Optimize Emby caching and image resize concurrency

internal/service/emby_compat.go

@@ -61,6 +61,9 @@ type EmbyService struct {
 
 	libraryCoverMu    sync.Mutex
 	libraryCoverCache map[string]embyArtworkCacheEntry
+
+	peopleMu    sync.RWMutex
+	peopleCache map[string]embyPeopleCacheEntry
 }
 
 // NewEmbyService is the constructor.
@@ -143,6 +146,13 @@ type embyVisibilityCacheEntry struct {
 	expiresAt  time.Time
 }
 
+// embyPeopleCacheEntry avoids re-statting/decoding the same NFO for every
+// list refresh. TV clients commonly request the same posters/items repeatedly.
+type embyPeopleCacheEntry struct {
+	people    []map[string]any
+	expiresAt time.Time
+}
+
 // Items paginates media in Emby's hierarchy. Episodic libraries are exposed as
 // Series -> Season -> Episode so Infuse/Vidhub/SenPlayer stop treating every
 // episode as a separate movie card. 带 embyremote~ 前缀的 ParentID / 搜索自动

internal/service/emby_counts.go

@@ -3,12 +3,24 @@ package service
 import (
 	"context"
 	"strings"
+	"time"
 
 	"github.com/truewhile/MeBox/internal/model"
 	"gorm.io/gorm"
 )
 
 func (e *EmbyService) ItemCounts(ctx context.Context, userID string) (map[string]any, error) {
+	cacheKey := e.embyItemsCacheKey("counts-v1", ItemsParams{UserID: userID})
+	var cached embyCountsCacheValue
+	if e.cache != nil && e.cache.GetJSON(ctx, cacheKey, &cached) {
+		return map[string]any{
+			"MovieCount":   cached.MovieCount,
+			"SeriesCount":  int(cached.SeriesCount),
+			"EpisodeCount": cached.EpisodeCount,
+			"ItemCount":    cached.ItemCount,
+		}, nil
+	}
+
 	base := func() *gorm.DB {
 		q := e.repo.DB.WithContext(ctx).Model(&model.Media{}).Where("deleted_at IS NULL")
 		return e.applyUserMediaVisibility(ctx, q, userID)
@@ -34,6 +46,14 @@ func (e *EmbyService) ItemCounts(ctx context.Context, userID string) (map[string
 		return nil, err
 	}
 
+	if e.cache != nil {
+		e.cache.SetJSON(ctx, cacheKey, embyCountsCacheValue{
+			MovieCount:   movieCount,
+			SeriesCount:  int64(seriesCount),
+			EpisodeCount: episodeCount,
+			ItemCount:    itemCount,
+		}, time.Duration(e.mediaCacheTTLSeconds())*time.Second)
+	}
 	return map[string]any{
 		"MovieCount":   movieCount,
 		"SeriesCount":  seriesCount,

internal/service/emby_items_cache.go

@@ -9,9 +9,10 @@ import (
 )
 
 type embyItemsCacheValue struct {
-	Items            []map[string]any `json:"items"`
-	TotalRecordCount int64            `json:"total_record_count"`
-	StartIndex       int              `json:"start_index"`
+	Items            []map[string]any          `json:"items"`
+	TotalRecordCount int64                     `json:"total_record_count"`
+	StartIndex       int                       `json:"start_index"`
+	Artwork          map[string]embyArtworkRef `json:"artwork,omitempty"`
 }
 
 type embyLatestCacheValue struct {
@@ -19,6 +20,13 @@ type embyLatestCacheValue struct {
 	Artwork map[string]embyArtworkRef `json:"artwork,omitempty"`
 }
 
+type embyCountsCacheValue struct {
+	MovieCount   int64 `json:"movie_count"`
+	SeriesCount  int64 `json:"series_count"`
+	EpisodeCount int64 `json:"episode_count"`
+	ItemCount    int64 `json:"item_count"`
+}
+
 func (e *EmbyService) embyItemsCacheKey(kind string, p ItemsParams) string {
 	includeTypes := append([]string(nil), p.IncludeItemTypes...)
 	filters := append([]string(nil), p.Filters...)

