aria-code 0.167.0__py3-none-win_amd64.whl

aria_code/apps/cli/pdf_report.py

@@ -0,0 +1,977 @@
+"""pdf_report.py — 通用"结构化报告 → 设计过的 PDF"渲染引擎。
+
+背景：`report_generator.py` 已经有 `export_pdf()`（weasyprint/wkhtmltopdf 通用
+HTML→PDF 转换，直接复用，不重写）和 `_md_to_html()`，但它的 `_build_html()`
+是写死给单股票分析报告用的；`ui.py` 的 `get_ui_css_base()` 是给 LLM 现场生成
+HTML 用的 Bloomberg 黑橙配色 token；`backtest_report.py` 又是第三套配色。三者
+互不复用，也都没有"任意报告 → 好看的 PDF"的通用能力。
+
+这个模块提供的是中间层：一个与具体报告内容无关的 `Document`（title + 若干
+`Section`），每个 Section 有一个 `kind`（今天先实现短线报告实际用到的几种：
+stats/badge_table/change_columns/card_grid/row_list/data_table/detail_cards/
+news_cards/text），配合一个 `Theme`（配色/字体/密度）渲染成 HTML，再交给
+report_generator.export_pdf() 转 PDF。新增一种报告类型，只需要写一个"解析成
+Document"的函数（参考 parse_shortterm_report），不需要重新设计 CSS。
+
+用法（命令行侧见 apps/cli/commands/pdf_export_cmds.py 的 /export-pdf）：
+    doc = parse_shortterm_report(md_text)
+    html = render_document(doc, THEMES["institutional"])
+    Path("out.html").write_text(html)
+    from report_generator import export_pdf
+    export_pdf(Path("out.html"))
+"""
+from __future__ import annotations
+
+import html as _html
+import re
+from dataclasses import dataclass, field
+from typing import Any, Callable
+
+
+def esc(s: str) -> str:
+    return _html.escape(str(s or ""), quote=False)
+
+
+def strip_md_bold(s: str) -> str:
+    return re.sub(r"\*\*(.*?)\*\*", r"\1", s or "")
+
+
+# weasyprint (Pango/cairo) has no color-emoji glyph support and no emoji font
+# in this theme's font stack — when an emoji leaks into plain rendered text,
+# cairo substitutes a glyph from whatever monochrome fallback font is on the
+# system, with different font metrics than the surrounding text, so it
+# renders visibly offset from the baseline ("floating" above/beside the
+# line). The fix everywhere in this file is either (a) convert the emoji to
+# its structured {tone, label} form and render it as a `.badge`/`.tag` pill
+# (see _signal_tone/_badge_html — the correct path, used by sections that
+# render cleanly), or (b) strip it outright from any freeform text that has
+# no such structured equivalent. This regex is the fallback for (b): a broad
+# Unicode emoji-block range rather than an enumerated character class, so it
+# doesn't silently miss new emoji the report starts using later.
+_EMOJI_RE = re.compile(
+    "["
+    "\U0001F300-\U0001FAFF"  # misc symbols/pictographs, emoticons, transport, supplemental symbols
+    "\U00002600-\U000027BF"  # misc symbols, dingbats (⚠️☀✅ etc.)
+    "\U00002B00-\U00002BFF"  # misc symbols and arrows — NOT the same block as the
+                              # single-arrow ↑/↓/→ (U+2190-21FF) this report's
+                              # "均线" trend column relies on; this block is stars/
+                              # squares (⭐U+2B50, ⬜U+2B1C) only, safe to strip
+    "\U0001F100-\U0001F1FF"  # enclosed alphanumeric supplement (🆕🆓🆙🆒🆗 etc.)
+                              # + regional indicators (flag pairs) — same block,
+                              # found missing 2026-07-21 when 🆕 (U+1F195) leaked
+                              # through into a section-column label and dragged
+                              # in an Apple-Color-Emoji-Bold font embed even
+                              # though every other emoji in the doc was clean
+    "\U0000FE0F"              # variation selector-16 (emoji presentation)
+    "\U0000200D"              # zero-width joiner
+    "]+"
+)
+
+
+def strip_emoji(s: str) -> str:
+    return _EMOJI_RE.sub("", str(s or "")).strip()
+
+
+# ── 数据模型 ──────────────────────────────────────────────────────────────────
+
+@dataclass
+class Section:
+    kind: str
+    title: str
+    icon: str = "•"
+    note: str = ""
+    content: Any = None
+    new_page: bool = True  # 大部分 section 独立起一页；小的（如 stats）可设 False
+
+
+@dataclass
+class Document:
+    title: str
+    meta: str = ""
+    warning: str = ""
+    sections: list[Section] = field(default_factory=list)
+    footer: str = ""
+
+    def select(self, include: list[str] | None = None, exclude: list[str] | None = None) -> "Document":
+        """按 section 标题子串过滤（内容取舍）。include 为空表示全选。"""
+        secs = self.sections
+        if include:
+            secs = [s for s in secs if any(k.lower() in s.title.lower() for k in include)]
+        if exclude:
+            secs = [s for s in secs if not any(k.lower() in s.title.lower() for k in exclude)]
+        return Document(title=self.title, meta=self.meta, warning=self.warning,
+                         sections=secs, footer=self.footer)
+
+
+@dataclass(frozen=True)
+class Theme:
+    name: str
+    bg: str
+    surface: str          # 卡片/表头以外的浅底色
+    text: str
+    text_muted: str
+    border: str
+    accent: str            # section 图标底色 / 表头底色
+    buy: str; sell: str; hold: str; reduce: str
+    buy_bg: str; sell_bg: str; hold_bg: str; reduce_bg: str
+    font: str
+    mono: str
+    radius: str = "3px"
+
+
+THEMES: dict[str, Theme] = {
+    # 浅色研报风格 —— 面向对外交付（尽调材料/合规文档同调性）
+    "institutional": Theme(
+        name="institutional",
+        bg="#ffffff", surface="#f8f9fb", text="#1c2230", text_muted="#6b7280", border="#e7e9ee",
+        accent="#1f3a5f",
+        buy="#16803c", sell="#c81e1e", hold="#5b6472", reduce="#b4650a",
+        buy_bg="#e8f8ee", sell_bg="#fde8e8", hold_bg="#eef0f3", reduce_bg="#fef3e0",
+        # STHeiti first, not PingFang SC: PingFang ships as a multi-face .ttc
+        # (TrueType Collection) — see fc-list — and weasyprint has to extract/
+        # resubset a single face out of that container into a standalone
+        # CID-keyed CFF font for PDF embedding. That conversion isn't something
+        # Apple's own font ever shipped as; poppler-based readers parse the
+        # result fine, but some other PDF viewers mis-render it as substituted/
+        # wrong glyphs (coherent-looking nonsense, not blank characters — the
+        # signature of a glyph-index mismatch, not missing content). STHeiti
+        # is a single-file font with no .ttc-extraction step, avoiding the
+        # whole failure class. Less refined-looking than PingFang, but correct
+        # everywhere beats good-looking in one viewer and broken in another.
