume_num.isnot(None),
+            MeituanCoupon.city_id == city_id,
+        )
         if req.platform is not None:
             base = base.where(MeituanCoupon.platform == req.platform)
         deduped = base.distinct(MeituanCoupon.dedup_key).order_by(
@@ -292,6 +321,14 @@ def top_sales(req: TopSalesRequest, db: Session = Depends(get_db)) -> CouponList
         except Exception:  # noqa: BLE001
             continue
         if card.product_view_sign:
+            # 不显示距离:库里的距离是相对城市默认点的(对用户无意义、且误导)。
+            # 置空后前端"距离 店名"那行只剩店名、自动顶到最左(店名移到原距离的位置)。
+            # 逻辑与推荐流保持一致
+            card.distance_text = None
+            card.distance_meters = None
             cards.append(card)
+    if not cards and req.page == 1:
+        # 命中城市却 0 券:可能该城确无券,也可能 ETL 灌的 city_id 与 city_dict 口径不一致(静默降级的隐患)。
+        logger.info("[top-sales] city_id=%s 命中 0 券(该城确无券?或 ETL/city_dict 的 city_id 口径不一致)", city_id)
     return CouponListResponse(items=cards, has_next=has_next, search_id=None,
                               status="ok" if cards else "empty")

app/api/v1/order.py

@@ -1,9 +1,14 @@
+import logging
+
 from fastapi import APIRouter
 
 from app.api.deps import CurrentUser, DbSession
+from app.repositories import invite as crud_invite
 from app.repositories import savings as crud_savings
 from app.schemas.order import OrderReportOut, OrderReportRequest
 
+logger = logging.getLogger("shagua.order")
+
 router = APIRouter(prefix="/api/v1/order", tags=["order"])
 
 
@@ -15,6 +20,20 @@ router = APIRouter(prefix="/api/v1/order", tags=["order"])
 def report_order(req: OrderReportRequest, user: CurrentUser, db: DbSession) -> OrderReportOut:
     # 记账唯一真相表是 savings_record(source='compare')。
     rec, duplicated = crud_savings.create_from_report(db, user.id, req)
+    # 邀请 v2 发奖:被邀请人【实际下单】(而非仅完成比价)才给邀请人发邀请奖励金(幂等,只发一次)。
+    # 触发点从"比价成功上报"挪到这里(冰:完成比价并实际下单才算成功)。仅首次真实上报触发,
+    # 重复上报(duplicated)跳过。best-effort:发奖异常不影响订单上报本身(rec 已 commit),只 log;
+    # try_reward_on_compare 自带 compare_reward_granted 幂等闸,漏发靠后续对账补。
+    if not duplicated:
+        try:
+            reward = crud_invite.try_reward_on_compare(db, user.id)
+            if reward.status == "granted":
+                logger.info(
+                    "invite order reward granted inviter=%s invitee=%s cents=%s",
+                    reward.inviter_user_id, user.id, reward.reward_cents,
+                )
+        except Exception as e:  # noqa: BLE001 best-effort,发奖失败不阻塞订单上报
+            logger.warning("invite order reward failed invitee=%s: %s", user.id, e)
     return OrderReportOut(
         id=rec.id,
         platform=rec.platform or req.platform,

app/api/v1/platform.py

@@ -63,8 +63,8 @@ def ad_config(db: DbSession) -> AdConfigPublicOut:
     return AdConfigPublicOut(
         app_id=c["app_id"],
         reward_code_id=c["reward_code_id"],
-        compare_feed_code_id=c["compare_feed_code_id"],
-        coupon_feed_code_id=c["coupon_feed_code_id"],
+        compare_draw_code_id=c["compare_draw_code_id"],
+        coupon_draw_code_id=c["coupon_draw_code_id"],
         reward_enabled=c["reward_enabled"],
         compare_ad_enabled=c["compare_ad_enabled"],
         coupon_ad_enabled=c["coupon_ad_enabled"],

app/api/v1/user.py

@@ -69,6 +69,18 @@ def complete_onboarding(
     return OkResponse()
 
 
+@router.post("/onboarding/reset", response_model=OkResponse, summary="重置新手引导(删该 设备+账号 完成标记,下次登录重走)")
+def reset_onboarding(
+    req: OnboardingCompleteRequest, user: CurrentUser, db: DbSession
+) -> OkResponse:
+    """删该 (当前账号, device_id) 的引导完成标记 → 下次登录 onboarding_completed=false,客户端重走。
+    与 /onboarding/complete 互逆。device_id 取客户端硬件级 ANDROID_ID(与登录/complete 一致)。
+    幂等:无记录也返回 ok。仅删自己(当前 JWT 用户)这台设备的记录,不影响别的账号/设备。"""
+    deleted = onboarding_repo.delete_completion(db, user_id=user.id, device_id=req.device_id)
+    logger.info("onboarding reset user_id=%d device_len=%d deleted=%d", user.id, len(req.device_id), deleted)
+    return OkResponse()
+
+
 @router.get(
     "/onboarding/status",
     response_model=OnboardingStatusResponse,

app/api/v1/wallet.py

@@ -200,7 +200,11 @@ def withdraw(req: WithdrawRequest, user: CurrentUser, db: DbSession) -> Withdraw
         raise HTTPException(status_code=status.HTTP_503_SERVICE_UNAVAILABLE, detail="wechat pay not configured")
     try:
         order = crud_wallet.create_withdraw(
-            db, user.id, req.amount_cents, user_name=req.user_name, out_bill_no=req.out_bill_no
+            db, user.id, req.amount_cents, source=req.source,
+            user_name=req.user_name, out_bill_no=req.out_bill_no,
+            # 0.01 元调试提现:放行低于最低额的小额。双闸——客户端仅 debug 包在「0.01 元提现」开关开时
+            # 连同 skip_review 一起下发;服务端仅非 prod 才认。生产恒 False,最低额校验照常。
+            allow_sub_min=(req.skip_review and not settings.is_prod),
         )
     except crud_wallet.InvalidWithdrawAmountError as e:
         raise HTTPException(
@@ -274,10 +278,11 @@ def withdraw_status(
 def withdraw_orders(
     user: CurrentUser,
     db: DbSession,
+    source: str | None = Query(None, description="按账户来源过滤:coin_cash / invite_cash;不传=全部"),
     limit: int = Query(20, ge=1, le=100),
     cursor: int | None = Query(None, description="上一页末条 id"),
 ) -> WithdrawOrderPage:
-    items, next_cursor = crud_wallet.list_withdraw_orders(db, user.id, limit=limit, cursor=cursor)
+    items, next_cursor = crud_wallet.list_withdraw_orders(db, user.id, source=source, limit=limit, cursor=cursor)
     return WithdrawOrderPage(
         items=[WithdrawOrderOut.model_validate(it) for it in items],
         next_cursor=next_cursor,

app/core/config.py

@@ -68,7 +68,7 @@ class Settings(BaseSettings):
 
     # 无障碍保护存活监控后台任务(pull 后置检测;本期不接推送)
     HEARTBEAT_MONITOR_ENABLED: bool = True       # 总开关
-    HEARTBEAT_TIMEOUT_MINUTES: int = 10          # 多久没心跳算掉线(≈3 个客户端心跳周期)
+    HEARTBEAT_TIMEOUT_MINUTES: int = 60          # 多久没心跳算掉线(1 小时,避免短暂离线误判被杀)
     HEARTBEAT_SCAN_INTERVAL_SEC: int = 60        # 扫描周期
 
     # ===== 短信 =====
@@ -81,7 +81,6 @@ class Settings(BaseSettings):
     SMS_SIGN_ID: int = 31729       # 极光短信签名 ID(非机密,可被 .env 覆盖)
     SMS_TEMPLATE_ID: int = 1       # 极光短信模板 ID(变量名 code,有效期 5 分钟)
     SMS_CODE_LENGTH: int = 6       # 验证码位数(本服务生成;前端 code 字段 4-8 位兼容)
-    SMS_DAILY_LIMIT_PER_PHONE: int = 10   # 单手机号每日发送上限(防刷 + 控费)
     SMS_MAX_VERIFY_ATTEMPTS: int = 5      # 单个验证码最多校验失败次数,超过即作废(防爆破)
 
     # ===== 测试账号(release 包全流程联调用)=====
@@ -113,6 +112,21 @@ class Settings(BaseSettings):
         """美团 CPS 凭证齐全(缺则接口返空,而非 502)。"""
         return bool(self.MT_CPS_APP_KEY and self.MT_CPS_APP_SECRET)
 
+    # ===== 京东联盟 CPS =====
+    # app_key/app_secret 来自京东联盟应用;site_id 是推广管理里的 APP/网站 ID;
+    # auth_key 是工具商授权 key,自有应用查询可留空。
+    JD_UNION_APP_KEY: str = ""
+    JD_UNION_APP_SECRET: str = ""
+    JD_UNION_SITE_ID: str = ""
+    JD_UNION_AUTH_KEY: str = ""
+    JD_UNION_GATEWAY: str = "https://api.jd.com/routerjson"
+    JD_UNION_TIMEOUT_SEC: int = 15
+
+    @property
+    def jd_union_configured(self) -> bool:
+        """京东联盟订单查询凭证齐全。"""
+        return bool(self.JD_UNION_APP_KEY and self.JD_UNION_APP_SECRET)
+
     # ===== 微信服务号(网页授权) =====
     # CPS 落地页在微信内拿用户 openid(base 静默)/昵称头像(userinfo),做用户级群统计。
     # ⚠️ 区别于 WECHAT_APP_ID(那是 App 移动应用,用于微信支付);这是【已认证服务号】。
@@ -211,6 +225,41 @@ class Settings(BaseSettings):
         """回调开关打开且至少配了一个验签密钥,才接受发奖回调。"""
         return bool(self.PANGLE_CALLBACK_ENABLED and self.pangle_reward_secrets)
 