internal/service/emby_items_detail.go

@@ -538,6 +538,14 @@ func (e *EmbyService) resolveMediaPeople(ctx context.Context, m *model.Media) []
 	if m == nil || strings.TrimSpace(m.Path) == "" {
 		return []map[string]any{}
 	}
+	cacheKey := strings.TrimSpace(m.ID)
+	if cacheKey == "" {
+		cacheKey = strings.ToLower(filepath.Clean(m.Path))
+	}
+	if people, ok := e.cachedMediaPeople(cacheKey); ok {
+		return people
+	}
+
 	dir := filepath.Dir(m.Path)
 	candidates := make([]string, 0, 6)
 	seenPath := map[string]struct{}{}
@@ -610,9 +618,48 @@ func (e *EmbyService) resolveMediaPeople(ctx context.Context, m *model.Media) []
 			}
 		}
 	}
+	e.rememberMediaPeople(cacheKey, people)
 	return people
 }
 
+func (e *EmbyService) cachedMediaPeople(key string) ([]map[string]any, bool) {
+	if e == nil || strings.TrimSpace(key) == "" {
+		return nil, false
+	}
+	now := time.Now()
+	e.peopleMu.RLock()
+	entry, ok := e.peopleCache[key]
+	e.peopleMu.RUnlock()
+	if !ok || now.After(entry.expiresAt) {
+		if ok {
+			e.peopleMu.Lock()
+			delete(e.peopleCache, key)
+			e.peopleMu.Unlock()
+		}
+		return nil, false
+	}
+	out := make([]map[string]any, len(entry.people))
+	copy(out, entry.people)
+	return out, true
+}
+
+func (e *EmbyService) rememberMediaPeople(key string, people []map[string]any) {
+	if e == nil || strings.TrimSpace(key) == "" {
+		return
+	}
+	e.peopleMu.Lock()
+	defer e.peopleMu.Unlock()
+	if e.peopleCache == nil || len(e.peopleCache) > 8000 {
+		e.peopleCache = make(map[string]embyPeopleCacheEntry, 128)
+	}
+	stored := make([]map[string]any, len(people))
+	copy(stored, people)
+	e.peopleCache[key] = embyPeopleCacheEntry{
+		people:    stored,
+		expiresAt: time.Now().Add(embyVirtualCacheTTL),
+	}
+}
+
 func embyPersonID(name, roleType string) string {
 	sum := sha256.Sum256([]byte(strings.ToLower(strings.TrimSpace(name)) + ":" + strings.ToLower(strings.TrimSpace(roleType))))
 	return "person-" + hex.EncodeToString(sum[:8])

internal/service/emby_items_list.go

@@ -75,9 +75,9 @@ func (e *EmbyService) mediaItems(ctx context.Context, p ItemsParams) (map[string
 		orderIncludesDirection = false
 	case "premieredate", "productionyear":
 		order = mediaReleaseOrderSQL(desc)
-		case "datecreated", "datelastmediaadded", "datelastcontentadded":
-			order = "media.created_at"
-			orderIncludesDirection = false
+	case "datecreated", "datelastmediaadded", "datelastcontentadded":
+		order = "media.created_at"
+		orderIncludesDirection = false
 	case "dateplayed":
 		order = "resume.watched_at"
 		orderIncludesDirection = false
@@ -225,6 +225,17 @@ func (e *EmbyService) collapseMediaVersionRows(ctx context.Context, rows []model
 }
 