+        font='"STHeiti", "PingFang SC", "Helvetica Neue", Arial, sans-serif',
+        mono='"SF Mono", Menlo, Consolas, monospace',
+    ),
+    # 深色 Bloomberg 终端风格 —— 沿用 apps/cli/prompts/ui.py 的黑橙配色 token，
+    # 跟 /ui 生成的仪表盘、dashboard_generator.py 保持视觉一致
+    "bloomberg": Theme(
+        name="bloomberg",
+        bg="#000000", surface="#111111", text="#E8E9EA", text_muted="#808080", border="#2A2A2A",
+        accent="#F5A623",
+        buy="#00CC66", sell="#FF3B3B", hold="#4A9EFF", reduce="#FFB800",
+        buy_bg="#00331a", sell_bg="#330d0d", hold_bg="#0d1a26", reduce_bg="#332400",
+        font='"IBM Plex Sans", "STHeiti", "PingFang SC", "Helvetica Neue", sans-serif',
+        mono='"IBM Plex Mono", "SF Mono", Menlo, monospace',
+    ),
+}
+
+
+# ── CSS（按 Theme 参数化，两套主题共用同一份布局规则） ──────────────────────────
+
+def build_css(t: Theme) -> str:
+    return f"""
+/* @page 的 margin 区域（页眉页脚跑马灯文字所在的地方）默认不继承 body 的
+   背景色 —— 深色主题下会在页面上下各留一道刺眼的白边，这里显式把整张纸
+   （html 根元素）的背景设成主题色，margin 区域才会跟内容区一致。 */
+html {{ background: {t.bg}; }}
+@page {{
+    size: A4;
+    margin: 20mm 16mm 18mm 16mm;
+    background: {t.bg};
+    @top-center {{ content: "{{TITLE}}"; font-family: {t.font}; font-size: 8.5pt; color: {t.text_muted}; }}
+    @bottom-right {{ content: counter(page) " / " counter(pages); font-family: {t.font}; font-size: 8.5pt; color: {t.text_muted}; }}
+}}
+* {{ box-sizing: border-box; }}
+body {{ font-family: {t.font}; color: {t.text}; background: {t.bg}; font-size: 9.5pt; line-height: 1.55; }}
+.mono {{ font-family: {t.mono}; }}
+
+.cover-title {{ font-size: 24pt; font-weight: 700; color: {t.text}; margin: 0 0 3mm 0; letter-spacing: -0.3px; }}
+.cover-meta {{ font-size: 9pt; color: {t.text_muted}; margin-bottom: 3mm; }}
+.warning-banner {{ background: {t.reduce_bg}; border-left: 3px solid {t.reduce}; color: {t.reduce};
+    padding: 2.5mm 4mm; border-radius: 2px; font-size: 8.8pt; margin-bottom: 6mm; }}
+
+.section {{ margin-top: 8mm; }}
+.section.new-page {{ break-before: page; }}
+.section-head {{ display: flex; align-items: center; gap: 2.5mm; border-bottom: 1.5px solid {t.accent};
+    padding-bottom: 2mm; margin-bottom: 4mm; }}
+.section-icon {{ width: 5.5mm; height: 5.5mm; border-radius: 50%; background: {t.accent}; color: {t.bg};
+    font-size: 7.5pt; display: flex; align-items: center; justify-content: center; flex-shrink: 0; }}
+.section-title {{ font-size: 13pt; font-weight: 700; color: {t.text}; }}
+.section-note {{ font-size: 8.3pt; color: {t.text_muted}; background: {t.surface}; border-radius: 2px;
+    padding: 2mm 3mm; margin-bottom: 4mm; border-left: 2px solid {t.border}; }}
+
+.stat-grid {{ display: flex; flex-wrap: wrap; gap: 3mm; margin-bottom: 4mm; }}
+.stat-card {{ flex: 1 1 27mm; min-width: 27mm; background: {t.surface}; border: 1px solid {t.border};
+    border-radius: {t.radius}; padding: 3mm 3.5mm; break-inside: avoid; }}
+.stat-card .label {{ font-size: 7.8pt; color: {t.text_muted}; margin-bottom: 1mm; }}
+.stat-card .value {{ font-size: 13pt; font-weight: 700; color: {t.text}; }}
+.stat-card.buy .value {{ color: {t.buy}; }}
+.stat-card.sell .value {{ color: {t.sell}; }}
+.meta-table {{ width: 100%; border-collapse: collapse; font-size: 8.6pt; margin-bottom: 2mm; }}
+.meta-table td {{ padding: 1.6mm 2mm; border-bottom: 1px solid {t.border}; vertical-align: top; }}
+.meta-table td:first-child {{ color: {t.text_muted}; width: 32mm; white-space: nowrap; }}
+
+.badge {{ display: inline-block; padding: 0.6mm 2.2mm; border-radius: 8px; font-size: 8pt; font-weight: 600; white-space: nowrap; }}
+.badge.buy {{ background: {t.buy_bg}; color: {t.buy}; }}
+.badge.buy2 {{ background: {t.buy}; color: {t.bg}; }}
+.badge.hold {{ background: {t.hold_bg}; color: {t.hold}; }}
+.badge.reduce {{ background: {t.reduce_bg}; color: {t.reduce}; }}
+.badge.sell {{ background: {t.sell_bg}; color: {t.sell}; }}
+.tag {{ display: inline-block; padding: 0.4mm 1.8mm; border-radius: 6px; font-size: 7.6pt; font-weight: 600; }}
+.tag.up {{ background: {t.sell_bg}; color: {t.sell}; }}
+.tag.down {{ background: {t.sell_bg}; color: {t.sell}; }}
+.tag.pos {{ background: {t.buy_bg}; color: {t.buy}; }}
+.tag.neutral {{ background: {t.hold_bg}; color: {t.hold}; }}
+.tag.confirm {{ background: {t.hold_bg}; color: {t.hold}; }}
+
+table.gtable {{ width: 100%; border-collapse: collapse; font-size: 8.5pt; }}
+table.gtable th {{ background: {t.accent}; color: {t.bg}; text-align: left; font-weight: 600; padding: 2mm 2.2mm; font-size: 8pt; }}
+table.gtable td {{ padding: 1.9mm 2.2mm; border-bottom: 1px solid {t.border}; vertical-align: middle; }}
+table.gtable tr:nth-child(even) td {{ background: {t.surface}; }}
+table.gtable tr {{ break-inside: avoid; }}
+table.gtable .name-cell {{ font-weight: 600; color: {t.text}; }}
+table.gtable .code {{ color: {t.text_muted}; font-size: 7.6pt; }}
+table.gtable.compact {{ font-size: 7.4pt; }}
+table.gtable.compact th {{ padding: 1.6mm 1.3mm; font-size: 7.1pt; text-align: center; }}
+table.gtable.compact td {{ padding: 1.4mm 1.3mm; text-align: center; }}
+table.gtable .up-txt {{ color: {t.sell}; }}
