From 9af2060259f8ccd6b5982c3eb89a8bfb755f1075 Mon Sep 17 00:00:00 2001
From: yangshilin <2157598560@qq.com>
Date: Thu, 16 Apr 2026 20:42:54 +0800
Subject: [PATCH] =?UTF-8?q?fix:=201.=20=E4=BF=AE=E5=A4=8D=E5=AE=89?=
 =?UTF-8?q?=E5=8D=93=E9=83=A8=E5=88=86=E6=9C=BA=E5=9E=8B=E9=A1=B6=E9=83=A8?=
 =?UTF-8?q?=E5=AE=89=E5=85=A8=E5=8C=BA=E9=81=AE=E6=8C=A1=E5=9B=9E=E5=BF=86?=
 =?UTF-8?q?=E5=BD=95=E6=A0=87=E9=A2=98=E7=9A=84=E9=97=AE=E9=A2=98=EF=BC=9B?=
 =?UTF-8?q?=202.=20=E9=99=8D=E4=BD=8E=E5=B0=81=E9=9D=A2=E5=9B=BE=E7=94=9F?=
 =?UTF-8?q?=E6=88=90=E9=98=88=E5=80=BC=E5=92=8C=E5=B1=95=E7=A4=BA=E9=80=BB?=
 =?UTF-8?q?=E8=BE=91=EF=BC=8C=E7=8B=AC=E7=AB=8B=E5=B0=81=E9=9D=A2=E5=9B=BE?=
 =?UTF-8?q?=E6=9C=AA=E7=94=9F=E6=88=90=E6=97=B6=EF=BC=8C=E4=BD=BF=E7=94=A8?=
 =?UTF-8?q?=E6=AD=A3=E6=96=87=E5=9B=BE=EF=BC=9B=203.=20=E5=8E=BB=E6=8E=89?=
 =?UTF-8?q?=E2=80=9C=E5=97=AF=E3=80=82=E2=80=9D=E7=94=9F=E7=A1=AC=E5=9B=9E?=
 =?UTF-8?q?=E7=AD=94=EF=BC=8C=E5=8E=BB=E6=8E=89=E4=B8=8D=E5=90=88=E7=90=86?=
 =?UTF-8?q?=E6=AE=B5=E9=A6=96=E6=89=BF=E6=8E=A5=E8=AF=8D=EF=BC=9B=204.=20?=
 =?UTF-8?q?=E6=96=B0=E5=A2=9E=E7=AB=A0=E8=8A=82=E5=B0=81=E9=9D=A2=E6=89=80?=
 =?UTF-8?q?=E9=9C=80=E6=9C=80=E5=B0=91=E6=8F=92=E5=9B=BE=E6=95=B0=E7=9A=84?=
 =?UTF-8?q?=E9=85=8D=E7=BD=AE=E9=A1=B9?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 api/.env.example                              |   2 +
 api/.env.production                           |   2 +
 api/.env.staging                              |   2 +
 api/app/agents/chat/output_rules.py           |   8 +-
 api/app/agents/chat/prompts_conversation.py   |   4 +-
 api/app/agents/chat/reply_limits.py           |  28 ++++-
 api/app/core/config.py                        |   2 +
 api/app/features/memoir/cover_eligibility.py  |  65 ++++++++---
 api/app/features/memoir/helpers.py            | 104 +++++++++++++-----
 api/app/tasks/chapter_cover_enqueue.py        |  14 +--
 .../test_chapter_cover_fallback_inline.py     |  57 ++++++++++
 ...st_cover_eligibility_effective_markdown.py |  54 +++++++++
 api/tests/test_reply_segments.py              |  13 +++
 app-expo/src/app/(main)/chapter/[id].tsx      |  18 ++-
 app-expo/src/components/screen-header.tsx     |  78 ++++++++++---
 15 files changed, 377 insertions(+), 74 deletions(-)
 create mode 100644 api/tests/test_chapter_cover_fallback_inline.py
 create mode 100644 api/tests/test_cover_eligibility_effective_markdown.py

