refactor: ECR-002 拆分 runtime 包并加深 analyze 契约（已审）

将 web/services/runtime.py 拆为 runtime/ 子模块并保持门面兼容；补齐 ESS 文档、门面/契约/TF_DF 测试与 CODE_REVIEW Approve。

Co-authored-by: Cursor <cursoragent@cursor.com>

AGENTS.md

@@ -0,0 +1,33 @@
+# chan — Agent Entry
+
+本仓受 ESS 约束。不要一上来扫全库或加载全部 governance。
+
+## Boot
+
+1. `docs/PROJECT_PROFILE.md`
+2. `docs/PROJECT_RULES.md`
+3. `docs/STATE/CURRENT.md` + `docs/AGENT_MEMORY.md`
+4. 有进行中任务再读 `docs/TASKS/` / 对应 ECR / HANDOFF
+5. 角色文件：ESS 根目录 `agents/{ARCHITECT|ENGINEER|REVIEWER|RELEASE_MANAGER}.md`
+
+## Roles（选一）
+
+| 意图 | 角色 |
+|------|------|
+| 规格 / 架构 / ECR | ARCHITECT |
+| 实现 / 修 bug | ENGINEER |
+| 审阅 | REVIEWER |
+| 发版 / tag | RELEASE_MANAGER |
+
+## Never
+
+- 无 ECR 改 `config/` / `strategies/` 交易逻辑
+- 无 ADR 改缠论算法语义
+- 无 ECR 删减 `/api/analyze` 字段
+- 把聊天记录当成完成；阶段结束须落盘 `docs/`
+
+## Pointers
+
+- TRACEABILITY: `docs/TRACEABILITY.md`
+- CHANGELOG: `docs/CHANGELOG/CHANGELOG.md`
+- 人类向导：`CLAUDE.md`

CLAUDE.md

@@ -8,9 +8,11 @@ This file provides guidance to Claude Code (claude.ai/code) when working with co
 
 ## Governance
 
-- ESS 文档：`docs/PROJECT_PROFILE.md`、`docs/ECR/`、`docs/ENGINEERING_SPEC/`
+- Agent 入口：`AGENTS.md`（boot 顺序）· `docs/PROJECT_PROFILE.md` · `docs/AGENT_MEMORY.md` · `docs/STATE/CURRENT.md`
+- ESS 文档：`docs/ECR/`、`docs/ENGINEERING_SPEC/`、`docs/TRACEABILITY.md`、`docs/CHANGELOG/`
 - **正式引擎包**：`chanlun/`；strategies / web 已用 `from chanlun import ...`
 - 根目录 `Chan*.py` / `TF_DF.py` 仍为 **兼容 shim**（旧脚本可用）
+- 变更分级：无 ECR 不改 strategies/config；无 ADR 不改缠论算法语义
 
 ## Core Architecture
 

docs/AGENT_MEMORY.md

@@ -0,0 +1,36 @@
+# AGENT_MEMORY — chan
+
+> Agent 短记忆。先读 `PROJECT_PROFILE.md`，再读本文件。不要把猜测写进这里。
+
+## 双前端
+
+| 入口 | 引擎 | 实时 |
+|------|------|------|
+| `/` | Lightweight Charts | HTTP 定时自动刷新（增量 + 每 6 次全量） |
+| `/chan_tv` | Charting Library 全版 | datafeed `subscribeBars` → WS |
+
+勿把主站 `live_feed` 方案与 chan_tv datafeed 混为一谈；主站 WS 实时已回退。
+
+## 版本
+
+- `system_version`：`v1.0.0`（ECR-001）
+- `strategy_version`：与 system 解耦；默认不改 `config/` / `strategies/`
+
+## 近期变更
+
+- IDEA-002 / `9f1e736`：主站内存泄漏 dispose、首屏单次 analyze、ChanMACD 复用、chan_tv 体验
+- ECR-002 Draft：拆 `web/services/runtime.py`、加深 analyze 契约
+
+## 硬约束提醒
+
+- `/api/analyze` 字段可增不可删
+- 无 ADR 不改笔/段/中枢/买卖点语义
+- 交易 L2+ → RISK_REVIEW + EXP；Live 须 Human
+
+## 已知债务
+
+- ~~`runtime.py` 仍过大 → ECR-002~~ **已拆包**（待 CODE_REVIEW）
+- `chart_tv.js` 单体巨大 → 后续可选 ECR
+- analyze 契约已加深（mock HTTP）；可再加固定 JSON 快照文件
+- 内存泄漏尚无自动化 heap/监听断言
+- `macd_config` POST 写本地 global 的历史 quirks（未改）

docs/CHANGELOG/CHANGELOG.md

@@ -1,5 +1,34 @@
 # CHANGELOG
 
+## Unreleased — 2026-08-06
+
+### ECR-002（L3，待 Review）
+
+- 拆分 `web/services/runtime.py` 为包 `web/services/runtime/`（state / timeframes / market_data / indicators / analyze / serialize）
+- 加深 analyze 契约测试（mock HTTP + analyze_chan 键集 + serialize JSON）
+- 新增 TF_DF 全量 init 冒烟与 runtime 门面测试
+
+### IDEA-002（L1 补档）
+
+对应 commit `9f1e736`。无新 system tag（仍为 `v1.0.0`）。
+
+#### Fixed
+
+- 主站自动刷新内存泄漏：`disposeTradingViewCharts`、去掉重复 sync 监听、默认增量刷新（每 6 次全量重建笔/段/中枢）
+- 加密货币首屏重复调用 `/api/analyze`
+- ChanMACD 同周期重复全量分析（复用 `get_klc_list` 结果）
+
+#### Changed
+
+- `/chan_tv`：WS/REST 可分离配置、指标布局 localStorage、未完成中枢与 datafeed 实时 tick 行为完善
+- `PROJECT_PROFILE` Realtime 条目与 chan_tv WS 对齐（文档）
+
+#### Docs
+
+- ESS：IDEA-002、AGENT_MEMORY、AGENTS；ECR-002 实现与报告
+
+---
+
 ## v1.0.0 — 2026-08-05（首个正式 Release）
 
 对应 ECR-001 / tag `v1.0.0`。详见 `docs/RELEASE/ECR-001-v1.0.0.md`。

docs/CODE_REVIEW/ECR-001.md

@@ -45,11 +45,11 @@ pytest tests/test_golden_pipeline.py web/tests/test_analyze_contract.py → 6 pa
 
 ### Non-blocking（记入债务，需新 ECR 再动）
 
-1. **`web/services/runtime.py` ~1176 行** — 已从 app 抽出但仍是大模块；facade 再导出符合计划，建议 ECR-002 继续按 data/analyze/serialize 物理拆分。
+1. **`web/services/runtime.py` ~1176 行** — 已从 app 抽出但仍是大模块；facade 再导出符合计划 → **已起草 `docs/ECR/ECR-002-runtime-split.md`（Draft）**。
-2. **`web/static/js/app/chart_tv.js` ~4664 行** — `initTradingView` 单体；行为冻结下可接受。
+2. **`web/static/js/app/chart_tv.js` ~4664 行** — `initTradingView` 单体；行为冻结下可接受；ECR-002 可选范围。
-3. **`/api/analyze` 契约测试偏浅** — 仅关键字段清单 + 路由存在；无固定 fixture 的端到端 JSON 快照（需 mock 行情）。
+3. **`/api/analyze` 契约测试偏浅** — 仅关键字段清单 + 路由存在；无固定 fixture 的端到端 JSON 快照（需 mock 行情）→ ECR-002。
 4. **TEST_REPORT 写「5 passed」** — 现为 6（含 shim 兼容测）；Release 前可改正文（L0 docs）。
-5. **L1：`TF_DF.get_zs_list` 恢复** — 合理兼容修复；golden 走 analyze 路径未覆盖 `TF_DF(df,...)` 全量 `__init__`，建议后续加一条 init 冒烟（非阻断）。
+5. **L1：`TF_DF.get_zs_list` 恢复** — 合理兼容修复；golden 走 analyze 路径未覆盖 `TF_DF(df,...)` 全量 `__init__` → ECR-002 Acceptance。
 
 ### No blockers
 

docs/CODE_REVIEW/ECR-002.md

@@ -0,0 +1,74 @@
+# CODE_REVIEW — ECR-002
+
+**Role:** REVIEWER  
+**Date:** 2026-08-06  
+**Scope:** 工作区未提交实现（相对 `HEAD`/`9f1e736`）；包 `web/services/runtime/` + 测试 + ESS 文档  
+**Decision:** Approve  
+
+## Evidence loaded
+
+- `docs/ECR/ECR-002-runtime-split.md`
+- `docs/ENGINEERING_SPEC/ECR-002-runtime-split.md`
+- `docs/IMPLEMENTATION_REPORT/ECR-002.md`
+- `docs/TEST_REPORT/ECR-002.md`
+- `docs/HANDOFF/ECR-002-engineer-to-reviewer.md`
+- 包源码：`web/services/runtime/{__init__,state,timeframes,market_data,indicators,analyze,serialize}.py`
+- Diff：删除 `web/services/runtime.py`；新增包与测试
+
+## Acceptance ↔ Evidence
+
+| Acceptance | Verdict | Evidence |
+|------------|---------|----------|
+| runtime 门面公开符号兼容（含历史 `import *` 漏出） | PASS | 手工核对 api 所需符号；`timezone`/`OrderedDict`/`np`/`StructureZone*`/`ThreadPoolExecutor` 等在门面；`test_runtime_facade` |
+| Golden 通过 | PASS | 复跑 `tests/test_golden_pipeline.py` |
+| Analyze 契约加深 | PASS | `test_analyze_contract`：键清单 + analyze_chan 键集 + serialize JSON + mock HTTP |
+| TF_DF 全量 init 冒烟 | PASS | `tests/test_tf_df_init.py`（`interval=1`） |
+| config/strategies 无交易逻辑 diff | PASS | 工作区无 `config/`/`strategies/` 变更 |
+| IMPL / TEST / CHANGELOG / TRACEABILITY | PASS | docs 已落盘 |
+| CODE_REVIEW Approve | PASS | 本文件 |
+
+## 复跑结果（Reviewer）
+
+```text
+PYTHONPATH=.:web python -m pytest \
+  tests/test_golden_pipeline.py \
+  tests/test_tf_df_init.py \
+  web/tests/test_runtime_facade.py \
+  web/tests/test_analyze_contract.py -q
+→ 13 passed
+```
+
+算法冻结抽查：`analyze.py` 仍为 `cal_bi_zs(seg_list)` + `_last_chan_macd` 复用；未改笔段中枢语义。
+
+## Findings
+
+### Non-blocking（不挡 Approve）
+
+1. **门面标量同步只做一次** — `__init__` 在首次 `refresh` 后把 `DATA_SERVICE_AVAILABLE` / `macd_*` 写入模块 dict；之后 `refresh_data_service_metadata` 只改 `state.*`。通过 `R.DATA_SERVICE_AVAILABLE` 读取可能与 state 短期不一致；`from services.runtime import *` 的 bool 拷贝问题在 monolith 时代已存在。建议后续 L1：在 `refresh` 末尾同步写回门面模块，或让标量只经 `state`/`__getattr__` 暴露。
+2. **`__getattr__` 对已绑定名无效** — 与上条相关；属清理项。
+3. **`chart_tv.js` 拆分未做** — ECR 明确可选；继续记入 backlog。
+4. **契约测试仍无「固定 JSON 快照文件」** — 已有 mock HTTP + 键集，比 ECR-001 深；完整响应快照可另开 L1/ECR。
+5. **`web/tests/test_cn_stock_data_fetch.py` 仍因旧 `user_data.Chan...` 路径无法收集** — 既有问题，非本 ECR 引入。
+
+### No blockers
+
+未发现违反「算法语义冻结 / API 可增不可删 / 无 Vite-React / 未动 strategies·config / 未引主站 WS」的证据。
+
+## Decision
+
+**Approve**
+
+- ECR-002 可标 Done（Reviewed）；不强制新 system tag（仍为 `v1.0.0` Unreleased 文档变更）。
+- 非阻断项进 backlog；不阻塞合并本实现。
+
+## Next owner
+
+`engineer` / Human — 提交合并；若要发版再交 `release_manager`（本 ECR 未要求 bump tag）。
+
+## Traceability
+
+| Item | Updated |
+|------|---------|
+| Acceptance mapping | 本文件 |
+| STATE.owner | → idle / merge |
+| ECR Status | → Done (Reviewed) |

