feat: 盘面洞察收官 — AI 盘面复盘 + 板块主力资金日历

- AI 盘面复盘（市场情绪页）：一键汇总当前情绪/量能/涨停梯队数据生成
  摘要 Prompt，走 /llm/chat/async 异步任务 + 轮询，回复页内展示并经
  _record_history 自动归档「AI 解读历史」
- 板块资金日历：FundFlowSampler 每交易日 14:45 后经
  get_board_ranking(main_net_amount) 采样行业主力净流入 Top10，
  SentimentStore 新增 board_fund 表（(date,rank) 幂等）；口径为
  "涨幅前 50 名中主力净流入最高 10 个"（逐板块 summary 太贵，非全市场
  严格排序）；/market/board-fund/history + 市场情绪页日历视图
- app.py lifespan 接线启停；路线图 ⑥⑩ 标记收官

docs/market-insights-roadmap.md

@@ -10,11 +10,11 @@
 | ② | **涨停生态 / 连板天梯** `/limitup` | 连板高度、首板/二板分布、炸板率、跌停 | 本地 vipdoc .day 文件（strength 扫描器同款读取器），close==涨停价 连续天数可回算 | ✅ 第二批 |
 | ① | **市场情绪时间线** `/sentiment` | 情绪处于冰点/回暖/高潮/退潮 | 涨跌家数、涨停跌停数逐分钟采样（新采样器 + sqlite） | ✅ 第三批 |
 | ⑨ | **市场宽度分时** `/sentiment` | 指数新高但上涨家数背离的顶部信号 | 依赖 ① 的采样器 | ✅ 第三批（随①） |
-| ⑥ | **板块资金日历** | 哪天钱涌向了哪个板块 | board summary 主力净额逐日采样 | ⏳ 后续批次 |
+| ⑥ | **板块资金日历** `/sentiment` | 哪天钱涌向了哪个板块 | FundFlowSampler 每交易日 14:45 后采一次行业主力净流入 Top10（涨幅前 50 名口径） | ✅ 收官 |
 | ⑤ | **板块相关性热力图** `/hotspots` | 哪些板块同涨同跌（抱团 vs 分散） | 热点滚动已缓存的 60 日涨跌矩阵求两两相关 | ✅ 第四批（热点滚动页内「相关性」视图） |
 | ③ | ~~异动雷达时间线~~ | ~~异动密度骤增 = 盘面转折点~~ | `/mac/unusual` 盘中数据源质量差（盘中几乎无记录），**整项取消**（页面/导航/端点已移除） | ❌ 已取消 |
 | ⑧ | **量能仪表盘** `/sentiment` | 放量/缩量（两市累计成交 vs 5日同期均值） | 指数 5 分钟线现成 | ✅ 第四批（并入市场情绪页） |
-| ⑩ | **AI 盘面早报/复盘** | 把以上所有数据"自动读"给你听 | LLM 管道 + ai-history 归档现成 | ⏳ 收尾（必须做） |
+| ⑩ | **AI 盘面复盘** `/sentiment` | 把以上所有数据"自动读"给你听 | LLM 管道 + ai-history 归档现成 | ✅ 收官（情绪页「生成 AI 复盘」按钮，异步任务 + 自动归档） |
 
 ## 批次
 

src/easy_tdx/web/app.py

@@ -173,6 +173,19 @@ async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]:
             ex_client = None
     app.state.ex_client = ex_client
 
+    # --- 板块资金采样器（交易日 14:45 后记一次行业主力净流入排行，供资金日历） ---
+    app.state.fund_flow_sampler = None
+    if mac_client is not None:
+        try:
+            from easy_tdx.web.sentiment_sampler import FundFlowSampler
+
+            fund_sampler = FundFlowSampler(mac_client)
+            fund_sampler.start()
+            app.state.fund_flow_sampler = fund_sampler
+            logger.info("FundFlowSampler 已启动")
+        except Exception:
+            logger.warning("FundFlowSampler 启动失败 — 资金日历不可用", exc_info=True)
+
     # --- 市场情绪采样器（交易时段每分钟落一条广度快照，供 /market/sentiment/*） ---
     # 依赖标准 TDX 客户端（get_market_stat），mock 模式缩短间隔让曲线快速成形。
     app.state.sentiment_sampler = None
@@ -191,6 +204,14 @@ async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]:
 