+table.gtable .down-txt {{ color: {t.buy}; }}
+
+.change-cols {{ display: flex; gap: 4mm; }}
+.change-col {{ flex: 1; }}
+.change-col h4 {{ font-size: 9.3pt; margin: 0 0 2mm 0; color: {t.text}; }}
+.change-col ul {{ margin: 0; padding-left: 4mm; font-size: 8.4pt; }}
+.change-col li {{ margin-bottom: 1.3mm; break-inside: avoid; }}
+
+/* 卡片网格用 CSS columns 而不是 flex-wrap —— weasyprint 的 flexbox 分页/
+   分片实现有个已验证过的真实 bug：同样内容前面只要有一段稍微复杂点的封面区块
+   （标题+meta+警示条），紧跟着的 flex-wrap 卡片网格就会被整体错误地推到下一页，
+   哪怕当前页还有大片空白。CSS columns 是印刷排版的标准工具，weasyprint 对它的
+   分页支持成熟得多，同样的内容不会触发这个问题（用 /tmp 下 15 轮最小复现实验
+   二分定位到的，不是瞎猜的）。*/
+.card-grid {{ columns: 2; column-gap: 2.5mm; }}
+.card {{ border: 1px solid {t.border}; border-left: 3px solid {t.border};
+    border-radius: 2px; padding: 2.2mm 3mm; break-inside: avoid; font-size: 8.2pt; background: {t.bg};
+    margin-bottom: 2.5mm; }}
+.card.buy {{ border-left-color: {t.buy}; }}
+.card.sell {{ border-left-color: {t.sell}; }}
+.card.reduce {{ border-left-color: {t.reduce}; }}
+.card .head {{ display: flex; justify-content: space-between; align-items: center; margin-bottom: 1.3mm; }}
+.card .head .name {{ font-weight: 700; color: {t.text}; }}
+.card .head .score {{ color: {t.text_muted}; font-size: 7.8pt; }}
+.card .line {{ color: {t.text_muted}; margin-top: 0.8mm; }}
+
+/* 同样避开 flex：当 .detail 文本长到要换行时，flex 行内的换行会在分页边界
+   被从中间切开（子元素各自独立分页，不是整行一起移动）。改用普通块级 +
+   inline-block 定宽标签，避免这个问题。 */
+.row-list {{ font-size: 8.2pt; }}
+.row-list .row {{ padding: 1.6mm 0; border-bottom: 1px solid {t.border}; break-inside: avoid; }}
+.row-list .row .name {{ font-weight: 600; display: inline-block; width: 30mm; vertical-align: top; }}
+.row-list .row .meta {{ color: {t.text_muted}; display: inline-block; width: 14mm; vertical-align: top; }}
+.row-list .row .detail {{ color: {t.text_muted}; display: inline; }}
+
+.detail-grid {{ columns: 2; column-gap: 3.5mm; }}
+.detail-card {{ border: 1px solid {t.border}; border-radius: 3px;
+    overflow: hidden; break-inside: avoid; margin-bottom: 3.5mm; }}
+.detail-card .dc-head {{ background: {t.buy_bg}; padding: 2.2mm 3.5mm; display: flex; justify-content: space-between;
+    align-items: center; border-bottom: 1px solid {t.border}; }}
+.detail-card .dc-head .name {{ font-size: 10.5pt; font-weight: 700; color: {t.text}; }}
+.detail-card .dc-head .code {{ font-size: 8pt; color: {t.text_muted}; margin-left: 1.5mm; }}
+.detail-card .dc-body {{ padding: 2.5mm 3.5mm; }}
+.kv-grid {{ display: flex; flex-wrap: wrap; }}
+.kv-grid .item {{ width: 50%; padding: 1mm 0; font-size: 8.3pt; }}
+.kv-grid .item .k {{ color: {t.text_muted}; font-size: 7.6pt; }}
+.kv-grid .item .v {{ color: {t.text}; font-weight: 600; }}
+.callout {{ margin-top: 1.5mm; background: {t.reduce_bg}; color: {t.reduce}; font-size: 7.9pt;
+    padding: 1.8mm 2.5mm; border-radius: 2px; }}
+
+.news-item {{ border: 1px solid {t.border}; border-radius: 2px; padding: 2.5mm 3.5mm; margin-bottom: 2.5mm; break-inside: avoid; }}
+.news-item .n-head {{ display: flex; align-items: center; gap: 2mm; margin-bottom: 1.3mm; }}
+.news-item .n-head .name {{ font-weight: 700; color: {t.text}; font-size: 9.3pt; }}
+.news-item .kw {{ font-size: 7.8pt; color: {t.text_muted}; margin-bottom: 1.3mm; }}
+.news-item .article {{ font-size: 7.9pt; margin: 0.8mm 0; color: {t.text}; }}
+.news-item .article .src {{ color: {t.text_muted}; }}
+.news-item .article a {{ color: {t.text}; }}
+
+.footer-note {{ text-align: center; color: {t.text_muted}; font-size: 8pt; margin-top: 10mm; font-style: italic; }}
+
+.text-block {{ color: {t.text}; font-size: 9pt; line-height: 1.7; }}
+.text-block strong {{ color: {t.text}; font-weight: 700; }}
+"""
+
+
+# ── 各 section kind 的渲染器 ──────────────────────────────────────────────────
+
+def _r_stats(content: dict, t: Theme) -> str:
+    cards = "".join(
+        f'<div class="stat-card {c.get("tone","")}"><div class="label">{esc(c["label"])}</div>'
+        f'<div class="value">{esc(c["value"])}</div></div>'
+        for c in content.get("cards", [])
+    )
+    meta = "".join(f"<tr><td>{esc(m['label'])}</td><td>{esc(m['value'])}</td></tr>" for m in content.get("meta", []))
+    return f'<div class="stat-grid">{cards}</div><table class="meta-table">{meta}</table>'
+
+
+def _badge_html(tone: str, label: str) -> str:
+    cls = "buy2" if tone == "buy2" else tone
+    return f'<span class="badge {cls}">{esc(label)}</span>'
+
+
+def _tone_cell_html(v: dict, as_badge: bool) -> str:
+    if as_badge:
+        return _badge_html(v["tone"], v["label"])
+    return f'<span class="tag {v["tone"]}">{esc(v["label"])}</span>'
+
+
+def _r_badge_table(content: dict, t: Theme) -> str:
+    cols = content["columns"]
+    ths = "".join(f'<th style="{c.get("style","")}">{esc(c["label"])}</th>' for c in cols)
+    trs = []
+    for row in content["rows"]:
+        tds = [f'<td class="name-cell">{esc(row["name"])} <span class="code">{esc(row.get("code",""))}</span></td>']
+        for c in cols[1:]:
+            v = row.get(c["key"])
+            if isinstance(v, dict) and "tone" in v:
+                tds.append(f'<td>{_tone_cell_html(v, bool(c.get("badge")))}</td>')