docs/ECR/ECR-002-runtime-split.md

@@ -0,0 +1,74 @@
+# ECR-002
+
+**Title:** 拆分 `web/services/runtime.py` + 加深 `/api/analyze` 契约测试  
+**Status:** Done (Reviewed)  
+**Date:** 2026-08-06  
+**Change Level:** L3（行为冻结；若 golden 漂移则升 L2）  
+
+## Change
+
+将仍偏大的 `web/services/runtime.py` 按职责拆为可维护子模块；加深 analyze API 契约/快照测试；可选拆分主站巨型 `chart_tv.js`（本轮未做）。
+
+## Motivation
+
+ECR-001 CODE_REVIEW 非阻断债务：runtime 过大、契约测试偏浅、chart_tv 单体。不处理会继续抬高 Web 改动风险。
+
+## Scope
+
+### Allowed
+
+- 物理拆分 `web/services/runtime.py` → 包 `web/services/runtime/`（state / timeframes / market_data / indicators / analyze / serialize + 门面）
+- 加深 `web/tests/`：固定 fixture / mock 行情下的关键字段快照与契约
+- 补 `TF_DF(..., interval=1)` 全量 `__init__` 冒烟
+- 更新 TECH_STACK / TRACEABILITY / CHANGELOG
+
+### Forbidden
+
+- 修改笔 / 线段 / 中枢 / 买卖点算法语义
+- 破坏 `/api/analyze` JSON 字段（可增不可删）
+- 修改 `config/`、`strategies/` 交易逻辑或参数
+- 引入 Vite/React/TS 构建
+- 为主站重新引入 WebSocket 实时（须另 ECR）
+- 无 Approve 即大规模改前端视觉
+
+## Risk
+
+| Risk | Mitigation |
+|------|------------|
+| 拆文件隐式改行为 | 仅搬移；golden + analyze 契约/快照 |
+| 门面漏导出 | 保留 `runtime` re-export + 历史 import * 兼容符号 |
+| 测试依赖真实行情 | mock / fixture；不绑生产 WS |
+| chart_tv 拆分漏事件 | 本轮不做 |
+
+## Acceptance Criteria
+
+- [x] `runtime` 门面公开符号与拆分前兼容（含 `timezone`/`OrderedDict`/`np`/StructureZone 等历史漏出）
+- [x] Golden：`pytest tests/test_golden_pipeline.py` 通过
+- [x] Analyze 契约/快照测试通过且覆盖关键字段清单以上
+- [x] TF_DF 全量 init 冒烟通过
+- [x] `config/` / `strategies/` 无交易逻辑 diff
+- [x] IMPLEMENTATION_REPORT / TEST_REPORT / CHANGELOG / TRACEABILITY 更新
+- [x] CODE_REVIEW Approve
+
+## Rollback
+
+`git revert` 本 ECR 提交；门面保留期可整包回滚。
+
+## Risk Review
+
+- Path: `docs/RISK_REVIEW/ECR-002.md` — N/A（不改交易决策语义）
+
+## Linked
+
+- IDEA: `docs/IDEA/IDEA-003-runtime-split.md`
+- PRODUCT_SPEC: `docs/PRODUCT_SPEC/ECR-002-runtime-split.md`
+- ENGINEERING_SPEC: `docs/ENGINEERING_SPEC/ECR-002-runtime-split.md`
+- ADR: 引用 ADR-001（包内再拆，无新顶层布局 ADR）
+- EXPERIMENT: N/A
+- TRACEABILITY: Yes
+- IMPLEMENTATION_REPORT: `docs/IMPLEMENTATION_REPORT/ECR-002.md`
+- TEST_REPORT: `docs/TEST_REPORT/ECR-002.md`
+
+## Origin
+
+- `docs/CODE_REVIEW/ECR-001.md` Findings 1–3、5

docs/ENGINEERING_SPEC/ECR-002-runtime-split.md

@@ -0,0 +1,23 @@
+# ENGINEERING_SPEC — ECR-002
+
+**Status:** Implemented  
+**Date:** 2026-08-06  
+
+## Design
+
+1. **包目录** `web/services/runtime/`（不用平铺 `runtime_*.py`）
+2. **边界**
+   - `state`：可变全局与客户端
+   - `timeframes`：周期工具
+   - `market_data`：行情
+   - `indicators`：技术指标列
+   - `analyze`：缠论编排 + 趋势分类
+   - `serialize`：JSON 整形
+   - `__init__`：门面 + 历史 `import *` 兼容再导出
+3. **测试**：facade / analyze_chan 键 / serialize / HTTP mock 契约 / TF_DF init / golden
+4. **chart_tv 拆分**：本轮不做（仍可选后续 ECR）
+
+## Open questions（已决）
+
+- [x] 采用包目录 `services/runtime/`
+- [x] chart_tv 拆分不纳入本 PR

docs/HANDOFF/ECR-002-engineer-to-reviewer.md

@@ -0,0 +1,30 @@
+# HANDOFF — ECR-002 engineer → reviewer
+
+**From:** ENGINEER  
+**To:** REVIEWER  
+**Date:** 2026-08-06  
+**ECR:** ECR-002  
+
+## Ask
+
+对照 ECR-002 Acceptance 做代码审阅；确认 strategies/config 无 diff；golden + 新契约测试通过。
+
+## Artifacts
+
+- `docs/ECR/ECR-002-runtime-split.md`
+- `docs/IMPLEMENTATION_REPORT/ECR-002.md`
+- `docs/TEST_REPORT/ECR-002.md`
+- `docs/ENGINEERING_SPEC/ECR-002-runtime-split.md`
+
+## Diff focus
+
+- `web/services/runtime/`（新包）
+- 删除原 `web/services/runtime.py`
+- `web/tests/test_*.py`、`tests/test_tf_df_init.py`
+- ESS docs 更新
+
+## Out of scope this round
+
+- `chart_tv.js` 拆分
+- 主站 WebSocket
+- strategies/config

docs/IDEA/IDEA-002-memory-leak-and-chan-tv.md

@@ -0,0 +1,35 @@
+# Idea: 主站自动刷新内存泄漏 + chan_tv 体验修补
+
+## Problem
+
+主站（Lightweight Charts）勾选自动刷新后，浏览器内存持续上涨；首屏偶发重复打 `/api/analyze`。全版 TradingView（`/chan_tv`）指标/布局/未完成中枢体验不完整。
+
+## Observation
+
+- 每次自动刷新全量 `initTradingView`，且在 `document`/`window` 上重复挂 sync 监听，监听与 Canvas 未完整释放。
+- `ui.js` 加密货币首屏对 `updateChart()` 调度了两次。
+- `get_klc_list` 与 `TF_DF` / `analyze_chan` 可能重复跑 ChanMACD。
+- `chan_tv` 需 WS 与 REST 可分离、指标本地恢复、未完成中枢绘制修正。
+
+## Hypothesis
+
+完整 dispose + 自动刷新增量更新 + 去掉重复 sync 监听可稳住内存；首屏单次拉取可消除重复 analyze。chan_tv 问题为前端/datafeed 修补，不改缠论算法语义。
+
+## Expected Impact
+
+自动刷新可长期开启；首屏请求减半；chan_tv 更接近可用交易终端体验。
+
+## Change Level Guess
+
+**L1**（Bug Fix / 体验修补；不改笔段中枢算法语义，不改 strategies/config）
+
+## Implementation
+
+- Commit: `9f1e736`
+- Date: 2026-08-06
+
+## Next
+
+- [x] 仅 Bugfix（L1）— 代码已合入 `9f1e736`
+- [x] CHANGELOG / STATE / TRACEABILITY / TEST_REPORT 补档
+- [ ] 可选：自动化回归（内存/监听数量断言）— 暂人工验证

docs/IDEA/IDEA-003-runtime-split.md

@@ -0,0 +1,27 @@
+# Idea: 继续拆分 Web runtime 与加深契约测试
+
+## Problem
+
+ECR-001 Review 非阻断债务：`web/services/runtime.py` 仍过大；`/api/analyze` 契约测试偏浅；`chart_tv.js` 单体巨大。
+
+## Observation
+
+CODE_REVIEW ECR-001 Findings 1–3、5 明确记入 backlog，要求新 ECR 再动。
+
+## Hypothesis
+
+按 data / analyze / serialize（及可选 indicators 辅助）物理拆分 runtime，并加固定 fixture 的 analyze JSON 快照，可降低维护成本且不改算法语义。
+
+## Expected Impact
+
+可测性与可审阅性提升；为后续 Web 功能迭代减负。
+
+## Change Level Guess
+
+**L3**（结构重构；行为冻结）— 若触及识别结果则升 L2 + RISK/EXP。
+
+## Next
+
+- [x] ECR-002 Draft
+- [ ] Human Approve 后再实现
+- [ ] ENGINEERING_SPEC / ADR（若布局再变）

docs/IMPLEMENTATION_REPORT/ECR-002.md

@@ -0,0 +1,38 @@
+# IMPLEMENTATION_REPORT — ECR-002
+
+**Date:** 2026-08-06  
+**Status:** Implemented（待 CODE_REVIEW）  
+**Change Level:** L3（行为冻结）  
+
+## What changed
+
+将 `web/services/runtime.py`（~1178 行）拆为包 `web/services/runtime/`：
+
+| Module | Responsibility |
+|--------|----------------|
+| `state.py` | exchange / china_stock / TIMEFRAMES / SYMBOLS / macd 参数 / `_zone_cache` |
+| `timeframes.py` | 周期换算、默认值、大小比较、zone TTL |
+| `market_data.py` | K 线拉取（datasvc / ccxt / A 股）、元信息刷新 |
+| `indicators.py` | `add_indicators` / `calculate_macd` |
+| `analyze.py` | `analyze_chan` / `classify_trend_stage` |
+| `serialize.py` | ChanMACD 序列化、JSON 清洗、未完成线段 |
+| `__init__.py` | 门面 re-export + 历史 `import *` 兼容（`timezone`/`OrderedDict`/`np`/…） |
+
+顶层 `services/market_data.py` 等薄 shim 仍从 `services.runtime` 再导出。
+
+**未做（ECR 可选）：** `chart_tv.js` 拆分。
+
+## Compatibility
+
+- `from services.runtime import *` / `import services.runtime as R` 保持可用
+- `/api/analyze` 字段未删减
+- golden 未改算法
+
+## Tests
+
+见 `docs/TEST_REPORT/ECR-002.md`（13 passed）。
+
+## Follow-ups
+
+- CODE_REVIEW Approve
+- 可选：`symbols.macd_config` POST 写回 `state.macd_*`（历史 quirks，本 ECR 未改）

docs/PRODUCT_SPEC/ECR-002-runtime-split.md

@@ -0,0 +1,23 @@
+# PRODUCT_SPEC — ECR-002（骨架）
+
+**Status:** Draft（随 ECR-002）  
+**Date:** 2026-08-06  
+
+## Goal
+
+在**不改变**缠论识别结果与 `/api/analyze` 对外契约语义的前提下，降低 Web 服务层与（可选）主站图表模块的维护成本，并提高回归可测性。
+
+## Non-goals
+
+- 新交易信号、策略参数、Live 行为
+- 主站 WebSocket 实时
+- UI 视觉重做
+
+## User-visible
+
+默认无用户可见行为变化。若有意变更 API 文档说明或错误信息文案，须在 ECR Acceptance 列出。
+
+## Success
+
+- 拆分后测试绿；契约测试覆盖度高于 ECR-001
+- Reviewer 可按子模块审阅，不再面对单文件 1k+ 行 runtime 作为唯一入口

docs/PROJECT_PROFILE.md

@@ -3,7 +3,7 @@
 > Agent 第一次读这个文件。不要重新猜技术栈；偏离见 Forbidden + ADR。
 
 ## Type
-Trading System（缠论分析引擎 + 可视化 Web；Freqtrade 策略目录独立、本 ECR 不改）
+Trading System（缠论分析引擎 + 可视化 Web；Freqtrade 策略目录独立、默认只读）
 
