feat(editor): refine toc hierarchy and search scope

tools/epub_review_editor/MAINTENANCE.md

@@ -40,8 +40,8 @@ epub_review_sessions/
 结构接口 `/api/structure` 应保持向后兼容：
 
 - text chapter: `kind: "text"`、`parts`、`row_count`、`touched_count`、`marked_count`
-- image chapter: `kind: "image"`、`images`、`parts: []`、`row_count: 0`
-- 统计字段同时保留总数和细分：`chapter_count`、`text_chapter_count`、`image_chapter_count`、`image_count`、`asset_count`
+- image item: 优先作为 text chapter 的 `items` 子项返回，字段为 `kind: "image"`、`images`、`parts: []`、`row_count: 0`；没有所属文字章时才保留顶层 image chapter
+- 统计字段同时保留总数和细分：`chapter_count`、`text_chapter_count`、`image_chapter_count`、`top_level_image_chapter_count`、`image_count`、`asset_count`。其中 `image_chapter_count` 表示目录里的图片条目数，不一定是顶层章节数
 - part 标题默认使用 `part0000` 形式，真实标题可保留在 `source_title`
 - 插图通过 `/api/asset/<entry>` 读取，必须继续使用 `extracted_path()` 防止路径穿越；只把 raster 图片纳入 `/api/structure` 与 `/api/asset` 白名单，不能任意读取解包目录文件
 
@@ -49,16 +49,20 @@ epub_review_sessions/
 
 1. EPUB 内部目录页的中文条目
 2. `nav.xhtml` / `toc.xhtml`
-3. 正文标题
+3. 正文 `<h1-6>` 标题
 4. 文件名 fallback
 
+不要把正文首段或第一句对白作为章节标题 fallback。宁可显示 `part0000`，也不要让目录被正文内容污染。
+
 ## 阅读与侧栏交互
 
 验收口径：
 
 - 初始进入审校时，左侧侧栏隐藏，正文占主要空间。
 - 顶部“目录”只打开目录，“检索”只打开搜索/段落列表，不合并成一个常驻面板。
-- 目录可展开/收起章节，插图章节可点击阅读，文字 part 可点击定位。
+- 目录可展开/收起章节，插图作为所属章节的次级目录条目可点击阅读，文字 part 可点击定位。
+- 点击目录章节、part 或插图条目后，侧栏不得自动收起；用户仍可点“隐藏”或按 `Escape` 关闭。
+- 检索必须保留“当前范围 / 全书全局”切换；当前范围随当前章节或 part 更新，全局检索可跨章节返回并跳转。
 - 阅读模式不显示每段编号、快速编辑、快速标记按钮。
 - 点击正文段落打开右侧精修与重翻面板。
 - 右侧面板打开或关闭导致正文宽度变化时，应恢复用户原来正在看的段落位置。

tools/epub_review_editor/README.md

@@ -1,6 +1,6 @@
 # 双语 EPUB 审校编辑器
 
-当前版本：`0.4.1`
+当前版本：`0.5.0`
 
 这个工具用于把生成后的中日双语 EPUB 打开成浏览器审校界面。用户可以直接修改中文译文，也可以标记“哪里翻得不好”，并把所有记录沉淀为可交给 Codex 的反馈文件。
 
@@ -24,6 +24,8 @@
 
 `0.4.1` 修复后端/API 安全与结构细节：章节内插图与文字按 spine 顺序 flush，结构统计补充 text/image 细分字段；内部目录识别增加目录页/链接密度判定；结构和 `/api/asset` 仅允许已识别的 raster 图片并加 `nosniff`；重翻接口不再允许单次请求临时覆盖 Base URL 或模型后复用已保存密钥。
 
+`0.5.0` 改进长篇目录与检索：目录章节标题不再用正文首段猜测，章节内插图作为次级条目挂在所属章节下，阅读时可按原 spine 顺序穿插显示；点击目录不会自动收起侧栏，桌面端正文会为常驻侧栏让位；检索新增“当前范围 / 全书全局”切换。
+
 ## 独立安装
 
 这个审校器现在可以脱离 Codex 使用，只需要 Python 和浏览器。把 `tools/epub_review_editor` 目录复制到其他电脑或服务器后，在该目录里安装依赖：
@@ -81,8 +83,9 @@ python .\server.py --host 0.0.0.0 --port 5177 --review-root .\epub_review_sessio
 
 - 阅读模式下按章节连续阅读日文原文与中文译文，不需要逐句点击。
 - 左侧目录和检索默认隐藏，通过顶部“目录”“检索”两个按钮分别打开，不会把两种入口混在同一个常驻侧栏里。
-- 目录优先按 EPUB 内部目录页或 nav/toc 组织章节；章节下显示内部 `part0000` 形式的 part，可展开/收起，点击 part 后只看该 part，点击章节则看整章连续正文。
-- 插图页会像文字章节一样出现在目录里，点击后可在阅读模式中直接查看原 EPUB 图片。
+- 目录优先按 EPUB 内部目录页或 nav/toc 组织章节；章节下显示内部 `part0000` 形式的 part 与次级插图条目，可展开/收起，点击 part 后只看该 part，点击章节则按原 spine 顺序连续阅读正文和插图。
+- 插图页会作为所属章节的次级条目出现在目录里，点击后可在阅读模式中直接查看原 EPUB 图片；点击目录项不会自动隐藏左侧侧栏。
+- 检索支持“当前范围”和“全书全局”两种范围，适合在单章精查和整本书排查术语时切换。
 - 在阅读模式中点击段落即可打开右侧精修与重翻面板，直接编辑中文译文、标记问题或调用 GPT 重翻；打开/关闭面板时会尽量保持原来的阅读位置。
 - 精修模式保留原有逐段编辑界面，用于较细的逐段复审。
 - 可在网页中配置 OpenAI 兼容的 `base_url`、`model` 与 API Key，对当前段落生成重翻候选；候选不会自动覆盖正文，需要点击“应用重翻”并保存。

tools/epub_review_editor/server.py

@@ -799,9 +799,18 @@ def collect_asset_paths_from_structure(structure: dict[str, Any]) -> set[str]:
             asset_path = image.get("asset_path", "")
             if asset_path:
                 paths.add(asset_path)
+        for item in chapter.get("items", []):
+            for image in item.get("images", []):
+                asset_path = image.get("asset_path", "")
+                if asset_path:
+                    paths.add(asset_path)
     return paths
 