diff --git a/api/.env.example b/api/.env.example
index 9cca3f7..8c1484f 100644
--- a/api/.env.example
+++ b/api/.env.example
@@ -273,6 +273,8 @@ MEMOIR_IMAGE_MAX_ATTEMPTS=20
 MEMOIR_IMAGE_PROVIDER=liblib
 MEMOIR_IMAGE_STYLE_DEFAULT=watercolor
 MEMOIR_IMAGE_SIZE_DEFAULT=1280x720
+# 章节正文内至少多少张 asset:// 插图才生成/展示章节封面（默认 1=有一张正文图即可）
+MEMOIR_MIN_INLINE_IMAGES_FOR_CHAPTER_COVER=1
 # Story 正文至少多少字才生成主图 intent / 调图（0=不限制）
 STORY_IMAGE_MIN_BODY_CHARS=400
 # 叙事模型输出相对口述过短则回退为口述原文
diff --git a/api/.env.production b/api/.env.production
index 5d7a0a9..d909f42 100644
--- a/api/.env.production
+++ b/api/.env.production
@@ -235,6 +235,8 @@ MEMOIR_IMAGE_MAX_ATTEMPTS=20
 MEMOIR_IMAGE_PROVIDER=liblib
 MEMOIR_IMAGE_STYLE_DEFAULT=watercolor
 MEMOIR_IMAGE_SIZE_DEFAULT=1280x720
+# 章节正文内至少多少张 asset:// 插图才生成/展示章节封面（≥1 即有一张图可出封面）
+MEMOIR_MIN_INLINE_IMAGES_FOR_CHAPTER_COVER=1
 # Story 正文至少多少字才生成主图 intent / 调图（0=不限制）
 STORY_IMAGE_MIN_BODY_CHARS=800
 # 叙事模型输出相对口述过短则回退为口述原文
diff --git a/api/.env.staging b/api/.env.staging
index 1274ebb..5a8ab42 100644
--- a/api/.env.staging
+++ b/api/.env.staging
@@ -168,6 +168,8 @@ MEMOIR_IMAGE_MAX_ATTEMPTS=20
 MEMOIR_IMAGE_PROVIDER=liblib
 MEMOIR_IMAGE_STYLE_DEFAULT=watercolor
 MEMOIR_IMAGE_SIZE_DEFAULT=1280x720
+# 章节正文内至少多少张 asset:// 插图才生成/展示章节封面（≥1 即有一张图可出封面）
+MEMOIR_MIN_INLINE_IMAGES_FOR_CHAPTER_COVER=1
 # 可选，Liblib 返回图片域名不在默认白名单时（逗号分隔）
 # MEMOIR_IMAGE_DOWNLOAD_HOSTS=liblib.cloud,liblibai.cloud
 
diff --git a/api/app/agents/chat/output_rules.py b/api/app/agents/chat/output_rules.py
index 9aa7376..cbfeece 100644
--- a/api/app/agents/chat/output_rules.py
+++ b/api/app/agents/chat/output_rules.py
@@ -9,14 +9,17 @@ def chat_output_rules() -> str:
         "`[SPLIT]`（仅此一处方括号用法）；**禁止**输出全角或半角括号及其中任何内容，包括："
         "策略/舞台说明（如「（先接住情绪）」「（共情）」），以及**表演性、声效、动作描写**"
         "（如「（轻轻笑）」「（笑）」「（叹气）」「（顿了顿）」「（低声）」「（咳嗽）」「（清了清嗓子）」等——对用户说话就当口播，不要剧本括注）；"
-        "若需停顿或语气，用口语里的「嗯」「唉」或省略号自然写出，**不要**用括号包装动作或旁白；"
+        "**禁止**以「嗯。」**起头**（含「嗯。」后立刻接任何正文——一律不得用这种停顿起手）、禁止单独成泡只有「嗯。」——生硬、像生冷打字机；"
+        "若需停顿或语气，优先用省略号、或把承接半句直接钉在对方原词上；可用「唉」等；**避免**每条消息都以「好。」「对。」单独打头再接一大段（易像程式客服）；"
+        "**不要**用括号包装动作或旁白；"
         "思考过程或任何元注释同样**绝不可**出现在对用户说的话中；"
         "主持人口吻与播报腔（「那么接下来」「让我们」「首先」「感谢您的分享」类串联或晚会导语感）；"
         "课文式硬切话题（「下面我们聊聊」「接下来我想了解」「换个话题」「让我们把话题转向…」等未承接就上段话的起手或硬转向）；"
         "推白话轮与总结腔（空泛的「听起来你…」「听起来当时…」「听起来挺…」「听你这么说…」「照你这么说…」"
         "等阶段总结或程序性过渡，而非贴着对方上一轮话头半句并肩地往下长）；"
         "强行搭话式「这让我想起…」接**与当前画面不沾边**的自己的故事或常识，制造虚假亲密；"
-        "采访腔（「我注意到」「我想了解」）；尤其在用户长段倾诉或情绪很重时，**勿**整条回复仅单个语气词（孤立的「嗯」「好」「明白」等），须至少有半句贴着对方原词的承接；"
+        "采访腔（「我注意到」「我想了解」）；尤其在用户长段倾诉或情绪很重时，**勿**整条回复仅单个语气词（孤立的「好」「明白」等），须至少有半句贴着对方原词的承接；"
+        "连续多轮都以「好，……」「对，……」式**同一套路起句**（发语词后接泛共情），须主动轮换——尽量**直接**从对方刚说的物象、人或半句并肩起笔；"
         "书面评介腔（「值得一提的是」「总的来说」「从某种意义上」）；"
         "空话铺垫（「这确实是个好问题」类）；**以核对为名**重复对方已明确说过的基础信息（如「所以您是……对吗」「刚才您说的是……吗」），"
         "对方已交代清楚的事实应直接当作前提，在其上深化、延伸或关联提问；"
@@ -31,6 +34,7 @@ def chat_voice_style() -> str:
         "语气像**温暖的谈话场主持人**：口语、自然、能接住人，但心里始终为**回忆录口述**服务——"
         "不是冷冰冰盘问，也不是无底洞式的日常闲聊；更像懂行的老友在帮你把故事讲清楚。"
         "接话允许带一点画面感或感官细节（一两句即可，不要堆砌）；对方情绪重时别让整段只剩一个字。"
+        "起句尽量从对方**原词或具体画面**带入；**不要**用「嗯。」开场（**含**「嗯。」后立刻接正文），也不要「好。」「对。」单独一顿再接长句当习惯起手。"
         "用对方刚说的**那个具体细节**回应，不要写成泛泛的总结。"
         "不要用总结腔（'听起来你的童年很快乐'），要用对话腔（'那种……的感觉，现在想起来都觉得……'）。"
         "追问优先顺着对方刚说的具体细节往里走一层，不要跳到泛泛的新问题。"
diff --git a/api/app/agents/chat/prompts_conversation.py b/api/app/agents/chat/prompts_conversation.py
index a2c358d..7f7e9db 100644
--- a/api/app/agents/chat/prompts_conversation.py
+++ b/api/app/agents/chat/prompts_conversation.py
@@ -438,7 +438,7 @@ def get_guided_conversation_prompt(
 