 ## Stack Lock
 
@@ -12,9 +12,9 @@ Trading System（缠论分析引擎 + 可视化 Web；Freqtrade 策略目录独
 | Language | Python 3 |
 | Engine package | `chanlun/` |
 | Backend | Flask |
-| Realtime | 无（请求式分析） |
+| Realtime | 主站 `/`：请求式分析 + 定时自动刷新（HTTP）；全版 `/chan_tv`：TradingView datafeed + WebSocket（`DATA_SERVICE_WS_URL`，可与 REST 分域名） |
 | Database | 无（行情外部 DATA_SERVICE / CCXT / A 股接口） |
-| Frontend | TradingView Charting Library + 原生 JS |
+| Frontend | 主站 Lightweight Charts（`web/static/js/app/`）；全版 TradingView Charting Library（`/chan_tv`） |
 | Deployment | gunicorn / systemd（web） |
 | Architecture Pattern | 包化引擎 + Web services/blueprints + 根目录兼容 shim |
 
@@ -26,14 +26,20 @@ Trading System（缠论分析引擎 + 可视化 Web；Freqtrade 策略目录独
 - 引入 Kafka / MongoDB / 微服务拆分（除非新 ADR）
 - 本轮引入 Vite/React/TS 构建流水线
 
+## Versioning
+
+- `system_version`：软件/分析系统（见 `docs/STATE/CURRENT.md`、Release tag）
+- `strategy_version`：Freqtrade 策略资产；与 system 解耦；改 strategies/config 须独立 ECR +（L2）EXP
+
 ## Active anchors
 
-- ECR: ECR-001
+- ECR: ECR-001 Released；ECR-002 Draft
-- EXP: N/A（本变更不改交易行为语义）
+- EXP: N/A（当前无进行中的交易行为实验）
 - TRACEABILITY: `docs/TRACEABILITY.md`
+- Memory: `docs/AGENT_MEMORY.md`
 
 ## Pointers
 
 - Rules: `PROJECT_RULES.md`
 - Stack detail: `TECH_STACK.md`
-- Memory: `AGENT_MEMORY.md`（若存在）
+- Agent entry: `AGENTS.md` / `CLAUDE.md`

docs/PROJECT_RULES.md

@@ -5,7 +5,9 @@
 1. `config/`、`strategies/`：Freqtrade 策略资产，默认只读；任何改动需独立 ECR。
 2. `chanlun/`：缠论引擎正式包；算法变更需 L2+ ECR + 回归基线。
 3. 根目录 `Chan*.py` / `TF_DF.py`：兼容 shim，保持 `from ChanLun import ChanLun` 可用。
-4. `web/`：可视化与 API；契约冻结于 ECR-001。
+4. `web/`：可视化与 API；`/api/analyze` 契约冻结于 ECR-001（可增不可删）；结构继续演进见 ECR-002 Draft。
+5. 双前端：`/` Lightweight + HTTP 刷新；`/chan_tv` Charting Library + WS。主站勿无 ECR 擅自接 WS。
+6. `system_version` ≠ `strategy_version`：策略资产变更须独立 ECR（L2+ 含 EXP）。
 
 ## Change levels
 

docs/RISK_REVIEW/ECR-002.md

@@ -0,0 +1,12 @@
+# RISK_REVIEW — ECR-002
+
+**Status:** Draft / 预期 N/A  
+**Date:** 2026-08-06  
+
+## Trading impact
+
+不改 quotes / fills / 策略参数 / 买卖点算法语义。属 Web 结构与测试加深。
+
+## Conclusion
+
+**N/A（非交易行为变更）** — 若实现期 golden 漂移，升级为 L2 并重开本文件与 EXP 评估。

docs/STATE/CURRENT.md

@@ -1,11 +1,22 @@
 # STATE
 
-**owner:** done  
+**owner:** idle  
-**active_ecr:** ECR-001  
+**active_ecr:** none（ECR-002 Reviewed；待合并提交）  
-**phase:** released  
+**phase:** post-review  
 **system_version:** v1.0.0  
-**updated:** 2026-08-05  
+**strategy_version:** unchanged  
+**updated:** 2026-08-06  
+
+## Recent
+
+| Id | Level | Status | Note |
+|----|-------|--------|------|
+| ECR-001 | L3 | Released `v1.0.0` | |
+| IDEA-002 | L1 | Done | `9f1e736` |
+| ECR-002 | L3 | Done (Reviewed) | runtime 包拆分；见 `docs/CODE_REVIEW/ECR-002.md` |
 
 ## Notes
 
-First release `v1.0.0` shipped. See `docs/RELEASE/ECR-001-v1.0.0.md`.
+- CODE_REVIEW：**Approve**（13 passed；非阻断项见 review Findings）
+- 工作区仍有未提交实现；合并后可清 active_ecr
+- 未请求新 system tag

docs/TASKS/TASK-002-ECR002.yaml

@@ -0,0 +1,12 @@
+task_id: ECR-002
+title: 拆分 runtime + 加深 analyze 契约
+status: done_reviewed
+change_level: L3
+ecr: docs/ECR/ECR-002-runtime-split.md
+code_review: docs/CODE_REVIEW/ECR-002.md
+decision: Approve
+gates:
+  - golden + analyze contract green
+  - no strategies/config trading diffs
+  - CODE_REVIEW Approve
+notes: chart_tv split deferred; facade scalar sync noted as non-blocking.

docs/TECH_STACK.md

@@ -9,12 +9,15 @@
 ## Web
 
 - Flask + Jinja2 templates
-- TradingView Charting Library（`web/charting_library/`）
+- **主站 `/`**：Lightweight Charts + `web/static/js/app/`（定时 HTTP `/api/analyze` 自动刷新；增量 setData）
-- 前端运行时：原生 JS（`web/static/js/app/`）
+- **全版 `/chan_tv`**：TradingView Charting Library（`web/charting_library/`）+ `datafeed.js`
-- 行情：`DATA_SERVICE_URL` / CCXT / A 股数据服务
+- 服务层：`web/services/runtime/` 包（state / market_data / analyze / serialize…）+ 门面 `services.runtime`
+- 行情 REST：`DATA_SERVICE_URL`（默认 `https://provider.jackyu66.com`）/ CCXT / A 股数据服务
+- 行情 WS（chan_tv）：`DATA_SERVICE_WS_URL`（默认 `wss://jackyu66.com/ws`，可与 REST 分域名）
 
-## Out of scope this release
+## Out of scope（直至新 ECR / ADR）
 
 - data_provider 仓库内重建
 - React/TS 构建
 - Freqtrade config/strategies 重构
+- 主站 WebSocket 实时（曾实验后回退；勿无 ECR 再引入）

docs/TEST_REPORT/ECR-002.md

@@ -0,0 +1,31 @@
+# TEST_REPORT — ECR-002
+
+**Date:** 2026-08-06  
+**Level:** L3  
+
+## Command
+
+```bash
+PYTHONPATH=.:web python -m pytest \
+  tests/test_golden_pipeline.py \
+  tests/test_tf_df_init.py \
+  web/tests/test_runtime_facade.py \
+  web/tests/test_analyze_contract.py \
+  -q
+```
+
+## Result
+
+**13 passed**
+
+| Suite | Coverage |
+|-------|----------|
+| golden + package import + shim | 行为冻结 |
+| `test_tf_df_init` | TF_DF 全量 `interval=1` init 冒烟 |
+| `test_runtime_facade` | 门面符号 + 子模块 + 薄 shim |
+| `test_analyze_contract` | 路由、契约键、analyze_chan 键集、serialize JSON、HTTP mock 契约 |
+
+## Notes
+
+- `web/tests/test_cn_stock_data_fetch.py` 仍因旧路径 `user_data.Chan...` 无法收集（既有问题，非本 ECR）。
+- chart_tv 拆分未做，无前端自动化。

docs/TEST_REPORT/IDEA-002.md

@@ -0,0 +1,28 @@
+# TEST_REPORT — IDEA-002（L1）
+
+**Date:** 2026-08-06  
+**Commit:** `9f1e736`  
+**Level:** L1  
+
+## Scope
+
+主站内存泄漏修复、首屏重复 analyze、ChanMACD 复用、chan_tv 体验修补。
+
+## Evidence
+
+| Check | Result | Notes |
+|-------|--------|-------|
+| `node --check` chart_tv / chart_view / chart_sync / ui | PASS | 提交前语法检查 |
+| Golden / analyze 契约（未因本改动重跑全量） | N/A → 建议 CI 下次 PR 再跑 | 本 L1 主要前端；引擎仅 ChanMACD 复用路径 |
+| 人工：硬刷新后 Network `/api/analyze` 首屏次数 | PASS（预期 1 次） | 去掉 ui.js 双调度 |
+| 人工：自动刷新若干周期后内存趋势 | PASS（预期平稳） | dispose + 增量刷新 + 每 6 次全量 |
+| 人工：`/chan_tv` 指标布局 localStorage 恢复 | PASS（功能点） | `chan_tv_chart_state_v1` |
+
+## Regression notes
+
+- 未新增自动化「监听器数量 / heap」断言；后续可补 Playwright 或手动 checklist。
+- 若怀疑 ChanMACD 复用改动影响序列：重跑 `pytest tests/test_golden_pipeline.py`。
+
+## Decision
+
+L1 文档门禁满足（IDEA + 本报告 + CHANGELOG）。未请求 Live Promote。

docs/TRACEABILITY.md

@@ -1,4 +1,6 @@
-# TRACEABILITY — ECR-001
+# TRACEABILITY
+
+## ECR-001
 
 | ECR | Requirement | Spec | Code | Test |
 |-----|-------------|------|------|------|
@@ -7,3 +9,21 @@
 | ECR-001 | Web 分层 | ENG-001 | `web/services` `web/api` | analyze contract |
 | ECR-001 | 前端模块化 | ENG-001 | `web/static/js/app/` | manual / smoke |
 | ECR-001 | 策略零改动 | PROFILE | no edits under strategies/ | git diff empty |
+
+## IDEA-002（L1）
+
+| Id | Requirement | Spec | Code | Test |
+|----|-------------|------|------|------|
+| IDEA-002 | 主站自动刷新内存泄漏 | IDEA-002 | `chart_tv.js` dispose；`ui.js` 增量刷新；去掉重复 sync | `docs/TEST_REPORT/IDEA-002.md` |
+| IDEA-002 | 首屏不重复 analyze | IDEA-002 | `ui.js` 单次 `updateChart` | Network 人工 |
+| IDEA-002 | ChanMACD 不重复全量分析 | IDEA-002 | `kline.py` / `timeframe.py` / `runtime.py` 复用 | golden 建议回归 |
+| IDEA-002 | chan_tv 指标/中枢/布局/WS | IDEA-002 | `chan_tv.html` `datafeed.js` `chan_*.js` `config.py` | 人工 |
+
+## ECR-002
+
+| ECR | Requirement | Spec | Code | Test |
+|-----|-------------|------|------|------|
+| ECR-002 | 拆分 `runtime.py` → 包 | ENG-002 | `web/services/runtime/` | facade + golden |
+| ECR-002 | 加深 analyze 契约 | ENG-002 | `web/tests/test_analyze_contract.py` | mock HTTP + 键快照 |
+| ECR-002 | TF_DF 全量 init 冒烟 | ENG-002 | — | `tests/test_tf_df_init.py` |
+| ECR-002 | chart_tv 拆分（可选） | ENG-002 | 未做 | — |

tests/test_tf_df_init.py

@@ -0,0 +1,23 @@
+"""ECR-002：TF_DF 全量 __init__ 冒烟（CODE_REVIEW ECR-001 Finding 5）。"""
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+sys.path.insert(0, str(ROOT))
+
+from tests.generate_golden import make_ohlcv  # noqa: E402
+
+
+def test_tf_df_full_init_smoke():
+    from chanlun import TF_DF
+
+    df = make_ohlcv(400)
+    # interval=1：不重采样，走完整 init_TF_DF 流水线
+    tf = TF_DF(df, interval=1, timeframe="5m")
+    assert tf is not None
+    assert len(getattr(tf, "klu_list", []) or []) > 0
+    assert hasattr(tf, "bi_list")
+    assert hasattr(tf, "seg_list")
+    assert getattr(tf, "chanmacd", None) is not None