 func (e *EmbyService) seriesItemsForLibrary(ctx context.Context, libraryID string, p ItemsParams) (map[string]any, error) {
+	cacheKey := e.embyItemsCacheKey("series-items-v1", p)
+	var cached embyItemsCacheValue
+	if e.cache != nil && e.cache.GetJSON(ctx, cacheKey, &cached) {
+		e.rememberArtworkRefs(cached.Artwork)
+		return map[string]any{
+			"Items":            cached.Items,
+			"TotalRecordCount": int(cached.TotalRecordCount),
+			"StartIndex":       cached.StartIndex,
+		}, nil
+	}
+
 	q := e.repo.DB.WithContext(ctx).Model(&model.Media{}).Where("season_num > 0 OR episode_num > 0")
 	q = e.applyUserMediaVisibility(ctx, q, p.UserID)
 	if libraryID != "" {
@@ -246,9 +257,19 @@ func (e *EmbyService) seriesItemsForLibrary(ctx context.Context, libraryID strin
 	groups := e.seriesGroupsFromMedia(ctx, rows)
 	sortSeriesGroups(groups, p)
 	total := len(groups)
-	items := make([]map[string]any, 0, minInt(p.Limit, len(groups)))
-	for _, group := range pageSlice(groups, p.StartIndex, p.Limit) {
+	pageGroups := pageSlice(groups, p.StartIndex, p.Limit)
+	items := make([]map[string]any, 0, len(pageGroups))
+	for _, group := range pageGroups {
 		items = append(items, e.seriesPayload(group))
 	}
-	return map[string]any{"Items": items, "TotalRecordCount": total, "StartIndex": p.StartIndex}, nil
+	out := map[string]any{"Items": items, "TotalRecordCount": total, "StartIndex": p.StartIndex}
+	if e.cache != nil {
+		e.cache.SetJSON(ctx, cacheKey, embyItemsCacheValue{
+			Items:            items,
+			TotalRecordCount: int64(total),
+			StartIndex:       p.StartIndex,
+			Artwork:          e.artworkRefsForSeriesGroups(pageGroups),
+		}, time.Duration(e.mediaCacheTTLSeconds())*time.Second)
+	}
+	return out, nil
 }

internal/service/emby_movie_items.go

@@ -35,6 +35,17 @@ func (e *EmbyService) movieLibraryHasEpisodicContent(ctx context.Context, librar
 // 与 mediaItems 的区别: 后者会把剧集结构行当散装 Episode 漏出;这里改为聚合成
 // Series,从根本上消除「电影库里整部剧被拆成单集」的现象。
 func (e *EmbyService) movieLibraryItems(ctx context.Context, p ItemsParams) (map[string]any, error) {
+	cacheKey := e.embyItemsCacheKey("movie-library-items-v1", p)
+	var cached embyItemsCacheValue
+	if e.cache != nil && e.cache.GetJSON(ctx, cacheKey, &cached) {
+		e.rememberArtworkRefs(cached.Artwork)
+		return map[string]any{
+			"Items":            cached.Items,
+			"TotalRecordCount": int(cached.TotalRecordCount),
+			"StartIndex":       cached.StartIndex,
+		}, nil
+	}
+
 	libIDs := e.mergedLibraryIDs(ctx, p.ParentID)
 	apply := func(q *gorm.DB) *gorm.DB {
 		q = e.applyUserMediaVisibility(ctx, q, p.UserID)
@@ -78,33 +89,72 @@ func (e *EmbyService) movieLibraryItems(ctx context.Context, p ItemsParams) (map
 	if err := movieQ.Find(&movieRows).Error; err != nil {
 		return nil, err
 	}
-	movieItems, err := e.payloadsForMedia(ctx, movieRows, p.UserID)
-	if err != nil {
-		return nil, err
-	}
+	// 先按版本去重，再参与排序。这里不立即构建 payload：大电影库可能有
+	// 数万行，而客户端一页通常只要几十条，提前构建会触发大量 NFO / 数据
+	// 查询并把响应时间浪费在用户根本看不到的条目上。
+	movieRows = e.collapseMediaVersionRows(ctx, movieRows)
 