     yield
 
+    # --- 停止板块资金采样器 ---
+    fund_svc = getattr(app.state, "fund_flow_sampler", None)
+    if fund_svc is not None:
+        try:
+            await fund_svc.stop()
+        except Exception:
+            logger.warning("FundFlowSampler stop failed", exc_info=True)
+
     # --- 停止市场情绪采样器 ---
     sampler_svc = getattr(app.state, "sentiment_sampler", None)
     if sampler_svc is not None:

src/easy_tdx/web/routers/market.py

@@ -195,6 +195,21 @@ async def limitup_history(
     return DictResponse.from_dict(payload)
 
 
+@router.get("/market/board-fund/history", response_model=DictResponse)
+async def board_fund_history(
+    days: int = Query(15, ge=1, le=90, description="返回交易日数"),
+) -> DictResponse:
+    """行业主力净流入逐日排行（FundFlowSampler 每交易日 14:45 后采样一条）。
+
+    口径：涨幅前 50 名行业中主力净流入最高的 10 个（逐板块 summary 太贵，
+    非全市场严格排序）。数据需采样积累，页面空态有明示。
+    """
+    from easy_tdx.web.sentiment_store import get_sentiment_store
+
+    days_rows = get_sentiment_store().list_fund_days(days)
+    return DictResponse.from_dict({"count": len(days_rows), "days": days_rows})
+
+
 @router.get("/fund-flow", response_model=DataFrameResponse)
 async def fund_flow(
     market: str = Query(..., description="市场: SZ, SH"),

src/easy_tdx/web/sentiment_sampler.py

@@ -97,3 +97,81 @@ class SentimentSampler:
             }
         )
         self.samples += 1
+
+
+class FundFlowSampler:
+    """每日收盘前记录一次行业主力净流入排行（板块资金日历数据源）。
+
+    采样窗口：交易时段内 14:45 之后（临近收盘的净流入已基本定型），
+    每日只采一次（``latest_fund_date`` 幂等）。数据走 MAC
+    ``get_board_ranking(sort_by="main_net_amount")``——该实现先按涨幅
+    取候选池再聚合 summary，因此口径是"涨幅前 ``top_n`` 名中主力净流入
+    最高的 ``keep`` 个行业"，并非全市场严格排序（逐板块 summary 太贵）。
+    """
+
+    def __init__(
+        self,
+        client: Any,
+        store: SentimentStore | None = None,
+        interval: float = 300.0,
+        top_n: int = 50,
+        keep: int = 10,
+    ):
+        self._client = client
+        self._store = store or get_sentiment_store()
+        self._interval = interval
+        self._top_n = top_n
+        self._keep = keep
+        self._task: asyncio.Task | None = None
+
+    def start(self) -> None:
+        if self._task is None or self._task.done():
+            self._task = asyncio.create_task(self._run())
+
+    async def stop(self) -> None:
+        if self._task is not None:
+            self._task.cancel()
+            try:
+                await self._task
+            except asyncio.CancelledError:
+                pass
+            self._task = None
+
+    async def _run(self) -> None:
+        logger.info("FundFlowSampler 启动（间隔 %ss，交易日 14:45 后每日一条）", self._interval)
+        while True:
+            try:
+                now = datetime.now()
+                if is_trading_time(now) and (now.hour * 100 + now.minute) >= 1445:
+                    await self._sample_once()
+            except asyncio.CancelledError:
+                raise
+            except Exception:  # noqa: BLE001 — 采样器永不退出
+                logger.warning("板块资金采样失败", exc_info=True)
+            await asyncio.sleep(self._interval)
+
+    async def _sample_once(self) -> None:
+        from easy_tdx.mac.enums import BoardType
+
+        today = int(datetime.now().strftime("%Y%m%d"))
+        if self._store.latest_fund_date() == today:
+            return  # 当日已采样
+        df = await self._client.get_board_ranking(
+            board_type=BoardType.HY,
+            top_n=self._top_n,
+            sort_by="main_net_amount",
+            ascending=False,
+        )
+        if df is None or df.empty:
+            return
+        ranked = df.sort_values("main_net_amount", ascending=False).head(self._keep)
+        boards = [
+            {
+                "code": str(r["code"]),
+                "name": str(r.get("name", r["code"])),
+                "main_net": round(float(r["main_net_amount"]), 0),
+            }
+            for _, r in ranked.iterrows()
+        ]
+        self._store.upsert_fund_day(today, boards)
+        logger.info("板块资金采样完成：%s，Top1 %s", today, boards[0]["name"] if boards else "-")

