fix: restore markdown-based summary hero extraction

utils/markdown_to_image.py

@@ -10,6 +10,8 @@ import asyncio
 import re
 from loguru import logger
 
+META_KEYWORDS = ["群", "群名", "时间", "日期", "成员", "消息", "统计", "总结", "来源", "生成", "记录"]
+
 
 async def safe_close_browser(browser, timeout: float = 4.0) -> None:
     if not browser:
@@ -67,35 +69,44 @@ async def safe_close_browser(browser, timeout: float = 4.0) -> None:
             logger.warning(f"force kill failed: {e}")
 
 
+def _clean_text(html: str) -> str:
+    return re.sub(r'\s+', ' ', re.sub(r'<.*?>', ' ', html)).strip()
+
+
+def _looks_like_meta(html: str) -> bool:
+    clean = _clean_text(html)
+    if not clean:
+        return False
+    if any(k in clean for k in META_KEYWORDS):
+        return True
+    return len(clean) <= 80
+
+
 def _split_hero(html_body: str):
     title_match = re.search(r'<h1>(.*?)</h1>', html_body, re.S | re.I)
-    hero_title = title_match.group(1).strip() if title_match else "聊天总结"
+    hero_title = _clean_text(title_match.group(1)) if title_match else "聊天总结"
     remain = re.sub(r'<h1>.*?</h1>', '', html_body, count=1, flags=re.S | re.I).strip()
 
-    paragraphs = re.findall(r'<p>(.*?)</p>', remain, re.S | re.I)
-    meta_parts = []
-    used = 0
-    for para in paragraphs[:3]:
-        clean = re.sub(r'<.*?>', '', para).strip()
-        if not clean:
-            continue
-        if len(clean) <= 80 or any(k in clean for k in ["群", "时间", "日期", "成员", "消息", "统计", "总结", "来源"]):
-            meta_parts.append(para.strip())
-            used += 1
-        else:
+    block_pattern = re.compile(r'^\s*(<(?:p|blockquote|ul|ol)[^>]*>.*?</(?:p|blockquote|ul|ol)>)', re.S | re.I)
+    meta_blocks = []
+    for _ in range(4):
+        m = block_pattern.match(remain)
+        if not m:
             break
+        block = m.group(1)
+        if not _looks_like_meta(block):
+            break
+        meta_blocks.append(block.strip())
+        remain = remain[m.end():].strip()
 
-    hero_meta = "<br/>".join(meta_parts) if meta_parts else "群聊总结 / 自动生成"
-    if used > 0:
-        remain = re.sub(r'^(\s*<p>.*?</p>){' + str(used) + r'}', '', remain, count=1, flags=re.S | re.I).strip()
-
-    return hero_title, hero_meta, remain
+    hero_meta = ''.join(meta_blocks)
+    hero_enabled = bool(title_match or meta_blocks)
+    return hero_title, hero_meta, remain, hero_enabled
 
 
 async def md_str_to_html_content(md_content):
-    """将 Markdown 字符串转换为更美观的 HTML 内容。"""
     html_body = markdown.markdown(md_content, extensions=['extra', 'codehilite'])
-    hero_title, hero_meta, remain_html = _split_hero(html_body)
+    hero_title, hero_meta, remain_html, hero_enabled = _split_hero(html_body)
 
