fix: reduce duplicate home recommendations

91VideoSpider/spider_91porn.py

@@ -68,6 +68,7 @@ import time
 import random
 import json
 import os
+import socket
 import sys
 import html
 from urllib.parse import urljoin, unquote, urlparse
@@ -80,6 +81,28 @@ except ImportError:
     print("请运行: pip install beautifulsoup4 lxml")
     sys.exit(1)
 
+
+def prefer_ipv4_for_plain_socks5_proxy():
+    """PySocks may pick IPv6 first for socks5://; some SOCKS5 servers only accept IPv4."""
+    proxy_envs = (
+        os.environ.get("HTTPS_PROXY", ""),
+        os.environ.get("HTTP_PROXY", ""),
+        os.environ.get("https_proxy", ""),
+        os.environ.get("http_proxy", ""),
+    )
+    uses_plain_socks5 = any(v.strip().lower().startswith("socks5://") for v in proxy_envs)
+    if not uses_plain_socks5 or getattr(socket, "_spider91_ipv4_first", False):
+        return
+
+    original_getaddrinfo = socket.getaddrinfo
+
+    def getaddrinfo_ipv4_first(*args, **kwargs):
+        infos = original_getaddrinfo(*args, **kwargs)
+        return sorted(infos, key=lambda info: 0 if info[0] == socket.AF_INET else 1)
+
+    socket.getaddrinfo = getaddrinfo_ipv4_first
+    socket._spider91_ipv4_first = True
+
 # ===================== 配置区域 =====================
 BASE_URL = "https://www.91porn.com/v.php"
 LIST_PARAMS = {
@@ -757,13 +780,15 @@ def main():
                              "日志改走 stderr。配合 backend 边读边下载使用。")
 
     args, _ = parser.parse_known_args()
+    cli_out = sys.stderr if args.stream_output else sys.stdout
+    prefer_ipv4_for_plain_socks5_proxy()
 
     print("""
 ================================================
     91porn 视频爬虫启动中...
 ================================================
 按 Ctrl+C 可随时中断并保存进度
-""")
+""", file=cli_out)
 
     # 加载已知 ID（来自 backend 的 catalog 已入库列表；兼容旧参数名）
     seen_viewkeys = []
@@ -775,9 +800,9 @@ def main():
                     if line:
                         seen_viewkeys.append(line)
         except FileNotFoundError:
-            print(f"警告: --seen-viewkeys-file 不存在: {args.seen_viewkeys_file}")
+            print(f"警告: --seen-viewkeys-file 不存在: {args.seen_viewkeys_file}", file=cli_out)
         except Exception as e:
-            print(f"警告: 读取 --seen-viewkeys-file 失败: {e}")
+            print(f"警告: 读取 --seen-viewkeys-file 失败: {e}", file=cli_out)
 
     # 决定运行模式
     if args.target_new is not None:

Dockerfile

@@ -36,6 +36,7 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
     python3-bs4 \
     python3-lxml \
     python3-requests \
+    python3-socks \
     tar \
     tzdata \
     && rm -rf /var/lib/apt/lists/*

README.md

@@ -20,7 +20,7 @@
 ## 功能特性
 
 - **多后端支持** — 兼容 115 云盘、PikPak 云盘、OneDrive、Google Drive 和本地存储
-- **低带宽播放** — 支持 302 的云盘可直连播放；Google Drive 等需鉴权直链的来源走后端代理
+- **低带宽播放** — 115 云盘、PikPak 云盘、OneDrive 都支持302模式，在线播放视频时，不占用服务器带宽，播放体验不受服务器带宽影响；Google Drive 不支持302模式，走服务器中转，观看体验会受服务器带宽影响
 - **封面 & 预览片段** — 自动为每个视频生成封面图和预览片段，首页快速选片
 - **91 爬虫** — 内置爬虫，支持抓取 91 本月最热视频
 - **双主题** — 黑黄经典主题 / 粉白清新主题，随时切换

backend/cmd/server/main.go

@@ -182,6 +182,9 @@ func main() {
 		OnRegenFailedThumbnails: func(driveID string) {
 			go app.regenFailedThumbnails(ctx, driveID)
 		},
+		OnRegenFailedFingerprints: func(driveID string) {
+			go app.regenFailedFingerprints(ctx, driveID)
+		},
 		GetDriveGenerationStatuses: func() map[string]api.DriveGenerationStatuses {
 			return app.driveGenerationStatuses()
 		},
@@ -854,13 +857,14 @@ func (a *App) attachSpider91Crawler(d *catalog.Drive, drv *spider91.Driver) {
 	a.spider91Crawlers[driveID] = c
 	a.mu.Unlock()
 