 ### 第一步：先接住——让对方觉得你真的听进了情绪与细节
 - 用对方刚说的**那个具体细节**回应，不要写成泛泛的"听起来很好"。
-- **节奏**：用户刚说完**一大段**或字里行间**情绪很重**（委屈、哽咽、后怕、赌气、突然抒情）时，**禁止**整条回复就只有孤立的一个「嗯」「好」「明白」——会显得敷衍、支持感不够；承接仍要**短**，但至少要**半句并肩**或**钉住对方原词的一点点展开**（例如「是哦，讲到那儿换谁都会堵得慌」「那种……光听你描述就挺沉的」），不必写成工整小结再接问。若用 `[SPLIT]`：**前一泡也须有质感**，不能单字凑数，后一泡再追问或继续陪聊。
+- **节奏**：用户刚说完**一大段**或字里行间**情绪很重**（委屈、哽咽、后怕、赌气、突然抒情）时，**禁止**整条回复就只有孤立的一个「嗯。」「嗯」「好」「明白」——会显得敷衍、支持感不够；承接仍要**短**，但至少要**半句并肩**或**钉住对方原词的一点点展开**（例如「是哦，讲到那儿换谁都会堵得慌」「那种……光听你描述就挺沉的」），不必写成工整小结再接问。若用 `[SPLIT]`：**前一泡也须有质感**，不能单字凑数，后一泡再追问或继续陪聊。
 - **跟随—沉浸**：长段叙述后，可插入**极短**一两句**并肩式画面或体感**（像朋友旁听时轻轻嘬一口气），打破纯「一问一答」节拍；须**贴着对方刚讲的物象/动作**，可用「那种…」「换我可能也会…」等**泛指**，**禁止**宣称自己身上发生的具体人名地名事件，**禁止**用「这让我想起…」硬接无关轶事抢戏。
 - **留白**：用户抛出**强烈情绪**或**金句式人生总结**（如「爱是流动的」一类）时，允许**本轮不接任何问题**，只作更深一层的、贴着原句的共情，让情绪**淌一小会儿**；若用 `[SPLIT]`，第一泡可以**只有共情不讲题**，但仍须是**有内容的短句**（贴原词或并肩），第二泡仍可不问，或只在末尾留极轻的一句勾子，勿赶着交卷。
 - 好的接法：借用对方话里的意象往下走一步，例如对方说"烤红薯"，你可以说"那种外面焦焦的、掰开冒热气的感觉"。
@@ -467,11 +467,13 @@ def get_guided_conversation_prompt(
 - 不要每轮都像第一次见面。
 
 ## 语言与文笔（隐性执行，勿念给用户听）
+- **句首习惯**：**禁止**「嗯。」起头（**含**「嗯。」后立刻接正文，一律不要）；**禁止**单独成泡只有「嗯。」。「好。」「对。」也少当每轮固定发语词；更像真人时**直接**咬对方原词往下长——短停顿用省略号或半句并肩即可。
 - 长短句掺着来；能少说一个字就不堆「很、特别、真的」。
 - 同一个意思别用排比或同义词连打三遍；留一点空白，像聊天不像文章。
 - 共情与小总结像朋友捎一句，不要像晚会主持人收口或卷首语（但仍要在恰当的轮次把话头**勾回人生故事**，见「主持人职责」）。
 