+    # ===== 穿山甲 GroMore 数据 API(报表收益拉取,T+1)=====
+    # ⚠️ 与上面发奖回调的 m-key 是【两套完全不同的凭证】:这三样在穿山甲后台
+    # 「接入中心 → GroMore-API → 聚合数据报告 API」文档页领取(user_id / role_id / Security Key),
+    # 仅用于按天拉 GroMore 收益报表(revenue 预估收益 + api_revenue 收益Api),不参与发奖。
+    # 该 API 只能查【GroMore 聚合代码位】的数据(=我们 useMediation 的口径),非穿山甲 SDK 数据;
+    # 且不提供用户/设备维度(官方明确),故收益只能落到 日期×代码位 汇总,不能挂到逐条事件。
+    # 子账号(role_id≠user_id)需主账号在「角色管理」授予「查看全部数据」权限,否则查不到
+    # ecpm/revenue(接口返回 118);role_id 填成与 user_id 一致 = 查主账号数据。
+    PANGLE_REPORT_USER_ID: int = 0        # 媒体账号 user_id
+    PANGLE_REPORT_ROLE_ID: int = 0        # 子账号 role_id(=user_id 时查主账号)
+    PANGLE_REPORT_SECURITY_KEY: str = ""  # 该账号的 Security Key(secure_key,≠ 发奖 m-key)
+    # GroMore 聚合 AppId(报表 site_id 维度)→ 我们的应用环境。报表按 site_id 区分两个穿山甲应用,
+    # 用它把每行归到 prod(傻瓜比价正式)/ test(测试应用)。默认值取自现网两个应用,部署时按需覆盖。
+    PANGLE_REPORT_SITE_ID_PROD: str = "5830519"   # 傻瓜比价正式应用 AppId
+    PANGLE_REPORT_SITE_ID_TEST: str = "5832303"   # 测试应用 AppId
+
+    @property
+    def pangle_report_configured(self) -> bool:
+        """三样齐全(user_id / role_id / security_key)才能拉 GroMore 报表;缺任一 → 同步脚本 no-op。"""
+        return bool(
+            self.PANGLE_REPORT_USER_ID
+            and self.PANGLE_REPORT_ROLE_ID
+            and self.PANGLE_REPORT_SECURITY_KEY
+        )
+
+    @property
+    def pangle_report_site_id_to_env(self) -> dict[str, str]:
+        """site_id(穿山甲 AppId)→ app_env(prod/test);供同步脚本把报表行归到我们的应用。留空项忽略。"""
+        out: dict[str, str] = {}
+        if self.PANGLE_REPORT_SITE_ID_PROD.strip():
+            out[self.PANGLE_REPORT_SITE_ID_PROD.strip()] = "prod"
+        if self.PANGLE_REPORT_SITE_ID_TEST.strip():
+            out[self.PANGLE_REPORT_SITE_ID_TEST.strip()] = "test"
+        return out
+
     # ===== Pricebot 上游 (领券/比价业务透传目标) =====
     # pricebot-backend 默认跑在 8000。/api/v1/coupon/step 会透传到这里的 /api/coupon/step
     PRICEBOT_BASE_URL: str = "http://localhost:8000"

app/core/config_schema.py

@@ -11,7 +11,10 @@ from typing import Any
 
 from app.core import rewards as r
 
-# type 约定(给前端渲染编辑控件用):int / int_list / dict_str_int / bool
+# type 约定(给前端渲染编辑控件用):int / int_list / dict_str_int / bool / enum
+# hidden=True:仍是合法可配项(业务照常 get_value / admin 可经专用端点读写),但**不在通用
+#   「系统配置」页渲染**(admin/routers/config.py:list_config 按此过滤)。用于把已下线/已改由
+#   专用页管理的项从福利页 Tab 收起,同时保留后端默认值与写入能力。
 CONFIG_DEFS: dict[str, dict[str, Any]] = {
     "signin_rewards": {
         "default": list(r.SIGNIN_REWARDS), "label": "签到 7 天金币档位",
@@ -30,18 +33,21 @@ CONFIG_DEFS: dict[str, dict[str, Any]] = {
         "default": r.WITHDRAW_MAX_CENTS, "label": "提现最高额(分)",
         "group": "钱包", "type": "int",
     },
+    # 从福利页 Tab 收起(hidden)的项:任务 / 里程碑 整组 + 看广告组里的「单次金币(遗留展示值)/
+    # 每轮次数 / 比价领券信息流广告开关 comparing_ad_enabled」。key 与默认值保留、后端业务照常读取,
+    # 仅不在配置页渲染。看广告组保留可见的:每日上限 / 单次金币上限 / 关闭后冷却。
     "task_rewards": {
         "default": dict(r.TASK_REWARDS), "label": "一次性任务奖励",
-        "group": "任务", "type": "dict_str_int", "help": "task_key → 金币。",
+        "group": "任务", "type": "dict_str_int", "help": "task_key → 金币。", "hidden": True,
     },
     "record_milestones": {
         "default": list(r.RECORD_MILESTONES), "label": "比价里程碑金币档位",
-        "group": "里程碑", "type": "int_list",
+        "group": "里程碑", "type": "int_list", "hidden": True,
         "help": "累计成功比价第 1~N 档解锁发的金币。",
     },
     "ad_reward_coin": {
         "default": r.AD_REWARD_COIN, "label": "看广告单次金币",
-        "group": "看广告", "type": "int",
+        "group": "看广告", "type": "int", "hidden": True,
         "help": "历史兼容/测试展示值;正式发放按 eCPM 公式计算。",
     },
     "ad_daily_limit": {
@@ -54,7 +60,7 @@ CONFIG_DEFS: dict[str, dict[str, Any]] = {
     },
     "ad_round_count": {
         "default": r.VIDEO_ROUND_REQUIRED_COUNT, "label": "每轮看广告次数",
-        "group": "看广告", "type": "int", "help": "当前为 1,表示每次广告关闭后触发短冷却。",
+        "group": "看广告", "type": "int", "hidden": True, "help": "当前为 1,表示每次广告关闭后触发短冷却。",
     },
     "ad_cooldown_sec": {
         "default": r.VIDEO_ROUND_COOLDOWN_SECONDS, "label": "广告关闭后冷却(秒)",
@@ -67,7 +73,7 @@ CONFIG_DEFS: dict[str, dict[str, Any]] = {
     },
     "comparing_ad_enabled": {
         "default": True, "label": "比价/领券期信息流广告",
-        "group": "看广告", "type": "bool",
+        "group": "看广告", "type": "bool", "hidden": True,
         "help": (
             "开启后,比价进行中 + 领券等候期会在悬浮窗展示穿山甲信息流广告(变现行为);"
             "关闭则全程不出广告。客户端按 app 启动 / 每场比价开始时拉取并缓存,故为「最终一致」的"
@@ -84,4 +90,10 @@ CONFIG_DEFS: dict[str, dict[str, Any]] = {
             "提现页「批量对账」手动按钮不受影响。"
         ),
     },
+    # 首页轮播数据源(hidden:不进通用配置 Tab,由「首页轮播种子」页的专用端点 /marquee-seeds/mode 读写)。
+    "marquee_feed_mode": {
+        "default": "mixed", "label": "首页轮播数据源",
+        "group": "首页轮播", "type": "enum", "hidden": True,
+        "help": "mixed=真实优先+种子补位(默认);real=只用真实比价记录;seed=只用种子/合成(演示)。",
+    },
 }

app/core/logging.py

@@ -2,9 +2,15 @@
 
 业务代码用 `logger = logging.getLogger("shagua.xxx")` 即可, 本模块在 main.py 启动时调一次。
 
-- 控制台(stdout): 人类可读文本, 给 systemd / 本地看。
-- 文件 `logs/app-server.log`: 单行 JSON, 供阿里云 SLS/Logtail 采集(JSON 模式零正则);
-  异常栈作为字段内嵌不换行 → 每条日志一行。
+- 控制台(stdout): 人类可读文本, 给 systemd / 本地看; 有 trace 时行尾附 `trace=xxx`。
+- 文件 `logs/app-server.log`: 单行 JSON, 供**阿里云 SLS / Logtail** 采集(JSON 模式零正则);
+  异常栈内嵌为字段不换行 → 每条日志一行。
+- **结构化字段(SLS 可直接查/聚合)**:
+  - `trace_id`: 请求级贯穿——在入口 `trace_id_ctx.set(...)` 后, 本请求内**每一行日志**(含
+    run_in_threadpool 里的 harvest, contextvars 自动拷进线程)都自动带上, 无需手写。
+    SLS 里 `trace_id: "xxx"` 一查即得整条比价链路, 按 time 升序即请求顺序。
+  - 任意 `logger.info(msg, extra={"phase": ..., "step": ..., "command": ...})` 的 extra
+    键都会平铺进 JSON 顶层 → SLS 可按 phase/step/command/cost_ms 等过滤聚合。
 - 环境变量:
   - LOG_JSON_CONSOLE=1   控制台也输出 JSON
   - LOG_DIR / LOG_FILE   改落盘路径(默认 logs/app-server.log)
@@ -17,13 +23,36 @@ import json
 import logging
 import os
 import sys
+from contextvars import ContextVar
 from datetime import datetime
 from logging.handlers import RotatingFileHandler
 from pathlib import Path
 