-	// 确保 "91porn" 系统标签存在，并把已入库的 spider91 视频按 author 字段
-	// 匹配补打这个标签（CreateTagAndClassify 内部对所有视频走一遍 classify）。
-	// 重复调用是幂等的：tags 用 INSERT OR IGNORE，video_tags 也是 INSERT OR IGNORE。
+	// 确保 "91porn" 系统标签存在，并按 spider91 来源前缀给历史视频补打。
+	// 不能只靠文本匹配：老版本入库的视频可能没有 author/tags 字段，但 id 前缀
+	// "spider91-<driveID>-" 会一直保留，即使后续迁移到 PikPak/115 也不变。
 	bgCtx, cancel := context.WithTimeout(context.Background(), 30*time.Second)
 	go func() {
 		defer cancel()
-		if _, err := a.cat.CreateTagAndClassify(bgCtx, spider91.DefaultTag, nil, "system"); err != nil {
+		prefix := "spider91-" + driveID + "-"
+		if _, err := a.cat.EnsureTagForVideoIDPrefix(bgCtx, prefix, spider91.DefaultTag, nil, "system"); err != nil {
 			log.Printf("[spider91] ensure %q tag: %v", spider91.DefaultTag, err)
 		}
 	}()
@@ -1551,6 +1555,42 @@ func (a *App) regenFailedThumbnails(ctx context.Context, driveID string) {
 	log.Printf("[thumb] enqueued failed thumbnails for regen drive=%s queued=%d", driveID, queued)
 }
 
+func (a *App) regenFailedFingerprints(ctx context.Context, driveID string) {
+	items, err := a.cat.ListVideosByFingerprintStatus(ctx, driveID, "failed", 0)
+	if err != nil {
+		log.Printf("[fingerprint] list failed videos for regen drive=%s: %v", driveID, err)
+		return
+	}
+	a.mu.Lock()
+	fingerprintWorker := a.fingerprintWorkers[driveID]
+	a.mu.Unlock()
+	if fingerprintWorker == nil {
+		log.Printf("[fingerprint] regen failed drive=%s skipped: fingerprint worker not found", driveID)
+		return
+	}
+	log.Printf("[fingerprint] enqueue failed videos for regen drive=%s count=%d", driveID, len(items))
+	queued := 0
+	for _, v := range items {
+		if err := ctx.Err(); err != nil {
+			log.Printf("[fingerprint] enqueue failed canceled drive=%s queued=%d: %v", driveID, queued, err)
+			return
+		}
+		if err := a.cat.UpdateVideoFingerprint(ctx, v.ID, "", "pending", ""); err != nil {
+			log.Printf("[fingerprint] reset failed video %s drive=%s: %v", v.ID, driveID, err)
+			continue
+		}
+		v.SampledSHA256 = ""
+		v.FingerprintStatus = "pending"
+		v.FingerprintError = ""
+		if !fingerprintWorker.EnqueueBlocking(ctx, v) {
+			log.Printf("[fingerprint] enqueue failed canceled drive=%s queued=%d", driveID, queued)
+			return
+		}
+		queued++
+	}
+	log.Printf("[fingerprint] enqueued failed videos for regen drive=%s queued=%d", driveID, queued)
+}
+
 // listScanTargetIDs 返回 nightly Phase 1 应扫描的所有 drive ID
 // （非 spider91、非 localupload）。它直接读 catalog，而不是 registry，这样
 // 进程刚启动、云盘还在后台挂载时，nightly 也不会漏掉配置过的 drive。