+            else:
+                tds.append(f'<td>{esc(v)}</td>')
+        trs.append(f"<tr>{''.join(tds)}</tr>")
+    return f'<table class="gtable"><tr>{ths}</tr>{"".join(trs)}</table>'
+
+
+def _r_change_columns(content: dict, t: Theme) -> str:
+    cols = []
+    for col in content["columns"]:
+        items = col.get("items", [])
+        if not items:
+            continue
+        lis = "".join(f"<li>{esc(it)}</li>" for it in items)
+        cols.append(f'<div class="change-col"><h4>{esc(col["label"])} ({len(items)})</h4><ul>{lis}</ul></div>')
+    if not cols:
+        return '<div class="section-note">无显著变化</div>'
+    return f'<div class="change-cols">{"".join(cols)}</div>'
+
+
+def _r_card_grid(content: dict, t: Theme) -> str:
+    cards = []
+    for it in content["cards"]:
+        lines = "".join(f'<div class="line">{esc(ln)}</div>' for ln in it.get("lines", []))
+        cards.append(f"""<div class="card {it.get('tone','')}">
+            <div class="head"><span class="name">{esc(it['title'])} <span class="code">{esc(it.get('code',''))}</span></span>
+            <span class="score">{esc(it.get('subtitle',''))}</span></div>{lines}</div>""")
+    return f'<div class="card-grid">{"".join(cards)}</div>'
+
+
+def _r_row_list(content: dict, t: Theme) -> str:
+    rows = []
+    for it in content["rows"]:
+        badge = _badge_html(it["tone"], it["badge_label"]) if it.get("badge_label") else ""
+        rows.append(f"""<div class="row">
+            <span class="name">{esc(it['name'])} {badge}</span>
+            <span class="meta">{esc(it.get('meta',''))}</span>
+            <span class="detail">{esc(it.get('detail',''))}</span></div>""")
+    return f'<div class="row-list">{"".join(rows)}</div>'
+
+
+def _r_data_table(content: dict, t: Theme) -> str:
+    cols = content["columns"]
+    compact = " compact" if content.get("compact") else ""
+    ths = "".join(f"<th>{esc(c['label'])}</th>" for c in cols)
+    trs = []
+    for row in content["rows"]:
+        tds = []
+        for c in cols:
+            v = row.get(c["key"], "")
+            cls = []
+            if c.get("numeric"):
+                cls.append("mono")
+            if c.get("align") == "left":
+                cls.append("name-cell")
+            if c.get("color_rule") == "cn_updown":
+                sv = str(v).strip()
+                if sv.startswith("+"):
+                    cls.append("up-txt")
+                elif sv.startswith("-"):
+                    cls.append("down-txt")
+            if c.get("badge") and _has_signal_emoji(str(v)):
+                tone, label = _signal_tone(str(v))
+                cell = _badge_html(tone, label)
+            elif c.get("badge"):
+                # badge column but this particular row has no recognized signal
+                # emoji (e.g. a subtotal/footer row mixed into a signal table) —
+                # forcing it through _signal_tone would default to a fake "持有
+                # 观望" badge that misrepresents a row that was never a signal.
+                cell = esc(strip_emoji(str(v)))
+            else:
+                cell = esc(v)
+            tds.append(f'<td class="{" ".join(cls)}">{cell}</td>')
+        trs.append(f"<tr>{''.join(tds)}</tr>")
+    return f'<table class="gtable{compact}"><tr>{ths}</tr>{"".join(trs)}</table>'
+
+
+def _r_detail_cards(content: dict, t: Theme) -> str:
+    cards = []
+    for it in content["cards"]:
+        items = "".join(
+            f'<div class="item"><div class="k">{esc(kv["k"])}</div><div class="v">{esc(kv["v"])}</div></div>'
+            for kv in it.get("kv", [])
+        )
+        callout = f'<div class="callout">{esc(it["callout"])}</div>' if it.get("callout") else ""
+        badge = _badge_html(it["tone"], it["badge_label"])
+        cards.append(f"""<div class="detail-card">
+            <div class="dc-head"><span><span class="name">{esc(it['title'])}</span><span class="code">{esc(it.get('code',''))}</span></span>{badge}</div>
+            <div class="dc-body"><div class="kv-grid">{items}</div>{callout}</div></div>""")
+    return f'<div class="detail-grid">{"".join(cards)}</div>'
+
+
+def _r_news_cards(content: dict, t: Theme) -> str:
+    cards = []
+    for it in content["cards"]:
+        arts = "".join(
+            f'<div class="article"><a href="{esc(a["url"])}">{esc(a["title"])}</a> <span class="src">{esc(a.get("meta",""))}</span></div>'
+            for a in it.get("articles", [])
+        )
+        kw = f'<div class="kw">{esc(it["kw"])}</div>' if it.get("kw") else ""
+        cards.append(f"""<div class="news-item">
+            <div class="n-head"><span class="name">{esc(it['title'])}</span><span class="code">{esc(it.get('code',''))}</span>
+            <span class="tag {it['tone']}">{esc(it['tag_label'])}</span></div>{kw}{arts}</div>""")
+    return "".join(cards)
+
+
+def _r_text(content: dict, t: Theme) -> str:
+    # Split on **bold** runs first so each piece can be escaped independently,
+    # then wrap the bold pieces in <strong> — escaping after adding the tags
+    # would turn the tags themselves into visible text.