 
+# 请求级 trace_id:入口(如 compare.py 透传壳)set 之后, 本请求上下文(含 run_in_threadpool
+# 拷贝出去的线程)内所有日志自动带上。默认空串 = 非请求上下文(启动/后台 worker)。
+trace_id_ctx: ContextVar[str] = ContextVar("trace_id", default="")
+
+
+# 标准 LogRecord 属性 + 格式化期附加项:凡不在此集合的 record 属性都视为业务 extra, 平铺进 JSON。
+_RESERVED = set(
+    logging.LogRecord("", 0, "", 0, "", (), None).__dict__
+) | {"message", "asctime", "trace_id", "taskName"}
+
+
+class _ContextFilter(logging.Filter):
+    """把 trace_id_ctx 注入每条 record(供两个 formatter 取用)。挂在 handler 上,
+    命中每条(含 propagate 上来的)记录, 在 format 之前置好 record.trace_id。"""
+
+    def filter(self, record: logging.LogRecord) -> bool:
+        if not hasattr(record, "trace_id"):
+            record.trace_id = trace_id_ctx.get()
+        return True
+
+
 class JsonFormatter(logging.Formatter):
-    """把 LogRecord 序列化成单行 JSON(SLS/Logtail 友好)。异常栈内嵌为字段, 整条仍是一行。"""
+    """LogRecord → 单行 JSON(SLS/Logtail 友好)。trace_id 提到顶层、extra 键平铺, 异常栈内嵌。"""
+
     def __init__(self, service: str = "app-server"):
         super().__init__()
         self.service = service
@@ -35,10 +64,17 @@ class JsonFormatter(logging.Formatter):
             "level": record.levelname,
             "service": self.service,
             "logger": record.name,
-            "func": record.funcName,
-            "line": record.lineno,
-            "message": record.getMessage(),
         }
+        tid = getattr(record, "trace_id", "") or trace_id_ctx.get()
+        if tid:
+            data["trace_id"] = tid
+        # 业务 extra 字段(phase / step / endpoint / command / cost_ms / status ...)平铺进顶层
+        for k, v in record.__dict__.items():
+            if k not in _RESERVED and not k.startswith("_"):
+                data[k] = v
+        data["func"] = record.funcName
+        data["line"] = record.lineno
+        data["message"] = record.getMessage()
         if record.exc_info:
             data["exception"] = self.formatException(record.exc_info)
         if record.stack_info:
@@ -46,6 +82,15 @@ class JsonFormatter(logging.Formatter):
         return json.dumps(data, ensure_ascii=False, default=str)
 
 
+class TextFormatter(logging.Formatter):
+    """控制台文本:标准行 + 有 trace_id 时行尾附 `trace=xxx`(无 trace 的启动/后台日志不加噪)。"""
+
+    def format(self, record: logging.LogRecord) -> str:
+        base = super().format(record)
+        tid = getattr(record, "trace_id", "") or trace_id_ctx.get()
+        return f"{base} trace={tid}" if tid else base
+
+
 _CONFIGURED = False
 
 
@@ -63,14 +108,17 @@ def setup_logging(debug: bool = False) -> None:
     for h in list(root.handlers):
         root.removeHandler(h)
 
+    ctx_filter = _ContextFilter()
+
     # 控制台: 默认文本(systemd/本地看); LOG_JSON_CONSOLE=1 时输出 JSON
     console = logging.StreamHandler(sys.stdout)
     if os.getenv("LOG_JSON_CONSOLE") == "1":
         console.setFormatter(JsonFormatter(service))
     else:
         console.setFormatter(
-            logging.Formatter("%(asctime)s %(levelname)s %(name)s: %(message)s")
+            TextFormatter("%(asctime)s %(levelname)s %(name)s: %(message)s")
         )
+    console.addFilter(ctx_filter)
     root.addHandler(console)
 
     # 文件: 单行 JSON, 供 Logtail 采集(自动轮转, 单文件 10MB, 保留 5 个)
@@ -82,10 +130,15 @@ def setup_logging(debug: bool = False) -> None:
         log_file, maxBytes=10 * 1024 * 1024, backupCount=5, encoding="utf-8",
     )
     file_handler.setFormatter(JsonFormatter(service))
+    file_handler.addFilter(ctx_filter)
     root.addHandler(file_handler)
 
     # 第三方库降噪
     logging.getLogger("httpx").setLevel(logging.WARNING)
     logging.getLogger("httpcore").setLevel(logging.WARNING)
+    # uvicorn --reload 的文件监视器: DEBUG 下会把每次"检测到变更"打出来。我们的 root 文件
+    # handler 又把这行写回 logs/app-server.log → watchfiles 再次检测 → 自我喂食死循环。
+    # 降到 WARNING 同时消除噪音和这个回环(真正的 .py 热重载不受影响)。
+    logging.getLogger("watchfiles").setLevel(logging.WARNING)
 
     _CONFIGURED = True

app/core/pricebot_client.py

@@ -0,0 +1,35 @@
+"""透传到 pricebot 的共享 httpx.AsyncClient 单例。
+
+为什么不能每请求新建(coupon.py / compare.py 老写法 async with httpx.AsyncClient(...)):
+  ① 每次构造都重建一套 SSL 上下文(httpx.create_ssl_context 加载 certifi CA),实测
+     ~1s+/次;而 pricebot 是纯 http 透传,根本用不到 TLS → 纯浪费,且每帧重交一次。
+  ② trust_env 默认 True 会读进程 HTTP_PROXY,把 http://localhost:8000 这条本地透传整个
+     塞进本机代理(如 Clash 7897),恒定再多几秒。
+单例:启动只建一次(SSL/连接池一次性),keep-alive 复用 TCP,每帧降到个位数 ms。
+trust_env=False:对齐 integrations/meituan.py 的既有约定,不被进程代理误导,直连 pricebot。
+"""
+from __future__ import annotations
+
+import httpx
+
+_client: httpx.AsyncClient | None = None
+
+
+def get_pricebot_client() -> httpx.AsyncClient:
+    """取透传单例。lifespan 启动会预热;未预热(如测试态)懒建兜底。
+
+    超时不在此固化(coupon 30s / compare 60s 不同),由调用点 client.post(timeout=...) 传。
+    懒建无 await,asyncio 单线程下不会有并发竞态。
+    """
+    global _client
+    if _client is None:
+        _client = httpx.AsyncClient(trust_env=False)
+    return _client
+
+
+async def aclose_pricebot_client() -> None:
+    """lifespan 关停时调,优雅关连接池。"""
+    global _client
+    if _client is not None:
+        await _client.aclose()
+        _client = None

app/core/ratelimit.py

@@ -57,3 +57,29 @@ def rate_limit(limit: int, window_sec: float, scope: str):
             )
 
     return _dep
+
+
+def enforce_rate_limit(
+    request: Request,
+    scope: str,
+    subject: str,
+    limit: int,
+    window_sec: float,
+    *,
+    detail: str = "操作过于频繁,请稍后再试",
+) -> None:
+    """在路由内部手动限流,按 (subject, 客户端 IP) 计数。
+
+    用于限流 key 需要请求体字段(如手机号)、Depends 阶段还拿不到 body 时
+    —— 此时无法用 [rate_limit] 依赖,改在 handler 解析完 body 后调用本函数。
+    key = `scope:subject:client_ip`;同一 (subject, IP) 在 window_sec 内超过 limit 次 → 抛 429。
+    受 [settings.RATE_LIMIT_ENABLED] 总开关控制(与 [rate_limit] 一致)。
+    """
+    if not settings.RATE_LIMIT_ENABLED:
+        return
+    key = f"{scope}:{subject}:{_client_ip(request)}"
+    if not _hit(key, limit, window_sec):
+        raise HTTPException(
+            status_code=status.HTTP_429_TOO_MANY_REQUESTS,
+            detail=detail,
+        )

app/core/rewards.py

@@ -113,10 +113,9 @@ PRICE_REPORT_REWARD_COINS: int = 1000
 FEEDBACK_REWARD_MAX_COINS: int = 10000
 
 
-# ===== 邀请好友(注册即生效,邀请人 + 被邀请人各发金币)=====
-# 10000 金币 = 1 元,双方各得 1 元。MVP 先用固定常量(不走 app_config)。
-INVITE_INVITER_COINS: int = 10000
-INVITE_INVITEE_COINS: int = 10000
+# ===== 邀请好友(绑定即生效,绑定只建归因关系、双方都不发钱)=====
+# v3(冰 2026-06-26):废除 v1 的"绑定双方各发金币"——被邀请人无奖励、邀请人改比价后发现金
+# (见下方 INVITE_COMPARE_REWARD_CENTS)。原 INVITE_INVITER_COINS / INVITE_INVITEE_COINS 已删。
 # "新用户闸":被邀请人必须在注册后此窗口内绑定才发奖(挡存量老用户互相填码薅羊毛)。
 # 自动绑(剪贴板)在首次注册登录后几秒内发生;留 72h 给手动填码兜底。
 INVITE_NEW_USER_WINDOW_HOURS: int = 72
@@ -127,6 +126,13 @@ INVITE_NEW_USER_WINDOW_HOURS: int = 72
 INVITE_FP_WINDOW_DAYS: int = 7
 