 	// 合并: Series 卡片 + Movie 项, 统一按首播/上映日期倒序。
 	type entry struct {
-		sortAt  time.Time
-		payload map[string]any
+		sortAt time.Time
+		media  *model.Media
+		group  *embySeriesGroup
 	}
-	entries := make([]entry, 0, len(seriesGroups)+len(movieItems))
-	for _, g := range seriesGroups {
-		entries = append(entries, entry{sortAt: embySeriesReleaseSortTime(g), payload: e.seriesPayload(g)})
+	entries := make([]entry, 0, len(seriesGroups)+len(movieRows))
+	for i := range seriesGroups {
+		group := &seriesGroups[i]
+		entries = append(entries, entry{sortAt: embySeriesReleaseSortTime(*group), group: group})
 	}
-	for _, item := range movieItems {
-		entries = append(entries, entry{sortAt: embyPayloadReleaseSortTime(item), payload: item})
+	for i := range movieRows {
+		media := &movieRows[i]
+		entries = append(entries, entry{sortAt: embyMediaReleaseSortTime(*media), media: media})
 	}
 	sort.SliceStable(entries, func(i, j int) bool {
 		return entries[i].sortAt.After(entries[j].sortAt)
 	})
 	total := len(entries)
 	paged := pageSlice(entries, p.StartIndex, p.Limit)
-	items := make([]map[string]any, 0, len(paged))
+
+	pageMovies := make([]model.Media, 0, len(paged))
 	for _, en := range paged {
-		items = append(items, en.payload)
+		if en.media != nil {
+			pageMovies = append(pageMovies, *en.media)
+		}
 	}
-	return map[string]any{"Items": items, "TotalRecordCount": total, "StartIndex": p.StartIndex}, nil
+	moviePayloads, err := e.payloadsForMedia(ctx, pageMovies, p.UserID)
+	if err != nil {
+		return nil, err
+	}
+	payloadByID := make(map[string]map[string]any, len(moviePayloads))
+	for _, item := range moviePayloads {
+		if id, ok := item["Id"].(string); ok {
+			payloadByID[id] = item
+		}
+	}
+
+	items := make([]map[string]any, 0, len(paged))
+	pageGroups := make([]embySeriesGroup, 0, len(paged))
+	for _, en := range paged {
+		switch {
+		case en.group != nil:
+			pageGroups = append(pageGroups, *en.group)
+			items = append(items, e.seriesPayload(*en.group))
+		case en.media != nil:
+			if item := payloadByID[en.media.ID]; item != nil {
+				items = append(items, item)
+			}
+		}
+	}
+	out := map[string]any{"Items": items, "TotalRecordCount": total, "StartIndex": p.StartIndex}
+	if e.cache != nil {
+		e.cache.SetJSON(ctx, cacheKey, embyItemsCacheValue{
+			Items:            items,
+			TotalRecordCount: int64(total),
+			StartIndex:       p.StartIndex,
+			Artwork:          e.artworkRefsForSeriesGroups(pageGroups),
+		}, time.Duration(e.mediaCacheTTLSeconds())*time.Second)
+	}
+	return out, nil
 }
 
 // embyPayloadCreatedAt 从 item payload 里取 DateCreated(time.Time),用于合并排序。

internal/service/emby_series_hierarchy_test.go

@@ -578,3 +578,45 @@ func TestEmbySeriesSortByDateLastMediaAdded(t *testing.T) {
 		t.Fatalf("DateLastMediaAdded = %v, want %v", items[0]["DateLastMediaAdded"], tNew)
 	}
 }
+
+func TestEmbySeriesLibraryListUsesRuntimeCache(t *testing.T) {
+	svc := newTestEmbyService(t)
+	svc.cache = NewRuntimeCacheService(nil, nil)
+	lib := model.Library{Name: "番剧", Path: `/media/anime`, Type: "anime", Enabled: true}
+	if err := svc.repo.Library.Create(t.Context(), &lib); err != nil {
+		t.Fatalf("create library: %v", err)
+	}
+	for i := 1; i <= 2; i++ {
+		media := model.Media{
+			Base:       model.Base{ID: fmt.Sprintf("cache-ep-%d", i)},
+			LibraryID:  lib.ID,
+			Title:      "缓存测试番",
+			Path:       fmt.Sprintf(`/media/anime/缓存测试番/Season 01/缓存测试番.S01E%02d.mkv`, i),
+			SeasonNum:  1,
+			EpisodeNum: i,
+		}
+		if err := svc.repo.DB.Create(&media).Error; err != nil {
+			t.Fatalf("create media: %v", err)
+		}
+	}
+
+	first, err := svc.Items(t.Context(), ItemsParams{ParentID: lib.ID, Limit: 20})
+	if err != nil {
+		t.Fatalf("first items call: %v", err)
+	}
+	if first["TotalRecordCount"] != 1 {
+		t.Fatalf("first series total = %#v, want 1", first["TotalRecordCount"])
+	}
+	if err := svc.repo.DB.Unscoped().Where("library_id = ?", lib.ID).Delete(&model.Media{}).Error; err != nil {
+		t.Fatalf("delete media: %v", err)
+	}
+
+	second, err := svc.Items(t.Context(), ItemsParams{ParentID: lib.ID, Limit: 20})
+	if err != nil {
+		t.Fatalf("second items call: %v", err)
+	}
+	items, _ := second["Items"].([]map[string]any)
+	if second["TotalRecordCount"] != 1 || len(items) != 1 {
+		t.Fatalf("cached series list = %#v, want the first response", second)
+	}
+}