+    text = content.get("text", "")
+    parts = re.split(r"(\*\*.+?\*\*)", text)
+    html_parts = [
+        f"<strong>{esc(p[2:-2])}</strong>" if p.startswith("**") and p.endswith("**")
+        else esc(p)
+        for p in parts
+    ]
+    body = "".join(html_parts).replace("\n\n", "<br><br>").replace("\n", "<br>")
+    return f'<div class="text-block">{body}</div>'
+
+
+RENDERERS: dict[str, Callable[[Any, Theme], str]] = {
+    "stats": _r_stats,
+    "badge_table": _r_badge_table,
+    "change_columns": _r_change_columns,
+    "card_grid": _r_card_grid,
+    "row_list": _r_row_list,
+    "data_table": _r_data_table,
+    "detail_cards": _r_detail_cards,
+    "news_cards": _r_news_cards,
+    "text": _r_text,
+}
+
+
+def render_document(doc: Document, theme: Theme) -> str:
+    css = build_css(theme).replace("{TITLE}", esc(doc.title))
+    parts = [f"""<div class="cover">
+        <div class="cover-title">{esc(doc.title)}</div>
+        <div class="cover-meta">{esc(doc.meta)}</div>
+        {f'<div class="warning-banner">{esc(doc.warning)}</div>' if doc.warning else ""}
+    </div>"""]
+    for i, sec in enumerate(doc.sections):
+        body = RENDERERS.get(sec.kind, _r_text)(sec.content, theme)
+        note_html = f'<div class="section-note">{esc(sec.note)}</div>' if sec.note else ""
+        page_cls = " new-page" if (sec.new_page and i > 0) else ""
+        parts.append(f"""<div class="section{page_cls}">
+            <div class="section-head"><span class="section-icon">{esc(sec.icon)}</span><span class="section-title">{esc(sec.title)}</span></div>
+            {note_html}{body}</div>""")
+    if doc.footer:
+        parts.append(f'<div class="footer-note">{esc(doc.footer)}</div>')
+    return f'<!doctype html><html><head><meta charset="utf-8"><style>{css}</style></head><body>{"".join(parts)}</body></html>'
+
+
+# ── 报告类型解析器：短线分析报告（阿萨阿股短线报告的 .md 结构） ────────────────
+# 新增报告类型时，照这个函数的样子写一个新的 parse_xxx_report(text) -> Document，
+# 不用碰上面的渲染逻辑。
+
+def _parse_md_table(block: str) -> list[dict]:
+    lines = [ln for ln in block.strip().splitlines() if ln.strip().startswith("|")]
+    if len(lines) < 2:
+        return []
+    headers = [c.strip() for c in lines[0].strip("|").split("|")]
+    rows = []
+    for ln in lines[2:]:
+        cells = [c.strip() for c in ln.strip("|").split("|")]
+        if len(cells) == len(headers):
+            rows.append(dict(zip(headers, cells)))
+    return rows
+
+
+def _parse_name_code(cell: str) -> tuple[str, str]:
+    m = re.match(r"\*\*(.+?)\*\*\s*\((\d+)\)", cell.strip())
+    return (m.group(1), m.group(2)) if m else (cell.strip(), "")
+
+
+def _signal_tone(cell: str) -> tuple[str, str]:
+    if "🟢🟢" in cell:
+        return "buy2", "强力买入"
+    if "🟢" in cell:
+        return "buy", "买入"
+    if "🟡" in cell:
+        return "reduce", "减仓"
+    if "🔴" in cell:
+        return "sell", "卖出"
+    return "hold", "持有观望"
+
+
+def _has_signal_emoji(cell: str) -> bool:
+    return any(m in cell for m in ("🟢", "🟡", "🔴", "⬜"))
+
+
+def parse_shortterm_report(md_text: str) -> Document:
+    title_m = re.search(r"^# (.+)$", md_text, re.M)
+    meta_m = re.search(r"^> \*\*生成时间\*\*:(.+)$", md_text, re.M)
+    warn_m = re.search(r"^> ⚠️ (.+)$", md_text, re.M)
+
+    sections_raw = re.split(r"^## ", md_text, flags=re.M)[1:]
+    sec_map = {}
+    for s in sections_raw:
+        heading, _, body = s.partition("\n")
+        sec_map[heading.strip()] = body
+
+    doc = Document(
+        title=title_m.group(1).strip() if title_m else "分析报告",
+        meta=strip_md_bold(meta_m.group(1)).strip() if meta_m else "",
+        warning=warn_m.group(1).strip() if warn_m else "",
+    )
+
+    # ① 今日概览
+    ov_key = next((k for k in sec_map if "今日概览" in k), None)
+    if ov_key:
+        rows = _parse_md_table(sec_map[ov_key])
+        stat_keys = {"分析标的": "hold", "🟢 买入信号": "buy", "⬜ 持有观望": "hold", "🔴 减仓/离场": "sell",
+                     "平均综合评分": "hold", "深度数据覆盖": "hold"}
+        cards, meta = [], []
+        for r in rows:
+            k, v = r.get("项目", ""), r.get("数值", "")
+            if k in stat_keys:
+                cards.append({"label": strip_emoji(k), "value": v, "tone": stat_keys[k]})
+            else:
+                meta.append({"label": strip_emoji(k), "value": strip_emoji(v)})
+        doc.sections.append(Section("stats", "今日概览", "1", content={"cards": cards, "meta": meta}, new_page=False))
+
+    # ② 综合研判
+    vd_key = next((k for k in sec_map if "综合研判" in k), None)
+    if vd_key:
+        raw = _parse_md_table(sec_map[vd_key])
+        rows = []
+        for r in raw:
+            name, code = _parse_name_code(r.get("标的", ""))
+            tone, label = _signal_tone(r.get("🚦 综合研判", r.get("综合研判", "")))
+            news_cell = r.get("新闻面", "").strip()
+            score_m = re.search(r"[+-][\d.]+", news_cell)
+            score_txt = score_m.group(0) if score_m else ""
+            if "利好" in news_cell:
+                news = {"tone": "pos", "label": f"利好 {score_txt}".strip()}
+            elif "利空" in news_cell:
+                news = {"tone": "up", "label": f"利空 {score_txt}".strip()}
+            elif "中性" in news_cell:
+                news = {"tone": "neutral", "label": "中性 +0.00"}
+            else:
+                news = {"tone": "neutral", "label": "无"}
+            confirm_cell = r.get("印证", "").strip()
+            confirm = {"tone": "confirm" if "共振" in confirm_cell else "neutral",
+                       "label": "共振" if "共振" in confirm_cell else ("中性" if "中性" in confirm_cell else "—")}
+            rows.append({"name": name, "code": code, "signal": {"tone": tone, "label": label}, "news": news,
+                         "confirm": confirm, "action": strip_md_bold(r.get("一句话操作", "")).strip()})
+        doc.sections.append(Section("badge_table", "综合研判 · 一眼看清买卖", "2", content={
+            "columns": [
+                {"key": "name", "label": "标的", "style": "width:26%"},