 ## 绝对不要做的
+- **禁止**以「嗯。」起头，**即使**后面还有长正文也不行（不要用「嗯。」当停顿再接句）；禁止单独成泡只有「嗯。」。
 - 不要为了赶大纲无视用户刚露出来的情绪。
 - 不要在用户**长段倾诉或情绪很满**时，用**整条消息只有单个语气词**打发；要短可以，但须有贴肉的半句承接。
 - 不要用主持人口吻、晚会串联语、课文式硬切话题，或在已答信息上做复述式确认。
diff --git a/api/app/agents/chat/reply_limits.py b/api/app/agents/chat/reply_limits.py
index 89db164..5d8761a 100644
--- a/api/app/agents/chat/reply_limits.py
+++ b/api/app/agents/chat/reply_limits.py
@@ -62,6 +62,19 @@ def strip_parenthetical_asides_for_chat(text: str) -> str:
     return s.strip()
 
 
+def strip_leading_en_period_ack_for_chat(text: str) -> str:
+    """
+    去掉段首生硬的「嗯。」（可重复），即使后面还有正文；只剥字符串开头，不误伤句中「嗯。」。
+    支持全角/半角句号。
+    """
+    s = (text or "").strip()
+    if not s:
+        return s
+    # 允许多次「嗯。」/「嗯嗯。」叠在段首；句号仅匹配全角 。、． 与 ASCII `.`
+    s2 = re.sub(r"^(?:嗯+(?:。|．|\.)+\s*)+", "", s)
+    return s2.strip()
+
+
 def segments_from_llm_response(
     response_text: str,
     *,
@@ -76,13 +89,22 @@ def segments_from_llm_response(
     text = strip_parenthetical_asides_for_chat(text)
     if not text:
         return []
-    primary = [p.strip() for p in text.split("[SPLIT]") if p.strip()]
+    primary = [
+        strip_leading_en_period_ack_for_chat(p)
+        for p in text.split("[SPLIT]")
+        if strip_leading_en_period_ack_for_chat(p).strip()
+    ]
     if len(primary) > 1:
         return primary[:max_segments]
-    blob = primary[0] if primary else text
+    blob = primary[0] if primary else strip_leading_en_period_ack_for_chat(text)
+    blob = strip_leading_en_period_ack_for_chat(blob)
     if "\n" not in blob:
         return [blob]
-    paras = [p.strip() for p in re.split(r"\n\s*\n+", blob) if p.strip()]
+    paras = [
+        strip_leading_en_period_ack_for_chat(p)
+        for p in re.split(r"\n\s*\n+", blob)
+        if strip_leading_en_period_ack_for_chat(p).strip()
+    ]
     if len(paras) < 2:
         return [blob]
     paras = [p for p in paras if len(p) >= min_paragraph_chars]
diff --git a/api/app/core/config.py b/api/app/core/config.py
index c868f02..f0595b5 100644
--- a/api/app/core/config.py
+++ b/api/app/core/config.py
@@ -264,6 +264,8 @@ class Settings(BaseSettings):
     memoir_image_style_default: str = "watercolor"
     memoir_image_size_default: str = "1280x720"
     memoir_image_download_hosts: str = ""
+    # 章节 canonical_markdown 中至少含多少张 asset:// 正文插图才生成/展示章节封面（≥ 该值即满足；0 表示不以此条件拦截）
+    memoir_min_inline_images_for_chapter_cover: int = Field(default=1, ge=0, le=100)
     # Story 正文至少多少字才创建主图 intent / 调图（0 表示不限制）
     story_image_min_body_chars: int = 400
     # generate_story_image 入队去重（Redis SET NX，秒）
diff --git a/api/app/features/memoir/cover_eligibility.py b/api/app/features/memoir/cover_eligibility.py
index fffd6f8..16a09db 100644
--- a/api/app/features/memoir/cover_eligibility.py
+++ b/api/app/features/memoir/cover_eligibility.py
@@ -4,15 +4,16 @@ from __future__ import annotations
 
 from typing import Any
 
-from app.features.memoir.asset_resolver import parse_asset_refs
+from app.core.config import settings
+from app.features.memoir.asset_resolver import (
+    parse_asset_refs,
+    strip_image_placeholders,
+)
 from app.features.memoir.memoir_images.schema import (
     IMAGE_STATUS_FAILED,
     IMAGE_STATUS_PENDING,
 )
 
-# 正文内 ![...](asset://...) 数量需 **大于** 此值才生成/展示章节封面（与故事头图、正文配图任务独立）
-MIN_INLINE_BODY_IMAGES_FOR_CHAPTER_COVER = 3
-
 
 def chapter_has_story_links(chapter: Any) -> bool:
     return any(
@@ -21,18 +22,47 @@ def chapter_has_story_links(chapter: Any) -> bool:
     )
 