internal/service/image_proxy.go

@@ -17,6 +17,7 @@ import (
 	"net"
 	"net/http"
 	"path/filepath"
+	"runtime"
 	"strings"
 	"sync"
 	"syscall"
@@ -35,6 +36,12 @@ type ImageProxy struct {
 	cacheDir string
 	mu       sync.Mutex
 
+	// resizeSem bounds concurrent decode/resize jobs. Emby TV clients request
+	// poster grids in bursts; letting every request decode a source image at
+	// once causes CPU and memory spikes that make the whole UI feel sluggish.
+	resizeSemMu sync.Mutex
+	resizeSem   chan struct{}
+
 	// libraryRootsFn returns the configured media library roots so that
 	// sidecar poster/artwork files stored alongside media (under arbitrary
 	// per-library paths) are allowed by isAllowedLocalPath. It is provided
@@ -55,6 +62,7 @@ type ImageProxy struct {
 const (
 	imageBrowserCacheControl     = "public, max-age=2592000, immutable"
 	imagePlaceholderCacheControl = "no-store"
+	imageMaxResizeConcurrency    = 4
 )
 
 // NewImageProxy is the constructor.
@@ -64,6 +72,7 @@ func NewImageProxy(cfg *config.Config, log *zap.Logger) *ImageProxy {
 		log:      log,
 		cacheDir: filepath.Join(cfg.Cache.CacheDir, "images"),
 	}
+	proxy.resizeSem = make(chan struct{}, imageResizeConcurrency())
 
 	// Honor HTTP(S)_PROXY env vars so deployments behind GFW can pull
 	// from image.tmdb.org via their HTTP proxy without extra config. On
@@ -181,6 +190,20 @@ func (p *ImageProxy) isAllowedRemoteHost(host string) bool {
 	return p.allowedHostsCache[host]
 }
 
+// imageResizeConcurrency keeps decode/resize concurrency within the number
+// of CPU threads the process is allowed to use, capped to avoid large
+// temporary RGBA buffers on tiny hosts.
+func imageResizeConcurrency() int {
+	n := runtime.GOMAXPROCS(0)
+	if n < 1 {
+		n = 1
+	}
+	if n > imageMaxResizeConcurrency {
+		n = imageMaxResizeConcurrency
+	}
+	return n
+}
+
 // Prune removes oldest cached images until disk usage is within the configured limit.
 func (p *ImageProxy) Prune() (PruneImageCacheResult, error) {
 	if p.cfg == nil || p.cfg.Cache.ImagesMaxSizeMB <= 0 {

internal/service/image_proxy_remote.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"context"
 	"errors"
+	"io"
 	"net/http"
 	"os"
 	"path/filepath"
@@ -134,12 +135,38 @@ func (p *ImageProxy) serveCachedImage(w http.ResponseWriter, r *http.Request, ke
 }
 