 
+# ===== 邀请好友 v2(好友"下载+登录+比价一次"→ 给邀请人发邀请奖励金·现金)=====
+# v2 新规则:不再注册即发金币,改"好友完成首次成功比价"才给【邀请人】发奖,发的是【现金·分】
+# 进独立的邀请奖励金账户(coin_account.invite_cash_balance_cents),与金币体系物理隔离。
+# 200 分 = 2 元。⚠️ 金额待产品定准:邀请主页=2元 / 福利入口=3.5元 不一致,定后改此处。
+INVITE_COMPARE_REWARD_CENTS: int = 200
+
+
 # ===== 看激励视频 / 信息流广告发金币 =====
 # eCPM 取自穿山甲 SDK getShowEcpm().getEcpm(),官方口径单位是【分/千次展示】(不是元!
 # csjplatform 文档原文"通过 getEcpm 获取的单位是分")。计算时先 ÷100 转成元;

app/integrations/jd_union.py

@@ -0,0 +1,163 @@
+"""京东联盟 OpenAPI 客户端。
+
+当前只接数据大盘需要的订单明细接口:
+`jd.union.open.order.row.query`。京东要求订单查询时间窗最长 1 小时,
+调用方负责切窗分页。
+"""
+from __future__ import annotations
+
+import hashlib
+import json
+import logging
+from datetime import datetime, timedelta, timezone
+from typing import Any
+
+import httpx
+
+from app.core.config import settings
+
+logger = logging.getLogger(__name__)
+
+_BEIJING = timezone(timedelta(hours=8))
+
+
+class JdUnionError(RuntimeError):
+    """京东联盟 API 调用失败。"""
+
+
+def _parse_json_maybe(value: Any) -> Any:
+    if not isinstance(value, str):
+        return value
+    text = value.strip()
+    if not text:
+        return value
+    try:
+        return json.loads(text)
+    except json.JSONDecodeError:
+        return value
+
+
+def _sign(params: dict[str, Any], secret: str) -> str:
+    pieces = [secret]
+    for key in sorted(k for k in params if k != "sign"):
+        value = params[key]
+        if value is None:
+            continue
+        pieces.append(f"{key}{value}")
+    pieces.append(secret)
+    raw = "".join(pieces)
+    return hashlib.md5(raw.encode("utf-8")).hexdigest().upper()
+
+
+def _unwrap_response(data: dict[str, Any]) -> dict[str, Any]:
+    if "error_response" in data:
+        err = data["error_response"] or {}
+        msg = err.get("zh_desc") or err.get("en_desc") or err.get("msg") or err
+        raise JdUnionError(f"京东 API 错误: {msg}")
+
+    body: Any = data
+    for key, value in data.items():
+        if key.endswith("_responce") or key.endswith("_response"):
+            body = value
+            break
+
+    body = _parse_json_maybe(body)
+    if not isinstance(body, dict):
+        raise JdUnionError("京东 API 返回格式异常")
+
+    result = body.get("queryResult", body.get("result", body))
+    result = _parse_json_maybe(result)
+    if not isinstance(result, dict):
+        raise JdUnionError("京东 API 业务结果格式异常")
+
+    code = str(result.get("code", result.get("resultCode", "200")))
+    if code not in {"0", "200"}:
+        msg = result.get("message") or result.get("msg") or result.get("resultMsg") or result
+        raise JdUnionError(f"京东 API 业务错误: {msg}")
+    return result
+
+
+def call(method: str, payload: dict[str, Any], *, version: str = "1.0") -> dict[str, Any]:
+    if not settings.jd_union_configured:
+        raise JdUnionError("京东联盟凭证未配置")
+
+    biz_json = json.dumps(payload, ensure_ascii=False, separators=(",", ":"))
+    params: dict[str, Any] = {
+        "method": method,
+        "app_key": settings.JD_UNION_APP_KEY,
+        "timestamp": datetime.now(_BEIJING).strftime("%Y-%m-%d %H:%M:%S"),
+        "format": "json",
+        "v": version,
+        "sign_method": "md5",
+        "360buy_param_json": biz_json,
+    }
+    params["sign"] = _sign(params, settings.JD_UNION_APP_SECRET)
+
+    try:
+        with httpx.Client(timeout=settings.JD_UNION_TIMEOUT_SEC, trust_env=False) as client:
+            resp = client.post(settings.JD_UNION_GATEWAY, data=params)
+            resp.raise_for_status()
+            data = resp.json()
+    except httpx.HTTPError as e:
+        raise JdUnionError(f"京东 API 网络错误: {e}") from e
+    except json.JSONDecodeError as e:
+        raise JdUnionError("京东 API 返回非 JSON") from e
+
+    return _unwrap_response(data)
+
+
+def _extract_rows(result: dict[str, Any]) -> tuple[list[dict[str, Any]], bool]:
+    payload = _parse_json_maybe(result.get("data", result.get("result", result)))
+    has_more = bool(result.get("hasMore") or result.get("has_more"))
+
+    if isinstance(payload, dict):
+        for key in ("orderRowResp", "orderRows", "orderList", "orders", "list", "rows"):
+            rows = _parse_json_maybe(payload.get(key))
+            if isinstance(rows, list):
+                return [r for r in rows if isinstance(r, dict)], bool(
+                    payload.get("hasMore") or payload.get("has_more") or has_more
+                )
+        return [], bool(payload.get("hasMore") or payload.get("has_more") or has_more)
+    if isinstance(payload, list):
+        return [r for r in payload if isinstance(r, dict)], has_more
+    return [], has_more
+
+
+def query_order_rows(
+    *,
+    start_time: datetime,
+    end_time: datetime,
+    query_time_type: int = 3,
+    page_index: int = 1,
+    page_size: int = 200,
+) -> dict[str, Any]:
+    """查询京东 CPS 订单行。
+
+    query_time_type: 1 下单时间, 2 完成时间, 3 更新时间。
+    start_time/end_time 用北京时间展示给京东;调用方需保证窗口不超过 1 小时。
+    """
+    start_bj = start_time.astimezone(_BEIJING)
+    end_bj = end_time.astimezone(_BEIJING)
+    if end_bj <= start_bj:
+        return {"rows": [], "has_more": False}
+    if end_bj - start_bj > timedelta(hours=1):
+        raise JdUnionError("京东订单查询单次时间窗不能超过 1 小时")
+
+    order_req: dict[str, Any] = {
+        "pageIndex": page_index,
+        "pageSize": min(max(page_size, 1), 200),
+        "type": query_time_type,
+        "startTime": start_bj.strftime("%Y-%m-%d %H:%M:%S"),
+        "endTime": end_bj.strftime("%Y-%m-%d %H:%M:%S"),
+    }
+    if settings.JD_UNION_AUTH_KEY:
+        order_req["key"] = settings.JD_UNION_AUTH_KEY
+    result = call("jd.union.open.order.row.query", {"orderReq": order_req})
+    rows, has_more = _extract_rows(result)
+    logger.info(
+        "jd.union.open.order.row.query fetched rows=%s page=%s has_more=%s",
+        len(rows),
+        page_index,
+        has_more,
+    )
+    return {"rows": rows, "has_more": has_more, "raw": result}