 
-def count_chapter_inline_body_images(chapter: Any) -> int:
-    """统计章节 canonical_markdown 中正文插图（asset:// 图片引用）次数。"""
-    md = getattr(chapter, "canonical_markdown", None) or ""
-    return len(parse_asset_refs(md))
+def effective_chapter_markdown_for_cover_gates(chapter: Any) -> str:
+    """
+     用于封面闸门计数：优先 DB canonical；若为空且已挂 stories，则用内存物化串
+    （与列表/详情在 compose_dirty 时的临时正文对齐，避免「有图但 canonical 未落库」导致永不出封面）。
+    """
+    md = (getattr(chapter, "canonical_markdown", None) or "").strip()
+    if md:
+        return md
+    if chapter_has_story_links(chapter):
+        from app.features.memoir.chapter_markdown_compose import (
+            materialize_chapter_markdown_from_loaded_chapter,
+        )
+
+        try:
+            alt = (
+                materialize_chapter_markdown_from_loaded_chapter(chapter) or ""
+            ).strip()
+        except Exception:
+            return ""
+        return alt
+    return ""
 
 
-def chapter_eligible_for_cover_by_inline_body_image_count(chapter: Any) -> bool:
-    """仅当正文内插图数量 > MIN_INLINE_BODY_IMAGES_FOR_CHAPTER_COVER 时才生成/展示章节封面。"""
-    return (
-        count_chapter_inline_body_images(chapter)
-        > MIN_INLINE_BODY_IMAGES_FOR_CHAPTER_COVER
+def count_chapter_inline_body_images(
+    chapter: Any, *, markdown: str | None = None
+) -> int:
+    """统计 asset:// 插图次数；未传 markdown 时用 effective_chapter_markdown_for_cover_gates。"""
+    source = (
+        markdown
+        if markdown is not None
+        else effective_chapter_markdown_for_cover_gates(chapter)
     )
+    return len(parse_asset_refs(source))
+
+
+def chapter_eligible_for_cover_by_inline_body_image_count(
+    chapter: Any, *, markdown: str | None = None
+) -> bool:
+    """正文内 asset:// 数量 ≥ 配置阈值时允许封面；markdown 非 None 时仅用该串计数。"""
+    min_required = int(settings.memoir_min_inline_images_for_chapter_cover)
+    return count_chapter_inline_body_images(chapter, markdown=markdown) >= min_required
 
 
 def primary_chapter_memoir_image(chapter: Any) -> Any | None:
@@ -45,17 +75,18 @@ def primary_chapter_memoir_image(chapter: Any) -> Any | None:
 
 
 def chapter_needs_cover_enqueue(chapter) -> bool:
-    """尚无 cover_asset、有正文、且正文内 asset 插图多于阈值时，可派发 generate_chapter_cover。"""
+    """尚无 cover_asset、有正文、且正文内 asset 插图达到 env 阈值时，可派发 generate_chapter_cover。"""
     if not chapter:
         return False
     if not chapter_has_story_links(chapter):
         return False
     if getattr(chapter, "cover_asset_id", None):
         return False
-    md = (getattr(chapter, "canonical_markdown", None) or "").strip()
-    if not md:
+    view = effective_chapter_markdown_for_cover_gates(chapter)
+    body = strip_image_placeholders(view).strip()
+    if not body:
         return False
-    return chapter_eligible_for_cover_by_inline_body_image_count(chapter)
+    return chapter_eligible_for_cover_by_inline_body_image_count(chapter, markdown=view)
 
 
 def chapter_has_cover_to_generate(chapter) -> bool:
diff --git a/api/app/features/memoir/helpers.py b/api/app/features/memoir/helpers.py
index f048d30..1dbb089 100644
--- a/api/app/features/memoir/helpers.py
+++ b/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)
diff --git a/api/app/tasks/chapter_cover_enqueue.py b/api/app/tasks/chapter_cover_enqueue.py
index ce1173a..4fec856 100644
--- a/api/app/tasks/chapter_cover_enqueue.py
+++ b/api/app/tasks/chapter_cover_enqueue.py
@@ -18,6 +18,7 @@ from app.features.memoir.cover_eligibility import (
     chapter_eligible_for_cover_by_inline_body_image_count,
     chapter_has_story_links,
     chapter_needs_cover_enqueue,
+    effective_chapter_markdown_for_cover_gates,
     primary_chapter_memoir_image,
 )
 from app.features.memoir.models import Chapter, ChapterStoryLink
@@ -42,14 +43,13 @@ def _chapter_eligible_for_http_enqueue(chapter: Chapter | None) -> bool:
         return False
     if getattr(chapter, "cover_asset_id", None):
         return False
-    md = (chapter.canonical_markdown or "").strip()
-    body = md or ""
-    if not body.strip():
-        return False
-    body = strip_image_placeholders(body).strip()
+    view = effective_chapter_markdown_for_cover_gates(chapter)
+    body = strip_image_placeholders(view).strip()
     if not body:
         return False
-    if not chapter_eligible_for_cover_by_inline_body_image_count(chapter):
+    if not chapter_eligible_for_cover_by_inline_body_image_count(
+        chapter, markdown=view
+    ):
         return False
     cover_rec = primary_chapter_memoir_image(chapter)
     if cover_rec and (cover_rec.status or "").strip() == "completed":
@@ -58,7 +58,7 @@ def _chapter_eligible_for_http_enqueue(chapter: Chapter | None) -> bool:
 