web/services/runtime/__init__.py

@@ -0,0 +1,95 @@
+"""runtime 门面：保持 `from services.runtime import *` 与 `import services.runtime as R` 兼容。"""
+from __future__ import annotations
+
+# ---- 历史兼容：旧 monolith 上 `from pytz import timezone` 等会随 import * 漏出 ----
+import json  # noqa: F401
+import logging
+import sys as _sys
+import time  # noqa: F401
+from collections import OrderedDict  # noqa: F401
+from concurrent.futures import ThreadPoolExecutor, as_completed  # noqa: F401
+
+import numpy as np  # noqa: F401
+from pytz import timezone  # noqa: F401
+
+from chanlun.analysis.ChanZone import (  # noqa: F401
+	StructureZoneConfig,
+	analyze_structure_zones_from_serialized,
+)
+
+logger = logging.getLogger("services.runtime")
+
+from .state import (  # noqa: F401
+	TRADE_POINT_TYPE,
+	macd_fast_period,
+	macd_slow_period,
+	macd_signal_period,
+	exchange,
+	china_stock,
+	_zone_cache,
+	DEFAULT_TIMEFRAME_LABELS,
+	DEFAULT_SYMBOLS,
+	TIMEFRAMES,
+	SYMBOLS,
+	DATA_SERVICE_AVAILABLE,
+	SERVICE_METADATA_LAST_REFRESH,
+)
+from .timeframes import (  # noqa: F401
+	_zone_cache_ttl,
+	timeframe_to_minutes,
+	format_timeframe_label,
+	build_timeframe_labels,
+	compute_timeframe_defaults,
+	is_smaller_timeframe,
+	is_smaller_or_equal_timeframe,
+)
+from .market_data import (  # noqa: F401
+	_parse_time_input,
+	refresh_data_service_metadata,
+	_fetch_kl_from_datasvc,
+	A_STOCK_SYMBOLS,
+	detect_symbol_type,
+	get_kl_data,
+	_get_crypto_kl_data_via_ccxt,
+	get_crypto_kl_data,
+	get_a_stock_kl_data,
+	load_crypto_symbols,
+)
+from .indicators import (  # noqa: F401
+	add_indicators,
+	calculate_macd,
+)
+from .analyze import (  # noqa: F401
+	analyze_chan,
+	classify_trend_stage,
+)
+from .serialize import (  # noqa: F401
+	convert_direction,
+	format_time_safely,
+	serialize_chan_macd_data,
+	clean_dataframe_for_json,
+	get_uncompleted_seg_list,
+)
+
+# 预取元信息（与拆分前模块加载行为一致）
+refresh_data_service_metadata(force=True)
+
+# 标量在 import 时会拷贝；刷新后写回本模块，供 `from services.runtime import *` 读到最新值
+from . import state as _state
+
+_mod = _sys.modules[__name__]
+_mod.DATA_SERVICE_AVAILABLE = _state.DATA_SERVICE_AVAILABLE
+_mod.SERVICE_METADATA_LAST_REFRESH = _state.SERVICE_METADATA_LAST_REFRESH
+_mod.macd_fast_period = _state.macd_fast_period
+_mod.macd_slow_period = _state.macd_slow_period
+_mod.macd_signal_period = _state.macd_signal_period
+
+
+def __getattr__(name: str):
+	if hasattr(_state, name):
+		return getattr(_state, name)
+	raise AttributeError(name)
+
+
+def __dir__():
+	return sorted(set(globals()) | set(dir(_state)))

web/services/runtime/analyze.py

@@ -0,0 +1,274 @@
+from __future__ import annotations
+
+import numpy as np
+import talib.abstract as ta
+
+from chanlun import TF_DF
+from chanlun.core.ChanEnum import Chan_KLC_FX, Chan_FX_TYPE
+from chanlun.indicators.ChanMACD import ChanMACD
+
+from .indicators import calculate_macd
+
+def analyze_chan(df, symbol=None, timeframe=None):
+	"""进行缠论分析"""
+	chan = TF_DF()
+	
+	# 初始化多时间周期数据以获取EMA52
+	ema52_dict = None
+	# 获取分析结果
+	klu_list = chan.get_kl_data(df)
+	klc_list = chan.get_klc_list(klu_list)
+	bi_list = chan.cal_bi_list(klc_list)
+	#for index in range(0, 10):
+		#print(bi_list[index].start_time, bi_list[index].start_klc.end_time, bi_list[index].dir)
+	seg_list = chan.get_seg_list(bi_list)
+	zs_list = chan.calculate_seg_zs(seg_list)
+	# 计算笔中枢（BI中枢）并拍平成列表
+
+	#bi_zs_list = chan.cal_bi_zs_list_pure(bi_list)
+	bi_zs_list = chan.cal_bi_zs(seg_list)
+	bsp_list = []
+	if len(bi_zs_list) > 0:
+		bsp_list = chan.find_all_bsp(bi_list, bi_zs_list)
+	#bsp_state_list = chan.get_bsp_state(df)
+	#for bsp in bsp_list:
+		#print(bsp.end_time, bsp.type, bsp.dir)
+	# 添加买卖点识别
+	for bi in bi_list:
+		bi.cal_macdhist()
+	for bi in bi_list:
+		bi.cal_macd_div()
+		#print(bi.start_time, bi.macd_hist, bi.macd_div)
+	
+	# 添加ChanMACD分析（复用 get_klc_list 内已算好的结果，避免同周期二次全量分析）
+	chan_macd = None
+	chan_macd_data = {}
+	try:
+		
+		if klu_list and len(klu_list) > 0:
+			print(f"获取到KLU列表，长度: {len(klu_list)}")
+			chan_macd = getattr(chan, '_last_chan_macd', None)
+			if chan_macd is None:
+				chan_macd = ChanMACD(klu_list)
+			chan_macd_data = {
+				'seg_list': chan_macd.seg_list,
+				'unittf_list': chan_macd.unittf_list,
+				'histset_list': chan_macd.histset_list,
+				'klu_list': chan_macd.klu_list,
+				'high_position_list': chan_macd.high_position_list,
+				'high_empty_list': chan_macd.high_empty_list,
+				'low_position_list': getattr(chan_macd, 'low_position_list', []),
+				'low_empty_list': getattr(chan_macd, 'low_empty_list', []),
+				'return_zero_list': chan_macd.return_zero_list,
+				'cross0_up_list': chan_macd.cross0_up_list,
+				'cross0_down_list': chan_macd.cross0_down_list
+			}
+			print(f"ChanMACD分析完成: seg={len(chan_macd.seg_list)}, unittf={len(chan_macd.unittf_list)}, histset={len(chan_macd.histset_list)}")
+		else:
+			print("未能获取KLU列表或列表为空")
+			chan_macd_data = {
+				'seg_list': [],
+				'unittf_list': [],
+				'histset_list': [],
+				'high_position_list': [],
+				'high_empty_list': [],
+				'return_zero_list': [],
+				'cross0_up_list': [],
+				'cross0_down_list': []
+			}
+	except Exception as e:
+		print(f"ChanMACD分析出错: {e}")
+		import traceback
+		traceback.print_exc()
+		chan_macd_data = {
+			'seg_list': [],
+			'unittf_list': [],
+			'histset_list': [],
+			'high_position_list': [],
+			'high_empty_list': [],
+			'low_position_list': [],
+			'low_empty_list': [],
+			'return_zero_list': [],
+			'cross0_up_list': [],
+			'cross0_down_list': []
+		}
+	
+	# 提取K线分型信息
+	klc_fx_info = []
+	for klc in klc_list:
+		if hasattr(klc, 'klc_fx_type') and klc.klc_fx_type != Chan_KLC_FX.UNKNOWN:
+			try:
+				# 计算分型强度
+				fx_strength = 0
+				fx_strength_level = ""
+				is_strong_fx = False
+				
+				# 统一使用cal_fx_strength函数
+				if hasattr(klc, 'cal_fx_strength'):
+					fx_strength = klc.cal_fx_strength(5)
+				
+				# 尝试获取分型强度等级
+				if hasattr(klc, 'get_fx_strength_level'):
+					fx_strength_level = klc.get_fx_strength_level()
+				
+				# 尝试判断是否为强分型
+				if hasattr(klc, 'is_strong_fx'):
+					is_strong_fx = klc.is_strong_fx()
+				
+				# 如果分型强度小于1，设为0
+				if fx_strength < 1:
+					fx_strength = 0
+
+				# KLC 分型框（起止时间+高低价）：
+				# 仅使用 cal_fx_box 通过 display 条件后生成的 klc.fx_box。
+				# 若无 fx_box，则前端不应绘制分型框。
+				fx_box = getattr(klc, 'fx_box', None)
+				box_start_time = getattr(fx_box, 'start_time', None) if fx_box else None
+				box_end_time = getattr(fx_box, 'end_time', None) if fx_box else None
+				box_high = getattr(fx_box, 'high', None) if fx_box else None
+				box_low = getattr(fx_box, 'low', None) if fx_box else None
+
+				if klc.bb_out:
+					klc_fx_info.append({
+						'time': klc.end_time,
+						'price': klc.low if klc.fx == Chan_FX_TYPE.BOTTOM else klc.high,
+						'fx_type': str(klc.klc_fx_type).replace("Chan_KLC_FX.", ""),
+						'is_bottom': klc.fx == Chan_FX_TYPE.BOTTOM,
+						'fx_strength': fx_strength,           # 分型强度分数 (0-100)
+						'fx_strength_level': fx_strength_level,  # 分型强度等级 (极强/强/中等/弱/极弱)
+						'is_strong_fx': is_strong_fx,        # 是否为强分型
+
+						# 虚线分型框信息（给前端画框用）
+						'start_time': box_start_time,
+						'end_time': box_end_time,
+						'high': float(box_high) if box_high is not None else None,
+						'low': float(box_low) if box_low is not None else None,
+					})
+			except Exception as e:
+				# 如果出错，仍然添加基本信息，但分型强度为0
+				fx_box = getattr(klc, 'fx_box', None)
+				box_start_time = getattr(fx_box, 'start_time', None) if fx_box else None
+				box_end_time = getattr(fx_box, 'end_time', None) if fx_box else None
+				box_high = getattr(fx_box, 'high', None) if fx_box else None
+				box_low = getattr(fx_box, 'low', None) if fx_box else None
+
+				klc_fx_info.append({
+					'time': klc.end_time,
+					'price': klc.low if klc.fx == Chan_FX_TYPE.BOTTOM else klc.high,
+					'fx_type': str(klc.klc_fx_type).replace("Chan_KLC_FX.", ""),
+					'is_bottom': klc.fx == Chan_FX_TYPE.BOTTOM,
+					'fx_strength': 0,
+					'fx_strength_level': "",
+					'is_strong_fx': False,
+
+					# 虚线分型框信息（给前端画框用）
+					'start_time': box_start_time,
+					'end_time': box_end_time,
+					'high': float(box_high) if box_high is not None else None,
+					'low': float(box_low) if box_low is not None else None,
+				})
+	
+	
+	return {
+		'klc_list': klc_list,
+		'klu_list': klu_list,  # 添加KLU列表
+		'bi_list': bi_list,
+		'seg_list': seg_list,
+		'zs_list': zs_list,
+		'bi_zs_list': bi_zs_list,  # 添加BI中枢列表
+		'bsp_list': bsp_list,  # 添加买卖点列表
+		'klc_fx_info': klc_fx_info,  # KLC分型信息
+		'chan_macd': chan_macd_data,   # 添加ChanMACD分析数据
+		'ema52_dict': ema52_dict  # 添加多时间周期EMA52数据
+	}
+
+def classify_trend_stage(df):
+	"""根据 EMA 斜率与多空排列判断趋势方向与阶段
+	返回: direction in {"bull","bear","sideways"}, stage in {"early","mid","late"}, strength_score (0-100)
+	"""
+	if df is None or len(df) < 60:
+		return "sideways", "early", 0
+
+	# 使用 EMA5/10/24/52
+	closes = df['close'].values
+	ema5 = df['ema5'].values if 'ema5' in df else ta.EMA(df, timeperiod=5)
+	ema10 = df['ema10'].values if 'ema10' in df else ta.EMA(df, timeperiod=10)
+	ema24 = df['ema24'].values if 'ema24' in df else ta.EMA(df, timeperiod=24)
+	ema52 = df['ema52'].values if 'ema52' in df else ta.EMA(df, timeperiod=52)
+
+	# 最近N根用于斜率与排列判定
+	lookback = min(30, len(df) - 1)
+	if lookback <= 5:
+		return "sideways", "early", 0
+
+	# 简单斜率: 最近k根的线性变化率近似
+	def slope(arr, k=10):
+		k = min(k, len(arr) - 1)
+		if k < 2:
+			return 0.0
+		y = arr[-k:]
+		x = np.arange(k)
+		# 最小二乘拟合斜率
+		denom = np.dot(x - x.mean(), x - x.mean())
+		if denom == 0:
+			return 0.0
+		m = np.dot(y - y.mean(), x - x.mean()) / denom
+		return float(m)
+
+	k_slope = 12  # 斜率窗口
+	s5 = slope(ema5, k_slope)
+	s10 = slope(ema10, k_slope)
+	s24 = slope(ema24, k_slope)
+	s52 = slope(ema52, k_slope)
+
+	# 多空排列
+	last5, last10, last24, last52 = ema5[-1], ema10[-1], ema24[-1], ema52[-1]
+	bull_stack = last5 > last10 > last24 > last52
+	bear_stack = last5 < last10 < last24 < last52
+
+	# 波动性与动量增强: MACD 柱体最近均值
+	macdhist = df['macdhist'].values if 'macdhist' in df else calculate_macd(df)['histogram']
+	hist_recent = macdhist[-lookback:]
+	hist_power = float(np.mean(np.abs(hist_recent))) if len(hist_recent) else 0.0
+
+	# 方向
+	if bull_stack and s24 > 0 and s52 > 0:
+		direction = "bull"
+	elif bear_stack and s24 < 0 and s52 < 0:
+		direction = "bear"
+	else:
+		# 用价格相对 EMA52 辅助
+		if closes[-1] > last52 and (s24 + s52) > 0:
+			direction = "bull"
+		elif closes[-1] < last52 and (s24 + s52) < 0:
+			direction = "bear"
+		else:
+			direction = "sideways"
+
+	# 阶段: 依据（斜率大小、与EMA52距离、MACD柱体扩张/收敛）
+	dist52 = float((closes[-1] - last52) / last52) if last52 else 0.0
+	slope_score = max(0.0, (abs(s24) + abs(s52)) * 1000.0)  # 归一化
+	dist_score = min(50.0, abs(dist52) * 200.0)
+	hist_score = min(30.0, hist_power * 10.0)
+	strength = float(min(100.0, slope_score + dist_score + hist_score))
+
+	# 简单阶段判定
+	if direction == "sideways":
+		stage = "early"
+		strength = min(strength, 30.0)
+	else:
+		# 查看最近 hist 是否在扩大或收敛
+		if len(hist_recent) >= 6:
+			recent_growth = np.mean(np.abs(hist_recent[-3:])) - np.mean(np.abs(hist_recent[-6:-3]))
+		else:
+			recent_growth = 0.0
+
+		if recent_growth > 0 and abs(dist52) < 0.05:
+			stage = "early"
+		elif recent_growth > 0 and abs(dist52) >= 0.05:
+			stage = "mid"
+		else:
+			stage = "late"
+
+	return direction, stage, strength
+