app/integrations/pangle_report.py

@@ -0,0 +1,148 @@
+"""穿山甲 GroMore「聚合数据报告 API」客户端 —— 按天拉取收益报表(只读、T+1)。
+
+⚠️ 这是 **GroMore 数据拉取 API**,与发奖回调验签(integrations/pangle.py 的 m-key)是
+两套完全不同的凭证与用途。凭证在后台「接入中心 → GroMore-API → 聚合数据报告 API」领取:
+媒体账号 user_id、子账号 role_id、Security Key(secure_key)。
+
+鉴权(文档 v2.x「方法一」):
+  1. 去掉请求参数里的 `sign` 字段与值为空的字段;
+  2. 其余参数按 key 字典序升序,拼成 `k1=v1&k2=v2&...&kn=vn`;
+  3. 末尾直接拼接 security_key(无分隔符),对整串做 MD5,取 32 位小写十六进制 = sign。
+  签名有 3 分钟过期(天级用 `timestamp` 秒级时间戳;小时级才用 `current_time` 字符串)。
+
+数据口径要点(来自官方文档):
+  - 只返回【GroMore 聚合代码位】在 GroMore 内的数据(=我们 useMediation 的口径),
+    查不到穿山甲 SDK 自身的数据;
+  - **不提供分用户/设备维度**(官方 FAQ 明确拒绝),最细到 日期×应用×代码位×广告源;
+  - `revenue` = 预估收益(元,所有 ADN 都有);`api_revenue` = 收益Api(各 ADN 经 Reporting
+    回传、按实时汇率折算账号币种,更接近结算),需后台为该 ADN 配置 Reporting 才有、且不支持当天;
+  - 「今天」与「今天以前」必须分开查;天级跨度 ≤ 1 个月、不早于 12 个月。
+"""
+from __future__ import annotations
+
+import hashlib
+import logging
+import time
+from collections.abc import Mapping
+from typing import Any
+
+import httpx
+
+from app.core.config import settings
+
+logger = logging.getLogger("shagua.pangle_report")
+
+HOST = "https://www.csjplatform.com"
+# 天级收益报表(路径与文档代码示例一致;另有小时级 get_hour_report_data,本服务只用天级)。
+DAILY_PATH = "/union_media/open/api/mediation/get_daily_income_report_data"
+VERSION = "2.0"
+SIGN_TYPE = "MD5"
+PAGE_LIMIT = 5000          # 文档上限;一次尽量多取,减少翻页
+DEFAULT_TIMEOUT = 30.0
+_MAX_PAGES = 200           # 翻页安全阀(5000×200=100w 行,远超我们规模),防异常 has_next 死循环
+
+
+class PangleReportError(Exception):
+    """GroMore 报表接口调用失败(未配置 / 网络 / 业务码非 100)。"""
+
+
+def build_sign(params: Mapping[str, Any], security_key: str) -> str:
+    """按文档「方法一」生成 sign:去 sign/空值 → key 升序 → k=v& 拼接 → 末尾接 secure_key → MD5(32 位小写)。
+
+    与官方 Python/Java 示例逐字节一致(见 tests/test_pangle_report.py 的两个测试向量)。
+    """
+    items = [
+        (str(k), str(v))
+        for k, v in params.items()
+        if k != "sign" and v is not None and str(v) != ""
+    ]
+    items.sort(key=lambda kv: kv[0])
+    raw = "&".join(f"{k}={v}" for k, v in items)
+    return hashlib.md5((raw + security_key).encode("utf-8")).hexdigest()
+
+
+def fetch_daily_report(
+    *,
+    start_date: str,
+    end_date: str,
+    # ⚠️ 用 ad_unit_id(GroMore 广告位ID = 我们的 104xxx),不要用 code_id(底层各 ADN 代码位,对不上)
+    dimensions: str = "date,site_id,ad_unit_id",
+    metrics: str = "revenue,api_revenue,ecpm,imp_cnt",
+    site_ids: str | None = None,
+    code_ids: str | None = None,
+    os: str | None = None,
+    network: str | None = None,
+    time_zone: int = 8,
+    limit: int = PAGE_LIMIT,
+    timeout: float = DEFAULT_TIMEOUT,
+) -> list[dict[str, Any]]:
+    """拉一个日期区间(北京时间,time_zone=8)的天级收益报表,自动翻页,返回 report_list 全量行。
+
+    每行是字符串字典(接口原值),形如
+    {"start_date": "2026-06-27", "site_id": "5830519", "ad_unit_id": "104142227",
+     "revenue": "1.23", "api_revenue": "1.05", "ecpm": "0.80", "imp_cnt": "1537", ...}。
+    业务码非 100 直接抛 PangleReportError(由调用方/脚本兜底,不静默吞)。
+    """
+    if not settings.pangle_report_configured:
+        raise PangleReportError(
+            "PANGLE_REPORT_USER_ID / ROLE_ID / SECURITY_KEY 未配置,无法拉取 GroMore 报表"
+        )
+
+    base_params: dict[str, Any] = {
+        "user_id": settings.PANGLE_REPORT_USER_ID,
+        "role_id": settings.PANGLE_REPORT_ROLE_ID,
+        "version": VERSION,
+        "sign_type": SIGN_TYPE,
+        "start_date": start_date,
+        "end_date": end_date,
+        "dimensions": dimensions,
+        "metrics": metrics,
+        "time_zone": time_zone,
+        "limit": limit,
+    }
+    # 可选过滤(空则不传,避免进签名串)
+    for key, val in (
+        ("site_ids", site_ids), ("code_ids", code_ids),
+        ("os", os), ("network", network),
+    ):
+        if val:
+            base_params[key] = val
+
+    rows: list[dict[str, Any]] = []
+    offset = 0
+    try:
+        with httpx.Client(timeout=timeout) as client:
+            for _ in range(_MAX_PAGES):
+                params = dict(base_params)
+                params["offset"] = offset
+                # timestamp 每页临请求时取最新(3 分钟过期),并参与签名
+                params["timestamp"] = int(time.time())
+                params["sign"] = build_sign(params, settings.PANGLE_REPORT_SECURITY_KEY)
+
+                resp = client.get(HOST + DAILY_PATH, params=params)
+                resp.raise_for_status()
+                body = resp.json()
+
+                code = str(body.get("code"))
+                if code != "100":
+                    raise PangleReportError(
+                        f"GroMore 报表业务失败 code={code} message={body.get('message')!r} "
+                        f"(101=验签失败/102=userid无效/107=无权限/118=无收益查看权限,详见文档状态码)"
+                    )
+
+                data = body.get("data") or {}
+                page = data.get("report_list") or []
+                rows.extend(page)
+
+                # has_next=1 还有下一页;无该字段则按本页是否取满判断
+                has_next = str(data.get("has_next", "")) == "1"
+                if not page or not has_next:
+                    break
+                offset += len(page)
+    except httpx.HTTPError as e:
+        raise PangleReportError(f"GroMore 报表请求异常: {e}") from e
+
+    logger.info(
+        "GroMore 天级报表拉取完成 %s~%s dims=%s 行数=%d", start_date, end_date, dimensions, len(rows)
+    )
+    return rows

app/integrations/sms.py

@@ -8,13 +8,16 @@
     校验)→ 鉴权复用极光一键登录的 `JG_APP_KEY`/`JG_MASTER_SECRET`(同一极光应用)。
 
 验证码存储:**进程内存**(单 worker uvicorn 够用)。重启丢失(用户重发即可)。多
-worker / 多机时内存不共享 → 冷却、每日上限、校验都会失效,届时迁移到 DB/Redis。
+worker / 多机时内存不共享 → 冷却、校验都会失效,届时迁移到 DB/Redis。
 见 docs/待办与技术债.md。
 
-防刷三层(短信花钱 + `/sms/send` 在登录前无法 JWT 鉴权):
+防刷两层(短信花钱 + `/sms/send` 在登录前无法 JWT 鉴权):
   1. 单号 `SMS_SEND_INTERVAL_SEC` 冷却(本文件)
-  2. 单号每日 `SMS_DAILY_LIMIT_PER_PHONE` 条上限(本文件)
-  3. 单 IP 频控(api 层 rate_limit 依赖)+ 极光控制台 IP 白名单/防轰炸(运维侧)
+  2. 单设备(device_id)每小时频控(api 层 auth.sms_send 内 enforce_rate_limit)+ 极光控制台 IP 白名单/防轰炸(运维侧)。
+     ⚠️ 原「单 IP 频控(rate_limit 依赖)」2026-06-26 按产品要求删除、改设备维度;但 device_id 客户端可伪造/轮换,
+        脚本轮换 id 能绕过本层 → 挡脚本狂发主要靠极光控制台侧(+ 可选 nginx 限流)。
+     ⚠️ 原「单号每日上限」2026-07-03 按精简要求删除(mentor 定:登录风控只留单号冷却 + 单设备频控);
+        单号维度现仅剩 60s 冷却,「换号轰炸」由单设备频控封顶。
 另:单码校验失败 `SMS_MAX_VERIFY_ATTEMPTS` 次即作废(防爆破),验过即作废(一次性)。
 """
 from __future__ import annotations
@@ -24,7 +27,6 @@ import logging
 import secrets
 import time
 from dataclasses import dataclass
-from datetime import datetime
 from threading import Lock
 
 import httpx
@@ -54,22 +56,17 @@ class _CodeRecord:
 # 进程内存(单 worker 有效;多 worker 不共享,见模块 docstring)
 _codes: dict[str, _CodeRecord] = {}            # phone -> 当前有效验证码
 _last_sent: dict[str, float] = {}              # phone -> 上次发送 epoch(冷却)
-_daily_count: dict[str, tuple[str, int]] = {}  # phone -> (date_str, 当日发送数)
 _lock = Lock()
 _GC_THRESHOLD = 10000  # 任一内存 dict 超此阈值,send 时顺手清过期项(防无限增长,仿 ratelimit)
 
 
-def _today() -> str:
-    return datetime.now().strftime("%Y-%m-%d")
-
-
 def _gen_code() -> str:
     """生成 N 位数字验证码(用 secrets 而非 random;允许前导 0)。"""
     return "".join(secrets.choice("0123456789") for _ in range(settings.SMS_CODE_LENGTH))
 
 
 def _gc(now: float) -> None:
-    """顺手清理过期内存项,防三个 dict 无限增长。仅在持锁时调用,且某 dict 超
+    """顺手清理过期内存项,防两个 dict 无限增长。仅在持锁时调用,且某 dict 超
     _GC_THRESHOLD 才扫它(低频,开销可忽略)。"""
     if len(_codes) > _GC_THRESHOLD:
         for p in [p for p, r in _codes.items() if now > r.expires_at]:
@@ -78,10 +75,6 @@ def _gc(now: float) -> None:
         cutoff = now - settings.SMS_SEND_INTERVAL_SEC
         for p in [p for p, ts in _last_sent.items() if ts < cutoff]:
             _last_sent.pop(p, None)
-    if len(_daily_count) > _GC_THRESHOLD:
-        today = _today()
-        for p in [p for p, (d, _c) in _daily_count.items() if d != today]:
-            _daily_count.pop(p, None)
 
 
 def send_code(phone: str) -> int:
@@ -100,17 +93,9 @@ def send_code(phone: str) -> int:
             remain = int(settings.SMS_SEND_INTERVAL_SEC - elapsed)
             raise SmsError(f"发送过于频繁,请 {remain}s 后再试")
 
-        today = _today()
-        day, cnt = _daily_count.get(phone, ("", 0))
-        if day != today:
-            cnt = 0
-        if cnt >= settings.SMS_DAILY_LIMIT_PER_PHONE:
-            raise SmsError("今日验证码发送次数已达上限,请明天再试")
-
         code = _gen_code()
-        # 预占:先记冷却/计数/存码,释放锁后再发网络(发失败保留冷却+计数,见下)
+        # 预占:先记冷却/存码,释放锁后再发网络(发失败保留冷却,见下)
         _last_sent[phone] = now
-        _daily_count[phone] = (today, cnt + 1)
         _codes[phone] = _CodeRecord(code=code, expires_at=now + settings.SMS_CODE_TTL_SEC)
 