+                {"key": "signal", "label": "综合研判", "badge": True, "style": "width:12%"},
+                {"key": "news", "label": "新闻面", "style": "width:14%"},
+                {"key": "confirm", "label": "印证", "style": "width:9%"},
+                {"key": "action", "label": "一句话操作"},
+            ],
+            "rows": rows,
+        }, note="综合研判以已回测的量化技术/基本面信号为主导，新闻面情绪仅作共振/分歧提示，分歧不等于反向。"))
+
+    # ③ 信号变化
+    ch_key = next((k for k in sec_map if "信号变化" in k), None)
+    if ch_key:
+        body = sec_map[ch_key]
+        cols = []
+        for pat, label in [(r"\*\*🆕 新增买入\*\*.*?(?=\n\*\*|\Z)", "🆕 新增买入"),
+                            (r"\*\*🚪 退出买入\*\*.*?(?=\n\*\*|\Z)", "🚪 退出买入"),
+                            (r"\*\*⬇️ 信号下调\*\*.*?(?=\n\*\*|\Z)", "⬇️ 信号下调"),
+                            (r"\*\*⬆️ 信号上调\*\*.*?(?=\n\*\*|\Z)", "⬆️ 信号上调")]:
+            m = re.search(pat, body, re.S)
+            if not m:
+                continue
+            items = [strip_emoji(strip_md_bold(ln.strip()[2:])) for ln in m.group(0).splitlines() if ln.strip().startswith("- ")]
+            if items:
+                cols.append({"label": strip_emoji(label), "items": items})
+        if cols:
+            date_m = re.search(r"vs 上一交易日 ([\d\-]+)", ch_key)
+            doc.sections.append(Section("change_columns", f"信号变化（vs 上一交易日 {date_m.group(1) if date_m else ''}）",
+                                         "3", content={"columns": cols}, new_page=False))
+
+    # ④ 今日特殊行情
+    sp_key = next((k for k in sec_map if "今日特殊行情" in k), None)
+    if sp_key:
+        blocks = re.split(r"\n(?=- \*\*)", sec_map[sp_key].strip())
+        cards = []
+        for b in blocks:
+            hm = re.match(r"- \*\*(.+?)\*\* \((\d+)\) (\S+)\s*评分(\d+)", b.strip())
+            if not hm:
+                continue
+            name, code, badge_emoji, score = hm.groups()
+            tone, _ = _signal_tone(badge_emoji)
+            # 第一行是标题行本身（- **name** (code) ...），之后每一行缩进的
+            # "  - 详情" 才是风险提示，用 splitlines 按行处理，不猜整块的缩进规律
+            alerts = [strip_emoji(ln.strip()[2:]) for ln in b.splitlines()[1:] if ln.strip().startswith("- ")]
+            cards.append({"title": name, "code": code, "subtitle": f"评分 {score}", "tone": tone,
+                          "lines": [a.strip() for a in alerts]})
+        if cards:
+            doc.sections.append(Section("card_grid", "今日特殊行情 & 风险提示", "4", content={"cards": cards}))
+
+    # ⑤ 今日优选短线标的
+    pk_key = next((k for k in sec_map if "今日优选短线标的" in k), None)
+    if pk_key:
+        rows = _parse_md_table(sec_map[pk_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"}, {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "分", "label": "分", "numeric": True}, {"key": "信号", "label": "信号", "badge": True},
+            {"key": "现价", "label": "现价", "numeric": True}, {"key": "今涨", "label": "今涨", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "5日", "label": "5日", "numeric": True, "color_rule": "cn_updown"}, {"key": "量比", "label": "量比", "numeric": True},
+            {"key": "目标/参考", "label": "目标/参考", "numeric": True}, {"key": "止损/参考", "label": "止损/参考", "numeric": True},
+            {"key": "R/R", "label": "R/R", "numeric": True}, {"key": "RSI", "label": "RSI", "numeric": True},
+            {"key": "MACD", "label": "MACD"}, {"key": "均线", "label": "均线"}, {"key": "形态", "label": "形态"},
+        ]
+        # "信号" carries its raw 🟢/⬜/🔴 emoji through untouched — _r_data_table's
+        # badge column path converts it via _signal_tone()/_badge_html() the same
+        # way the ②⑥⑦ sections already do. Every other column gets emoji-stripped:
+        # source data appends decorative suffix emoji straight onto plain values
+        # (e.g. "建设银行⭐", RSI "77🔴") that have no badge equivalent and were
+        # rendering as floating glyphs (see strip_emoji's docstring above).
+        clean_rows = [
+            {
+                c["key"]: strip_md_bold(r.get(c["key"], "")) if c["key"] == "信号"
+                else strip_emoji(strip_md_bold(r.get(c["key"], "")))
+                for c in cols_spec
+            }
+            for r in rows
+        ]
+        doc.sections.append(Section("data_table", "今日优选短线标的", "5", content={
+            "columns": cols_spec, "rows": clean_rows, "compact": True,
+        }, note="~¥ 前缀 = 持有标的参考区间，非买入入场价。"))
+
+    # ⑥ 买入/加仓候选
+    buy_key = next((k for k in sec_map if k.startswith("🟢 买入")), None)
+    if buy_key:
+        blocks = re.split(r"\n(?=### )", sec_map[buy_key].strip())
+        cards = []
+        kv_order = ["现价", "建议入场", "目标价", "止损价", "盈亏比", "今日涨幅", "RSI(14)", "MACD状态", "信号类型"]
+        for b in blocks:
+            hm = re.match(r"### (.+?) \((\d+)\)\s*(\S+)", b)
+            if not hm:
+                continue
+            name, code, badge = hm.groups()
+            tone, label = _signal_tone(badge)
+            rows = _parse_md_table(b)
+            kv = {r["项目"]: r["数值"] for r in rows if "项目" in r}
+            risk = strip_emoji(kv.get("⚠️ 追高风险", ""))
+            cards.append({"title": name, "code": code, "tone": tone, "badge_label": label,
+                          "kv": [{"k": strip_emoji(k), "v": strip_emoji(kv[k])} for k in kv_order if k in kv],
+                          "callout": f"注意：{risk}" if risk else None})
+        if cards:
+            doc.sections.append(Section("detail_cards", "买入/加仓候选", "6", content={"cards": cards}))
+
+    # ⑦ 减仓/观察
+    rd_key = next((k for k in sec_map if k.startswith("🔴 减仓")), None)
+    if rd_key:
+        rows = []
+        for line in sec_map[rd_key].splitlines():
+            m = re.match(r"- \*\*(.+?)\*\* \((\d+)\) — (\S+)\s*评分\s*(\d+)\s*\|\s*(.+)", line.strip())
+            if not m:
+                continue
+            name, code, badge, score, detail = m.groups()
+            tone, label = _signal_tone(badge)
+            rows.append({"name": name, "tone": tone, "badge_label": label, "meta": f"评分{score}", "detail": strip_emoji(detail)})
+        if rows:
+            doc.sections.append(Section("row_list", "减仓/观察", "7", content={"rows": rows}))
+
+    # ⑧ 新闻面预测
+    nw_key = next((k for k in sec_map if "新闻面预测" in k), None)
+    if nw_key:
+        blocks = re.split(r"\n(?=### )", sec_map[nw_key].strip())
+        cards = []
+        for b in blocks:
+            hm = re.match(r"### (\S+) (.+?) \((\d+)\) · (\S+) ([+\-][\d.]+)（(\d+) 条相关）", b)
+            if not hm:
+                continue
+            _e, name, code, label, score, count = hm.groups()
+            tone = {"利好": "pos", "利空": "up", "中性": "neutral"}.get(label, "neutral")
+            kw_m = re.search(r"- 情绪关键词: (.+)$", b, re.M)
+            articles = [{"title": t, "url": u, "meta": f"{s} · {d}"}
+                       for t, u, s, d in re.findall(r"- \[(.+?)\]\((.+?)\)\s*<sub>(.+?) · (.+?)</sub>", b)]
+            cards.append({"title": name, "code": code, "tone": tone, "tag_label": f"{label} {score}（{count}条）",
+                          "kw": f"情绪关键词: {kw_m.group(1).strip()}" if kw_m else "", "articles": articles})
+        if cards:
+            doc.sections.append(Section("news_cards", "新闻面预测（近两周资讯情绪）", "8", content={"cards": cards},
+                                        note="情绪分 ∈ [−1, +1]：> +0.15 偏多、< −0.15 偏空，其间为中性。"))
+
+    # ⑨ 短线交易纪律
+    rl_key = next((k for k in sec_map if "短线交易纪律" in k), None)
+    if rl_key:
+        raw = _parse_md_table(sec_map[rl_key])
+        rows = [{"规则": strip_emoji(strip_md_bold(r.get("规则", ""))),
+                 "要求": strip_emoji(strip_md_bold(r.get("要求", "")))} for r in raw]
+        doc.sections.append(Section("data_table", "短线交易纪律", "9", content={
+            "columns": [{"key": "规则", "label": "规则", "align": "left"}, {"key": "要求", "label": "要求", "align": "left"}],
+            "rows": rows,
+        }))
+
+    doc.footer = "Arthera Quant — 短线分析，严格止损，风险自担 · 每日盘后自动更新"
+    return doc
+
+
+def parse_longterm_report(md_text: str) -> Document:
+    title_m = re.search(r"^# (.+)$", md_text, re.M)
+    meta_m = re.search(r"^> \*\*生成日期\*\*:(.+)$", md_text, re.M)
+    warn_m = re.search(r"^> ⚠️ (.+)$", md_text, re.M)
+
+    sections_raw = re.split(r"^## ", md_text, flags=re.M)[1:]
+    sec_map = {}
+    for s in sections_raw:
+        heading, _, body = s.partition("\n")
+        sec_map[heading.strip()] = body
+
+    doc = Document(
+        title=title_m.group(1).strip() if title_m else "长线分析报告",
+        meta=strip_md_bold(meta_m.group(1)).strip() if meta_m else "",
+        warning=warn_m.group(1).strip() if warn_m else "",
+    )
+
+    # ① 本期概览
+    ov_key = next((k for k in sec_map if "本期概览" in k), None)
+    if ov_key:
+        rows = _parse_md_table(sec_map[ov_key])
+        stat_keys = {"分析标的": "hold", "🟢 买入信号": "buy", "⬜ 持有/观望": "hold", "🔴 减仓/卖出": "sell"}
+        cards, meta = [], []
+        for r in rows:
+            k, v = r.get("项目", ""), r.get("数值", "")
+            if k in stat_keys:
+                cards.append({"label": strip_emoji(k), "value": v, "tone": stat_keys[k]})
+            else:
+                meta.append({"label": strip_emoji(k), "value": strip_emoji(v)})
+        doc.sections.append(Section("stats", "本期概览", "1", content={"cards": cards, "meta": meta}, new_page=False))
+
+    # ② 市场环境 — free-text paragraph, not a table. The section body runs up
+    # to the next "^## " heading, which includes the "---" hr separator line
+    # before it — that's markdown page-break decoration, not content, strip it.
+    mk_key = next((k for k in sec_map if "市场环境" in k), None)
+    if mk_key:
+        body = re.sub(r"\n-{3,}\s*$", "", sec_map[mk_key].strip())
+        text = strip_emoji(body.strip())
+        if text:
+            doc.sections.append(Section("text", "市场环境", "2", content={"text": text}, new_page=False))
+
+    # ③ 本期信号变化（买卖候选进出）— richer table form of the same data the
+    # bullet-list "🆕 新增买入" section shows; skip the bullet-list version to
+    # avoid rendering the same 14 stocks twice.
+    ch_key = next((k for k in sec_map if "本期信号变化" in k), None)
+    if ch_key:
+        raw = _parse_md_table(sec_map[ch_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"},
+            {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "当前信号", "label": "当前信号", "badge": True},
+            {"key": "上期信号", "label": "上期信号"},
+            {"key": "当前评分", "label": "当前评分", "numeric": True},
+        ]
+        clean_rows = [
+            {
+                c["key"]: strip_md_bold(r.get(c["key"], "")) if c["key"] == "当前信号"
+                else strip_emoji(strip_md_bold(r.get(c["key"], "")))
+                for c in cols_spec
+            }
+            for r in raw
+        ]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "本期信号变化（买卖候选进出）", "3",
+                                        content={"columns": cols_spec, "rows": clean_rows}))
+
+    # ④ 行业景气度
+    ig_key = next((k for k in sec_map if "行业景气度" in k), None)
+    if ig_key:
+        raw = _parse_md_table(sec_map[ig_key])
+        cols_spec = [
+            {"key": "行业", "label": "行业", "align": "left"},
+            {"key": "标的数", "label": "标的数", "numeric": True},
+            {"key": "平均分", "label": "平均分", "numeric": True},
+            {"key": "景气度", "label": "景气度"},
+            {"key": "3月均涨跌", "label": "3月均涨跌", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "看多/看空", "label": "看多/看空"},
+        ]
+        # "平均分" 单元格里带一个反引号包住的进度条(`███████░░░`)——那是 markdown
+        # 行内代码语法，不是数据，得先去掉反引号本身，保留方块字符(区块元素
+        # U+2580-259F，跟 emoji 完全不同的 Unicode 区块，正常渲染)。
+        clean_rows = [
+            {c["key"]: strip_emoji(r.get(c["key"], "").replace("`", "")) for c in cols_spec}
+            for r in raw
+        ]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "行业景气度", "4",
+                                        content={"columns": cols_spec, "rows": clean_rows}))
+
+    # ⑤ 优选长线标的（评分前列）— the flagship table, same treatment as
+    # shortterm's "今日优选短线标的": 信号 column badge, everything else stripped.