src/easy_tdx/web/sentiment_store.py

@@ -56,6 +56,18 @@ class SentimentStore:
                     """
                 )
                 conn.execute("CREATE INDEX IF NOT EXISTS idx_samples_date ON samples(date)")
+                conn.execute(
+                    """
+                    CREATE TABLE IF NOT EXISTS board_fund (
+                        date INTEGER NOT NULL,          -- YYYYMMDD
+                        rank INTEGER NOT NULL,          -- 主力净流入名次（1 起）
+                        code TEXT NOT NULL,
+                        name TEXT NOT NULL,
+                        main_net REAL NOT NULL,         -- 主力净流入（元）
+                        PRIMARY KEY (date, rank)
+                    )
+                    """
+                )
                 conn.commit()
             finally:
                 conn.close()
@@ -94,6 +106,62 @@ class SentimentStore:
             finally:
                 conn.close()
 
+    def latest_fund_date(self) -> int:
+        """最近有板块资金采样的交易日（YYYYMMDD），无数据返回 0。"""
+        conn = self._connect()
+        try:
+            row = conn.execute("SELECT MAX(date) AS d FROM board_fund").fetchone()
+            return int(row["d"] or 0)
+        finally:
+            conn.close()
+
+    def upsert_fund_day(self, date: int, boards: list[dict[str, Any]]) -> None:
+        """覆盖写入某日行业主力净流入排行（rank 按列表顺序 1 起）。"""
+        with _write_lock:
+            conn = self._connect()
+            try:
+                conn.execute("DELETE FROM board_fund WHERE date = ?", (int(date),))
+                conn.executemany(
+                    "INSERT INTO board_fund (date, rank, code, name, main_net) VALUES (?,?,?,?,?)",
+                    [
+                        (int(date), i + 1, str(b["code"]), str(b["name"]), float(b["main_net"]))
+                        for i, b in enumerate(boards)
+                    ],
+                )
+                conn.commit()
+            finally:
+                conn.close()
+
+    def list_fund_days(self, days: int = 15) -> list[dict[str, Any]]:
+        """近 N 个有采样的交易日（降序），每日主力净流入排行。"""
+        conn = self._connect()
+        try:
+            rows = conn.execute(
+                """
+                SELECT date, rank, code, name, main_net
+                FROM board_fund
+                WHERE date IN (
+                    SELECT DISTINCT date FROM board_fund ORDER BY date DESC LIMIT ?
+                )
+                ORDER BY date DESC, rank
+                """,
+                (int(days),),
+            ).fetchall()
+            grouped: dict[int, dict[str, Any]] = {}
+            for r in rows:
+                g = grouped.setdefault(int(r["date"]), {"date": int(r["date"]), "boards": []})
+                g["boards"].append(
+                    {
+                        "rank": int(r["rank"]),
+                        "code": str(r["code"]),
+                        "name": str(r["name"]),
+                        "main_net": float(r["main_net"]),
+                    }
+                )
+            return list(grouped.values())
+        finally:
+            conn.close()
+
     def day_samples(self, date: int) -> list[dict[str, Any]]:
         """某交易日的全部分钟采样（按时间升序）。"""
         conn = self._connect()

web-ui/src/types.ts

@@ -748,6 +748,13 @@ export interface LimitUpHistoryRow {
   limit_down: number
 }
 