backend/go.mod

@@ -7,15 +7,17 @@ toolchain go1.23.4
 require (
 	github.com/OpenListTeam/wopan-sdk-go v0.2.0
 	github.com/SheltonZhu/115driver v1.3.2
+	github.com/aliyun/aliyun-oss-go-sdk v3.0.2+incompatible
 	github.com/go-chi/chi/v5 v5.1.0
 	github.com/go-resty/resty/v2 v2.14.0
+	golang.org/x/net v0.27.0
+	golang.org/x/sys v0.30.0
 	gopkg.in/yaml.v3 v3.0.1
 	modernc.org/sqlite v1.33.1
 )
 
 require (
 	github.com/aead/ecdh v0.2.0 // indirect
-	github.com/aliyun/aliyun-oss-go-sdk v3.0.2+incompatible // indirect
 	github.com/andreburgaud/crypt2go v1.1.0 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/google/uuid v1.6.0 // indirect
@@ -28,8 +30,6 @@ require (
 	github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
 	github.com/skip2/go-qrcode v0.0.0-20200617195104-da1b6568686e // indirect
 	golang.org/x/crypto v0.25.0 // indirect
-	golang.org/x/net v0.27.0 // indirect
-	golang.org/x/sys v0.30.0 // indirect
 	golang.org/x/time v0.8.0 // indirect
 	modernc.org/gc/v3 v3.0.0-20240107210532-573471604cb6 // indirect
 	modernc.org/libc v1.55.3 // indirect

backend/internal/api/admin.go

@@ -7,6 +7,7 @@ import (
 	"errors"
 	"fmt"
 	"net/http"
+	"net/url"
 	"os"
 	"strconv"
 	"strings"
@@ -46,6 +47,7 @@ type AdminServer struct {
 	OnRegenAllPreviews         func()
 	OnRegenFailedPreviews      func(driveID string)
 	OnRegenFailedThumbnails    func(driveID string)
+	OnRegenFailedFingerprints  func(driveID string)
 	GetDriveGenerationStatuses func() map[string]DriveGenerationStatuses
 	// OnTeaserEnabledChanged 在 per-drive teaser 开关被切换后调用。
 	// enabled=true 时上层应该重新把 pending teaser 入队（类似旧的全局开关从关到开）；
@@ -121,6 +123,7 @@ func (a *AdminServer) Register(r chi.Router) {
 			r.Get("/drives/{id}/dirtree", a.handleListDriveDirTree)
 			r.Post("/drives/{id}/previews/failed/regenerate", a.handleRegenFailedPreviews)
 			r.Post("/drives/{id}/thumbnails/failed/regenerate", a.handleRegenFailedThumbnails)
+			r.Post("/drives/{id}/fingerprints/failed/regenerate", a.handleRegenFailedFingerprints)
 
 			// 视频
 			r.Get("/videos", a.handleAdminListVideos)
@@ -393,6 +396,7 @@ func (a *AdminServer) handleListDrives(w http.ResponseWriter, r *http.Request) {
 		SkipDirIDs []string `json:"skipDirIds"`
 		// LastCrawlAt 是 spider91 上次成功爬取的 unix 秒（来自 credentials.last_crawl_at）。
 		// 其它 kind 留 0；前端用它显示"上次抓取: N 小时前"。
+		Spider91Proxy                 string           `json:"spider91Proxy,omitempty"`
 		LastCrawlAt                   int64            `json:"lastCrawlAt,omitempty"`
 		ThumbnailGenerationStatus     GenerationStatus `json:"thumbnailGenerationStatus"`
 		PreviewGenerationStatus       GenerationStatus `json:"previewGenerationStatus"`
@@ -451,6 +455,7 @@ func (a *AdminServer) handleListDrives(w http.ResponseWriter, r *http.Request) {
 			HasCredential:                 hasCred,
 			TeaserEnabled:                 d.TeaserEnabled,
 			SkipDirIDs:                    append([]string{}, d.SkipDirIDs...),
+			Spider91Proxy:                 spider91ProxyForDrive(d),
 			LastCrawlAt:                   lastCrawlAt,
 			ThumbnailGenerationStatus:     generation.Thumbnail,
 			PreviewGenerationStatus:       generation.Preview,
@@ -503,7 +508,14 @@ func (a *AdminServer) handleUpsertDrive(w http.ResponseWriter, r *http.Request)
 	if existingDrive, err := a.Catalog.GetDrive(r.Context(), body.ID); err == nil {
 		existing = existingDrive
 	}
-	if len(body.Credentials) == 0 && existing != nil && len(existing.Credentials) > 0 {
+	if body.Kind == "spider91" {
+		credentials, err := mergeSpider91Credentials(existing, body.Credentials)
+		if err != nil {
+			http.Error(w, err.Error(), http.StatusBadRequest)
+			return
+		}
+		body.Credentials = credentials
+	} else if len(body.Credentials) == 0 && existing != nil && len(existing.Credentials) > 0 {
 		body.Credentials = existing.Credentials
 	}
 