 func (p *ImageProxy) removeUnusableImageCache(cachePath, failPath string) {
-	data, err := os.ReadFile(cachePath) // #nosec G304 -- cachePath is SHA-derived under cacheDir.
+	// 只读取文件头判断缓存是否可用。旧实现每次命中远程图片缓存都会把整个
+	// 原图读进内存再丢弃，电视端批量加载海报时会产生大量无意义的磁盘 I/O。
+	file, err := os.Open(cachePath) // #nosec G304 -- cachePath is SHA-derived under cacheDir.
 	if err != nil {
 		return
 	}
-	ctype := detectContentType(data)
-	if len(data) > 0 && isImageContentType(ctype) && !isTransparentPlaceholderData(data) {
+	stat, err := file.Stat()
+	if err != nil || stat.IsDir() || stat.Size() <= 0 {
+		_ = file.Close()
+		_ = os.Remove(cachePath)
+		_ = os.Remove(failPath)
+		return
+	}
+	headerSize := 512
+	if stat.Size() < int64(headerSize) {
+		headerSize = int(stat.Size())
+	}
+	header := make([]byte, headerSize)
+	n, readErr := io.ReadFull(file, header)
+	_ = file.Close()
+	if readErr != nil && readErr != io.ErrUnexpectedEOF {
+		_ = os.Remove(cachePath)
+		_ = os.Remove(failPath)
+		return
+	}
+	header = header[:n]
+	// A transparent placeholder is exactly 67 bytes; checking the header alone
+	// is enough for the normal image cache entries (they are much larger but
+	// detectContentType only inspects the same leading 512 bytes anyway).
+	// Close the handle before deleting: Windows refuses to delete an open file.
+	if n > 0 && isImageContentType(detectContentType(header)) &&
+		!(n == len(transparent1x1PNG) && bytes.Equal(header, transparent1x1PNG)) {
 		return
 	}
 	_ = os.Remove(cachePath)

internal/service/image_resize.go

@@ -2,6 +2,7 @@ package service
 
 import (
 	"bytes"
+	"context"
 	"crypto/sha256"
 	"encoding/hex"
 	"errors"
@@ -203,6 +204,28 @@ func (p *ImageProxy) resizeCachePath(key string) string {
 	return filepath.Join(p.cacheDir, imageResizeCacheSubdir, key+".img")
 }
 
+// acquireResizeSlot bounds CPU-heavy decode/resize work. Returning false means
+// the caller should fall back to the original image instead of blocking after
+// the request has already been canceled.
+func (p *ImageProxy) acquireResizeSlot(ctx context.Context) (func(), bool) {
+	if p == nil {
+		return func() {}, true
+	}
+	p.resizeSemMu.Lock()
+	if p.resizeSem == nil {
+		p.resizeSem = make(chan struct{}, imageResizeConcurrency())
+	}
+	sem := p.resizeSem
+	p.resizeSemMu.Unlock()
+
+	select {
+	case sem <- struct{}{}:
+		return func() { <-sem }, true
+	case <-ctx.Done():
+		return nil, false
+	}
+}
+
 // serveResizedFromFile 从 srcPath 读取图片，按选项缩放后写出，并把结果缓存
 // 到磁盘以免每次请求都重新解码。原图已满足目标尺寸时直接输出原文件。
 // 返回 false 表示缩放不可用，调用方应回退到原图直出。
@@ -211,6 +234,25 @@ func (p *ImageProxy) serveResizedFromFile(w http.ResponseWriter, r *http.Request
 	if err != nil || stat.IsDir() || stat.Size() <= 0 {
 		return false
 	}
+	// 缓存命中必须发生在读原图和解码之前。否则电视端每次刷新海报墙都会
+	// 把已经是缩略图缓存的原图重新解码、缩放一遍，造成明显的 CPU 抖动。
+	key := o.resizeCacheKey(srcPath, stat)
+	cachePath := p.resizeCachePath(key)
+	if serveCachedImageFile(w, r, key, cachePath) {
+		return true
+	}
+
+	release, ok := p.acquireResizeSlot(r.Context())
+	if !ok {
+		return false
+	}
+	defer release()
+
+	// 等待并发槽期间，别的请求可能已经生成了同一张缩略图。
+	if serveCachedImageFile(w, r, key, cachePath) {
+		return true
+	}
+
 	data, err := os.ReadFile(srcPath) // #nosec G304 -- srcPath comes from an allowed local path or a SHA-derived cache path.
 	if err != nil {
 		return false
@@ -224,11 +266,6 @@ func (p *ImageProxy) serveResizedFromFile(w http.ResponseWriter, r *http.Request
 		return serveImageFile(w, r, filepath.Base(srcPath), srcPath, imageBrowserCacheControl)
 	}
 