 
+def image_item_count(chapters: list[dict[str, Any]]) -> int:
+    return sum(1 for chapter in chapters for item in chapter.get("items", []) if item.get("kind") == "image")
+
+
 def nav_target_key(base_dir: str, href: str) -> tuple[str, str]:
     path, fragment = normalize_href_with_fragment(base_dir, href)
     return path, fragment
@@ -826,6 +835,7 @@ def parse_internal_toc(session_root: Path) -> list[dict[str, Any]]:
         candidate_items: list[dict[str, Any]] = []
         i = 0
         while i < len(matches):
+            source_attrs = matches[i].group(1)[2:-1]
             source_inner = matches[i].group(2)
             link_match = re.search(r"<a\b[^>]*href=['\"]([^'\"]+)['\"][^>]*>(.*?)</a>", source_inner, flags=re.S | re.I)
             if not link_match:
@@ -840,13 +850,20 @@ def parse_internal_toc(session_root: Path) -> list[dict[str, Any]]:
                 continue
             source_title = compact_text(strip_tags(link_match.group(2))) or compact_text(strip_tags(source_inner))
             title = source_title
-            if i + 1 < len(matches):
+            if is_gray_source(source_attrs, source_inner) and i + 1 < len(matches):
+                next_attrs = matches[i + 1].group(1)[2:-1]
                 next_inner = matches[i + 1].group(2)
-                if not is_gray_source(matches[i + 1].group(1)[2:-1], next_inner):
-                    next_title = compact_text(strip_tags(next_inner))
+                if not is_gray_source(next_attrs, next_inner):
+                    next_link = re.search(r"<a\b[^>]*href=['\"]([^'\"]+)['\"][^>]*>(.*?)</a>", next_inner, flags=re.S | re.I)
+                    same_target = True
+                    if next_link:
+                        next_target_file, next_target_fragment = nav_target_key(base_dir, next_link.group(1))
+                        same_target = next_target_file == target_file and next_target_fragment == target_fragment
+                    next_title = compact_text(strip_tags(next_link.group(2) if next_link else next_inner))
                     if next_title:
-                        title = next_title
-                        i += 1
+                        if same_target:
+                            title = next_title
+                            i += 1
             if title:
                 candidate_items.append(
                     {
@@ -1000,13 +1017,31 @@ def build_structure(session_root: Path) -> dict[str, Any]:
         if start and item.get("title") and start not in start_titles:
             start_titles[start] = item["title"]
 
-    for entry_name in ordered_files:
-        heading = read_body_heading(session_root, entry_name)
-        if heading and entry_name not in start_titles and len(strip_tags(heading)) <= 80:
-            start_titles[entry_name] = heading
+    if not start_titles:
+        for item in nav_flat:
+            href = item.get("href", "")
+            if href not in ordered_index:
+                continue
+            start = first_content_file_at_or_after(href)
+            if start and item.get("title") and start not in start_titles:
+                start_titles[start] = item["title"]
+
+    if not start_titles:
+        for entry_name in ordered_files:
+            heading = read_body_heading(session_root, entry_name)
+            if heading and entry_name not in start_titles and len(strip_tags(heading)) <= 80:
+                start_titles[entry_name] = heading
 
     if start_titles:
         starts = sorted(start_titles, key=lambda name: ordered_index.get(name, 10**9))
+        first_start_pos = ordered_index.get(starts[0], len(all_ordered_files)) if starts else len(all_ordered_files)
+        preface_files = [
+            entry_name
+            for entry_name in all_ordered_files[:first_start_pos]
+            if entry_name in rows_by_file or entry_name in image_by_file
+        ]
+        if preface_files:
+            chapter_specs.append({"title": "卷首", "files": preface_files, "start": preface_files[0]})
         for index, start in enumerate(starts):
             start_pos = ordered_index.get(start, -1)
             next_pos = ordered_index.get(starts[index + 1], len(all_ordered_files)) if index + 1 < len(starts) else len(all_ordered_files)
@@ -1057,8 +1092,6 @@ def build_structure(session_root: Path) -> dict[str, Any]:
             "source_title": title_override
             or nav_titles.get(entry_name)
             or read_body_heading(session_root, entry_name)
-            or first_content_text(session_root, entry_name)
-            or read_document_title(session_root, entry_name)
             or Path(entry_name).stem,
             "file": entry_name,
             "file_label": Path(entry_name).name,
@@ -1070,21 +1103,23 @@ def build_structure(session_root: Path) -> dict[str, Any]:
             "last_row_id": file_rows[-1]["id"] if file_rows else "",
         }
 
-    def make_image_chapter(
+    def make_image_item(
         entry_name: str,
         image: dict[str, Any],
         title_override: str = "",
         image_index: int = 1,
         image_total: int = 1,
     ) -> dict[str, Any]:
-        nonlocal chapter_serial, image_serial
-        chapter_serial += 1
+        nonlocal image_serial
         image_serial += 1
         suffix = f"-{image_index:02d}" if image_total > 1 else ""
         label = part_display_label(entry_name, image_serial - 1)
-        default_title = "封面" if Path(entry_name).stem.lower() in {"titlepage", "cover"} else f"插图 {label}{suffix}"
+        is_cover = Path(entry_name).stem.lower() in {"titlepage", "cover"}
+        default_title = "封面" if is_cover else f"插图 {label}{suffix}"
         base_title = title_override or nav_titles.get(entry_name) or read_body_heading(session_root, entry_name)
-        if title_override and base_title:
+        if is_cover:
+            title = default_title
+        elif title_override and base_title:
             title = f"{base_title} · 插图{suffix}"
         else:
             title = base_title or default_title
@@ -1093,7 +1128,7 @@ def build_structure(session_root: Path) -> dict[str, Any]:
             "id": f"IMG{image_serial:04d}-{image['index']:02d}",
         }
         return {
-            "id": f"C{chapter_serial:04d}",
+            "id": image_item["id"],
             "kind": "image",
             "title": title,
             "file": entry_name,
@@ -1103,10 +1138,21 @@ def build_structure(session_root: Path) -> dict[str, Any]:
             "touched_count": 0,
             "marked_count": 0,
             "parts": [],
-            "_sort": ordered_index.get(entry_name, 10**9) + image_index / 100,
+            "first_image_id": image_item["id"],
+            "last_image_id": image_item["id"],
         }
 
-    def chapter_from_parts(chapter_id: str, title: str, parts: list[dict[str, Any]]) -> dict[str, Any]:
+    def image_chapter_from_item(chapter_id: str, item: dict[str, Any]) -> dict[str, Any]:
+        return {
+            **item,
+            "id": chapter_id,
+            "items": [item],
+            "_sort": ordered_index.get(item.get("file", ""), 10**9),
+        }
+
+    def chapter_from_items(chapter_id: str, title: str, items: list[dict[str, Any]]) -> dict[str, Any]:
+        parts = [item for item in items if item.get("kind") == "text"]
+        images = [image for item in items if item.get("kind") == "image" for image in item.get("images", [])]
         return {
             "id": chapter_id,
             "kind": "text",
@@ -1115,9 +1161,34 @@ def build_structure(session_root: Path) -> dict[str, Any]:
             "touched_count": sum(part["touched_count"] for part in parts),
             "marked_count": sum(part["marked_count"] for part in parts),
             "parts": parts,
-            "_sort": min((ordered_index.get(part["file"], 10**9) for part in parts), default=10**9),
+            "images": images,
+            "items": items,
+            "_sort": min((ordered_index.get(item.get("file", ""), 10**9) for item in items), default=10**9),
         }
 