@@ -552,6 +564,58 @@ func (a *AdminServer) handleUpsertDrive(w http.ResponseWriter, r *http.Request)
 	writeJSON(w, http.StatusOK, map[string]any{"ok": true})
 }
 
+func spider91ProxyForDrive(d *catalog.Drive) string {
+	if d == nil || d.Kind != "spider91" || d.Credentials == nil {
+		return ""
+	}
+	return strings.TrimSpace(d.Credentials["proxy"])
+}
+
+func mergeSpider91Credentials(existing *catalog.Drive, incoming map[string]string) (map[string]string, error) {
+	merged := map[string]string{}
+	if existing != nil {
+		for k, v := range existing.Credentials {
+			merged[k] = v
+		}
+	}
+	for k, v := range incoming {
+		if strings.TrimSpace(k) == "" {
+			continue
+		}
+		if k == "proxy" {
+			proxy, err := normalizeSpider91ProxyURL(v)
+			if err != nil {
+				return nil, err
+			}
+			if proxy == "" {
+				delete(merged, "proxy")
+			} else {
+				merged["proxy"] = proxy
+			}
+			continue
+		}
+		merged[k] = v
+	}
+	return merged, nil
+}
+
+func normalizeSpider91ProxyURL(raw string) (string, error) {
+	proxy := strings.TrimSpace(raw)
+	if proxy == "" {
+		return "", nil
+	}
+	u, err := url.Parse(proxy)
+	if err != nil || u.Scheme == "" || u.Host == "" {
+		return "", fmt.Errorf("91Spider 代理地址格式无效，请填写类似 http://127.0.0.1:7890 的地址")
+	}
+	switch strings.ToLower(u.Scheme) {
+	case "http", "https", "socks5", "socks5h":
+		return proxy, nil
+	default:
+		return "", fmt.Errorf("91Spider 代理地址仅支持 http://、https://、socks5:// 或 socks5h://")
+	}
+}
+
 func (a *AdminServer) handleDeleteDrive(w http.ResponseWriter, r *http.Request) {
 	id := chi.URLParam(r, "id")
 	if err := a.Catalog.DeleteDrive(r.Context(), id); err != nil {
@@ -907,6 +971,16 @@ func (a *AdminServer) handleRegenFailedThumbnails(w http.ResponseWriter, r *http
 	writeJSON(w, http.StatusAccepted, map[string]any{"ok": true})
 }
 
+// handleRegenFailedFingerprints triggers regeneration for all failed sampled
+// fingerprints on a drive. It mirrors the failed teaser/thumbnail retry endpoints.
+func (a *AdminServer) handleRegenFailedFingerprints(w http.ResponseWriter, r *http.Request) {
+	id := chi.URLParam(r, "id")
+	if a.OnRegenFailedFingerprints != nil {
+		a.OnRegenFailedFingerprints(id)
+	}
+	writeJSON(w, http.StatusAccepted, map[string]any{"ok": true})
+}
+
 // ---------- Settings ----------
 
 // settingsDTO 是 GET/PUT /admin/api/settings 的入参/出参。

backend/internal/api/admin_test.go

@@ -439,6 +439,190 @@ func TestHandleUpsertDriveReplacesExistingCredentialsWhenProvided(t *testing.T)
 	}
 }
 