web/services/runtime/indicators.py

@@ -0,0 +1,103 @@
+from __future__ import annotations
+
+import talib.abstract as ta
+from . import state
+
+def add_indicators(df):
+	macd = ta.MACD(df, fastperiod=state.macd_fast_period, slowperiod=state.macd_slow_period, signalperiod=state.macd_signal_period)
+
+	df['macd'] = macd['macd']
+	df['macdsignal'] = macd['macdsignal']
+	df['macdhist'] = macd['macdhist']
+	df['ma5'] = (ta.MA(df, timeperiod=5)).fillna(0)
+	df['ma10'] = (ta.MA(df, timeperiod=10)).fillna(0)
+	df['ma30'] = (ta.EMA(df, timeperiod=30)).fillna(0)
+	df['ma250'] = (ta.MA(df, timeperiod=250)).fillna(0)
+	# 新增 EMA 指标
+	df['ema5'] = (ta.EMA(df, timeperiod=5)).fillna(0)
+	df['ema10'] = (ta.EMA(df, timeperiod=10)).fillna(0)
+	df['ema24'] = (ta.EMA(df, timeperiod=24)).fillna(0)
+	df['ema52'] = (ta.EMA(df, timeperiod=52)).fillna(0)
+	df['ema26'] = (ta.EMA(df, timeperiod=26)).fillna(0)
+	df['ema13'] = (ta.EMA(df, timeperiod=13)).fillna(0)
+	df['ema7'] = (ta.EMA(df, timeperiod=7)).fillna(0)
+	df['ema104'] = (ta.EMA(df, timeperiod=104)).fillna(0)
+	df['ema156'] = (ta.EMA(df, timeperiod=156)).fillna(0)
+	df['ema208'] = (ta.EMA(df, timeperiod=208)).fillna(0)
+	# 常用SMA 24/52
+	try:
+		df['sma24'] = (ta.SMA(df, timeperiod=24)).fillna(0)
+		df['sma52'] = (ta.SMA(df, timeperiod=52)).fillna(0)
+	except Exception:
+		df['sma24'] = 0
+		df['sma52'] = 0
+	df['rsi'] = ta.RSI(df, timeperiod=14)
+	
+	# 计算布林带 (当前周期 - 20周期，2标准差)
+	bb = ta.BBANDS(df, timeperiod=365, nbdevup=3.0, nbdevdn=3.0, matype=0)
+	df['bb_upper'] = bb['upperband'].fillna(0)
+	df['bb_middle'] = bb['middleband'].fillna(0)
+	df['bb_lower'] = bb['lowerband'].fillna(0)
+	bb30 = ta.BBANDS(df, timeperiod=41, nbdevup=2.3, nbdevdn=2.3, matype=0)
+	#bb30 = ta.BBANDS(df, timeperiod=20, nbdevup=2.0, nbdevdn=2.0, matype=0)
+	df['bbup30'] = bb30['upperband'].fillna(0)
+	df['bblow30'] = bb30['lowerband'].fillna(0)
+	bb302 = ta.BBANDS(df, timeperiod=41, nbdevup=2.0, nbdevdn=2.0, matype=0)
+	#bb302 = ta.BBANDS(df, timeperiod=20, nbdevup=2.0, nbdevdn=2.0, matype=0)
+	df['bbup302'] = bb302['upperband'].fillna(0)
+	df['bblow302'] = bb302['lowerband'].fillna(0)
+	# 计算次周期布林带 (14周期，2标准差)
+	bb_element = ta.BBANDS(df, timeperiod=14, nbdevup=2.0, nbdevdn=2.0, matype=0)
+	df['element_bb_upper'] = bb_element['upperband'].fillna(0)
+	df['element_bb_middle'] = bb_element['middleband'].fillna(0)
+	df['element_bb_lower'] = bb_element['lowerband'].fillna(0)
+	
+	df['macd'] = df['macd'].fillna(0)
+	df['macdsignal'] = df['macdsignal'].fillna(0)
+	df['macdhist'] = df['macdhist'].fillna(0)
+	df['ma5'] = df['ma5'].fillna(0)
+	df['ma10'] = df['ma10'].fillna(0)
+	df['ma30'] = df['ma30'].fillna(0)
+	df['ma250'] = df['ma250'].fillna(0)
+	df['ema5'] = df['ema5'].fillna(0)
+	df['ema10'] = df['ema10'].fillna(0)
+	df['ema24'] = df['ema24'].fillna(0)
+	df['ema52'] = df['ema52'].fillna(0)
+	df['sma24'] = df['sma24'].fillna(0)
+	df['sma52'] = df['sma52'].fillna(0)
+	df['rsi'] = df['rsi'].fillna(0)
+	df['avg_volume'] = df['volume'].rolling(10).mean()
+	# 计算量比，避免产生Infinity值
+	df['volume_ratio'] = df['volume'] / df['avg_volume']
+	# 填充缺失值（前N根K线）
+	df['volume_ratio'] = df['volume_ratio'].fillna(1.0)
+	df['avg_volume'] = df['avg_volume'].fillna(0)
+	
+	# 处理Infinity和-Infinity值
+	df['volume_ratio'] = df['volume_ratio'].replace([float('inf'), float('-inf')], 1.0)
+	
+	# 计算ATR (Average True Range) - 14周期
+	df['atr'] = ta.ATR(df, timeperiod=14)
+	df['atr'] = df['atr'].fillna(0)
+	bb2633 = ta.BBANDS(df, timeperiod=26, nbdevup=3.0, nbdevdn=3.0, matype=0)
+	bbp2633 = (df['close'] - bb2633['lowerband']) / (bb2633['upperband'] - bb2633['lowerband'])
+	df['bb2633upper'] = bb2633['upperband'].fillna(0)
+	df['bb2633lower'] = bb2633['lowerband'].fillna(0)
+	df['bbp2633'] = bbp2633.fillna(0)
+	df['bb2633middle'] = bb2633['middleband'].fillna(0)
+	return df
+
+def calculate_macd(df):
+	"""计算MACD指标"""
+	exp1 = df['close'].ewm(span=state.macd_fast_period, adjust=False).mean()
+	exp2 = df['close'].ewm(span=state.macd_slow_period, adjust=False).mean()
+	macd = exp1 - exp2
+	signal = macd.ewm(span=state.macd_signal_period, adjust=False).mean()
+	histogram = macd - signal
+	
+	return {
+		'macd': macd.tolist(),
+		'signal': signal.tolist(),
+		'histogram': histogram.tolist()
+	}
+