+    def chapter_from_parts(chapter_id: str, title: str, parts: list[dict[str, Any]]) -> dict[str, Any]:
+        return chapter_from_items(chapter_id, title, parts)
+
+    def fold_image_chapters_into_text(input_chapters: list[dict[str, Any]]) -> list[dict[str, Any]]:
+        folded: list[dict[str, Any]] = []
+        last_text_chapter: dict[str, Any] | None = None
+        for chapter in input_chapters:
+            if chapter.get("kind") != "image":
+                folded.append(chapter)
+                last_text_chapter = chapter if chapter.get("kind") == "text" else last_text_chapter
+                continue
+            if last_text_chapter is None:
+                folded.append(chapter)
+                continue
+            image_items = chapter.get("items") or [chapter]
+            last_text_chapter.setdefault("items", []).extend(image_items)
+            last_text_chapter.setdefault("images", []).extend(
+                image
+                for item in image_items
+                for image in item.get("images", [])
+            )
+        return folded
+
     def flatten_node_files(node: dict[str, Any]) -> list[dict[str, str]]:
         items: list[dict[str, str]] = []
         href = node.get("href", "")
@@ -1136,7 +1207,9 @@ def build_structure(session_root: Path) -> dict[str, Any]:
         if entry_name in image_by_file:
             images = image_by_file.get(entry_name, [])
             for image_index, image in enumerate(images, start=1):
-                chapters.append(make_image_chapter(entry_name, image, title_override, image_index, len(images)))
+                chapter_serial += 1
+                image_item = make_image_item(entry_name, image, title_override, image_index, len(images))
+                chapters.append(image_chapter_from_item(f"C{chapter_serial:04d}", image_item))
             return
         if entry_name in rows_by_file:
             chapter_serial += 1
@@ -1145,28 +1218,36 @@ def build_structure(session_root: Path) -> dict[str, Any]:
 
     def append_grouped_files(title: str, files: list[str]) -> None:
         nonlocal chapter_serial
-        text_parts: list[dict[str, Any]] = []
+        chapter_items: list[dict[str, Any]] = []
         text_title = title
 
-        def flush_text_parts() -> None:
-            nonlocal chapter_serial, text_parts, text_title
+        def flush_chapter_items() -> None:
+            nonlocal chapter_serial, chapter_items, text_title
+            if not chapter_items:
+                return
+            text_parts = [item for item in chapter_items if item.get("kind") == "text"]
             if not text_parts:
+                for item in chapter_items:
+                    if item.get("kind") == "image":
+                        chapter_serial += 1
+                        chapters.append(image_chapter_from_item(f"C{chapter_serial:04d}", item))
+                chapter_items = []
+                text_title = ""
                 return
             chapter_serial += 1
-            chapters.append(chapter_from_parts(f"C{chapter_serial:04d}", text_title or text_parts[0]["source_title"] or text_parts[0]["title"], text_parts))
-            text_parts = []
+            chapters.append(chapter_from_items(f"C{chapter_serial:04d}", text_title or text_parts[0]["source_title"] or text_parts[0]["title"], chapter_items))
+            chapter_items = []
             text_title = ""
 
         for file_index, entry_name in enumerate(files):
             if entry_name in image_by_file:
-                flush_text_parts()
                 images = image_by_file.get(entry_name, [])
                 for image_index, image in enumerate(images, start=1):
-                    image_title = title if file_index == 0 and not text_parts else ""
-                    chapters.append(make_image_chapter(entry_name, image, image_title, image_index, len(images)))
+                    image_title = title if file_index == 0 and not chapter_items else ""
+                    chapter_items.append(make_image_item(entry_name, image, image_title, image_index, len(images)))
             elif entry_name in rows_by_file:
-                text_parts.append(make_part(entry_name))
-        flush_text_parts()
+                chapter_items.append(make_part(entry_name))
+        flush_chapter_items()
 
     for spec in chapter_specs:
         node_files = [entry_name for entry_name in spec.get("files", []) if entry_name not in assigned_files]
@@ -1207,17 +1288,33 @@ def build_structure(session_root: Path) -> dict[str, Any]:
         append_file_as_chapter(entry_name)
 
     chapters.sort(key=lambda chapter: chapter.get("_sort", 10**9))
+    chapters = fold_image_chapters_into_text(chapters)
     for index, chapter in enumerate(chapters, start=1):
         chapter["id"] = f"C{index:04d}"
         chapter.pop("_sort", None)
+        if chapter.get("kind") == "text":
+            for item_index, item in enumerate(chapter.get("items", []), start=1):
+                item["parent_chapter_id"] = chapter["id"]
+                item["display_level"] = 1
+                item["toc_title"] = item.get("title") or item.get("file_label") or item.get("id") or f"item{item_index:04d}"
+        else:
+            chapter.setdefault("display_level", 0)
 
     return {
         "chapters": chapters,
         "chapter_count": len(chapters),
         "text_chapter_count": sum(1 for chapter in chapters if chapter.get("kind") != "image"),
-        "image_chapter_count": sum(1 for chapter in chapters if chapter.get("kind") == "image"),
-        "part_count": sum(len(chapter["parts"]) for chapter in chapters),
-        "image_count": sum(len(chapter.get("images", [])) for chapter in chapters),
+        "image_chapter_count": image_item_count(chapters) or sum(1 for chapter in chapters if chapter.get("kind") == "image"),
+        "top_level_image_chapter_count": sum(1 for chapter in chapters if chapter.get("kind") == "image"),
+        "part_count": sum(len(chapter.get("parts", [])) for chapter in chapters),
+        "image_count": sum(
+            len(chapter.get("images", [])) if chapter.get("kind") == "image" else sum(
+                len(item.get("images", []))
+                for item in chapter.get("items", [])
+                if item.get("kind") == "image"
+            )
+            for chapter in chapters
+        ),
         "asset_count": len(collect_asset_paths_from_structure({"chapters": chapters})),
         "row_count": len(rows),
     }

tools/epub_review_editor/static/app.js

@@ -5,6 +5,7 @@ let currentIndex = -1;
 let currentRow = null;
 let currentChapter = null;
 let currentPart = null;
+let currentImageItem = null;
 let currentMode = "reading";
 let showTouchedOnly = false;
 let dirty = false;
@@ -12,6 +13,7 @@ let quickDirty = false;
 let hasSession = false;
 let expandedChapters = new Set();
 let sidePanelMode = null;
+let searchScope = "current";
 let gptConfigLoaded = false;
 let gptCandidateHtml = "";
 
@@ -37,6 +39,8 @@ const dom = {
   tocPanel: $("tocPanel"),
   searchPanel: $("searchPanel"),
   searchInput: $("searchInput"),
+  searchScopeCurrentBtn: $("searchScopeCurrentBtn"),
+  searchScopeBookBtn: $("searchScopeBookBtn"),
   stats: $("stats"),
   expandTocBtn: $("expandTocBtn"),
   collapseTocBtn: $("collapseTocBtn"),
@@ -238,14 +242,18 @@ function resetReviewState() {
   currentRow = null;
   currentChapter = null;
   currentPart = null;
+  currentImageItem = null;
   showTouchedOnly = false;
   dirty = false;
   quickDirty = false;
   expandedChapters = new Set();
   sidePanelMode = null;
+  searchScope = "current";
   gptCandidateHtml = "";
   gptConfigLoaded = false;
   dom.showTouchedBtn.classList.remove("primary");
+  dom.searchScopeCurrentBtn.classList.add("active");
+  dom.searchScopeBookBtn.classList.remove("active");
   dom.searchInput.value = "";
   dom.sidebar.hidden = true;
   dom.tocPanel.hidden = true;
@@ -352,10 +360,33 @@ function rowById(rowId) {
   return rows.find((item) => item.id === rowId) || null;
 }
 