+func TestHandleUpsertSpider91ProxyPreservesRuntimeCredentials(t *testing.T) {
+	ctx := context.Background()
+	cat, err := catalog.Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	if err := cat.UpsertDrive(ctx, &catalog.Drive{
+		ID:     "spider91-main",
+		Kind:   "spider91",
+		Name:   "91 Spider",
+		RootID: "/",
+		Credentials: map[string]string{
+			"last_crawl_at": "1800000000",
+			"proxy":         "http://old-proxy.local:7890",
+			"script_path":   "/opt/video-site-91/91VideoSpider/spider_91porn.py",
+		},
+		Status: "ok",
+	}); err != nil {
+		t.Fatalf("seed drive: %v", err)
+	}
+
+	req := httptest.NewRequest(http.MethodPost, "/admin/api/drives", strings.NewReader(`{
+		"id": "spider91-main",
+		"kind": "spider91",
+		"name": "91 Spider",
+		"rootId": "/",
+		"credentials": {"proxy": " socks5h://proxy-user:proxy-pass@127.0.0.1:7891 "}
+	}`))
+	rr := httptest.NewRecorder()
+	(&AdminServer{Catalog: cat}).handleUpsertDrive(rr, req)
+	if rr.Code != http.StatusOK {
+		t.Fatalf("status = %d, body = %s", rr.Code, rr.Body.String())
+	}
+
+	got, err := cat.GetDrive(ctx, "spider91-main")
+	if err != nil {
+		t.Fatalf("get drive: %v", err)
+	}
+	if got.Credentials["proxy"] != "socks5h://proxy-user:proxy-pass@127.0.0.1:7891" {
+		t.Fatalf("proxy = %q, want trimmed new proxy", got.Credentials["proxy"])
+	}
+	if got.Credentials["last_crawl_at"] != "1800000000" {
+		t.Fatalf("last_crawl_at = %q, want preserved", got.Credentials["last_crawl_at"])
+	}
+	if got.Credentials["script_path"] == "" {
+		t.Fatalf("script_path should be preserved")
+	}
+
+	req = httptest.NewRequest(http.MethodPost, "/admin/api/drives", strings.NewReader(`{
+		"id": "spider91-main",
+		"kind": "spider91",
+		"name": "91 Spider",
+		"rootId": "/",
+		"credentials": {"proxy": " "}
+	}`))
+	rr = httptest.NewRecorder()
+	(&AdminServer{Catalog: cat}).handleUpsertDrive(rr, req)
+	if rr.Code != http.StatusOK {
+		t.Fatalf("clear status = %d, body = %s", rr.Code, rr.Body.String())
+	}
+
+	got, err = cat.GetDrive(ctx, "spider91-main")
+	if err != nil {
+		t.Fatalf("get cleared drive: %v", err)
+	}
+	if _, ok := got.Credentials["proxy"]; ok {
+		t.Fatalf("proxy should be removed after empty save, got %q", got.Credentials["proxy"])
+	}
+	if got.Credentials["last_crawl_at"] != "1800000000" {
+		t.Fatalf("last_crawl_at after clear = %q, want preserved", got.Credentials["last_crawl_at"])
+	}
+}
+
+func TestHandleUpsertSpider91RejectsUnsupportedProxyScheme(t *testing.T) {
+	cat, err := catalog.Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	req := httptest.NewRequest(http.MethodPost, "/admin/api/drives", strings.NewReader(`{
+		"id": "spider91-main",
+		"kind": "spider91",
+		"name": "91 Spider",