 
 def _chapter_eligible_for_pipeline_enqueue(chapter: Chapter | None) -> bool:
-    """尚无 cover_asset、正文插图数 > 3（与 HTTP 闸门共用 chapter_needs_cover_enqueue 核心）。"""
+    """尚无 cover_asset、正文插图数达 MEMOIR_MIN_INLINE_IMAGES_FOR_CHAPTER_COVER（与 HTTP 闸门共用）。"""
     return bool(chapter_needs_cover_enqueue(chapter))
 
 
diff --git a/api/tests/test_chapter_cover_fallback_inline.py b/api/tests/test_chapter_cover_fallback_inline.py
new file mode 100644
index 0000000..c4dbecc
--- /dev/null
+++ b/api/tests/test_chapter_cover_fallback_inline.py
@@ -0,0 +1,57 @@
+"""章节封面：无 cover_asset_id 时用正文首张 asset:// 作列表封面。"""
+
+from unittest.mock import MagicMock
+
+from app.features.memoir.helpers import chapter_cover_to_dict
+
+
+def test_cover_falls_back_to_first_inline_asset_url() -> None:
+    ch = MagicMock(spec=[])
+    ch.canonical_markdown = "正文\n\n![场景](asset://img-1)"
+    ch.cover_asset_id = None
+    ch.story_links = []
+    ch.images = []
+    m = chapter_cover_to_dict(
+        ch,
+        asset_url_map={"img-1": "https://cos.example.com/signed-1"},
+        markdown_for_response=None,
+    )
+    assert m is not None
+    assert m["url"] == "https://cos.example.com/signed-1"
+    assert m["status"] == "completed"
+
+
+def test_cover_from_reading_segments_when_canonical_has_no_asset() -> None:
+    """分段快照里有 asset://，章节 canonical 未带图时仍能出封面 URL。"""
+    ch = MagicMock(spec=[])
+    ch.canonical_markdown = "只有文字没有图" * 20
+    ch.cover_asset_id = None
+    ch.story_links = []
+    ch.images = []
+    ch.reading_segments_json = [
+        {"story_id": "s1", "body_markdown": "![景](asset://seg-1)"}
+    ]
+    m = chapter_cover_to_dict(
+        ch,
+        asset_url_map={"seg-1": "https://cos.example.com/seg"},
+        markdown_for_response=None,
+    )
+    assert m is not None
+    assert m["url"] == "https://cos.example.com/seg"
+
+
+def test_cover_prefers_cover_asset_id_over_inline() -> None:
+    ch = MagicMock(spec=[])
+    ch.canonical_markdown = "![a](asset://inline-1)"
+    ch.cover_asset_id = "cover-99"
+    ch.story_links = []
+    ch.images = []
+    m = chapter_cover_to_dict(
+        ch,
+        asset_url_map={
+            "inline-1": "https://cos.example.com/inline",
+            "cover-99": "https://cos.example.com/cover",
+        },
+        markdown_for_response=None,
+    )
+    assert m["url"] == "https://cos.example.com/cover"
diff --git a/api/tests/test_cover_eligibility_effective_markdown.py b/api/tests/test_cover_eligibility_effective_markdown.py
new file mode 100644
index 0000000..363d068
--- /dev/null
+++ b/api/tests/test_cover_eligibility_effective_markdown.py
@@ -0,0 +1,54 @@
+"""封面闸门：canonical 未落库时须用物化正文计数 asset://。"""
+
+from unittest.mock import MagicMock, patch
+
+from app.features.memoir.cover_eligibility import (
+    chapter_eligible_for_cover_by_inline_body_image_count,
+    chapter_needs_cover_enqueue,
+    count_chapter_inline_body_images,
+    effective_chapter_markdown_for_cover_gates,
+)
+
+
+def test_effective_markdown_falls_back_to_materialize_when_canonical_empty() -> None:
+    ch = MagicMock()
+    ch.canonical_markdown = ""
+    ch.story_links = [MagicMock()]
+    with patch(
+        "app.features.memoir.chapter_markdown_compose.materialize_chapter_markdown_from_loaded_chapter",
+        return_value="正文\n\n![x](asset://a1)",
+    ):
+        assert "asset://" in effective_chapter_markdown_for_cover_gates(ch)
+
+
+def test_count_uses_effective_when_canonical_empty() -> None:
+    ch = MagicMock()
+    ch.canonical_markdown = ""
+    ch.story_links = [MagicMock()]
+    with patch(
+        "app.features.memoir.chapter_markdown_compose.materialize_chapter_markdown_from_loaded_chapter",
+        return_value="![alt](asset://id1)",
+    ):
+        assert count_chapter_inline_body_images(ch) == 1
+
+
+def test_eligible_with_explicit_markdown_override() -> None:
+    ch = MagicMock()
+    ch.canonical_markdown = ""
+    assert chapter_eligible_for_cover_by_inline_body_image_count(
+        ch, markdown="![a](asset://x)"
+    )
+
+
+def test_needs_cover_enqueue_uses_materialized_body() -> None:
+    ch = MagicMock()
+    ch.canonical_markdown = ""
+    ch.cover_asset_id = None
+    ch.story_links = [MagicMock(story=MagicMock())]
+    link = ch.story_links[0]
+    link.story = MagicMock()
+    with patch(
+        "app.features.memoir.chapter_markdown_compose.materialize_chapter_markdown_from_loaded_chapter",
+        return_value="故事\n\n![a](asset://z)",
+    ):
+        assert chapter_needs_cover_enqueue(ch) is True
diff --git a/api/tests/test_reply_segments.py b/api/tests/test_reply_segments.py
index 4caeb7c..410a92c 100644
--- a/api/tests/test_reply_segments.py
+++ b/api/tests/test_reply_segments.py
@@ -3,6 +3,7 @@
 from app.agents.chat.reply_limits import (
     nonempty_segments_or_fallback,
     segments_from_llm_response,
+    strip_leading_en_period_ack_for_chat,
     strip_markdown_for_chat,
     strip_parenthetical_asides_for_chat,
 )
@@ -58,3 +59,15 @@ def test_segments_strip_parentheticals_before_split():
 