+function itemKind(item) {
+  return item && item.kind === "image" ? "image" : "text";
+}
+
+function isImageItem(item) {
+  return itemKind(item) === "image";
+}
+
 function allParts() {
   return (structure.chapters || []).flatMap((chapter) => chapter.parts || []);
 }
 
+function chapterItems(chapter) {
+  if (!chapter) return [];
+  if (Array.isArray(chapter.items) && chapter.items.length) return chapter.items;
+  if (isImageChapter(chapter)) return [{ ...chapter, id: `${chapter.id}:image`, parent_chapter_id: chapter.id, display_level: 1 }];
+  return chapter.parts || [];
+}
+
+function chapterImages(chapter) {
+  if (!chapter) return [];
+  if (isImageChapter(chapter)) return chapter.images || [];
+  return chapterItems(chapter)
+    .filter(isImageItem)
+    .flatMap((item) => item.images || []);
+}
+
 function allChapterScopes() {
   return (structure.chapters || []).map((chapter) => ({
     type: "chapter",
@@ -364,11 +395,20 @@ function allChapterScopes() {
   }));
 }
 
+function activeChapterScopeId() {
+  if (currentImageItem && currentChapter) return currentChapter.id;
+  return currentChapter ? currentChapter.id : chapterForPart(currentPart)?.id;
+}
+
 function chapterIsExpanded(chapter) {
   if (!chapter) return false;
   return expandedChapters.has(chapter.id);
 }
 
+function chapterHasChildren(chapter) {
+  return chapterItems(chapter).length > 0;
+}
+
 function toggleChapterExpanded(chapter) {
   if (!chapter) return;
   if (expandedChapters.has(chapter.id)) {
@@ -386,7 +426,8 @@ function expandAllChapters() {
 
 function collapseAllChapters() {
   expandedChapters = new Set();
-  if (currentChapter) expandedChapters.add(currentChapter.id);
+  const activeId = activeChapterScopeId();
+  if (activeId) expandedChapters.add(activeId);
   renderToc();
 }
 
@@ -424,10 +465,34 @@ function chapterForPart(part) {
 
 function chapterRows(chapter) {
   if (!chapter) return [];
+  if (isImageChapter(chapter)) return [];
   const ids = new Set((chapter.parts || []).flatMap((part) => part.row_ids || []));
   return rows.filter((row) => ids.has(row.id));
 }
 
+function searchBaseRows() {
+  return searchScope === "book" ? rows : scopeRows();
+}
+
+function visibleRowsForScope() {
+  return filteredRows;
+}
+
+function updateSearchScopeButtons() {
+  dom.searchScopeCurrentBtn.classList.toggle("active", searchScope === "current");
+  dom.searchScopeBookBtn.classList.toggle("active", searchScope === "book");
+}
+
+function setSearchScope(scope) {
+  searchScope = scope === "book" ? "book" : "current";
+  updateSearchScopeButtons();
+  if (searchScope === "current" && (currentImageItem || (isImageChapter(currentChapter) && !currentPart))) {
+    currentRow = null;
+    currentIndex = -1;
+  }
+  applyFilter(true);
+}
+
 function scopeCoversRow(row) {
   if (!row) return false;
   if (currentPart) return (currentPart.row_ids || []).includes(row.id);
@@ -442,6 +507,7 @@ function syncScopeToRow(row, forcePart = false) {
   if (!nextPart) return;
   const nextChapter = chapterForPart(nextPart);
   if (!nextChapter) return;
+  currentImageItem = null;
   if (forcePart || !currentChapter || currentChapter.id !== nextChapter.id || !scopeCoversRow(row)) {
     currentChapter = nextChapter;
     currentPart = nextPart;
@@ -458,12 +524,20 @@ function partVisibleRows(part) {
   return filteredRows.filter((row) => ids.has(row.id));
 }
 
-function chapterVisibleRows(chapter) {
-  if (!chapter) return [];
+function chapterVisibleRowsInFilter(chapter) {
+  if (!chapter) return filteredRows;
   const ids = new Set((chapter.parts || []).flatMap((part) => part.row_ids || []));
   return filteredRows.filter((row) => ids.has(row.id));
 }
 
+function chapterVisibleRows(chapter) {
+  if (!chapter) return filteredRows;
+  if (searchScope === "book") {
+    return chapterVisibleRowsInFilter(chapter);
+  }
+  return filteredRows;
+}
+
 function scopeRows() {
   if (currentPart) return partRows(currentPart);
   if (currentChapter) return chapterRows(currentChapter);
@@ -471,13 +545,16 @@ function scopeRows() {
 }
 
 function scopeVisibleRows() {
-  if (currentPart) return partVisibleRows(currentPart);
-  if (currentChapter) return chapterVisibleRows(currentChapter);
+  if (searchScope === "book") {
+    if (currentPart) return partVisibleRows(currentPart);
+    if (currentChapter) return chapterVisibleRows(currentChapter);
+    return filteredRows;
+  }
   return filteredRows;
 }
 
 function imageChapterMatchesSearch(chapter) {
-  if (!isImageChapter(chapter)) return false;
+  if (!isImageChapter(chapter) && !isImageItem(chapter)) return false;
   if (showTouchedOnly) return false;
   const q = dom.searchInput.value.trim().toLowerCase();
   if (!q) return true;
@@ -533,7 +610,13 @@ async function loadRows(keepSelection = false) {
 
   if (!currentPart) {
     currentChapter = (structure.chapters || [])[0] || null;
-    currentPart = currentChapter && !isImageChapter(currentChapter) && (currentChapter.parts || []).length === 1 ? currentChapter.parts[0] : null;
+    currentPart = currentChapter
+      && !isImageChapter(currentChapter)
+      && (currentChapter.parts || []).length === 1
+      && chapterImages(currentChapter).length === 0
+      ? currentChapter.parts[0]
+      : null;
+    applyFilter(false);
   }
   if (filteredRows.length > 0 && currentIndex < 0) {
     currentRow = filteredRows[0];
@@ -596,7 +679,13 @@ function updateStructureCountsFromRows() {
     chapter.row_count = 0;
     chapter.touched_count = 0;
     chapter.marked_count = 0;
-    if (isImageChapter(chapter)) continue;
+    if (isImageChapter(chapter)) {
+      chapter.row_count = 0;
+      chapter.touched_count = 0;
+      chapter.marked_count = 0;
+      chapter.items = chapterItems(chapter);
+      continue;
+    }
     for (const part of chapter.parts || []) {
       const partItems = partRows(part);
       part.row_count = partItems.length;
@@ -606,6 +695,12 @@ function updateStructureCountsFromRows() {
       chapter.touched_count += part.touched_count;
       chapter.marked_count += part.marked_count;
     }
+    for (const item of chapter.items || []) {
+      if (!isImageItem(item)) {
+        const match = (chapter.parts || []).find((part) => part.id === item.id);
+        if (match) Object.assign(item, match);
+      }
+    }
   }
 }
 