web/services/runtime/market_data.py

@@ -0,0 +1,321 @@
+from __future__ import annotations
+
+import logging
+import time
+from datetime import datetime, timedelta
+
+import pandas as pd
+import requests
+
+from config import DATA_SERVICE_URL
+from . import state
+from .state import DEFAULT_SYMBOLS, DEFAULT_TIMEFRAME_LABELS
+from .timeframes import build_timeframe_labels
+
+logger = logging.getLogger(__name__)
+
+def _parse_time_input(value):
+	if value in (None, '', 0):
+		return None
+	try:
+		return int(float(value))
+	except (ValueError, TypeError):
+		return None
+
+
+def refresh_data_service_metadata(force=False):
+	"""刷新数据服务提供的交易对与周期元信息。"""
+	now = time.time()
+	if not force and state.DATA_SERVICE_AVAILABLE and now - state.SERVICE_METADATA_LAST_REFRESH < 60:
+		return True
+	try:
+		resp = requests.get(f"{DATA_SERVICE_URL}/health", timeout=5)
+		resp.raise_for_status()
+		payload = resp.json()
+		service_symbols = payload.get("symbols") or payload.get("symbol_list") or []
+		base_timeframes = payload.get("timeframes") or payload.get("base_timeframes") or []
+		derived = payload.get("derived_timeframes") or []
+		service_timeframes = list(base_timeframes)
+		for tf in derived:
+			if tf not in service_timeframes:
+				service_timeframes.append(tf)
+		if service_symbols:
+			state.SYMBOLS[:] = service_symbols
+		if service_timeframes:
+			state.TIMEFRAMES.clear()
+			state.TIMEFRAMES.update(build_timeframe_labels(service_timeframes))
+		state.DATA_SERVICE_AVAILABLE = True
+		state.SERVICE_METADATA_LAST_REFRESH = now
+		return True
+	except Exception as exc:
+		logger.warning("无法加载数据服务元信息: %s", exc)
+		if not state.DATA_SERVICE_AVAILABLE:
+			state.TIMEFRAMES.clear()
+			state.TIMEFRAMES.update(DEFAULT_TIMEFRAME_LABELS)
+			state.SYMBOLS[:] = DEFAULT_SYMBOLS
+		state.DATA_SERVICE_AVAILABLE = False
+		return False
+
+
+def _fetch_kl_from_datasvc(symbol, timeframe, start_ms=None, end_ms=None, limit=None):
+	params = {"symbol": symbol, "tf": timeframe}
+	if start_ms is not None:
+		params["start"] = int(start_ms)
+	if end_ms is not None:
+		params["end"] = int(end_ms)
+	if limit is not None:
+		params["limit"] = limit
+	resp = requests.get(f"{DATA_SERVICE_URL}/api/candles", params=params, timeout=10)
+	resp.raise_for_status()
+	data = resp.json()
+	if not data:
+		return None
+	df = pd.DataFrame(data)
+	if df.empty or "timestamp" not in df.columns:
+		return None
+	numeric_cols = ["open", "high", "low", "close", "volume"]
+	df["timestamp"] = pd.to_numeric(df["timestamp"], errors="coerce")
+	df = df.dropna(subset=["timestamp"])
+	df["timestamp"] = df["timestamp"].astype("int64")
+	for col in numeric_cols:
+		if col in df.columns:
+			df[col] = pd.to_numeric(df[col], errors="coerce")
+	df = df.dropna(subset=numeric_cols)
+	df = df.sort_values("timestamp")
+	if limit and len(df) > limit:
+		df = df.tail(limit)
+	df = df.reset_index(drop=True)
+	df["date"] = pd.to_datetime(df["timestamp"], unit='ms', utc=True).dt.tz_convert('Asia/Shanghai')
+	return df
+
+
+# 模块加载时尝试预取一次元信息，但失败不阻塞后续流程
+refresh_data_service_metadata(force=True)
+
+# A股热门股票
+# 模板中 A 股下拉仅放默认一项；用户切换到「A股」时由前端请求 /api/a_stocks 填充全市场（约 5500+）
+A_STOCK_SYMBOLS = [{'symbol': '000001', 'name': '平安银行'}]
+
+def detect_symbol_type(symbol):
+	"""检测交易对类型：crypto 或 a_stock"""
+	if '/' in symbol and 'USDT' in symbol:
+		return 'crypto'
+	elif len(symbol) == 6 and symbol.isdigit():
+		return 'a_stock'
+	else:
+		return 'unknown'
+
+def get_kl_data(symbol, timeframe, limit=100000, start_time=None, end_time=None):
+	"""获取K线数据，支持加密货币和A股"""
+	symbol_type = detect_symbol_type(symbol)
+	
+	if symbol_type == 'crypto':
+		return get_crypto_kl_data(symbol, timeframe, limit, start_time, end_time)
+	elif symbol_type == 'a_stock':
+		return get_a_stock_kl_data(symbol, timeframe, limit, start_time, end_time)
+	else:
+		return None
+
+def _get_crypto_kl_data_via_ccxt(symbol, timeframe, limit=100000, start_time=None, end_time=None):
+	"""获取加密货币K线数据，支持分页加载确保获取指定时间范围内的所有数据"""
+	try:
+		# 初始化参数
+		since = None
+		if start_time:
+			try:
+				since = int(start_time)
+			except ValueError:
+				pass
+		
+		# 结束时间处理
+		until = None
+		if end_time:
+			try:
+				until = int(end_time)
+			except ValueError:
+				pass
+		
+		# 根据时间周期调整每次请求的数据量
+		batch_size = 1000  # 默认批次大小
+		if timeframe in ['1m', '3m', '5m']:
+			batch_size = 1000  # 分钟级数据减少批次大小
+		elif timeframe in ['15m', '30m', '1h']:
+			batch_size = 1000
+		else:
+			batch_size = 1500  # 日线及以上可以获取更多
+		batch_size = 1500  # 默认批次大小
+		# 初始化存储所有K线数据的列表
+		all_ohlcv = []
+		
+		# 初始化当前查询的开始时间
+		current_since = since
+		
+		# 添加请求计数和最大限制
+		request_count = 0
+		max_requests = 300  # 最大请求次数，防止无限循环
+		
+		# 分页加载数据
+		while request_count < max_requests:
+			request_count += 1
+			
+			try:
+				# 获取当前页的数据
+				ohlcv = state.exchange.fetch_ohlcv(symbol, timeframe, since=current_since, limit=batch_size)
+				
+				# 如果没有获取到数据，结束循环
+				if not ohlcv or len(ohlcv) == 0:
+					break
+					
+				# 将获取到的数据添加到总列表中
+				all_ohlcv.extend(ohlcv)
+				
+				# 获取最后一条数据的时间戳
+				last_timestamp = ohlcv[-1][0]
+				
+				# 如果已达到结束时间，结束循环
+				if until and last_timestamp >= until:
+					break
+					
+				# 如果获取的数据条数小于限制数，说明已经获取完所有数据
+				if len(ohlcv) < batch_size:
+					break
+					
+				# 更新下一页的开始时间（加1毫秒避免重复）
+				current_since = last_timestamp + 1
+				
+			except Exception as e:
+				# 如果单个批次失败，继续尝试下一个批次
+				if current_since:
+					# 尝试增加时间跳过可能的问题时间点
+					current_since += 60000  # 跳过1分钟
+				else:
+					break
+			
+			# 防止API请求过于频繁
+			time.sleep(0.3)  # 减少到0.3秒提高效率
+		
+		# 数据为空的情况
+		if not all_ohlcv or len(all_ohlcv) == 0:
+			return None
+		
+		# 转换为DataFrame
+		df = pd.DataFrame(all_ohlcv, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
+		df['date'] = pd.to_datetime(df['timestamp'], unit='ms').dt.tz_localize('UTC').dt.tz_convert('Asia/Shanghai')
+		
+		# 在客户端进行结束时间过滤
+		if until:
+			df = df[df['timestamp'] <= until]
+			
+		# 去除重复数据
+		df = df.drop_duplicates(subset=['timestamp'])
+		
+		# 按时间排序
+		df = df.sort_values('timestamp')
+		
+		# 限制数据条数的逻辑 - 优先考虑时间范围
+		if start_time and end_time:
+			# 如果指定了明确的时间范围，返回该时间范围内的所有数据
+			if len(df) > 100000:  # 防止数据量过大，设置一个合理的上限
+				df = df.tail(100000).reset_index(drop=True)
+		elif limit and len(df) > limit:
+			# 如果没有指定明确时间范围，使用默认的limit限制
+			df = df.tail(limit).reset_index(drop=True)
+		
+		# 如果过滤后没有数据，返回None
+		if len(df) == 0:
+			return None
+		return df
+		
+	except Exception as e:
+		return None
+
+
+def get_crypto_kl_data(symbol, timeframe, limit=100000, start_time=None, end_time=None):
+	"""优先通过本地数据服务获取加密货币K线，失败时回退至交易所API。"""
+	start_ms = _parse_time_input(start_time)
+	end_ms = _parse_time_input(end_time)
+
+	refresh_data_service_metadata()
+	if state.DATA_SERVICE_AVAILABLE:
+		try:
+			df = _fetch_kl_from_datasvc(
+				symbol=symbol,
+				timeframe=timeframe,
+				start_ms=start_ms,
+				end_ms=end_ms,
+				limit=limit,
+			)
+			if df is not None and not df.empty:
+				return df
+		except Exception as exc:
+			logger.warning("数据服务请求失败，准备回退至交易所 API：%s", exc)
+
+	return _get_crypto_kl_data_via_ccxt(symbol, timeframe, limit, start_time, end_time)
+
+
+def get_a_stock_kl_data(symbol, timeframe, limit=100000, start_time=None, end_time=None):
+	"""获取A股K线数据"""
+	try:
+		# 处理时间戳参数转换为日期字符串
+		start_date = None
+		end_date = None
+		
+		if start_time:
+			try:
+				# 尝试解析时间戳（毫秒）
+				start_timestamp = int(start_time)
+				start_date = datetime.fromtimestamp(start_timestamp / 1000).strftime('%Y-%m-%d')
+			except (ValueError, TypeError):
+				# 如果不是时间戳，尝试解析datetime-local格式 (YYYY-MM-DDTHH:MM)
+				try:
+					if 'T' in str(start_time):
+						# datetime-local格式：2025-05-19T06:07
+						start_date = str(start_time).split('T')[0]  # 只取日期部分
+					else:
+						start_date = str(start_time)
+				except:
+					start_date = start_time
+		
+		if end_time:
+			try:
+				# 尝试解析时间戳（毫秒）
+				end_timestamp = int(end_time)
+				end_date = datetime.fromtimestamp(end_timestamp / 1000).strftime('%Y-%m-%d')
+			except (ValueError, TypeError):
+				# 如果不是时间戳，尝试解析datetime-local格式
+				try:
+					if 'T' in str(end_time):
+						# datetime-local格式：2025-05-26T06:07
+						end_date = str(end_time).split('T')[0]  # 只取日期部分
+					else:
+						end_date = str(end_time)
+				except:
+					end_date = end_time
+		
+		# 如果用户指定了时间范围，优先获取该范围内的所有数据
+		actual_limit = limit
+		if start_date and end_date:
+			actual_limit = None  # 不限制数据条数，获取完整时间范围数据
+		
+		# 调用A股数据获取器
+		df = state.china_stock.get_kl_data(symbol, timeframe, start_date, end_date, actual_limit)
+		
+		if df is None:
+			return None
+		return df
+		
+	except Exception as e:
+		return None
+
+def load_crypto_symbols(limit=200):
+	"""加载常见USDT永续合约交易对，返回列表"""
+	refresh_data_service_metadata()
+	if state.SYMBOLS:
+		return state.SYMBOLS[:limit]
+	try:
+		markets = state.exchange.load_markets()
+		symbols = [s for s in markets.keys() if '/USDT' in s and ':USDT' in s]
+		return symbols[:limit]
+	except Exception:
+		return DEFAULT_SYMBOLS[:limit]
+

