feat(cbz): adapt reader to real archive structure — PageIndex skips macOS junk (__MACOSX/ and ._ AppleDouble were polluting pages with 163-byte blacks), pages endpoint derives chapters[{title,start}] from folder layout, imageless zip now lands state=error instead of an empty reader, CbzReader gains 目录 sidebar + prev/next chapter (TextReader pattern, rd-* classes); pagesidx cache key bumped for one-time invalidation; e2e on 311MB 調教開關：第二季.zip: 15048→7524 real pages, 52 chapters, all JPEGs

backend/cmd/webui/handlers/content.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 	"net/http"
 	"os"
+	"path"
 	"path/filepath"
 	"strconv"
 	"strings"
@@ -118,7 +119,7 @@ func (h *H) openBook(c *gin.Context, b store.Book, root string) (*os.File, int64
 
 func (h *H) pageIndex(c *gin.Context, b store.Book, root string) ([]string, error) {
 	hash := bookfile.Hash(b.FileSize, b.ModTS)
-	key := fmt.Sprintf("pagesidx:%d:%s", b.ID, hash)
+	key := fmt.Sprintf("pagesidx2:%d:%s", b.ID, hash) // 前缀换代 = 索引逻辑变更时一次性作废旧缓存
 	if v, ok := h.rdb.Get(c, key); ok && v != "" {
 		return strings.Split(v, "\n"), nil
 	}
@@ -158,7 +159,48 @@ func (h *H) PagesCount(c *gin.Context) {
 		err(c, http.StatusUnprocessableEntity, "broken", e.Error())
 		return
 	}
-	c.JSON(http.StatusOK, gin.H{"count": len(idx)})
+	c.JSON(http.StatusOK, gin.H{"count": len(idx), "chapters": chaptersOf(idx)})
+}
+
+type cbzChapter struct {
+	Title string `json:"title"`
+	Start int    `json:"start"`
+}
+
+// chaptersOf 页索引已自然序排列,按父目录分组:每个新目录开一章,标题取目录名;扁平包或只有一组返回 nil
+func chaptersOf(idx []string) []cbzChapter {
+	type grp struct {
+		dir   string
+		start int
+	}
+	var grps []grp
+	last := "\x00"
+	for i, n := range idx {
+		d := path.Dir(n)
+		if d == last {
+			continue
+		}
+		last = d
+		if d == "." { // 根目录散页不开章
+			continue
+		}
+		grps = append(grps, grp{d, i})
+	}
+	if len(grps) < 2 {
+		return nil
+	}
+	titles := make(map[string]int)
+	out := make([]cbzChapter, len(grps))
+	for i, g := range grps {
+		out[i] = cbzChapter{Title: path.Base(g.dir), Start: g.start}
+		titles[out[i].Title]++
+	}
+	for i, g := range grps { // 同名目录(不同父级)撞车 → 用全路径消歧
+		if titles[out[i].Title] > 1 {
+			out[i].Title = g.dir
+		}
+	}
+	return out
 }
 
 func (h *H) Page(c *gin.Context) {