@@ -631,7 +726,7 @@ function rowMatchesSearch(row, q) {
 
 function applyFilter(shouldRender = true) {
   const q = dom.searchInput.value.trim().toLowerCase();
-  filteredRows = rows.filter((row) => rowMatchesSearch(row, q));
+  filteredRows = searchBaseRows().filter((row) => rowMatchesSearch(row, q));
   if (currentRow) {
     currentIndex = filteredRows.findIndex((row) => row.id === currentRow.id);
   }
@@ -656,7 +751,7 @@ function renderMode() {
   dom.editorPane.hidden = currentMode !== "polish";
   dom.readingModeBtn.classList.toggle("active", currentMode === "reading");
   dom.polishModeBtn.classList.toggle("active", currentMode === "polish");
-  dom.polishModeBtn.disabled = !hasSession || (isImageChapter(currentChapter) && !currentPart);
+  dom.polishModeBtn.disabled = !hasSession || ((currentImageItem || isImageChapter(currentChapter)) && !currentPart);
   document.body.classList.toggle("quickOpen", currentMode === "reading" && !dom.quickEditor.hidden);
   if (sidePanelMode) {
     dom.tocPanel.hidden = sidePanelMode !== "toc";
@@ -668,11 +763,13 @@ function renderToc() {
   dom.toc.innerHTML = "";
   const frag = document.createDocumentFragment();
   for (const chapter of structure.chapters || []) {
+    const items = chapterItems(chapter);
     const parts = chapter.parts || [];
+    const imageItems = items.filter(isImageItem);
     const chapterIsImage = isImageChapter(chapter);
     const isCurrentChapter = currentChapter && chapter.id === currentChapter.id;
     const containsCurrentPart = parts.some((part) => currentPart && part.id === currentPart.id);
-    const expanded = !chapterIsImage && (chapterIsExpanded(chapter) || containsCurrentPart);
+    const expanded = chapterIsExpanded(chapter) || containsCurrentPart || (isCurrentChapter && (!currentPart || currentImageItem));
     const group = document.createElement("section");
     group.className = "tocChapter";
     if (chapterIsImage) group.classList.add("imageChapter");
@@ -687,13 +784,12 @@ function renderToc() {
     const toggle = document.createElement("button");
     toggle.type = "button";
     toggle.className = "tocToggle";
-    toggle.textContent = chapterIsImage ? "图" : expanded ? "▾" : "▸";
-    toggle.disabled = chapterIsImage || parts.length === 0;
-    toggle.setAttribute("aria-label", chapterIsImage ? "插图章节" : expanded ? "收起章节" : "展开章节");
+    toggle.textContent = expanded ? "▾" : "▸";
+    toggle.disabled = !chapterHasChildren(chapter);
+    toggle.setAttribute("aria-label", expanded ? "收起章节" : "展开章节");
     toggle.setAttribute("aria-expanded", expanded ? "true" : "false");
     toggle.addEventListener("click", (event) => {
       event.stopPropagation();
-      if (chapterIsImage) return;
       toggleChapterExpanded(chapter);
     });
     row.appendChild(toggle);
@@ -709,7 +805,7 @@ function renderToc() {
     }
     title.innerHTML = `
       <span class="tocTitle">${escapeHtml(chapter.title || chapter.id)}</span>
-      <span class="tocCount">${chapterIsImage ? `${(chapter.images || []).length || 1} 图` : `${chapter.row_count || 0} 段`}</span>
+      <span class="tocCount">${chapterIsImage ? `${chapterImages(chapter).length || 1} 图` : `${chapter.row_count || 0} 段${imageItems.length ? ` · ${imageItems.length} 图` : ""}`}</span>
     `;
     title.addEventListener("click", () => selectChapter(chapter));
     row.appendChild(title);
@@ -718,54 +814,84 @@ function renderToc() {
     const partList = document.createElement("div");
     partList.className = "tocParts";
     partList.hidden = !expanded;
-    if (!chapterIsImage) for (const part of parts) {
-      const button = document.createElement("button");
-      button.type = "button";
-      button.className = "tocPartButton";
-      if (currentPart && part.id === currentPart.id) button.classList.add("active");
-      const badges = [];
-      if (part.touched_count) badges.push(`记 ${part.touched_count}`);
-      if (part.marked_count) badges.push(`标 ${part.marked_count}`);
-      button.innerHTML = `
-        <span class="tocTitle">${escapeHtml(part.title || part.file_label || part.id)}</span>
-        <span class="tocCount">${part.row_count || 0}${badges.length ? " · " + escapeHtml(badges.join(" · ")) : ""}</span>
-      `;
-      button.addEventListener("click", () => selectPart(part));
-      partList.appendChild(button);
+    for (const item of items) {
+      if (isImageItem(item)) {
+        const button = document.createElement("button");
+        button.type = "button";
+        button.className = "tocImageButton";
+        const imageActive = isCurrentChapter && !currentPart && (
+          currentImageItem ? currentImageItem.id === item.id : chapterIsImage
+        );
+        if (imageActive) button.classList.add("active");
+        button.innerHTML = `
+          <span class="tocTitle">${escapeHtml(item.toc_title || item.title || item.file_label || "插图")}</span>
+          <span class="tocCount">${(item.images || []).length || 1} 图</span>
+        `;
+        button.addEventListener("click", () => selectImageItem(chapter, item));
+        partList.appendChild(button);
+      } else {
+        const part = item;
+        const button = document.createElement("button");
+        button.type = "button";
+        button.className = "tocPartButton";
+        if (currentPart && part.id === currentPart.id) button.classList.add("active");
+        const badges = [];
+        if (part.touched_count) badges.push(`记 ${part.touched_count}`);
+        if (part.marked_count) badges.push(`标 ${part.marked_count}`);
+        button.innerHTML = `
+          <span class="tocTitle">${escapeHtml(part.title || part.file_label || part.id)}</span>
+          <span class="tocCount">${part.row_count || 0}${badges.length ? " · " + escapeHtml(badges.join(" · ")) : ""}</span>
+        `;
+        button.addEventListener("click", () => selectPart(part));
+        partList.appendChild(button);
+      }
     }
     group.appendChild(partList);
     frag.appendChild(group);
   }
   dom.toc.appendChild(frag);
+  requestAnimationFrame(() => {
+    dom.toc.querySelector(".tocPartButton.active, .tocImageButton.active, .tocChapterButton.active")?.scrollIntoView({ block: "nearest" });
+  });
 }
 