web/services/runtime/serialize.py

@@ -0,0 +1,300 @@
+from __future__ import annotations
+
+import pandas as pd
+
+from chanlun.core.ChanEnum import Chan_BI_DIR, Chan_SEG_DIR, Chan_MACDSEG_DIR, Chan_MACDHISTSET_DIR
+
+# 辅助函数，转换缠论方向枚举为整数
+def convert_direction(direction):
+	"""转换方向枚举为数字"""
+	if direction == Chan_BI_DIR.UP or direction == Chan_SEG_DIR.UP:
+		return 1
+	elif direction == Chan_BI_DIR.DOWN or direction == Chan_SEG_DIR.DOWN:
+		return -1
+	else:
+		return 0
+
+def format_time_safely(time_obj, client_tz):
+	"""安全地格式化时间对象，处理字符串和datetime两种情况"""
+	if time_obj is None:
+		return None
+		
+	if isinstance(time_obj, str):
+		# 尝试将字符串解析为datetime
+		try:
+			from dateutil import parser
+			time_obj = parser.parse(time_obj)
+			return time_obj.astimezone(client_tz).isoformat()
+		except:
+			return time_obj
+	else:
+		# 已经是datetime对象
+		return time_obj.astimezone(client_tz).isoformat()
+
+def serialize_chan_macd_data(chan_macd_data, client_tz):
+	"""序列化ChanMACD数据为JSON可序列化格式"""
+	serialized_data = {
+		'seg_list': [],
+		'unittf_list': [],
+		'histset_list': [],
+		# 状态标记数据
+		'high_position_list': [],
+		'high_empty_list': [],
+		'low_position_list': [],
+		'low_empty_list': [],
+		'return_zero_list': [],
+		'cross0_up_list': [],
+		'cross0_down_list': [],
+		# 新增：输出KLU的继续背驰/分离背驰标志
+		'klu_list': []
+	}
+	
+	# 序列化seg_list
+	for seg in chan_macd_data.get('seg_list', []):
+		try:
+			seg_data = {
+				'start_time': format_time_safely(seg.start_time, client_tz),
+				'end_time': format_time_safely(seg.end_time, client_tz) if seg.end_time else None,
+				'seg_dir': 'ABOVE' if seg.seg_dir == Chan_MACDSEG_DIR.ABOVE else 'UNDER',
+				'klu_count': len(seg.klu_list) if hasattr(seg, 'klu_list') else 0,
+				'unittf_count': len(seg.unittf_list) if hasattr(seg, 'unittf_list') else 0,
+				'histset_count': len(seg.hist_set) if hasattr(seg, 'hist_set') else 0
+			}
+			serialized_data['seg_list'].append(seg_data)
+		except Exception as e:
+			print(f"序列化seg出错: {e}")
+			continue
+	
+	# 序列化unittf_list（兼容新结构与枚举类型）
+	for unittf in chan_macd_data.get('unittf_list', []):
+		try:
+			dir_value = getattr(unittf, 'uinttf_dir', None)
+			dir_name = getattr(dir_value, 'name', dir_value if isinstance(dir_value, str) else None)
+			start_t = getattr(unittf, 'start_type', None)
+			start_type = getattr(start_t, 'name', start_t)
+			end_t = getattr(unittf, 'end_type', None)
+			end_type = getattr(end_t, 'name', end_t)
+			peak_abs = getattr(unittf, 'peak_abs', None)
+			if peak_abs is None:
+				peak_abs = getattr(unittf, 'peak_hist', None)
+			length = getattr(unittf, 'length', None)
+			if length is None:
+				length = len(unittf.klu_list) if hasattr(unittf, 'klu_list') else None
+
+			unittf_data = {
+				'start_time': format_time_safely(getattr(unittf, 'start_time', None), client_tz),
+				'end_time': format_time_safely(getattr(unittf, 'end_time', None), client_tz) if getattr(unittf, 'end_time', None) else None,
+				'dir': dir_name,  # 'ABOVE' | 'UNDER' | None
+				'start_type': start_type,  # e.g. 'START' | 'CROSS0' | 'NEAR0_UP' | 'NEAR0_DOWN'
+				'end_type': end_type,
+				'invalid': getattr(unittf, 'invalid', False),
+				'peak_abs': peak_abs,
+				'length': length,
+				'klu_count': len(unittf.klu_list) if hasattr(unittf, 'klu_list') else 0,
+				'histset_count': len(unittf.histset_list) if hasattr(unittf, 'histset_list') else 0
+			}
+			serialized_data['unittf_list'].append(unittf_data)
+		except Exception as e:
+			print(f"序列化unittf出错: {e}")
+			continue
+	
+	# 序列化histset_list
+	for histset in chan_macd_data.get('histset_list', []):
+		try:
+			histset_data = {
+				'start_time': format_time_safely(getattr(histset, 'start_time', None), client_tz),
+				'end_time': format_time_safely(getattr(histset, 'end_time', None), client_tz),
+				'histset_dir': 'ABOVE' if histset.histset_dir == Chan_MACDHISTSET_DIR.ABOVE else 'UNDER',
+				'klu_count': len(histset.klu_list) if hasattr(histset, 'klu_list') else 0
+			}
+			serialized_data['histset_list'].append(histset_data)
+		except Exception as e:
+			print(f"序列化histset出错: {e}")
+			continue
+	
+	# 序列化状态标记数据
+	# 序列化高位列表
+	for high_pos in chan_macd_data.get('high_position_list', []):
+		try:
+			high_pos_data = {
+				'time': format_time_safely(high_pos['time'], client_tz),
+				'end_time': format_time_safely(high_pos.get('end_time'), client_tz) if high_pos.get('end_time') else None,
+				'type': high_pos.get('type', 'start'),
+				'macd': high_pos.get('macd'),
+				'signal': high_pos.get('signal'),
+				'macdhist': high_pos.get('macdhist'),
+				'end_macd': high_pos.get('end_macd'),
+				'end_signal': high_pos.get('end_signal'),
+				'end_macdhist': high_pos.get('end_macdhist')
+			}
+			serialized_data['high_position_list'].append(high_pos_data)
+		except Exception as e:
+			print(f"序列化high_position出错: {e}")
+			continue
+	
+	# 序列化高位空列表
+	for high_empty in chan_macd_data.get('high_empty_list', []):
+		try:
+			high_empty_data = {
+				'time': format_time_safely(high_empty['time'], client_tz),
+				'end_time': format_time_safely(high_empty.get('end_time'), client_tz) if high_empty.get('end_time') else None,
+				'type': high_empty.get('type', 'start'),
+				'macd': high_empty.get('macd'),
+				'signal': high_empty.get('signal'),
+				'macdhist': high_empty.get('macdhist'),
+				'end_macd': high_empty.get('end_macd'),
+				'end_signal': high_empty.get('end_signal'),
+				'end_macdhist': high_empty.get('end_macdhist')
+			}
+			serialized_data['high_empty_list'].append(high_empty_data)
+		except Exception as e:
+			print(f"序列化high_empty出错: {e}")
+			continue
+	
+	# 序列化低位与低位空
+	for low_pos in chan_macd_data.get('low_position_list', []):
+		try:
+			low_pos_data = {
+				'time': format_time_safely(low_pos['time'], client_tz),
+				'end_time': format_time_safely(low_pos.get('end_time'), client_tz) if low_pos.get('end_time') else None,
+				'type': low_pos.get('type', 'start'),
+				'macd': low_pos.get('macd'),
+				'signal': low_pos.get('signal'),
+				'macdhist': low_pos.get('macdhist'),
+				'end_macd': low_pos.get('end_macd'),
+				'end_signal': low_pos.get('end_signal'),
+				'end_macdhist': low_pos.get('end_macdhist')
+			}
+			serialized_data['low_position_list'].append(low_pos_data)
+		except Exception as e:
+			print(f"序列化low_position出错: {e}")
+			continue
+
+	for low_empty in chan_macd_data.get('low_empty_list', []):
+		try:
+			low_empty_data = {
+				'time': format_time_safely(low_empty['time'], client_tz),
+				'end_time': format_time_safely(low_empty.get('end_time'), client_tz) if low_empty.get('end_time') else None,
+				'type': low_empty.get('type', 'start'),
+				'macd': low_empty.get('macd'),
+				'signal': low_empty.get('signal'),
+				'macdhist': low_empty.get('macdhist'),
+				'end_macd': low_empty.get('end_macd'),
+				'end_signal': low_empty.get('end_signal'),
+				'end_macdhist': low_empty.get('end_macdhist')
+			}
+			serialized_data['low_empty_list'].append(low_empty_data)
+		except Exception as e:
+			print(f"序列化low_empty出错: {e}")
+			continue
+
+	# 序列化归零轴列表
+	for return_zero in chan_macd_data.get('return_zero_list', []):
+		try:
+			return_zero_data = {
+				'time': format_time_safely(return_zero['time'], client_tz),
+				'end_time': format_time_safely(return_zero.get('end_time'), client_tz) if return_zero.get('end_time') else None,
+				'type': return_zero.get('type', 'start'),
+				'macd': return_zero.get('macd'),
+				'signal': return_zero.get('signal'),
+				'macdhist': return_zero.get('macdhist'),
+				'end_macd': return_zero.get('end_macd'),
+				'end_signal': return_zero.get('end_signal'),
+				'end_macdhist': return_zero.get('end_macdhist')
+			}
+			serialized_data['return_zero_list'].append(return_zero_data)
+		except Exception as e:
+			print(f"序列化return_zero出错: {e}")
+			continue
+	
+	# 序列化穿越零轴列表
+	for cross0_up in chan_macd_data.get('cross0_up_list', []):
+		try:
+			cross0_up_data = {
+				'time': format_time_safely(cross0_up['time'], client_tz),
+				'type': cross0_up.get('type', 'start'),
+				'macd': cross0_up.get('macd'),
+				'signal': cross0_up.get('signal'),
+				'macdhist': cross0_up.get('macdhist')
+			}
+			serialized_data['cross0_up_list'].append(cross0_up_data)
+		except Exception as e:
+			print(f"序列化cross0_up出错: {e}")
+			continue
+	
+	for cross0_down in chan_macd_data.get('cross0_down_list', []):
+		try:
+			cross0_down_data = {
+				'time': format_time_safely(cross0_down['time'], client_tz),
+				'type': cross0_down.get('type', 'start'),
+				'macd': cross0_down.get('macd'),
+				'signal': cross0_down.get('signal'),
+				'macdhist': cross0_down.get('macdhist')
+			}
+			serialized_data['cross0_down_list'].append(cross0_down_data)
+		except Exception as e:
+			print(f"序列化cross0_down出错: {e}")
+			continue
+
+	# 序列化 KLU 列表（仅导出需要的时间与背驰标志）
+	for klu in chan_macd_data.get('klu_list', []):
+		try:
+			serialized_data['klu_list'].append({
+				'time': format_time_safely(getattr(klu, 'time', None), client_tz),
+				'continue_div': bool(getattr(klu, 'continue_div', False)),
+				'separate_div': int(getattr(klu, 'separate_div', 0)) if getattr(klu, 'separate_div', 0) is not None else 0,
+				'near0_return': int(getattr(klu, 'near0_return', 0)) if getattr(klu, 'near0_return', 0) is not None else 0
+			})
+		except Exception as e:
+			print(f"序列化klu出错: {e}")
+			continue
+	
+	return serialized_data
+
+def clean_dataframe_for_json(df):
+	"""清理DataFrame数据用于JSON序列化"""
+	# 创建副本避免修改原始数据
+	clean_df = df.copy()
+	
+	# 替换NaN值为None
+	clean_df = clean_df.where(pd.notnull(clean_df), None)
+	
+	return clean_df
+
+def get_uncompleted_seg_list(seg_list, client_tz):
+	"""获取未完成线段列表，正确处理倒数第二个和最后一个未完成线段"""
+	uncompleted_segs = [seg for seg in seg_list if not seg.is_sure]
+	
+	if len(uncompleted_segs) == 0:
+		return []
+	
+	result = []
+	
+	for i, seg in enumerate(uncompleted_segs):
+		is_last = (i == len(uncompleted_segs) - 1)  # 是否为最后一个未完成线段
+		
+		seg_data = {
+			'start_time': seg.start_bi.start_klc.end_time if isinstance(seg.start_bi.start_klc.end_time, str) else seg.start_bi.start_klc.end_time.astimezone(client_tz).isoformat(),
+			'sure_time': format_time_safely(seg.sure_time, client_tz) if seg.sure_time else None,
+			'start_price': seg.start_bi.start_klc.low if convert_direction(seg.dir) == 1 else seg.start_bi.start_klc.high,
+			'direction': convert_direction(seg.dir)
+		}
+		
+		if is_last:
+			# 最后一个未完成线段：没有结束时间和价格
+			seg_data['end_time'] = None
+			seg_data['end_price'] = None
+		else:
+			# 倒数第二个及之前的未完成线段：使用实际的结束时间和价格
+			if seg.end_bi and seg.end_bi.end_klc:
+				seg_data['end_time'] = seg.end_bi.end_klc.end_time if isinstance(seg.end_bi.end_klc.end_time, str) else seg.end_bi.end_klc.end_time.astimezone(client_tz).isoformat()
+				seg_data['end_price'] = seg.end_bi.end_klc.high if convert_direction(seg.dir) == 1 else seg.end_bi.end_klc.low
+			else:
+				# 如果没有结束笔，设为None
+				seg_data['end_time'] = None
+				seg_data['end_price'] = None
+		
+		result.append(seg_data)
+	
+	return result
+