+// ── 板块主力资金日历（GET /api/v1/market/board-fund/history，每日采样） ─────
+
+export interface BoardFundDay {
+  date: number
+  boards: Array<{ rank: number; code: string; name: string; main_net: number }>
+}
+
 // ── Walk-Forward 样本外验证（v1.27 POST /backtest/wf/run/async）──────────────
 
 export interface WalkForwardWindow {

web-ui/src/views/SentimentView.vue

@@ -7,14 +7,23 @@ import { computed, nextTick, onBeforeUnmount, onMounted, ref } from 'vue'
 import echarts, { DOWN_COLOR, UP_COLOR } from '../echarts-setup'
 import {
   fetchBars,
+  fetchBoardFundHistory,
+  fetchLimitUpEcology,
   fetchLimitUpHistory,
   fetchMarketStat,
   fetchSentimentHistory,
   fetchSentimentToday,
   formatError,
+  runLlmChatWithPolling,
 } from '../api'
 import { fmtAmount, fmtPctSigned } from '../format'
-import type { LimitUpHistoryRow, MarketStat, SentimentDay, SentimentSample } from '../types'
+import type {
+  BoardFundDay,
+  LimitUpHistoryRow,
+  MarketStat,
+  SentimentDay,
+  SentimentSample,
+} from '../types'
 
 const today = ref<{ date: number; count?: number; samples: SentimentSample[] } | null>(null)
 const histDays = ref<SentimentDay[]>([])
@@ -212,6 +221,7 @@ const volEl = ref<HTMLDivElement>()
 let volChart: echarts.ECharts | null = null
 const volRatio = ref<number | null>(null)
 const volDate = ref('')
+const fundDays = ref<BoardFundDay[]>([])
 
 async function loadVolume() {
   try {
@@ -304,11 +314,91 @@ async function loadVolume() {
   }
 }
 
+async function loadFund() {
+  try {
+    fundDays.value = await fetchBoardFundHistory(15)
+  } catch {
+    fundDays.value = [] // 资金日历独立降级
+  }
+}
+
+
 let timer = 0
 
+// ── ⑩ AI 盘面复盘：自动汇总上方数据 → LLM 生成 → 自动归档「AI 解读历史」 ─────
+
+const aiReply = ref('')
+const aiBusy = ref(false)
+const aiError = ref('')
+const aiModel = ref('')
+
+async function buildDigest(): Promise<string> {
+  const lines: string[] = []
+  const s = stat.value
+  if (s) {
+    const denom = Math.max(s.up_count + s.down_count, 1)
+    lines.push(
+      `上涨 ${s.up_count} 家 / 下跌 ${s.down_count} 家（上涨占比 ${((100 * s.up_count) / denom).toFixed(1)}%），` +
+        `涨停 ${s.limit_up_count} 家，跌停 ${s.limit_down_count} 家，两市成交 ${fmtAmount(s.total_amount)}。`,
+    )
+  }
+  if (volRatio.value !== null) {
+    lines.push(`量能：当日两市累计成交较近 5 日同期均值 ${fmtPctSigned(volRatio.value)}。`)
+  }
+  try {
+    const eco = await fetchLimitUpEcology()
+    const sm = eco.summary
+    lines.push(`连板高度 ${sm.max_streak} 板（首板 ${sm.first_board}、二板 ${sm.second_board}、3 板以上 ${sm.plus3}），炸板率 ${sm.blown_rate ?? '-'}%。`)
+  } catch {
+    // 涨停生态不可用时跳过该维度
+  }
+  const lu = luHistory.value.slice(-5)
+  if (lu.length) {
+    lines.push(
+      `近 5 日涨停家数：${lu.map((r) => `${String(r.date).slice(4, 6)}-${String(r.date).slice(6, 8)} ${r.limit_up}`).join('；')}。`,
+    )
+  }
+  const sampled = histDays.value.filter((d) => d.n >= 10).slice(-5)
+  if (sampled.length) {
+    lines.push(
+      `采样上涨占比：${sampled.map((d) => `${String(d.date).slice(4, 6)}-${String(d.date).slice(6, 8)} ${d.up_ratio}%`).join('；')}。`,
+    )
+  }
+  if (lines.length === 0) return ''
+  return `以下是最新的 A 股盘面数据摘要：\n${lines.join('\n')}\n\n` +
+    '请以资深市场分析师的口吻写一段 200~400 字的盘面复盘，依次覆盖：1) 市场情绪与赚钱效应；' +
+    '2) 量能特征（放量/缩量及其含义）；3) 涨停梯队与炸板率反映的题材热度与分歧；4) 结尾一句风险提示。' +
+    '直接给观点和逻辑，不要复述数据。'
+}
+
+async function generateReview() {
+  aiBusy.value = true
+  aiError.value = ''
+  aiReply.value = ''
+  try {
+    const digest = await buildDigest()
+    if (!digest) {
+      aiError.value = '暂无盘面数据可生成复盘'
+      return
+    }
+    const state = await runLlmChatWithPolling(digest)
+    if (state.status === 'failed') {
+      throw new Error(String((state as { error?: string }).error ?? 'AI 解读任务失败'))
+    }
+    const result = state.result as { reply?: string; model?: string }
+    aiReply.value = result.reply ?? ''
+    aiModel.value = result.model ?? ''
+  } catch (e) {
+    aiError.value = formatError(e)
+  } finally {
+    aiBusy.value = false
+  }
+}
+
 onMounted(async () => {
   await load()
   loadVolume()
+  loadFund()
   timer = window.setInterval(() => {
     if (document.hidden) return
     load()
@@ -389,6 +479,41 @@ onBeforeUnmount(() => {
         </div>
       </div>
 