 function renderList() {
   const touched = rows.filter(rowTouched).length;
   const marked = rows.filter((row) => row.marked).length;
+  const visibleRows = visibleRowsForScope();
   const visibleInScope = scopeVisibleRows();
   const imageCount = structure.image_count || 0;
-  const isCurrentImage = isImageChapter(currentChapter) && !currentPart;
-  dom.stats.textContent = `显示 ${filteredRows.length} / ${rows.length} 段 · 当前 ${isCurrentImage ? `${(currentChapter.images || []).length} 图` : `${visibleInScope.length} 段`} · 插图 ${imageCount} · 已记录 ${touched} · 标记 ${marked}`;
+  const isBrowsingImage = !dom.searchInput.value.trim() && !showTouchedOnly;
+  const isCurrentImage = searchScope !== "book" && isBrowsingImage && !currentPart && (Boolean(currentImageItem) || isImageChapter(currentChapter));
+  const scopeLabel = searchScope === "book" ? "全书" : "当前范围";
+  dom.stats.textContent = `${scopeLabel}显示 ${visibleRows.length} / ${rows.length} 段 · 当前 ${isCurrentImage ? `${(currentImageItem ? currentImageItem.images || [] : chapterImages(currentChapter)).length} 图` : `${visibleInScope.length} 段`} · 插图 ${imageCount} · 已记录 ${touched} · 标记 ${marked}`;
   dom.rowList.innerHTML = "";
 
   const frag = document.createDocumentFragment();
   if (isCurrentImage) {
     const item = document.createElement("div");
     item.className = "rowItem imageRowItem active";
-    const images = currentChapter.images || [];
     item.innerHTML = `
       <div class="rowMeta">
         <span>${escapeHtml(currentChapter.id)}</span>
         <span class="rowBadge">插图</span>
         <span>${escapeHtml(currentChapter.file_label || "")}</span>
       </div>
-      <div class="rowPreview">${escapeHtml(images.map((image) => image.asset_name || image.asset_path).join(" / ") || "插图章节")}</div>
+      <div class="rowPreview">${escapeHtml((currentImageItem ? currentImageItem.images || [] : chapterImages(currentChapter)).map((image) => image.asset_name || image.asset_path).join(" / ") || "插图章节")}</div>
     `;
     frag.appendChild(item);
     dom.rowList.appendChild(frag);
     return;
   }
-  visibleInScope.forEach((row) => {
+  if (!visibleRows.length) {
+    const item = document.createElement("div");
+    item.className = "rowItem emptyRowItem";
+    item.textContent = dom.searchInput.value || showTouchedOnly ? "当前检索范围没有命中段落。" : "当前范围没有可检索段落。";
+    frag.appendChild(item);
+    dom.rowList.appendChild(frag);
+    return;
+  }
+  visibleRows.forEach((row) => {
     const button = document.createElement("button");
     button.type = "button";
     button.className = "rowItem";
@@ -789,7 +915,12 @@ function renderList() {
 }
 
 function renderImageChapter(chapter) {
-  const images = chapter.images || [];
+  const items = chapterItems(chapter).filter(isImageItem);
+  if (items.length > 1) {
+    renderMixedChapterImages(chapter, items);
+    return;
+  }
+  const images = chapterImages(chapter);
   dom.chapterKicker.textContent = chapter.file_label || "插图";
   dom.readerTitle.textContent = chapter.title || "插图";
   dom.readingFlow.innerHTML = "";
@@ -813,6 +944,73 @@ function renderImageChapter(chapter) {
   dom.readingFlow.appendChild(frag);
 }
 
+function createImageFigure(item) {
+  const frag = document.createDocumentFragment();
+  (item.images || []).forEach((image, index) => {
+    const figure = document.createElement("figure");
+    figure.className = "imageChapterFigure inlineImageFigure";
+    figure.innerHTML = `
+      <img src="${escapeHtml(image.asset_url || "")}" alt="${escapeHtml(image.alt || item.title || "")}" loading="lazy">
+      <figcaption>
+        <span>${escapeHtml(item.title || `插图 ${index + 1}`)}</span>
+        <span>${escapeHtml(image.asset_name || image.asset_path || "")}</span>
+      </figcaption>
+    `;
+    frag.appendChild(figure);
+  });
+  return frag;
+}
+
+function renderMixedChapterImages(chapter, imageItems, forceShow = false) {
+  dom.chapterKicker.textContent = chapter ? chapter.title : "插图";
+  dom.readerTitle.textContent = imageItems.length === 1
+    ? imageItems[0].title || "插图"
+    : chapter ? chapter.title : "插图";
+  dom.readingFlow.innerHTML = "";
+  const frag = document.createDocumentFragment();
+  imageItems.forEach((item) => {
+    if (forceShow || imageChapterMatchesSearch(item)) frag.appendChild(createImageFigure(item));
+  });
+  if (!frag.childNodes.length) {
+    dom.readingFlow.innerHTML = '<div class="emptyState">当前筛选条件下没有插图。</div>';
+    return;
+  }
+  dom.readingFlow.appendChild(frag);
+}
+
+function rowPassesCurrentFilters(row) {
+  const q = dom.searchInput.value.trim().toLowerCase();
+  return rowMatchesSearch(row, q);
+}
+
+function rowsForReadingPart(part) {
+  const visible = searchScope === "book"
+    ? partRows(part).filter(rowPassesCurrentFilters)
+    : partVisibleRows(part);
+  if (visible.length || dom.searchInput.value || showTouchedOnly) return visible;
+  return partRows(part);
+}
+
+function appendReadBlock(frag, row) {
+  const block = document.createElement("section");
+  block.className = "readBlock";
+  block.dataset.rowId = row.id;
+  if (currentRow && row.id === currentRow.id) block.classList.add("active");
+  if (rowTouched(row)) block.classList.add("touched");
+  if (row.marked) block.classList.add("marked");
+
+  block.innerHTML = `
+    <div class="readText">
+      <div class="readSource">${row.jp_html || escapeHtml(row.jp_text)}</div>
+      <div class="readCn">${row.current_html || row.cn_html || escapeHtml(row.cn_text)}</div>
+    </div>
+  `;
+  block.addEventListener("click", () => {
+    maybeSelectRow(row, { openQuick: true, scroll: false });
+  });
+  frag.appendChild(block);
+}
+
 function renderReading() {
   const selectedChapter = currentChapter || chapterForPart(currentPart);
   if (!currentPart && !selectedChapter) {
@@ -823,6 +1021,10 @@ function renderReading() {
   }
 
   const selectedPart = currentPart;
+  if (!selectedPart && currentImageItem) {
+    renderMixedChapterImages(selectedChapter, [currentImageItem], true);
+    return;
+  }
   if (!selectedPart && isImageChapter(selectedChapter)) {
     renderImageChapter(selectedChapter);
     return;
@@ -833,49 +1035,34 @@ function renderReading() {
     : selectedChapter.title || "未命名章节";
   dom.readingFlow.innerHTML = "";
 