+    pk_key = next((k for k in sec_map if "优选长线标的" in k), None)
+    if pk_key:
+        rows = _parse_md_table(sec_map[pk_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"}, {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "行业", "label": "行业"}, {"key": "综合分", "label": "分", "numeric": True},
+            {"key": "信号", "label": "信号", "badge": True},
+            {"key": "现价", "label": "现价", "numeric": True},
+            {"key": "技术参考价", "label": "参考价", "numeric": True},
+            {"key": "参考空间", "label": "空间", "numeric": True},
+            {"key": "止损", "label": "止损", "numeric": True},
+            {"key": "趋势", "label": "趋势"},
+            {"key": "PE", "label": "PE", "numeric": True}, {"key": "ROE", "label": "ROE", "numeric": True},
+            {"key": "3月涨跌", "label": "3月涨跌", "numeric": True, "color_rule": "cn_updown"},
+        ]
+        clean_rows = [
+            {
+                c["key"]: strip_md_bold(r.get(c["key"], "")) if c["key"] == "信号"
+                else strip_emoji(strip_md_bold(r.get(c["key"], "")))
+                for c in cols_spec
+            }
+            for r in rows
+        ]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "优选长线标的（评分前列）", "5",
+                                        content={"columns": cols_spec, "rows": clean_rows, "compact": True}))
+
+    # 完整分析明细(按行业拆的附录级子表)不纳入 PDF——⑤ 的评分前列表已经把
+    # 各行业最值得看的标的挑出来了，附录里的完整清单价值有限，不值得为一个
+    # h3 子标题嵌套结构再加一层解析复杂度。
+
+    # ⑥/⑦ 操作建议摘要 → 买入/加仓候选 + 减仓/观察候选（拆成两个 row_list）
+    op_key = next((k for k in sec_map if "操作建议摘要" in k), None)
+    if op_key:
+        body = sec_map[op_key]
+        buy_m = re.search(r"### 🟢 买入/加仓候选\s*\n(.*?)(?=\n### |\Z)", body, re.S)
+        sell_m = re.search(r"### 🔴 减仓/观察候选\s*\n(.*?)(?=\n### |\Z)", body, re.S)
+
+        def _rows_from_bullets(block: str) -> list[dict]:
+            out = []
+            for line in (block or "").splitlines():
+                m = re.match(r"-\s+\*\*(.+?)\*\*\s*\((\d+)\)\s*—\s*(\S+)\s*(.+)", line.strip())
+                if not m:
+                    continue
+                name, code, badge, rest = m.groups()
+                tone, label = _signal_tone(badge)
+                out.append({"name": f"{name} ({code})", "tone": tone, "badge_label": label,
+                            "detail": strip_emoji(rest.strip())})
+            return out
+
+        buy_rows = _rows_from_bullets(buy_m.group(1) if buy_m else "")
+        sell_rows = _rows_from_bullets(sell_m.group(1) if sell_m else "")
+        if buy_rows:
+            doc.sections.append(Section("row_list", "买入/加仓候选", "6", content={"rows": buy_rows}))
+        if sell_rows:
+            doc.sections.append(Section("row_list", "减仓/观察候选", "7", content={"rows": sell_rows}))
+
+    # ⑧ 分批建仓参考价
+    bd_key = next((k for k in sec_map if "分批建仓参考价" in k), None)
+    if bd_key:
+        rows = _parse_md_table(sec_map[bd_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"}, {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "信号", "label": "信号", "badge": True},
+            {"key": "当前价", "label": "当前价", "numeric": True},
+            {"key": "首仓(立即)", "label": "首仓", "numeric": True},
+            {"key": "二仓(-5%回撤)", "label": "二仓(-5%)", "numeric": True},
+            {"key": "三仓(-10%回撤)", "label": "三仓(-10%)", "numeric": True},
+            {"key": "建议仓位", "label": "建议仓位"},
+        ]
+        clean_rows = [
+            {
+                c["key"]: strip_md_bold(r.get(c["key"], "")) if c["key"] == "信号"
+                else strip_emoji(strip_md_bold(r.get(c["key"], "")))
+                for c in cols_spec
+            }
+            for r in rows
+        ]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "分批建仓参考价", "8",
+                                        content={"columns": cols_spec, "rows": clean_rows, "compact": True},
+                                        note="长线建仓策略：分 2–3 批次入场，止损统一为 -15%。"))
+
+    # ⑨ 止盈追踪（高涨幅持仓参考）— "建议操作"故意不走 badge：里面的具体指令
+    # ("建议减仓 50%")比单纯的买/卖徽章信息量更大，转成徽章会丢掉这个细节，
+    # 这里只清 emoji 前缀，保留完整文字。
+    tp_key = next((k for k in sec_map if "止盈追踪" in k), None)
+    if tp_key:
+        rows = _parse_md_table(sec_map[tp_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"}, {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "当前价", "label": "当前价", "numeric": True},
+            {"key": "1月涨", "label": "1月涨", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "3月涨", "label": "3月涨", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "年内涨", "label": "年内涨", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "技术参考价", "label": "参考价", "numeric": True},
+            {"key": "建议操作", "label": "建议操作"},
+        ]
+        clean_rows = [{c["key"]: strip_emoji(strip_md_bold(r.get(c["key"], ""))) for c in cols_spec} for r in rows]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "止盈追踪（高涨幅持仓参考）", "9",
+                                        content={"columns": cols_spec, "rows": clean_rows, "compact": True},
+                                        note="以下持仓已有显著盈利，建议逐步减仓 30–50% 锁定利润。"))
+
+    # ⑩ 下周核心关注（接近买点）
+    wk_key = next((k for k in sec_map if "下周核心关注" in k), None)
+    if wk_key:
+        rows = _parse_md_table(sec_map[wk_key])
+        cols_spec = [
+            {"key": "代码", "label": "代码"}, {"key": "名称", "label": "名称", "align": "left"},
+            {"key": "评分", "label": "评分", "numeric": True}, {"key": "现价", "label": "现价", "numeric": True},
+            {"key": "月RSI", "label": "月RSI", "numeric": True}, {"key": "趋势", "label": "趋势"},
+            {"key": "3月涨跌", "label": "3月涨跌", "numeric": True, "color_rule": "cn_updown"},
+            {"key": "上行空间", "label": "上行空间", "numeric": True},
+            {"key": "观察要点", "label": "观察要点", "align": "left"},
+        ]
+        clean_rows = [{c["key"]: strip_emoji(strip_md_bold(r.get(c["key"], ""))) for c in cols_spec} for r in rows]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "下周核心关注（接近买点）", "10",
+                                        content={"columns": cols_spec, "rows": clean_rows, "compact": True}))
+
+    # ⑪ 组合配置建议 — 末行是合计小计，"信号层级"列没有 emoji，_r_data_table
+    # 的 badge 分支会自动识别并原样显示文字，不会被硬套一个假的"持有观望"徽章。
+    pf_key = next((k for k in sec_map if "组合配置建议" in k), None)
+    if pf_key:
+        raw = _parse_md_table(sec_map[pf_key])
+        cols_spec = [
+            {"key": "信号层级", "label": "信号层级", "badge": True},
+            {"key": "数量", "label": "数量"},
+            {"key": "单只建议仓位", "label": "单只建议仓位"},
+            {"key": "小计（中值）", "label": "小计（中值）"},
+        ]
+        clean_rows = [
+            {
+                c["key"]: strip_md_bold(r.get(c["key"], "")) if c["key"] == "信号层级"
+                else strip_emoji(strip_md_bold(r.get(c["key"], "")))
+                for c in cols_spec
+            }
+            for r in raw
+        ]
+        if clean_rows:
+            doc.sections.append(Section("data_table", "组合配置建议", "11",
+                                        content={"columns": cols_spec, "rows": clean_rows}))
+
+    # ⑫ 风险与注意事项
+    rk_key = next((k for k in sec_map if "风险与注意事项" in k), None)
+    if rk_key:
+        # _r_card_grid's "line" divs are plain esc()'d text with no bold support
+        # (unlike _r_text above) — strip ** markers here rather than teach that
+        # renderer markdown, consistent with how every table cell in this file
+        # already handles bold (drop the markers, don't try to preserve emphasis).
+        lines = [strip_emoji(strip_md_bold(ln.strip()[2:])) for ln in sec_map[rk_key].splitlines()
+                if ln.strip().startswith("- ")]
+        if lines:
+            doc.sections.append(Section("card_grid", "风险与注意事项", "12",
+                                        content={"cards": [{"title": "重要提示", "lines": lines}]}))
+
+    doc.footer = "Arthera Quant — 长线分析，辅助决策，风险自担 · 下次更新: 下周同期"
+    return doc
+
+
+PARSERS: dict[str, Callable[[str], Document]] = {
+    "shortterm": parse_shortterm_report,
+    "longterm": parse_longterm_report,
+}