+      <!-- 板块主力资金日历 -->
+      <div class="section">
+        <div class="sec-title">行业主力资金 · 每日净流入 Top 10（交易日 14:45 后采样，需积累）</div>
+        <div class="card fund-card">
+          <div v-for="d in fundDays" :key="d.date" class="fund-row">
+            <span class="mono dim fund-date">{{ String(d.date).slice(4, 6) }}-{{ String(d.date).slice(6, 8) }}</span>
+            <span v-for="b in d.boards" :key="b.code" class="fund-chip mono">
+              {{ b.name }} <span class="up">+{{ (b.main_net / 1e8).toFixed(1) }}亿</span>
+            </span>
+          </div>
+          <div v-if="fundDays.length === 0" class="empty-hint dim">
+            尚无采样：每个交易日的 14:45 后自动记录一次行业主力净流入排行（涨幅前 50 名口径），持续运行后日历成形。
+          </div>
+        </div>
+      </div>
+
+      <!-- AI 盘面复盘 -->
+      <div class="section">
+        <div class="sec-title-ai">
+          AI 盘面复盘
+          <button class="gen-btn" :disabled="aiBusy" @click="generateReview">
+            {{ aiBusy ? '生成中…（约 1~3 分钟）' : aiReply ? '重新生成' : '生成 AI 复盘' }}
+          </button>
+          <span v-if="aiModel" class="dim">{{ aiModel }}</span>
+        </div>
+        <div class="card ai-card">
+          <div v-if="aiBusy" class="dim">模型基于上方情绪 / 量能 / 涨停数据生成中…</div>
+          <div v-else-if="aiError" class="up">{{ aiError }}</div>
+          <div v-else-if="aiReply" class="ai-reply">{{ aiReply }}</div>
+          <div v-else class="dim">
+            汇总本页情绪 / 量能 / 涨停数据交给已配置的模型生成复盘，自动归档到「AI 解读历史」。
+          </div>
+        </div>
+      </div>
+
       <!-- 近 60 日情绪 -->
       <div class="section">
         <div class="sec-title">近 60 日 · 涨停/跌停家数（vipdoc 回补）与上涨占比（采样积累）</div>
@@ -501,6 +626,49 @@ onBeforeUnmount(() => {
   font-weight: 600;
   color: var(--text-muted);
 }
+.sec-title-ai {
+  display: flex;
+  align-items: center;
+  gap: 10px;
+  font-size: 12.5px;
+  font-weight: 600;
+  color: var(--text-muted);
+}
+.gen-btn {
+  font-size: 11.5px;
+  padding: 3px 12px;
+}
+.gen-btn:disabled {
+  opacity: 0.6;
+}
+.fund-card {
+  display: flex;
+  flex-direction: column;
+  gap: 6px;
+  font-size: 12px;
+  padding: 10px 12px;
+}
+.fund-row {
+  display: flex;
+  align-items: center;
+  gap: 6px;
+  flex-wrap: wrap;
+}
+.fund-date {
+  width: 44px;
+  flex-shrink: 0;
+}
+.fund-chip {
+  padding: 2px 8px;
+  border-radius: 999px;
+  background: var(--bg-elevated);
+  border: 1px solid var(--border);
+}
+.ai-card {
+  font-size: 13px;
+  line-height: 1.8;
+  white-space: pre-wrap;
+}
 .chart-card {
   padding: 8px;
   position: relative;