     # --- lock 外:真正发送(网络 IO 不持锁)---
@@ -121,7 +106,7 @@ def send_code(phone: str) -> int:
             _send_via_jiguang(phone, code)
             logger.info("[SMS] sent to %s****", phone[:3])
     except Exception as e:
-        # 发送失败:**保留冷却 + 每日计数**(失败也限速,挡住余额不足/签名失效时
+        # 发送失败:**保留冷却**(失败也限速,挡住余额不足/签名失效时
         # 前端重试狂打极光),只清掉没发出去的码(用户收不到,留着无意义且占内存)。
         with _lock:
             _codes.pop(phone, None)

app/main.py

@@ -2,30 +2,32 @@
 
 通过 `uvicorn app.main:app --reload` 启动。
 """
+
 from __future__ import annotations
 
 import logging
 from collections.abc import AsyncIterator
 from contextlib import asynccontextmanager
-
 from pathlib import Path
 
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import FileResponse
 from fastapi.staticfiles import StaticFiles
 
+from app.api.internal.app_version import router as internal_app_version_router
+from app.api.internal.launch_confirm import router as internal_launch_confirm_router
+from app.api.internal.price import router as internal_price_router
+from app.api.internal.store import router as internal_store_router
 from app.api.v1.ad import router as ad_router
+from app.api.v1.analytics import router as analytics_router
 from app.api.v1.auth import router as auth_router
 from app.api.v1.compare import router as compare_router
 from app.api.v1.compare_milestone import router as compare_milestone_router
 from app.api.v1.compare_record import router as compare_record_router
 from app.api.v1.coupon import router as coupon_router
-from app.api.v1.device import router as device_router
 from app.api.v1.cps_redirect import router as cps_redirect_router
-from app.api.internal.app_version import router as internal_app_version_router
-from app.api.internal.launch_confirm import router as internal_launch_confirm_router
-from app.api.internal.price import router as internal_price_router
-from app.api.internal.store import router as internal_store_router
+from app.api.v1.device import router as device_router
 from app.api.v1.feedback import router as feedback_router
 from app.api.v1.invite import router as invite_router
 from app.api.v1.meituan import router as meituan_router
@@ -39,15 +41,16 @@ from app.api.v1.user import router as user_router
 from app.api.v1.wallet import router as wallet_router
 from app.api.v1.wxpay import router as wxpay_router
 from app.core.config import settings
-from app.core.heartbeat_monitor_worker import (
-    start_heartbeat_monitor,
-    stop_heartbeat_monitor,
-)
 from app.core.daily_exchange_worker import (
     start_daily_exchange_worker,
     stop_daily_exchange_worker,
 )
+from app.core.heartbeat_monitor_worker import (
+    start_heartbeat_monitor,
+    stop_heartbeat_monitor,
+)
 from app.core.logging import setup_logging
+from app.core.pricebot_client import aclose_pricebot_client, get_pricebot_client
 from app.core.withdraw_reconcile_worker import (
     start_withdraw_reconcile_worker,
     stop_withdraw_reconcile_worker,
@@ -67,6 +70,13 @@ async def lifespan(_: FastAPI) -> AsyncIterator[None]:
         settings.APP_DEBUG,
         settings.DATABASE_URL.split("://", 1)[0],
     )
+    get_pricebot_client()  # 预热透传 client:把建 SSL 上下文的一次性成本付在启动,首个领券请求即热
+    try:
+        # 预热离线地理库:首次加载 ~2.5M 行 CSV + 建 KDTree,摊到启动、不砸首个按城市过滤的请求
+        from app.utils import geo
+        geo.ensure_loaded()
+    except Exception:  # noqa: BLE001
+        logger.exception("reverse_geocoder 预热失败(城市反查将在首个请求时懒加载)")
     reconcile_task = start_withdraw_reconcile_worker()
     heartbeat_task = start_heartbeat_monitor()
     daily_exchange_task = start_daily_exchange_worker()
@@ -76,6 +86,7 @@ async def lifespan(_: FastAPI) -> AsyncIterator[None]:
         await stop_heartbeat_monitor(heartbeat_task)
         await stop_withdraw_reconcile_worker(reconcile_task)
         await stop_daily_exchange_worker(daily_exchange_task)
+        await aclose_pricebot_client()
         logger.info("shutting down")
 
 
@@ -105,6 +116,7 @@ def health() -> dict[str, str]:
 app.include_router(auth_router)
 app.include_router(user_router)
 app.include_router(feedback_router)
+app.include_router(analytics_router)
 app.include_router(invite_router)
 app.include_router(coupon_router)
 app.include_router(device_router)
@@ -132,6 +144,30 @@ app.include_router(cps_redirect_router)
 # 用户上传文件(头像)静态服务。生产可改由 nginx 直接 serve MEDIA_ROOT。
 _media_root = Path(settings.MEDIA_ROOT)
 _media_root.mkdir(parents=True, exist_ok=True)
+
+
+# 官网下载的 APK 直链(落地页 dl.html「官网下载」按钮指向 /media/shaguabijia.apk)。
+# 必须在 StaticFiles 挂载【之前】注册,否则被静态挂载吃掉。
+# StaticFiles 给 .apk 的 Content-Type 不对、且无 attachment 头 → 部分国产浏览器不触发下载、转甩应用市场;
+# 这里显式回 application/vnd.android.package-archive + Content-Disposition:attachment 强制浏览器下载。
+# 文件由 scripts/publish_apk.sh 编 release 包后放到 data/media/shaguabijia.apk(*.apk 不入 git,需部署时放)。
+_APK_PATH = _media_root / "shaguabijia.apk"
+
+
+@app.get(f"{settings.MEDIA_URL_PREFIX}/shaguabijia.apk", tags=["meta"], include_in_schema=False)
+def download_apk() -> FileResponse:
+    if not _APK_PATH.is_file():
+        from fastapi import HTTPException
+
+        raise HTTPException(status_code=404, detail="安装包未就绪")
+    return FileResponse(
+        _APK_PATH,
+        media_type="application/vnd.android.package-archive",
+        filename="shaguabijia.apk",
+        headers={"Content-Disposition": 'attachment; filename="shaguabijia.apk"'},
+    )
+
+
 app.mount(
     settings.MEDIA_URL_PREFIX,
     StaticFiles(directory=str(_media_root)),

app/models/__init__.py

@@ -1,9 +1,12 @@
 """所有 ORM model 必须在这里 import 一次,Alembic / metadata 才能扫到。"""
 from app.models.ad_ecpm import AdEcpmRecord  # noqa: F401
 from app.models.ad_feed_reward import AdFeedRewardRecord  # noqa: F401
+from app.models.ad_pangle_revenue import AdPangleDailyRevenue  # noqa: F401
 from app.models.ad_reward import AdRewardRecord  # noqa: F401
 from app.models.ad_watch_log import AdWatchLog  # noqa: F401
 from app.models.admin import AdminAuditLog, AdminUser  # noqa: F401
+from app.models.admin_role import AdminRole  # noqa: F401
+from app.models.analytics_event import AnalyticsEvent  # noqa: F401
 from app.models.app_config import AppConfig  # noqa: F401
 from app.models.comparison import ComparisonRecord  # noqa: F401
 from app.models.cps_activity import CpsActivity  # noqa: F401
@@ -17,6 +20,7 @@ from app.models.coupon_state import (  # noqa: F401
     CouponClaimRecord,
     CouponDailyCompletion,
     CouponPromptEngagement,
+    CouponSession,
 )
 from app.models.feedback import Feedback  # noqa: F401
 from app.models.invite import InviteRelation  # noqa: F401
@@ -37,5 +41,6 @@ from app.models.wallet import (  # noqa: F401
     CashTransaction,
     CoinAccount,
     CoinTransaction,
+    InviteCashTransaction,
     WithdrawOrder,
 )

app/models/ad_ecpm.py

@@ -29,6 +29,9 @@ class AdEcpmRecord(Base):
     )
     # 广告类型:reward_video(激励视频) / draw(Draw 信息流) 等;不强行统一代码位,各类型各自上报
     ad_type: Mapped[str] = mapped_column(String(32), nullable=False)
+    # 点位场景:comparison(比价) / coupon(领券) / welfare(福利),供收益报表区分比价/领券 Draw 收益;
+    # 仅信息流/Draw 上报(比价与领券共用同一代码位,只能客户端各调用点显式打标),激励视频为 NULL。
+    feed_scene: Mapped[str | None] = mapped_column(String(16), nullable=True)
     # 客户端生成的一次广告会话 id;激励视频 S2S 回调 extra 会透传同值
     ad_session_id: Mapped[str | None] = mapped_column(String(64), index=True, nullable=True)
     # 实际投放的 ADN(穿山甲 getShowEcpm().getSdkName(),如 pangle / gdt)

app/models/ad_feed_reward.py

@@ -30,6 +30,9 @@ class AdFeedRewardRecord(Base):
     ecpm_raw: Mapped[str] = mapped_column(String(32), nullable=False)
     adn: Mapped[str | None] = mapped_column(String(32), nullable=True)
     slot_id: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    # 广告类型:feed(信息流) / draw(Draw 信息流)。比价与领券共用同一 Draw 代码位,靠 feed_scene
+    # 区分收益;ad_type 区分广告形态。旧数据(未升级客户端)为 NULL,一律视为 feed,保持向后兼容。
+    ad_type: Mapped[str | None] = mapped_column(String(16), nullable=True, default="feed")
     # 点位场景:comparison(比价等待) / coupon(领券) / welfare(福利页)。比价与领券共用同一信息流
     # 代码位,slot_id/our_code_id 分不出,只能客户端各调用点显式打标;NULL=历史/未升级客户端=未分类。
     feed_scene: Mapped[str | None] = mapped_column(String(16), nullable=True)