+		"rootId": "/",
+		"credentials": {"proxy": "ftp://127.0.0.1:21"}
+	}`))
+	rr := httptest.NewRecorder()
+	(&AdminServer{Catalog: cat}).handleUpsertDrive(rr, req)
+
+	if rr.Code != http.StatusBadRequest {
+		t.Fatalf("status = %d, want 400; body = %s", rr.Code, rr.Body.String())
+	}
+	if !strings.Contains(rr.Body.String(), "socks5:// 或 socks5h://") {
+		t.Fatalf("body = %q, want supported schemes message", rr.Body.String())
+	}
+}
+
+func TestHandleListDrivesIncludesSpider91Proxy(t *testing.T) {
+	ctx := context.Background()
+	cat, err := catalog.Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	for _, d := range []*catalog.Drive{
+		{
+			ID:     "spider91-main",
+			Kind:   "spider91",
+			Name:   "91 Spider",
+			RootID: "/",
+			Credentials: map[string]string{
+				"last_crawl_at": "1800000000",
+				"proxy":         " http://127.0.0.1:7890 ",
+			},
+			Status: "ok",
+		},
+		{
+			ID:     "onedrive-main",
+			Kind:   "onedrive",
+			Name:   "OneDrive",
+			RootID: "root",
+			Credentials: map[string]string{
+				"proxy": "http://should-not-leak.local:7890",
+			},
+			Status: "ok",
+		},
+	} {
+		if err := cat.UpsertDrive(ctx, d); err != nil {
+			t.Fatalf("seed drive %s: %v", d.ID, err)
+		}
+	}
+
+	req := httptest.NewRequest(http.MethodGet, "/admin/api/drives", nil)
+	rr := httptest.NewRecorder()
+	(&AdminServer{Catalog: cat}).handleListDrives(rr, req)
+	if rr.Code != http.StatusOK {
+		t.Fatalf("status = %d, body = %s", rr.Code, rr.Body.String())
+	}
+
+	var got []struct {
+		ID            string `json:"id"`
+		Spider91Proxy string `json:"spider91Proxy"`
+		LastCrawlAt   int64  `json:"lastCrawlAt"`
+	}
+	if err := json.NewDecoder(rr.Body).Decode(&got); err != nil {
+		t.Fatalf("decode: %v", err)
+	}
+	byID := map[string]struct {
+		Spider91Proxy string
+		LastCrawlAt   int64
+	}{}
+	for _, d := range got {
+		byID[d.ID] = struct {
+			Spider91Proxy string
+			LastCrawlAt   int64
+		}{Spider91Proxy: d.Spider91Proxy, LastCrawlAt: d.LastCrawlAt}
+	}
+	if byID["spider91-main"].Spider91Proxy != "http://127.0.0.1:7890" {
+		t.Fatalf("spider91 proxy = %q, want trimmed proxy", byID["spider91-main"].Spider91Proxy)
+	}
+	if byID["spider91-main"].LastCrawlAt != 1800000000 {
+		t.Fatalf("lastCrawlAt = %d, want 1800000000", byID["spider91-main"].LastCrawlAt)
+	}
+	if byID["onedrive-main"].Spider91Proxy != "" {
+		t.Fatalf("onedrive spider91Proxy = %q, want empty", byID["onedrive-main"].Spider91Proxy)
+	}
+}
+
 func TestHandleListDrivesIncludesTeaserCounts(t *testing.T) {
 	ctx := context.Background()
 	cat, err := catalog.Open(t.TempDir() + "/catalog.db")
@@ -598,6 +782,28 @@ func TestHandleListDrivesIncludesTeaserCounts(t *testing.T) {
 	}
 }
 