-  const visibleRows = scopeVisibleRows();
-  const sourceRows = visibleRows.length || dom.searchInput.value || showTouchedOnly ? visibleRows : scopeRows();
-  if (!sourceRows.length) {
-    dom.readingFlow.innerHTML = '<div class="emptyState">当前筛选条件下没有段落。</div>';
+  const frag = document.createDocumentFragment();
+
+  if (selectedPart) {
+    rowsForReadingPart(selectedPart).forEach((row) => appendReadBlock(frag, row));
+  } else {
+    for (const item of chapterItems(selectedChapter)) {
+      if (isImageItem(item)) {
+        if (imageChapterMatchesSearch(item)) frag.appendChild(createImageFigure(item));
+        continue;
+      }
+      const partRowsToRender = rowsForReadingPart(item);
+      if (!partRowsToRender.length) continue;
+      const divider = document.createElement("h3");
+      divider.className = "partDivider";
+      divider.textContent = item.title || item.file_label || item.file || "part";
+      frag.appendChild(divider);
+      partRowsToRender.forEach((row) => appendReadBlock(frag, row));
+    }
+  }
+  if (!frag.childNodes.length) {
+    dom.readingFlow.innerHTML = '<div class="emptyState">当前筛选条件下没有内容。</div>';
     return;
   }
-
-  const frag = document.createDocumentFragment();
-  let lastFile = "";
-  sourceRows.forEach((row) => {
-    if (!selectedPart && row.file !== lastFile) {
-      const part = partForRow(row);
-      if (part) {
-        const divider = document.createElement("h3");
-        divider.className = "partDivider";
-        divider.textContent = part.title || part.file_label || row.file;
-        frag.appendChild(divider);
-      }
-      lastFile = row.file;
-    }
-    const block = document.createElement("section");
-    block.className = "readBlock";
-    block.dataset.rowId = row.id;
-    if (currentRow && row.id === currentRow.id) block.classList.add("active");
-    if (rowTouched(row)) block.classList.add("touched");
-    if (row.marked) block.classList.add("marked");
-
-    block.innerHTML = `
-      <div class="readText">
-        <div class="readSource">${row.jp_html || escapeHtml(row.jp_text)}</div>
-        <div class="readCn">${row.current_html || row.cn_html || escapeHtml(row.cn_text)}</div>
-      </div>
-    `;
-    block.addEventListener("click", () => {
-      maybeSelectRow(row, { openQuick: true, scroll: false });
-    });
-    frag.appendChild(block);
-  });
   dom.readingFlow.appendChild(frag);
 }
 
 function renderEditor() {
-  if (isImageChapter(currentChapter) && !currentPart) {
+  if ((currentImageItem || isImageChapter(currentChapter)) && !currentPart) {
     dom.emptyState.hidden = false;
     dom.emptyState.textContent = "插图章节没有可编辑译文，请切回阅读模式查看图片。";
     dom.editorCard.hidden = true;
@@ -997,7 +1184,7 @@ async function saveRow(row, data, silent = false) {
 
 async function saveCurrent(silent = false) {
   if (!hasSession) return true;
-  if (isImageChapter(currentChapter) && !currentPart) return true;
+  if ((currentImageItem || isImageChapter(currentChapter)) && !currentPart) return true;
   if (!currentRow) return true;
   return saveRow(currentRow, collectCurrent(), silent);
 }
@@ -1021,8 +1208,11 @@ async function maybeSelectRow(row, options = {}) {
 function selectRow(row, options = {}) {
   if (!row) return;
   currentRow = row;
-  currentIndex = filteredRows.findIndex((item) => item.id === row.id);
+  if (isImageChapter(currentChapter) && !currentPart) currentChapter = null;
+  currentImageItem = null;
   syncScopeToRow(row, Boolean(options.forcePart));
+  applyFilter(false);
+  currentIndex = filteredRows.findIndex((item) => item.id === row.id);
   dirty = false;
 
   if (options.openEditor) {
@@ -1042,24 +1232,24 @@ async function selectChapter(chapter) {
   if (!ok) return;
   currentChapter = chapter;
   currentPart = null;
+  currentImageItem = null;
   expandedChapters.add(chapter.id);
+  applyFilter(false);
   if (isImageChapter(chapter)) {
     currentRow = null;
     currentIndex = -1;
     closeQuickEditor(false);
     currentMode = "reading";
   } else {
-    const ids = new Set((chapter.parts || []).flatMap((part) => part.row_ids || []));
-    const visible = filteredRows.filter((row) => ids.has(row.id));
-    const fallback = rows.filter((row) => ids.has(row.id));
-    currentRow = visible[0] || fallback[0] || currentRow;
+    const visible = searchScope === "book" ? chapterVisibleRowsInFilter(chapter) : filteredRows;
+    const fallback = scopeRows();
+    currentRow = visible[0] || fallback[0] || null;
     currentIndex = currentRow ? filteredRows.findIndex((row) => row.id === currentRow.id) : -1;
   }
   renderAll();
   if (currentMode === "reading") {
     dom.readerPane.scrollTo({ top: 0, behavior: "smooth" });
   }
-  if (sidePanelMode === "toc") closeSidePanel();
 }
 
 async function selectPart(part) {
@@ -1068,16 +1258,34 @@ async function selectPart(part) {
   if (!ok) return;
   currentChapter = chapterForPart(part);
   currentPart = part;
+  currentImageItem = null;
   if (currentChapter) expandedChapters.add(currentChapter.id);
+  applyFilter(false);
   const visible = partVisibleRows(part);
   const fallback = partRows(part);
-  currentRow = visible[0] || fallback[0] || currentRow;
+  currentRow = visible[0] || fallback[0] || null;
   currentIndex = currentRow ? filteredRows.findIndex((row) => row.id === currentRow.id) : -1;
   renderAll();
   if (currentMode === "reading") {
     dom.readerPane.scrollTo({ top: 0, behavior: "smooth" });
   }
-  if (sidePanelMode === "toc") closeSidePanel();
+}
+
+async function selectImageItem(chapter, item) {
+  if (!chapter || !item) return;
+  const ok = await maybeSaveBeforeSwitch();
+  if (!ok) return;
+  currentChapter = chapter;
+  currentPart = null;
+  currentImageItem = item;
+  currentRow = null;
+  currentIndex = -1;
+  if (chapter) expandedChapters.add(chapter.id);
+  applyFilter(false);
+  closeQuickEditor(false);
+  currentMode = "reading";
+  renderAll();
+  dom.readerPane.scrollTo({ top: 0, behavior: "smooth" });
 }
 
 function scrollRowIntoView(rowId) {
@@ -1094,7 +1302,7 @@ async function setMode(mode, saveBefore = true) {
     const ok = await maybeSaveBeforeSwitch();
     if (!ok) return;
   }
-  if (mode === "polish" && isImageChapter(currentChapter) && !currentPart) {
+  if (mode === "polish" && (currentImageItem || isImageChapter(currentChapter)) && !currentPart) {
     toast("插图章节没有可编辑译文，已保持阅读模式。");
     mode = "reading";
   }
@@ -1105,7 +1313,7 @@ async function setMode(mode, saveBefore = true) {
   if (mode === "polish") {
     closeQuickEditor(false);
     if (!scopeCoversRow(currentRow)) currentRow = null;
-    if (!currentRow) currentRow = scopeVisibleRows()[0] || scopeRows()[0] || filteredRows[0] || null;
+    if (!currentRow) currentRow = visibleRowsForScope()[0] || scopeRows()[0] || filteredRows[0] || null;
   }
   withReaderAnchor(() => renderAll());
   if (mode === "reading" && currentRow) scrollRowIntoView(currentRow.id);
@@ -1114,8 +1322,11 @@ async function setMode(mode, saveBefore = true) {
 function openQuickEditor(row) {
   withReaderAnchor(() => {
     currentRow = row;
-    currentIndex = filteredRows.findIndex((item) => item.id === row.id);
+    if (isImageChapter(currentChapter) && !currentPart) currentChapter = null;
+    currentImageItem = null;
     syncScopeToRow(row);
+    applyFilter(false);
+    currentIndex = filteredRows.findIndex((item) => item.id === row.id);
     fillQuickEditor(row);
     dom.quickEditor.hidden = false;
     document.body.classList.add("quickOpen");
@@ -1274,17 +1485,18 @@ function clearMarks() {
 }
 