app/models/ad_pangle_revenue.py

@@ -0,0 +1,73 @@
+"""穿山甲 GroMore 天级收益报表(后台结算口径,定时拉取入库)。
+
+每行 = GroMore 数据 API 返回的一条「日期 × 应用 × 代码位」聚合收益(`integrations/pangle_report`
++ `scripts/sync_pangle_revenue` 落库)。**权威/预估收益的来源**,与 `ad_ecpm_record`(客户端自报
+eCPM 折算的预估)互为对照:
+
+- `revenue_yuan` ← 接口 `revenue`(预估收益,元;排序价×展示/1000,所有 ADN 都有);
+- `api_revenue_yuan` ← 接口 `api_revenue`(收益Api,元;各 ADN 经 Reporting 回传、更接近结算;
+  未配置该 ADN 的 Reporting 或查当天时为空)。
+
+⚠️ 穿山甲不提供分用户/设备维度,故本表最细只到 日期×应用×代码位,**无法挂到逐条广告事件**;
+广告收益报表里只用于汇总/趋势级的「穿山甲后台收益」,不改逐条行的客户端预估。
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from sqlalchemy import (
+    DateTime,
+    Float,
+    Integer,
+    String,
+    UniqueConstraint,
+    func,
+)
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+
+class AdPangleDailyRevenue(Base):
+    __tablename__ = "ad_pangle_daily_revenue"
+    __table_args__ = (
+        # 一行 = (日期, 应用, 代码位, 广告源);adn 用 "" 表示「未分广告源、该代码位汇总」,
+        # 避免 NULL 在唯一约束里被视为各不相同导致 upsert 重复(SQLite/PG 行为一致)。
+        UniqueConstraint(
+            "report_date", "app_env", "our_code_id", "adn",
+            name="uq_ad_pangle_daily",
+        ),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    # 北京时间日期串 'YYYY-MM-DD'(拉取时 time_zone=8),与 ad_ecpm_record.report_date 同口径,可直接 join。
+    report_date: Mapped[str] = mapped_column(String(10), index=True, nullable=False)
+    # 我们的应用环境:prod(傻瓜比价正式)/ test(测试);由 site_id 经 settings 映射而来。
+    app_env: Mapped[str] = mapped_column(String(16), nullable=False)
+    # 原始 GroMore AppId(site_id),留痕便于排查映射。
+    site_id: Mapped[str | None] = mapped_column(String(32), nullable=True)
+    # 广告位 ID(= 接口 ad_unit_id = 我们客户端配的 104xxx = ad_ecpm_record.our_code_id),join key。
+    # ⚠️ 不是接口的 code_id —— 那是底层各 ADN 的代码位(如 983674557),对不上我们的口径。
+    our_code_id: Mapped[str] = mapped_column(String(64), index=True, nullable=False)
+    # 广告源(接口 network 数字→名,如 pangle/gdt);"" = 未分广告源的代码位汇总行(当前默认口径)。
+    adn: Mapped[str] = mapped_column(String(16), nullable=False, default="")
+    # 预估收益(元)← 接口 revenue。
+    revenue_yuan: Mapped[float] = mapped_column(Float, nullable=False, default=0.0)
+    # 收益Api(元)← 接口 api_revenue;未配 Reporting / 当天 等情况接口不返回 → NULL。
+    api_revenue_yuan: Mapped[float | None] = mapped_column(Float, nullable=True)
+    # 预估 eCPM 原值(接口 ecpm,单位元/千次,**与客户端 getEcpm 的「分」不同**),参考用原样存。
+    ecpm: Mapped[str | None] = mapped_column(String(32), nullable=True)
+    # 展示次数 ← 接口 imp_cnt。
+    impressions: Mapped[int] = mapped_column(Integer, nullable=False, default=0)
+    # 货币类型(接口 currency,正常为 cny)。
+    currency: Mapped[str] = mapped_column(String(8), nullable=False, default="cny")
+    # 最近一次同步写入时间(同一行可被多次回补覆盖;T+1 数据穿山甲会订正)。
+    synced_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), server_default=func.now(), onupdate=func.now(), nullable=False
+    )
+
+    def __repr__(self) -> str:  # pragma: no cover
+        return (
+            f"<AdPangleDailyRevenue {self.report_date} {self.app_env} "
+            f"code={self.our_code_id} revenue={self.revenue_yuan}>"
+        )

app/models/admin.py

@@ -25,8 +25,14 @@ class AdminUser(Base):
     id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
     username: Mapped[str] = mapped_column(String(64), unique=True, index=True, nullable=False)
     password_hash: Mapped[str] = mapped_column(String(255), nullable=False)
-    # super_admin(全权+管账号)/ finance(钱:提现+金币)/ operator(用户+反馈+大盘)
+    # 明文登录密码:仅「后台 UI 创建/重置」的管理员留存,供超管在权限管理页复看转交。
+    # 脚本/起后台时建的超管账号不写(为 None → 前端「不显示密码」)。⚠️ 内部工具便利取舍,见 create/list。
+    plain_password: Mapped[str | None] = mapped_column(String(128), nullable=True)
+    # super_admin(全权+管账号)/ finance(钱:提现+金币)/ operator(用户+反馈+大盘)/ custom(按人自定义)
     role: Mapped[str] = mapped_column(String(20), nullable=False, default="operator")
+    # 「自定义」权限:仅当 role == "custom" 时有效,存这个人专属的可见页 key 列表(不共享给他人)。
+    # 非 custom 用户为 None → 可见页跟随角色。登录/`/me` 下发有效页时,非空即优先用它(见 auth._admin_out_with_pages)。
+    pages_override: Mapped[list | None] = mapped_column(_JSON, nullable=True)
     # active / disabled
     status: Mapped[str] = mapped_column(String(20), nullable=False, default="active")
 

app/models/admin_role.py

@@ -0,0 +1,42 @@
+"""admin 角色 → 可见页面(权限)映射表。
+
+RBAC 的「角色」侧:每个角色持有一组「页面 key」(= 左侧导航项),决定该角色登录后台后左边能看到
+哪些页。super_admin 是内建全权角色(is_builtin=True),恒可见全部页(effective_pages 里特判,
+不依赖本表存的 pages)、不可编辑/删除。其余角色(含 operator/finance)可由 super_admin 增删改。
+
+admin_user.role 存的是本表的 name(字符串弱引用,不建外键——与既有 require_role 字符串口径一致;
+删除在用角色由业务层拦截,见 routers/roles.py)。页面 key 清单见 app/admin/permissions.py。
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from sqlalchemy import JSON, Boolean, DateTime, Integer, String, func
+from sqlalchemy.dialects.postgresql import JSONB
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+_JSON = JSON().with_variant(JSONB(), "postgresql")
+
+
+class AdminRole(Base):
+    __tablename__ = "admin_role"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+    # 角色标识 key(不可变,承重):admin_user.role 引用它、require_role 按它鉴权。
+    # 内建 = super_admin/operator/finance/tech(英文,勿改);自定义 = 创建时的名称。
+    name: Mapped[str] = mapped_column(String(32), unique=True, index=True, nullable=False)
+    # 展示名(可改):内建 = 管理员/运营/财务/技术;自定义默认 = name。UI 一律展示 label。
+    label: Mapped[str] = mapped_column(String(32), nullable=False, default="")
+    # 该角色可见页面 key 列表(= 左侧导航项),见 app/admin/permissions.py 的 PERMISSION_CATALOG
+    pages: Mapped[list] = mapped_column(_JSON, nullable=False, default=list)
+    # 内建角色(当前仅 super_admin):不可编辑/删除,恒全权
+    is_builtin: Mapped[bool] = mapped_column(Boolean, nullable=False, default=False)
+
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), server_default=func.now(), nullable=False
+    )
+
+    def __repr__(self) -> str:  # pragma: no cover
+        return f"<AdminRole name={self.name} pages={len(self.pages)} builtin={self.is_builtin}>"