+func TestHandleRegenFailedFingerprintsInvokesHook(t *testing.T) {
+	called := ""
+	req := httptest.NewRequest(http.MethodPost, "/admin/api/drives/drive-one/fingerprints/failed/regenerate", nil)
+	rctx := chi.NewRouteContext()
+	rctx.URLParams.Add("id", "drive-one")
+	req = req.WithContext(context.WithValue(req.Context(), chi.RouteCtxKey, rctx))
+	rr := httptest.NewRecorder()
+
+	(&AdminServer{
+		OnRegenFailedFingerprints: func(driveID string) {
+			called = driveID
+		},
+	}).handleRegenFailedFingerprints(rr, req)
+
+	if rr.Code != http.StatusAccepted {
+		t.Fatalf("status = %d, body = %s", rr.Code, rr.Body.String())
+	}
+	if called != "drive-one" {
+		t.Fatalf("called drive = %q, want drive-one", called)
+	}
+}
+
 func TestHandleDriveStorageReportsLocalMediaUsage(t *testing.T) {
 	ctx := context.Background()
 	root := t.TempDir()

backend/internal/api/api.go

@@ -156,41 +156,60 @@ func (s *Server) handleGetTheme(w http.ResponseWriter, r *http.Request) {
 }
 
 func (s *Server) handleHome(w http.ResponseWriter, r *http.Request) {
-	// 首页优先展示封面已经生成好的视频，避免新盘扫盘时大量黑封面占满首页。
-	// 候选仍按发布时间覆盖最近 200 个，随后随机洗牌；封面不足时再用普通可见视频补齐。
-	const candidatePool = 200
-	readyItems, _, err := s.Catalog.ListVideos(r.Context(), catalog.ListParams{
-		Sort: "latest", Page: 1, PageSize: candidatePool, ThumbnailReadyOnly: true,
-	})
+	// 首页优先从全量已有封面的视频里随机抽取，避免只在最近一小段候选中反复出现。
+	excludeIDs := parseVideoIDQuery(r, "exclude", 120)
+	items, err := s.Catalog.RandomVideosWithReadyThumbnailsExcluding(r.Context(), excludeIDs, homePageSize)
 	if err != nil {
 		writeErr(w, http.StatusInternalServerError, err)
 		return
 	}
-	rand.Shuffle(len(readyItems), func(i, j int) {
-		readyItems[i], readyItems[j] = readyItems[j], readyItems[i]
-	})
-
-	items := appendUniqueVideos(nil, readyItems, homePageSize)
-	if len(items) > homePageSize {
-		items = items[:homePageSize]
-	}
 	if len(items) < homePageSize {
-		fallback, _, err := s.Catalog.ListVideos(r.Context(), catalog.ListParams{
-			Sort: "latest", Page: 1, PageSize: candidatePool,
-		})
+		fallbackExclude := append([]string{}, excludeIDs...)
+		for _, item := range items {
+			if item != nil {
+				fallbackExclude = append(fallbackExclude, item.ID)
+			}
+		}
+		fallback, err := s.Catalog.RandomVideosExcluding(r.Context(), fallbackExclude, homePageSize-len(items))
 		if err != nil {
 			writeErr(w, http.StatusInternalServerError, err)
 			return
 		}
-		rand.Shuffle(len(fallback), func(i, j int) {
-			fallback[i], fallback[j] = fallback[j], fallback[i]
-		})
 		items = appendUniqueVideos(items, fallback, homePageSize)
 	}
 	w.Header().Set("Cache-Control", "no-store")
 	writeJSON(w, http.StatusOK, mapVideos(items))
 }
 
+func parseVideoIDQuery(r *http.Request, key string, limit int) []string {
+	if r == nil {
+		return nil
+	}
+	values := r.URL.Query()[key]
+	if len(values) == 0 {
+		return nil
+	}
+	seen := map[string]struct{}{}
+	out := make([]string, 0, len(values))
+	for _, value := range values {
+		for _, id := range strings.Split(value, ",") {
+			id = strings.TrimSpace(id)
+			if id == "" {
+				continue
+			}
+			if _, ok := seen[id]; ok {
+				continue
+			}
+			seen[id] = struct{}{}
+			out = append(out, id)
+			if limit > 0 && len(out) >= limit {
+				return out
+			}
+		}
+	}
+	return out
+}
+
 func appendUniqueVideos(dst []*catalog.Video, candidates []*catalog.Video, limit int) []*catalog.Video {
 	if len(dst) >= limit {
 		return dst[:limit]

backend/internal/api/api_test.go

@@ -166,6 +166,59 @@ func TestHandleHomePrioritizesVideosWithReadyThumbnails(t *testing.T) {
 	}
 }
 