-	key := o.resizeCacheKey(srcPath, stat)
-	cachePath := p.resizeCachePath(key)
-	if serveCachedImageFile(w, r, key, cachePath) {
-		return true
-	}
 	p.writeResizeCache(cachePath, out)
 	w.Header().Set("Content-Type", ctype)
 	w.Header().Set("Cache-Control", imageBrowserCacheControl)

internal/service/image_resize_test.go

@@ -230,3 +230,47 @@ func TestServeResizedFromFileCachesScaledResult(t *testing.T) {
 		t.Fatal("expected a non-empty body")
 	}
 }
+
+func TestServeResizedFromFileUsesCacheBeforeDecodingSource(t *testing.T) {
+	dir := t.TempDir()
+	mediaDir := dir + string(os.PathSeparator) + "media"
+	if err := os.MkdirAll(mediaDir, 0o755); err != nil {
+		t.Fatalf("mkdir: %v", err)
+	}
+	src := mediaDir + string(os.PathSeparator) + "poster.png"
+	original := encodeTestPNG(t, 529, 911, 255)
+	if err := os.WriteFile(src, original, 0o644); err != nil {
+		t.Fatalf("write source: %v", err)
+	}
+
+	proxy := &ImageProxy{cacheDir: dir + string(os.PathSeparator) + "cache"}
+	opts := imageResizeOptions{MaxWidth: 400, Quality: 90}
+
+	first := httptest.NewRecorder()
+	if !proxy.serveResizedFromFile(first, httptest.NewRequest("GET", "/x?maxWidth=400", nil), src, opts) {
+		t.Fatal("expected first call to be served")
+	}
+
+	stat, err := os.Stat(src)
+	if err != nil {
+		t.Fatalf("stat source: %v", err)
+	}
+	// Same size and mtime keep the resize cache key stable, but the source is
+	// now invalid image data. A correct implementation serves the cached
+	// thumbnail before reading/decoding the source again.
+	broken := bytes.Repeat([]byte{0}, len(original))
+	if err := os.WriteFile(src, broken, 0o644); err != nil {
+		t.Fatalf("overwrite source: %v", err)
+	}
+	if err := os.Chtimes(src, stat.ModTime(), stat.ModTime()); err != nil {
+		t.Fatalf("restore mtime: %v", err)
+	}
+
+	second := httptest.NewRecorder()
+	if !proxy.serveResizedFromFile(second, httptest.NewRequest("GET", "/x?maxWidth=400", nil), src, opts) {
+		t.Fatal("expected second call to be served from resize cache")
+	}
+	if !bytes.Equal(first.Body.Bytes(), second.Body.Bytes()) {
+		t.Fatal("expected cached thumbnail to be reused without decoding the source")
+	}
+}

internal/service/media_release_sort.go

@@ -32,6 +32,18 @@ func mediaReleaseOrderSQL(desc bool) string {
 	return fmt.Sprintf("media.release_date %s, media.year %s, media.created_at %s, media.id %s", dir, dir, dir, dir)
 }
 
+// embyMediaReleaseSortTime matches the original payload-based ordering used by
+// the Emby movie-library merge: release date, then year, then created_at.
+func embyMediaReleaseSortTime(media model.Media) time.Time {
+	if t, ok := embyPremiereDate(media.ReleaseDate); ok {
+		return t
+	}
+	if media.Year > 0 {
+		return time.Date(media.Year, time.December, 31, 0, 0, 0, 0, time.UTC)
+	}
+	return media.CreatedAt
+}
+
 func embyPremiereDate(value string) (time.Time, bool) {
 	value = normalizeReleaseDate(value)
 	if value == "" {