     css = """
     <style>
@@ -164,9 +175,7 @@ async def md_str_to_html_content(md_content):
             height: 200px;
             border-radius: 50%;
             border: 1px solid rgba(255,255,255,0.12);
-            box-shadow:
-                0 0 0 24px rgba(255,255,255,0.04),
-                0 0 0 56px rgba(255,255,255,0.025);
+            box-shadow: 0 0 0 24px rgba(255,255,255,0.04), 0 0 0 56px rgba(255,255,255,0.025);
             opacity: 0.9;
             pointer-events: none;
         }
@@ -181,7 +190,6 @@ async def md_str_to_html_content(md_content):
             border: 1px solid rgba(255,255,255,0.18);
             margin-bottom: 14px;
             letter-spacing: .06em;
-            backdrop-filter: none;
         }
         .hero-title {
             position: relative;
@@ -203,13 +211,15 @@ async def md_str_to_html_content(md_content):
             font-size: 0.84em;
             line-height: 1.72;
         }
-        .hero-meta p {
+        .hero-meta p, .hero-meta blockquote, .hero-meta ul, .hero-meta ol {
             margin: 4px 0;
             color: inherit;
+            background: transparent;
+            border: none;
+            padding: 0;
         }
-        .content {
-            padding: 24px 34px 34px;
-        }
+        .hero-meta ul, .hero-meta ol { list-style: none; padding-left: 0; }
+        .content { padding: 24px 34px 34px; }
         h1, h2, h3, h4, h5, h6 {
             color: var(--text);
             margin-top: 24px;
@@ -218,7 +228,7 @@ async def md_str_to_html_content(md_content):
             line-height: 1.35;
             letter-spacing: -0.01em;
         }
-        h1 { display: none; }
+        .content.hero-active h1:first-of-type { display: none; }
         h2 {
             font-size: 1.42em;
             margin-top: 30px;
@@ -234,24 +244,11 @@ async def md_str_to_html_content(md_content):
             padding-left: 12px;
             border-left: 3px solid rgba(20,184,166,0.55);
         }
-        p {
-            margin: 14px 0;
-            color: #334155;
-            line-height: 1.88;
-        }
-        ul, ol {
-            padding-left: 26px;
-            margin: 14px 0 18px;
-        }
-        li {
-            margin: 8px 0;
-            color: #334155;
-        }
+        p { margin: 14px 0; color: #334155; line-height: 1.88; }
+        ul, ol { padding-left: 26px; margin: 14px 0 18px; }
+        li { margin: 8px 0; color: #334155; }
         li::marker { color: var(--primary); }
-        strong {
-            color: #1e293b;
-            font-weight: 700;
-        }
+        strong { color: #1e293b; font-weight: 700; }
         em { color: #5b6b84; }
         code {
             background: rgba(109,94,252,0.08);
@@ -270,12 +267,7 @@ async def md_str_to_html_content(md_content):
             border: 1px solid rgba(255,255,255,0.06);
             box-shadow: inset 0 1px 0 rgba(255,255,255,0.03);
         }
-        pre code {
-            background: transparent;
-            color: inherit;
-            border: none;
-            padding: 0;
-        }
+        pre code { background: transparent; color: inherit; border: none; padding: 0; }
         table {
             border-collapse: separate;
             border-spacing: 0;
@@ -287,17 +279,9 @@ async def md_str_to_html_content(md_content):
             overflow: hidden;
             box-shadow: 0 8px 24px rgba(15,23,42,0.05);
         }
-        th, td {
-            padding: 12px 14px;
-            text-align: left;
-            border-bottom: 1px solid rgba(148,163,184,0.12);
-        }
+        th, td { padding: 12px 14px; text-align: left; border-bottom: 1px solid rgba(148,163,184,0.12); }
         tr:last-child td { border-bottom: none; }
-        th {
-            background: linear-gradient(180deg, rgba(109,94,252,0.10), rgba(109,94,252,0.04));
-            color: #334155;
-            font-weight: 700;
-        }
+        th { background: linear-gradient(180deg, rgba(109,94,252,0.10), rgba(109,94,252,0.04)); color: #334155; font-weight: 700; }
         blockquote {
             margin: 18px 0;
             padding: 14px 18px;
@@ -307,27 +291,22 @@ async def md_str_to_html_content(md_content):
             border-radius: 14px;
             color: #355468;
         }
-        hr {
-            border: none;
-            height: 1px;
-            background: linear-gradient(90deg, transparent, rgba(148,163,184,0.35), transparent);
-            margin: 26px 0;
-        }
-        a {
-            color: var(--primary);
-            text-decoration: none;
-            border-bottom: 1px dashed rgba(109,94,252,0.35);
-        }
-        .signature {
-            margin-top: 34px;
-            text-align: right;
-            color: #73849c;
-            font-size: 0.95em;
-            font-style: italic;
-        }
+        hr { border: none; height: 1px; background: linear-gradient(90deg, transparent, rgba(148,163,184,0.35), transparent); margin: 26px 0; }
+        a { color: var(--primary); text-decoration: none; border-bottom: 1px dashed rgba(109,94,252,0.35); }
+        .signature { margin-top: 34px; text-align: right; color: #73849c; font-size: 0.95em; font-style: italic; }
     </style>
     """
 