 def test_strip_parenthetical_multiple_passes():
     assert strip_parenthetical_asides_for_chat("a（一）b（二）c") == "abc"
+
+
+def test_strip_leading_en_period_ack():
+    assert strip_leading_en_period_ack_for_chat("嗯。后面正文") == "后面正文"
+    assert strip_leading_en_period_ack_for_chat("嗯嗯。后面") == "后面"
+    assert strip_leading_en_period_ack_for_chat("  嗯。  第二句") == "第二句"
+    assert strip_leading_en_period_ack_for_chat("句中嗯。不打头") == "句中嗯。不打头"
+
+
+def test_segments_strip_leading_en_ack():
+    assert segments_from_llm_response("嗯。只有一句", max_segments=3) == ["只有一句"]
+    assert segments_from_llm_response("嗯。A[SPLIT]嗯。B", max_segments=3) == ["A", "B"]
diff --git a/app-expo/src/app/(main)/chapter/[id].tsx b/app-expo/src/app/(main)/chapter/[id].tsx
index c6ad3a2..43608dd 100644
--- a/app-expo/src/app/(main)/chapter/[id].tsx
+++ b/app-expo/src/app/(main)/chapter/[id].tsx
@@ -18,8 +18,13 @@ import { useTranslation } from 'react-i18next';
 
 import { Icon } from '@/components/ui/icon';
 import { Text } from '@/components/ui/text';
-import { ScreenHeader } from '@/components/screen-header';
+import {
+  getScreenHeaderLayoutMetrics,
+  ScreenHeader,
+} from '@/components/screen-header';
 import { ScreenGutter } from '@/constants/layout';
+import { useTypography } from '@/core/typography-context';
+import { useAppSettings } from '@/hooks/use-app-settings';
 import {
   MarkdownRenderer,
   ReadingMarkdownHorizontalRuleInColumn,
@@ -353,6 +358,8 @@ function ReadingSettingsModal({
 export default function ChapterScreen() {
   const { id } = useLocalSearchParams<{ id: string }>();
   const insets = useSafeAreaInsets();
+  const { largeText } = useAppSettings();
+  const typography = useTypography();
   const { width } = useWindowDimensions();
   const { t } = useTranslation('memoir');
   const { data: chapter, isLoading } = useChapterDetail(id ?? '');
@@ -429,8 +436,13 @@ export default function ChapterScreen() {
   const useReadingSegments =
     Array.isArray(readingSegments) && readingSegments.length > 0;
 
-  /** 与 ScreenHeader（reading、useSafeArea）可视高度对齐，避免返回栏与首屏内容之间出现空隙 */
-  const headerOccupiedHeight = Math.max(insets.top, 12) + 56;
+  /** 与 ScreenHeader（reading、useSafeArea）实际总高度一致，避免章节标题被顶栏或安全区遮挡 */
+  const headerOccupiedHeight = getScreenHeaderLayoutMetrics(insets, {
+    useSafeArea: true,
+    variant: 'reading',
+    largeText,
+    typography,
+  }).totalHeight;
 
   const handleDeletePress = () => {
     Alert.alert(
diff --git a/app-expo/src/components/screen-header.tsx b/app-expo/src/components/screen-header.tsx
index 1f5b376..0d7c69a 100644
--- a/app-expo/src/components/screen-header.tsx
+++ b/app-expo/src/components/screen-header.tsx
@@ -9,6 +9,8 @@ import { useTypography } from '@/core/typography-context';
 import { ScreenGutter } from '@/constants/layout';
 import { useAppSettings } from '@/hooks/use-app-settings';
 
