fix:

1. 修复安卓部分机型顶部安全区遮挡回忆录标题的问题；
2. 降低封面图生成阈值和展示逻辑，独立封面图未生成时，使用正文图；
3. 去掉“嗯。”生硬回答，去掉不合理段首承接词；
4. 新增章节封面所需最少插图数的配置项

api/app/features/memoir/helpers.py

@@ -2,9 +2,13 @@
 
 from app.core.config import settings
 from app.core.logging import get_logger
-from app.features.memoir.asset_resolver import resolve_asset_refs_in_markdown
+from app.features.memoir.asset_resolver import (
+    collect_asset_ids_from_markdown,
+    resolve_asset_refs_in_markdown,
+)
 from app.features.memoir.cover_eligibility import (
     chapter_eligible_for_cover_by_inline_body_image_count,
+    chapter_has_story_links,
     primary_chapter_memoir_image,
 )
 from app.features.memoir.memoir_images.schema import (
@@ -89,36 +93,82 @@ def is_image_permanently_unavailable(rec) -> bool:
     return False
 
 
+def _markdown_for_cover_asset_gate(
+    ch: Chapter, *, markdown_for_response: str | None = None
+) -> str:
+    """用于封面闸门与首张 asset 回落：canonical / override、物化 stories、分段快照 body 合并（去重 asset 计数由 parse 完成）。"""
+    md = _chapter_markdown(ch, override=markdown_for_response)
+    parts: list[str] = []
+    if (md or "").strip():
+        parts.append(md.strip())
+    elif chapter_has_story_links(ch):
+        from app.features.memoir.chapter_markdown_compose import (
+            materialize_chapter_markdown_from_loaded_chapter,
+        )
+
+        alt = (materialize_chapter_markdown_from_loaded_chapter(ch) or "").strip()
+        if alt:
+            parts.append(alt)
+    # 仅走 reading_segments 时：DB canonical 可能未写回或不含 asset://，从快照段补图
+    for row in getattr(ch, "reading_segments_json", None) or []:
+        b = (row.get("body_markdown") or "").strip()
+        if b:
+            parts.append(b)
+    return "\n\n".join(parts) if parts else (md or "")
+
+
+def _synthetic_cover_asset_dict(url: str, *, description: str) -> dict:
+    """列表/详情用：无 MemoirImage 行时，用 COS 签名 URL 拼一条 completed 封面 dict。"""
+    return {
+        "placeholder": "",
+        "description": description,
+        "index": 0,
+        "status": IMAGE_STATUS_COMPLETED,
+        "prompt": None,
+        "url": url,
+        "storage_key": None,
+        "provider": None,
+        "style": None,
+        "size": None,
+        "error": None,
+        "retryable": None,
+        "created_at": None,
+        "updated_at": None,
+    }
+
+
 def chapter_cover_to_dict(
-    ch: Chapter, asset_url_map: dict[str, str] | None = None
+    ch: Chapter,
+    asset_url_map: dict[str, str] | None = None,
+    *,
+    markdown_for_response: str | None = None,
 ) -> dict | None:
-    if not chapter_eligible_for_cover_by_inline_body_image_count(ch):
+    view_md = _markdown_for_cover_asset_gate(
+        ch, markdown_for_response=markdown_for_response
+    )
+    if not chapter_eligible_for_cover_by_inline_body_image_count(ch, markdown=view_md):
         return None
+    asset_url_map = asset_url_map or {}
+
+    # 1) 独立章节封面（Celery generate_chapter_cover 写入的 cover_asset_id）
+    aid = getattr(ch, "cover_asset_id", None)
+    if aid and asset_url_map.get(str(aid)):
+        return _synthetic_cover_asset_dict(
+            asset_url_map[str(aid)], description="章节封面"
+        )
+
+    # 2) 尚无独立封面时：用正文里首张 asset:// 的 URL 作卡片封面（故事主图已在正文内）
+    for asset_id in collect_asset_ids_from_markdown(view_md):
+        u = asset_url_map.get(str(asset_id))
+        if u:
+            return _synthetic_cover_asset_dict(u, description="章节封面")
+
+    # 3) 兼容旧数据：章节级 MemoirImage 首行
     m = primary_chapter_memoir_image(ch)
     if m and is_image_permanently_unavailable(m):
         m = None
     if m:
         return memoir_image_to_dict(m)
-    asset_url_map = asset_url_map or {}
-    aid = getattr(ch, "cover_asset_id", None)
-    if aid and asset_url_map.get(str(aid)):
-        url = asset_url_map[str(aid)]
-        return {
-            "placeholder": "",
-            "description": "章节封面",
-            "index": 0,
-            "status": IMAGE_STATUS_COMPLETED,
-            "prompt": None,
-            "url": url,
-            "storage_key": None,
-            "provider": None,
-            "style": None,
-            "size": None,
-            "error": None,
-            "retryable": None,
-            "created_at": None,
-            "updated_at": None,
-        }
     return None
 
 
@@ -139,7 +189,9 @@ def chapter_to_list_dict(
     markdown_for_response: str | None = None,
 ) -> dict:
     """列表视图：与详情字段对齐的最小子集。"""
-    cover = chapter_cover_to_dict(ch, asset_url_map=asset_url_map)
+    cover = chapter_cover_to_dict(
+        ch, asset_url_map=asset_url_map, markdown_for_response=markdown_for_response
+    )
     cover_normalized = first_normalized_image_for_api(cover)
     canonical_raw = _chapter_markdown(ch, override=markdown_for_response)
     wcount = len(canonical_raw.strip()) if canonical_raw else 0
@@ -170,7 +222,9 @@ def chapter_to_dict(
     asset_url_map = asset_url_map or {}
     resolve = lambda aid: asset_url_map.get(aid)  # noqa: E731
 
-    cover = chapter_cover_to_dict(ch, asset_url_map=asset_url_map)
+    cover = chapter_cover_to_dict(
+        ch, asset_url_map=asset_url_map, markdown_for_response=markdown_for_response
+    )
     cover_normalized = first_normalized_image_for_api(cover)
     # 正文真源：优先 canonical_markdown
     canonical_md = _chapter_markdown(ch, override=markdown_for_response)