app/models/analytics_event.py

@@ -0,0 +1,60 @@
+"""新手引导(及后续)埋点事件表。
+
+每行 = 客户端上报的一条行为埋点,按「Who / When / Where / What / How」五维组织:
+- What :event(事件名,如 video_play)+ props(事件专属属性,JSON)
+- Who  :device_id(硬件级设备标识)+ user_id(登录后才有,可空)
+- When :client_ts(端事件时间 epoch ms)+ session_id(本次引导会话)+ sent_at(端上报时间)
+         + created_at(服务端接收时间 = server_at)
+- Where:page(引导步/页面)+ client_ip(服务端从 X-Forwarded-For 取)
+- How  :oem / os / model / app_ver / network / channel(设备与环境)
+
+append-only,不更新;客户端批量上报(见 app/api/v1/analytics.py),admin 同库直接查(见
+app/admin/routers/event_logs.py)。未登录态也允许上报(user_id 为空),故 user_id 不设外键、只索引。
+"""
+from __future__ import annotations
+
+from datetime import datetime
+
+from sqlalchemy import JSON, BigInteger, DateTime, Integer, String, func
+from sqlalchemy.orm import Mapped, mapped_column
+
+from app.db.base import Base
+
+
+class AnalyticsEvent(Base):
+    __tablename__ = "analytics_event"
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+
+    # ---- What:做了什么 ----
+    event: Mapped[str] = mapped_column(String(64), index=True, nullable=False)
+    props: Mapped[dict | None] = mapped_column(JSON, nullable=True)
+
+    # ---- Who:谁 ----
+    device_id: Mapped[str] = mapped_column(String(64), index=True, nullable=False)
+    user_id: Mapped[int | None] = mapped_column(Integer, index=True, nullable=True)
+
+    # ---- When:何时 ----
+    session_id: Mapped[str | None] = mapped_column(String(64), index=True, nullable=True)
+    client_ts: Mapped[int] = mapped_column(BigInteger, nullable=False)  # 端事件时间 epoch ms
+    sent_at: Mapped[int | None] = mapped_column(BigInteger, nullable=True)  # 端上报时间 epoch ms
+
+    # ---- Where:何地 ----
+    page: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    client_ip: Mapped[str | None] = mapped_column(String(64), nullable=True)
+
+    # ---- How:用什么环境 ----
+    oem: Mapped[str | None] = mapped_column(String(32), nullable=True)
+    os: Mapped[str | None] = mapped_column(String(32), nullable=True)
+    model: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    app_ver: Mapped[str | None] = mapped_column(String(32), nullable=True)
+    network: Mapped[str | None] = mapped_column(String(16), nullable=True)
+    channel: Mapped[str | None] = mapped_column(String(32), nullable=True)
+
+    # 服务端接收时间(= When.server_at);客户端时间不可信,以此为权威落库时刻。
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), server_default=func.now(), index=True, nullable=False
+    )
+
+    def __repr__(self) -> str:  # pragma: no cover
+        return f"<AnalyticsEvent id={self.id} event={self.event} device={self.device_id}>"

app/models/comparison.py

@@ -39,16 +39,19 @@ _JSON = JSON().with_variant(JSONB(), "postgresql")
 class ComparisonRecord(Base):
     __tablename__ = "comparison_record"
     __table_args__ = (
-        # 同一用户同一次比价(trace_id)只存一条:客户端重试/误点重复上报时幂等覆盖。
-        UniqueConstraint("user_id", "trace_id", name="uq_comparison_user_trace"),
+        # trace_id 由 app-server 签发、全局唯一 → 一次比价一行,后端 harvest 按它 upsert。
+        # (原 (user_id,trace_id) 复合唯一改为 trace_id 单列:harvest 帧0 建行时 user_id 可能暂缺。)
+        UniqueConstraint("trace_id", name="uq_comparison_trace"),
         # 首页轮播 / 省钱战绩聚合都按 status='success' 过滤 + created_at 近期排序;
         # 复合索引避免随数据量增大退化成全表扫(单列 created_at 索引不含 status)。
         Index("ix_comparison_status_created", "status", "created_at"),
     )
 
     id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
-    user_id: Mapped[int] = mapped_column(
-        Integer, ForeignKey("user.id"), index=True, nullable=False
+    # 后端 harvest 在帧0(pricebot 出 trace_id)即建行,软鉴权下 user_id 可能暂缺(老客户端/匿名)→ 可空。
+    # C 端「我的比价记录」按 user_id 过滤天然排除 null-user 行;admin 全看(含孤儿行)。
+    user_id: Mapped[int | None] = mapped_column(
+        Integer, ForeignKey("user.id"), index=True, nullable=True
     )
     # 仍记录设备号(同一用户多设备的行为区分 / 与不鉴权期 device_id 数据对账)
     device_id: Mapped[str | None] = mapped_column(String(64), nullable=True)
@@ -83,6 +86,10 @@ class ComparisonRecord(Base):
 
     # ===== 订单概要 =====
     store_name: Mapped[str | None] = mapped_column(String(128), nullable=True)
+    # 下单商品名拼接串(顿号分隔,从 items[].name 去重派生),供 admin 列表「商品」列展示 + 商品名搜索。
+    # items 是 JSON(SQLite 下 json.dumps ensure_ascii 把中文转义,无法直接 CAST+LIKE),故另派生成普通
+    # 文本列——跨库 LIKE 一致、可加索引。写路径(upsert_record / harvest_done)落库时同步派生。
+    product_names: Mapped[str | None] = mapped_column(String(512), nullable=True)
     total_dish_count: Mapped[int | None] = mapped_column(Integer, nullable=True)
     skipped_dish_count: Mapped[int | None] = mapped_column(Integer, nullable=True)
 

app/models/coupon_state.py

@@ -181,3 +181,79 @@ class CouponPromptEngagement(Base):
             f"<CouponPromptEngagement device={self.device_id} "
             f"date={self.engage_date} type={self.engage_type}>"
         )
+
+
+class CouponSession(Base):
+    """一次领券任务的全程流水(admin「领券数据」看板数据源,2026-06-30)。
+
+    与 coupon_claim_record(按券一天一条去重)、coupon_daily_completion(按设备一天一条)都不同:
+    本表**一次领券一条**(trace_id 唯一),记从发起(started)到收尾(completed/failed/abandoned)的
+    全程耗时 + 各平台耗时 + 机型/ROM。客户端 POST /api/v1/coupon/session 两段上报:发起建行、
+    收尾按 trace_id 更新同一行。发起即落库 → admin 可算「发起数」与中途流失(started 无终态=未完成)。
+
+    口径:elapsed_ms 由客户端全程计时(点发起→收尾)、权威;started_at/finished_at 为时刻留痕。
+    started_date = started_at 的 Asia/Shanghai 自然日,供 admin 按天聚合 / 日期筛选(索引)。
+    """
+
+    __tablename__ = "coupon_session"
+    __table_args__ = (
+        # 一次领券一行:trace_id 幂等 upsert(发起建、收尾更新同一行)。
+        UniqueConstraint("trace_id", name="uq_coupon_session_trace"),
+        # admin 主聚合/筛选:按上海自然日 + 环境(报表默认只看 prod,避免测试数据串台)。
+        Index("ix_coupon_session_date_env", "started_date", "app_env"),
+    )
+
+    id: Mapped[int] = mapped_column(Integer, primary_key=True, autoincrement=True)
+
+    # 一次领券唯一 id(客户端 UUID,全程贯穿),upsert 键。
+    trace_id: Mapped[str] = mapped_column(String(64), nullable=False)
+    device_id: Mapped[str] = mapped_column(String(64), nullable=False)
+    # 登录态才带(admin join 用户表出手机号/昵称);匿名领券为空。
+    user_id: Mapped[int | None] = mapped_column(Integer, index=True, nullable=True)
+
+    # started / completed / failed / abandoned。started 无终态 = 中途流失(未完成)。
+    status: Mapped[str] = mapped_column(String(16), nullable=False)
+    # prod / dev(客户端 BuildConfig.DEBUG)。admin 报表默认只看 prod(对齐广告报表防串台口径)。
+    app_env: Mapped[str | None] = mapped_column(String(16), index=True, nullable=True)
+
+    # 发起勾选平台 ["meituan-waimai", ...](空=全领)。
+    platforms: Mapped[list | None] = mapped_column(_JSON, nullable=True)
+    # 发起来源外卖 App 包名;null=App 内(傻瓜比价首页)发起,非空=从美团/淘宝/京东弹券发起。
+    # admin「发起平台」列据此区分(空→傻瓜比价,包名→对应平台)。
+    origin_package: Mapped[str | None] = mapped_column(String(64), nullable=True)
+    # 机型(Build.MANUFACTURER + MODEL)与 ROM(OemDetector,如 "ColorOS 14")。明细「机型/ROM」列 + 维度筛选。
+    device_model: Mapped[str | None] = mapped_column(String(128), nullable=True)
+    rom: Mapped[str | None] = mapped_column(String(64), nullable=True)
+
+    # 发起时刻(客户端墙钟):明细「时间」列、趋势 X 轴。
+    started_at: Mapped[datetime] = mapped_column(DateTime(timezone=True), nullable=False)
+    # 发起的 Asia/Shanghai 自然日:按天聚合 / 日期范围筛选(索引)。
+    started_date: Mapped[date] = mapped_column(Date, nullable=False)
+    # 收尾时刻(服务端 now);未收尾(流失)为空。
+    finished_at: Mapped[datetime | None] = mapped_column(
+        DateTime(timezone=True), nullable=True
+    )
+
+    # 全程耗时(ms,客户端点发起→收尾):平均 / 分位都基于它(只统计 completed)。
+    elapsed_ms: Mapped[int | None] = mapped_column(Integer, nullable=True)
+    # 各平台领券耗时 {"meituan-waimai":3200,...}(ms)。明细美团/淘宝/京东耗时列。
+    platform_elapsed: Mapped[dict | None] = mapped_column(_JSON, nullable=True)
+    # 领到总张数(收尾帧带)。
+    claimed_count: Mapped[int | None] = mapped_column(Integer, nullable=True)
+    # pricebot done 帧回传的公网调试链接(price.shaguabijia.com/traces/{dir});含落盘时分秒、拼不出,只能存
+    # (同 ComparisonRecord.trace_url)。admin「领券数据」明细据此渲染可点 trace 链接;未到 done(failed/abandoned)为空。
+    trace_url: Mapped[str | None] = mapped_column(String(512), nullable=True)
+
+    created_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), server_default=func.now(), nullable=False
+    )
+    updated_at: Mapped[datetime] = mapped_column(
+        DateTime(timezone=True), server_default=func.now(), onupdate=func.now(),
+        nullable=False,
+    )
+
+    def __repr__(self) -> str:  # pragma: no cover
+        return (
+            f"<CouponSession trace={self.trace_id} status={self.status} "
+            f"elapsed_ms={self.elapsed_ms}>"
+        )