+import type { TypographyTokens } from '@/core/typography-context';
+
 /** 默认最小触控目标 48dp；大字模式下与标题字号匹配，略向左扩展便于够到边缘 */
 const BACK_HIT_MIN = 48;
 const BACK_HIT_MIN_LARGE = 56;
@@ -18,6 +20,52 @@ const BACK_EXTRA_HIT_LEFT = 4;
 
 export type ScreenHeaderVariant = 'default' | 'chat' | 'reading';
 
+export type ScreenHeaderLayoutOpts = {
+  useSafeArea: boolean;
+  variant: ScreenHeaderVariant;
+  largeText: boolean;
+  typography: TypographyTokens;
+};
+
+/**
+ * 与组件内布局一致的总高度（含顶部安全区内边距），供绝对定位顶栏下的 ScrollView paddingTop 等使用。
+ */
+export function getScreenHeaderLayoutMetrics(
+  insets: { top: number },
+  opts: ScreenHeaderLayoutOpts,
+) {
+  const barPaddingBottom = opts.largeText ? 18 : 16;
+  const titleRowPaddingV = opts.largeText ? 8 : 4;
+  const backTouchMin = opts.largeText ? BACK_HIT_MIN_LARGE : BACK_HIT_MIN;
+  const { variant, typography, largeText } = opts;
+  const titleFontSize =
+    variant === 'chat'
+      ? largeText
+        ? typography.headingMedium
+        : typography.headingSmall
+      : Math.max(typography.titleLarge, typography.headingSmall);
+  const titleLineMin =
+    variant === 'chat'
+      ? Math.ceil(titleFontSize * (largeText ? 1.45 : 1.38)) +
+        (largeText ? 14 : 10)
+      : Math.ceil(titleFontSize * 1.32) + (largeText ? 10 : 8);
+  const titleRowMinHeight = Math.max(backTouchMin, titleLineMin);
+  const titleRowOuterHeight = titleRowMinHeight + 2 * titleRowPaddingV;
+  const paddingTop = opts.useSafeArea ? Math.max(insets.top, 12) : 12;
+  const totalHeight = paddingTop + titleRowOuterHeight + barPaddingBottom;
+  return {
+    barPaddingBottom,
+    titleRowPaddingV,
+    titleFontSize,
+    titleLineMin,
+    titleRowMinHeight,
+    titleRowOuterHeight,
+    paddingTop,
+    totalHeight,
+    backTouchMin,
+  };
+}
+
 const VARIANT_COLORS = {
   default: {
     title: undefined, // use theme foreground
@@ -76,7 +124,6 @@ export function ScreenHeader({
   const typography = useTypography();
   const colors = VARIANT_COLORS[variant];
 
-  const backTouchMin = largeText ? BACK_HIT_MIN_LARGE : BACK_HIT_MIN;
   const backIconSize = largeText ? BACK_ICON_SIZE_LARGE : BACK_ICON_SIZE;
 
   const handleBack = onBack ?? (() => router.back());
@@ -88,27 +135,26 @@ export function ScreenHeader({
    * 不要用外层 minHeight「框死」整块栏：paddingTop 含安全区时会把内容区压扁。
    * 标题行最小高度随当前排版 token 计算（关大字也要留够，避免裁字 / Android font padding）。
    */
-  const barPaddingBottom = largeText ? 18 : 16;
-  const titleRowPaddingV = largeText ? 8 : 4;
-  const titleFontSize =
-    variant === 'chat'
-      ? largeText
-        ? typography.headingMedium
-        : typography.headingSmall
-      : Math.max(typography.titleLarge, typography.headingSmall);
-  const titleLineMin =
-    variant === 'chat'
-      ? Math.ceil(titleFontSize * (largeText ? 1.45 : 1.38)) +
-        (largeText ? 14 : 10)
-      : Math.ceil(titleFontSize * 1.32) + (largeText ? 10 : 8);
-  const titleRowMinHeight = Math.max(backTouchMin, titleLineMin);
+  const {
+    barPaddingBottom,
+    titleRowPaddingV,
+    titleFontSize,
+    titleRowMinHeight,
+    paddingTop: headerPaddingTop,
+    backTouchMin,
+  } = getScreenHeaderLayoutMetrics(insets, {
+    useSafeArea,
+    variant,
+    largeText,
+    typography,
+  });
 
   const containerStyle = {
     flexDirection: 'row' as const,
     alignItems: 'center' as const,
     justifyContent: 'space-between' as const,
     paddingHorizontal: Math.max(ScreenGutter, 16),
-    paddingTop: useSafeArea ? Math.max(insets.top, 12) : 12,
+    paddingTop: headerPaddingTop,
     paddingBottom: barPaddingBottom,
     ...(bgColor && { backgroundColor: bgColor }),
     ...(absolute && {