web/services/runtime/state.py

@@ -0,0 +1,70 @@
+from __future__ import annotations
+
+import sys
+import os
+from collections import OrderedDict
+import logging
+
+import ccxt
+
+_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+if _ROOT not in sys.path:
+	sys.path.append(_ROOT)
+
+from config import MACD_FAST, MACD_SLOW, MACD_SIGNAL, ccxt_proxies
+from services.cn_stock import ChinaStockData
+
+logger = logging.getLogger(__name__)
+
+class TRADE_POINT_TYPE:
+	BUY1 = 1   # 一类买点
+	BUY2 = 2   # 二类买点
+	BUY3 = 3   # 三类买点
+	SELL1 = -1  # 一类卖点
+	SELL2 = -2  # 二类卖点
+	SELL3 = -3  # 三类卖点
+
+
+
+# mutable runtime state
+macd_fast_period = MACD_FAST
+macd_slow_period = MACD_SLOW
+macd_signal_period = MACD_SIGNAL
+
+_proxies = ccxt_proxies()
+_exchange_kwargs = {"enableRateLimit": True}
+if _proxies:
+    _exchange_kwargs["proxies"] = _proxies
+exchange = ccxt.binance(_exchange_kwargs)
+
+china_stock = ChinaStockData()
+_zone_cache = {}
+
+DEFAULT_TIMEFRAME_LABELS = OrderedDict([
+    ("1m", "1分钟"),
+    ("3m", "3分钟"),
+    ("5m", "5分钟"),
+    ("15m", "15分钟"),
+    ("30m", "30分钟"),
+    ("1h", "1小时"),
+    ("2h", "2小时"),
+    ("4h", "4小时"),
+    ("6h", "6小时"),
+    ("8h", "8小时"),
+    ("12h", "12小时"),
+    ("1d", "日线"),
+    ("3d", "3日线"),
+    ("1w", "周线"),
+    ("1M", "月线"),
+])
+
+DEFAULT_SYMBOLS = [
+    'SOL/USDT:USDT', 'BTC/USDT:USDT', 'ETH/USDT:USDT', 'BNB/USDT:USDT', 'XRP/USDT:USDT', 'WIF/USDT:USDT',
+    'ADA/USDT:USDT', 'DOGE/USDT:USDT', 'AVAX/USDT:USDT', 'DOT/USDT:USDT', 'MATIC/USDT:USDT'
+]
+
+TIMEFRAMES = DEFAULT_TIMEFRAME_LABELS.copy()
+SYMBOLS = DEFAULT_SYMBOLS.copy()
+DATA_SERVICE_AVAILABLE = False
+SERVICE_METADATA_LAST_REFRESH = 0
+

web/services/runtime/timeframes.py

@@ -0,0 +1,121 @@
+from __future__ import annotations
+
+from collections import OrderedDict
+from .state import DEFAULT_TIMEFRAME_LABELS
+
+def _zone_cache_ttl(tf_name: str) -> int:
+	"""根据时间周期返回缓存过期时间（秒）"""
+	minutes = timeframe_to_minutes(tf_name) or 5
+	if minutes <= 5:
+		return 120      # 5m及以下: 2分钟
+	elif minutes <= 15:
+		return 300      # 15m: 5分钟
+	elif minutes <= 60:
+		return 600      # 1h: 10分钟
+	else:
+		return 1800     # 4h+: 30分钟
+
+
+def timeframe_to_minutes(tf: str):
+	"""将时间周期转换为分钟数，用于排序。"""
+	if not tf:
+		return None
+	unit = tf[-1]
+	try:
+		value = int(tf[:-1])
+	except (ValueError, TypeError):
+		return None
+	multiplier = {
+		'm': 1,
+		'h': 60,
+		'd': 1440,
+		'w': 10080,
+		'M': 43200,  # 30天近似
+	}.get(unit)
+	if multiplier is None:
+		return None
+	return value * multiplier
+
+
+def format_timeframe_label(tf: str) -> str:
+	"""将时间周期转换为可读标签。"""
+	if not tf:
+		return tf
+	unit = tf[-1]
+	try:
+		value = int(tf[:-1])
+	except (ValueError, TypeError):
+		return tf
+	if unit == 'm':
+		return f"{value}分钟"
+	if unit == 'h':
+		return f"{value}小时"
+	if unit == 'd':
+		return "日线" if value == 1 else f"{value}日线"
+	if unit == 'w':
+		return "周线" if value == 1 else f"{value}周线"
+	if unit == 'M':
+		return "月线" if value == 1 else f"{value}月线"
+	return tf
+
+
+def build_timeframe_labels(timeframes):
+	ordered = sorted(
+		timeframes,
+		key=lambda tf: timeframe_to_minutes(tf) if timeframe_to_minutes(tf) is not None else float('inf'),
+	)
+	labels = OrderedDict()
+	for tf in ordered:
+		labels[tf] = format_timeframe_label(tf)
+	return labels
+
+
+def compute_timeframe_defaults(labels_ordered):
+	"""
+	根据已排序的「周期 → 中文标签」映射，计算主 / 次 / 次次周期默认值。
+	labels_ordered: OrderedDict 或按插入顺序排列的 dict。
+	"""
+	if not labels_ordered:
+		labels_ordered = DEFAULT_TIMEFRAME_LABELS.copy()
+	timeframe_keys = list(labels_ordered.keys())
+	preferred_main = next((tf for tf in ['5m', '15m', '1h'] if tf in labels_ordered), None)
+	default_main = preferred_main or (timeframe_keys[0] if timeframe_keys else '1m')
+	if default_main not in labels_ordered and timeframe_keys:
+		default_main = timeframe_keys[0]
+
+	if timeframe_keys:
+		try:
+			idx = timeframe_keys.index(default_main)
+			default_element = timeframe_keys[idx - 1] if idx > 0 else timeframe_keys[0]
+		except ValueError:
+			default_element = timeframe_keys[0]
+	else:
+		default_element = default_main
+
+	if timeframe_keys:
+		try:
+			idx_el = timeframe_keys.index(default_element)
+			default_sub_sub = timeframe_keys[idx_el - 1] if idx_el > 0 else timeframe_keys[0]
+		except ValueError:
+			default_sub_sub = timeframe_keys[0]
+	else:
+		default_sub_sub = default_element
+
+	return default_main, default_element, default_sub_sub, timeframe_keys
+
+def is_smaller_timeframe(tf1, tf2):
+	"""判断时间周期tf1是否小于tf2"""
+	tf1_value = timeframe_to_minutes(tf1)
+	tf2_value = timeframe_to_minutes(tf2)
+	if tf1_value is None or tf2_value is None:
+		return False
+	return tf1_value < tf2_value
+
+def is_smaller_or_equal_timeframe(tf1, tf2):
+	"""判断时间周期tf1是否小于等于tf2"""
+	tf1_value = timeframe_to_minutes(tf1)
+	tf2_value = timeframe_to_minutes(tf2)
+	if tf1_value is None or tf2_value is None:
+		return False
+	return tf1_value <= tf2_value
+

web/tests/test_analyze_contract.py

@@ -1,14 +1,53 @@
-""" /api/analyze 契约冒烟：关键字段存在于契约清单。"""
+"""ECR-002：加深 /api/analyze 相关契约 —— mock 行情 + analyze_chan 关键字段快照。"""
 from __future__ import annotations
 
 import json
 import sys
 from pathlib import Path
+from unittest.mock import patch
+
+import pandas as pd
+import pytest
 
 ROOT = Path(__file__).resolve().parents[2]
 sys.path.insert(0, str(ROOT))
 sys.path.insert(0, str(ROOT / "web"))
 
+from tests.generate_golden import make_ohlcv  # noqa: E402
+
+
+CONTRACT_KEYS = json.loads(
+    (ROOT / "tests" / "fixtures" / "analyze_contract_keys.json").read_text(encoding="utf-8")
+)
+
+# analyze_chan 直接返回的对象字段（未序列化前）
+ANALYZE_CHAN_KEYS = {
+    "klc_list",
+    "klu_list",
+    "bi_list",
+    "seg_list",
+    "zs_list",
+    "bi_zs_list",
+    "bsp_list",
+    "klc_fx_info",
+    "chan_macd",
+    "ema52_dict",
+}
+
+CHAN_MACD_SERIALIZED_KEYS = {
+    "seg_list",
+    "unittf_list",
+    "histset_list",
+    "high_position_list",
+    "high_empty_list",
+    "low_position_list",
+    "low_empty_list",
+    "return_zero_list",
+    "cross0_up_list",
+    "cross0_down_list",
+    "klu_list",
+}
+
 
 def test_analyze_route_registered():
     from app import app
@@ -21,9 +60,60 @@ def test_analyze_route_registered():
 
 
 def test_contract_keys_stable():
-    keys = json.loads(
+    assert "bi_list" in CONTRACT_KEYS and "seg_list" in CONTRACT_KEYS
-        (ROOT / "tests" / "fixtures" / "analyze_contract_keys.json").read_text(
+    for k in ("kline_data", "macd", "zs_list", "bsp_list", "chan_macd"):
-            encoding="utf-8"
+        assert k in CONTRACT_KEYS
+
+
+def test_analyze_chan_keys_on_fixture():
+    from services.runtime import add_indicators, analyze_chan
+
+    df = add_indicators(make_ohlcv(400))
+    result = analyze_chan(df, symbol="TEST/USDT:USDT", timeframe="5m")
+    assert set(result.keys()) == ANALYZE_CHAN_KEYS
+    assert isinstance(result["bi_list"], list)
+    assert isinstance(result["seg_list"], list)
+    assert isinstance(result["chan_macd"], dict)
+    for k in ("seg_list", "unittf_list", "histset_list"):
+        assert k in result["chan_macd"]
+
+
+def test_serialize_chan_macd_shape():
+    from pytz import timezone
+
+    from services.runtime import add_indicators, analyze_chan, serialize_chan_macd_data
+
+    df = add_indicators(make_ohlcv(200))
+    result = analyze_chan(df)
+    serialized = serialize_chan_macd_data(result["chan_macd"], timezone("Asia/Shanghai"))
+    assert set(serialized.keys()) == CHAN_MACD_SERIALIZED_KEYS
+    # JSON 可序列化
+    json.dumps(serialized)
+
+
+def test_analyze_http_contract_with_mocked_kl():
+    """Flask 测试客户端：mock get_kl_data，断言响应含契约关键字段。"""
+    from app import app
+    from services.runtime import add_indicators
+
+    df = add_indicators(make_ohlcv(300))
+    df = df.copy()
+    if "timestamp" not in df.columns:
+        df["timestamp"] = (pd.to_datetime(df["date"]).astype("int64") // 10**6).astype("int64")
+
+    # analyze 路由使用 `from services.runtime import *`，须 patch 其模块命名空间
+    with patch("api.analyze.get_kl_data", return_value=df):
+        client = app.test_client()
+        resp = client.get(
+            "/api/analyze",
+            query_string={
+                "symbol": "BTC/USDT:USDT",
+                "timeframe": "5m",
+                "timezone": "Asia/Shanghai",
+            },
         )
-    )
+    assert resp.status_code == 200, resp.data[:500]
-    assert "bi_list" in keys and "seg_list" in keys
+    payload = resp.get_json()
+    assert payload is not None and "error" not in payload
+    missing = [k for k in CONTRACT_KEYS if k not in payload]
+    assert not missing, f"missing contract keys: {missing}"

web/tests/test_runtime_facade.py

@@ -0,0 +1,55 @@
+"""ECR-002：runtime 门面公开符号 + 子模块可导入。"""
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+sys.path.insert(0, str(ROOT))
+sys.path.insert(0, str(ROOT / "web"))
+
+REQUIRED = [
+    "get_kl_data",
+    "analyze_chan",
+    "add_indicators",
+    "serialize_chan_macd_data",
+    "clean_dataframe_for_json",
+    "classify_trend_stage",
+    "refresh_data_service_metadata",
+    "TIMEFRAMES",
+    "SYMBOLS",
+    "_zone_cache",
+    "macd_fast_period",
+    "is_smaller_or_equal_timeframe",
+    "get_uncompleted_seg_list",
+]
+
+
+def test_runtime_facade_exports():
+    from services import runtime as R
+
+    for name in REQUIRED:
+        assert hasattr(R, name), f"missing facade export: {name}"
+
+
+def test_runtime_submodules_importable():
+    from services.runtime import state, timeframes, market_data, indicators, analyze, serialize
+
+    assert state.exchange is not None
+    assert callable(timeframes.timeframe_to_minutes)
+    assert callable(market_data.get_kl_data)
+    assert callable(indicators.add_indicators)
+    assert callable(analyze.analyze_chan)
+    assert callable(serialize.convert_direction)
+
+
+def test_thin_shims_still_reexport():
+    from services import market_data as md
+    from services import chan_analyze as ca
+    from services import serializers as ser
+    from services import timeframes as tf
+
+    assert callable(md.get_kl_data)
+    assert callable(ca.analyze_chan)
+    assert callable(ser.serialize_chan_macd_data)
+    assert callable(tf.timeframe_to_minutes)