 async function go(delta) {
-  if (!filteredRows.length) return;
-  const index = currentIndex >= 0 ? currentIndex : filteredRows.findIndex((row) => currentRow && row.id === currentRow.id);
-  const next = Math.max(0, Math.min(filteredRows.length - 1, index + delta));
+  const candidates = visibleRowsForScope();
+  if (!candidates.length) return;
+  const index = Math.max(0, candidates.findIndex((row) => currentRow && row.id === currentRow.id));
+  const next = Math.max(0, Math.min(candidates.length - 1, index + delta));
   if (next === index) return;
-  await maybeSelectRow(filteredRows[next], { openEditor: currentMode === "polish", openQuick: currentMode === "reading" && !dom.quickEditor.hidden, scroll: true });
+  await maybeSelectRow(candidates[next], { openEditor: currentMode === "polish", openQuick: currentMode === "reading" && !dom.quickEditor.hidden, scroll: true });
 }
 
 async function goPart(delta) {
   const scopes = allChapterScopes();
   if (!scopes.length) return;
-  const currentScopeId = currentChapter ? currentChapter.id : chapterForPart(currentPart)?.id;
+  const currentScopeId = activeChapterScopeId();
   const index = Math.max(0, scopes.findIndex((scope) => scope.id === currentScopeId));
   const next = Math.max(0, Math.min(scopes.length - 1, index + delta));
   if (next === index) return;
@@ -1344,6 +1556,8 @@ function initEvents() {
     currentIndex = -1;
     applyFilter(true);
   });
+  dom.searchScopeCurrentBtn.addEventListener("click", () => setSearchScope("current"));
+  dom.searchScopeBookBtn.addEventListener("click", () => setSearchScope("book"));
   dom.tocPanelBtn.addEventListener("click", () => {
     if (sidePanelMode === "toc") {
       closeSidePanel();

tools/epub_review_editor/static/index.html

@@ -9,7 +9,7 @@
 <body>
   <header class="topbar">
     <div class="brandBlock">
-      <h1>双语 EPUB 审校编辑器 <span id="appVersion" class="versionBadge">v0.4.1</span></h1>
+      <h1>双语 EPUB 审校编辑器 <span id="appVersion" class="versionBadge">v0.5.0</span></h1>
       <p id="sessionMeta">正在载入……</p>
     </div>
     <div class="modeTabs" role="tablist" aria-label="审校模式">
@@ -75,6 +75,10 @@
       <section class="searchPanel" id="searchPanel" hidden>
         <div class="searchbox">
           <input id="searchInput" type="search" placeholder="搜索原文、译文、备注">
+          <div class="searchScope" role="group" aria-label="检索范围">
+            <button id="searchScopeCurrentBtn" type="button" class="active">当前范围</button>
+            <button id="searchScopeBookBtn" type="button">全书全局</button>
+          </div>
         </div>
         <div class="stats" id="stats"></div>
         <div class="rowList" id="rowList"></div>

tools/epub_review_editor/static/style.css

@@ -351,6 +351,8 @@ button.primary:hover,
 .searchbox {
   padding: 12px;
   border-bottom: 1px solid var(--line);
+  display: grid;
+  gap: 8px;
 }
 
 .searchbox input {
@@ -362,6 +364,30 @@ button.primary:hover,
   background: #ffffff;
 }
 
+.searchScope {
+  display: grid;
+  grid-template-columns: 1fr 1fr;
+  border: 1px solid var(--line);
+  border-radius: 6px;
+  overflow: hidden;
+  background: #ffffff;
+}
+
+.searchScope button {
+  min-height: 30px;
+  border: 0;
+  border-radius: 0;
+  background: transparent;
+  color: var(--muted);
+  padding: 3px 8px;
+  font-size: 13px;
+}
+
+.searchScope button.active {
+  background: var(--accent);
+  color: #ffffff;
+}
+
 .stats {
   padding: 8px 12px;
   color: var(--muted);
@@ -432,7 +458,8 @@ button.primary:hover,
 }
 
 .tocChapterButton,
-.tocPartButton {
+.tocPartButton,
+.tocImageButton {
   width: 100%;
   display: grid;
   grid-template-columns: minmax(0, 1fr) auto;
@@ -453,23 +480,37 @@ button.primary:hover,
   font-weight: 600;
 }
 
-.tocPartButton {
+.tocPartButton,
+.tocImageButton {
   margin-left: 24px;
   width: calc(100% - 24px);
   color: var(--muted);
 }
 
+.tocImageButton {
+  padding-left: 14px;
+  font-size: 13px;
+}
+
+.tocImageButton .tocTitle::before {
+  content: "插图 · ";
+  color: var(--accent);
+  font-weight: 600;
+}
+
 .tocChapterRow:hover,
 .tocChapterRow.active,
 .tocChapterButton:hover,
 .tocToggle:hover,
-.tocPartButton:hover {
+.tocPartButton:hover,
+.tocImageButton:hover {
   background: #eef3f6;
 }
 
 .tocChapterRow.active,
 .tocChapterButton.active,
-.tocPartButton.active {
+.tocPartButton.active,
+.tocImageButton.active {
   background: var(--accent-soft);
   color: var(--accent-strong);
 }
@@ -524,6 +565,15 @@ button.primary:hover,
   color: var(--warn);
 }
 
+.emptyRowItem {
+  color: var(--muted);
+  cursor: default;
+}
+
+.emptyRowItem:hover {
+  background: transparent;
+}
+
 .imageRowItem {
   cursor: default;
 }
@@ -570,6 +620,11 @@ button.primary:hover,
   background: #f4f6f8;
 }
 
+.readerPane,
+.editorPane {
+  transition: margin-left 160ms ease;
+}
+
 .readerHeader {
   position: sticky;
   top: 0;
@@ -667,6 +722,11 @@ button.primary:hover,
   text-align: center;
 }
 
+.inlineImageFigure {
+  margin: 18px auto 30px;
+  padding-top: 8px;
+}
+
 .imageChapterFigure img {
   display: block;
   max-width: 100%;
@@ -973,6 +1033,17 @@ textarea {
   display: none !important;
 }
 
+@media (min-width: 1041px) {
+  body.sidebarOpen .readerPane,
+  body.sidebarOpen .editorPane {
+    margin-left: 380px;
+  }
+
+  body.sidebarOpen .sidebar {
+    box-shadow: none;
+  }
+}
+
 @media (min-width: 1320px) {
   body.quickOpen .readingFlow {
     margin-left: 56px;

tools/epub_review_editor/version.py

@@ -1,4 +1,4 @@
-version = "0.4.1"
+version = "0.5.0"
 
 VERSION_RULES = {
     "PATCH": "修复 bug 或兼容性小修",