+    hero_html = ''
+    content_class = 'content hero-active' if hero_enabled else 'content'
+    if hero_enabled:
+        hero_html = f'''
+      <div class="hero">
+        <div class="hero-badge">AI 群聊总结</div>
+        <h1 class="hero-title">{hero_title}</h1>
+        <div class="hero-meta">{hero_meta}</div>
+      </div>'''
+
     full_html = f'''<html>
     <head>
     <meta charset="UTF-8">
@@ -335,14 +314,9 @@ async def md_str_to_html_content(md_content):
     {css}
     </head>
     <body>
-    <div class="wrap">
-      <div class="hero">
-        <div class="hero-badge">AI 群聊总结</div>
-        <h1 class="hero-title">{hero_title}</h1>
-        <div class="hero-meta">{hero_meta}</div>
-      </div>
-      <div class="content">
-        {remain_html}
+    <div class="wrap">{hero_html}
+      <div class="{content_class}">
+        {remain_html if hero_enabled else html_body}
       </div>
     </div>
     </body>
@@ -378,12 +352,7 @@ async def html_to_image(html_content, output_image):
                     browser_path = path
                     break
 
-        launch_args = [
-            "--no-sandbox",
-            "--disable-setuid-sandbox",
-            "--disable-dev-shm-usage",
-            "--disable-gpu"
-        ]
+        launch_args = ["--no-sandbox", "--disable-setuid-sandbox", "--disable-dev-shm-usage", "--disable-gpu"]
 
         if browser_path:
             logger.debug(f"Launch chromium with system chrome: {browser_path}")
@@ -393,29 +362,17 @@ async def html_to_image(html_content, output_image):
             browser = await p.chromium.launch(args=launch_args)
 
         try:
-            context = await browser.new_context(
-                viewport={"width": 780, "height": 960},
-                device_scale_factor=1.2
-            )
+            context = await browser.new_context(viewport={"width": 780, "height": 960}, device_scale_factor=1.2)
             page = await context.new_page()
-
             logger.debug("Set page content")
             await page.set_content(html_content, wait_until='domcontentloaded', timeout=15000)
-
             logger.debug("Wait for fonts ready")
             await page.evaluate("document.fonts.ready")
             await asyncio.sleep(0.2)
-
             logger.debug(f"Take screenshot: output={output_image}")
-            await page.screenshot(
-                path=output_image,
-                full_page=True,
-                timeout=15000,
-                animations="disabled"
-            )
+            await page.screenshot(path=output_image, full_page=True, timeout=15000, animations="disabled")
             if not os.path.exists(output_image):
                 raise RuntimeError(f"截图失败，输出文件不存在: {output_image}")
-
         finally:
             logger.debug("Closing browser")
             await safe_close_browser(browser)
@@ -431,24 +388,18 @@ async def convert_md_str_to_image(md_content: str, output_image: str, max_retrie
     output_image_path = temp_dir / output_image
 
     last_error = None
-
     for attempt in range(max_retries):
         try:
             logger.debug(f"尝试第 {attempt + 1}/{max_retries} 次生成图片")
-
             if output_image_path.exists():
                 os.remove(str(output_image_path))
-
             full_html = await md_str_to_html_content(md_content)
             await html_to_image(full_html, str(output_image_path))
-
             image_size = os.path.getsize(str(output_image_path))
             if image_size < 1024:
                 raise RuntimeError(f"图片生成异常，大小仅为: {image_size} bytes")
-
             logger.info(f"图片成功生成：{output_image_path}")
             return str(output_image_path.resolve())
-
         except Exception as e:
             last_error = e
             logger.warning(f"第 {attempt + 1} 次尝试失败: {e}")