+func TestHandleHomeExcludesRecentlyShownVideos(t *testing.T) {
+	ctx := context.Background()
+	cat, err := catalog.Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	now := time.Now()
+	for i := 0; i < homePageSize+4; i++ {
+		id := "ready-video-" + strconv.Itoa(i)
+		if err := cat.UpsertVideo(ctx, &catalog.Video{
+			ID:           id,
+			DriveID:      "drive",
+			FileID:       id,
+			Title:        id,
+			ThumbnailURL: "https://thumb.example/" + id + ".jpg",
+			PublishedAt:  now.Add(time.Duration(i) * time.Minute),
+			CreatedAt:    now.Add(time.Duration(i) * time.Minute),
+			UpdatedAt:    now.Add(time.Duration(i) * time.Minute),
+		}); err != nil {
+			t.Fatalf("seed ready video %s: %v", id, err)
+		}
+	}
+
+	rr := httptest.NewRecorder()
+	req := httptest.NewRequest(http.MethodGet, "/api/home?exclude=ready-video-0&exclude=ready-video-1", nil)
+	(&Server{Catalog: cat}).handleHome(rr, req)
+
+	if rr.Code != http.StatusOK {
+		t.Fatalf("status = %d, body = %s", rr.Code, rr.Body.String())
+	}
+	var got []VideoDTO
+	if err := json.NewDecoder(rr.Body).Decode(&got); err != nil {
+		t.Fatalf("decode response: %v", err)
+	}
+	if len(got) != homePageSize {
+		t.Fatalf("home items = %d, want %d", len(got), homePageSize)
+	}
+	for _, item := range got {
+		if item.ID == "ready-video-0" || item.ID == "ready-video-1" {
+			t.Fatalf("home returned excluded video %q; items=%#v", item.ID, got)
+		}
+		if !strings.HasPrefix(item.ID, "ready-video-") {
+			t.Fatalf("home returned %q without a ready thumbnail; items=%#v", item.ID, got)
+		}
+	}
+}
+
 func TestHandleListLatestPrefersReadyThumbnails(t *testing.T) {
 	ctx := context.Background()
 	cat, err := catalog.Open(t.TempDir() + "/catalog.db")

backend/internal/catalog/catalog.go

@@ -765,6 +765,37 @@ func (c *Catalog) ListVideosNeedingFingerprint(ctx context.Context, driveID stri
 	return out, rows.Err()
 }
 
+// ListVideosByFingerprintStatus lists visible videos on a drive by fingerprint status.
+// It is used by the admin "retry failed fingerprints" action to reset failed rows
+// back to pending and enqueue them again.
+func (c *Catalog) ListVideosByFingerprintStatus(ctx context.Context, driveID, status string, limit int) ([]*Video, error) {
+	if limit <= 0 {
+		limit = 10000
+	}
+	rows, err := c.db.QueryContext(ctx,
+		`SELECT `+allVideoCols+` FROM videos
+		 WHERE drive_id = ?
+		   AND COALESCE(sampled_sha256, '') = ''
+		   AND COALESCE(fingerprint_status, 'pending') = ?
+		   AND COALESCE(hidden, 0) = 0
+		 ORDER BY created_at ASC, id ASC
+		 LIMIT ?`,
+		driveID, status, limit)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var out []*Video
+	for rows.Next() {
+		v, err := scanVideo(rows)
+		if err != nil {
+			return nil, err
+		}
+		out = append(out, v)
+	}
+	return out, rows.Err()
+}
+
 func (c *Catalog) UpdateVideoFingerprint(ctx context.Context, id, sampledSHA256, status, errText string) error {
 	sampledSHA256 = normalizeContentHash(sampledSHA256)
 	if status == "" {
@@ -822,12 +853,7 @@ func (c *Catalog) ListVideos(ctx context.Context, p ListParams) ([]*Video, int,
 		args = append(args, p.DriveID)
 	}
 	if p.Tag != "" {
-		where = append(where, `EXISTS (
-			SELECT 1
-			FROM video_tags vt
-			JOIN tags t ON t.id = vt.tag_id
-			WHERE vt.video_id = videos.id AND t.label = ? COLLATE NOCASE
-		)`)
+		where = append(where, videoMatchesTagLabelSQL("videos"))
 		args = append(args, p.Tag)
 	}
 	if p.Category != "" && p.Category != "all" {
@@ -906,6 +932,14 @@ func (c *Catalog) CountVisibleVideos(ctx context.Context) (int, error) {
 // 如果剩余可选数量 < limit，就返回所有可选项；调用方负责判断是否需要开新一轮。
 // limit <= 0 时返回 nil, nil。
 func (c *Catalog) RandomVideosExcluding(ctx context.Context, excludeIDs []string, limit int) ([]*Video, error) {
+	return c.randomVideosExcluding(ctx, excludeIDs, limit, false)
+}
+
+func (c *Catalog) RandomVideosWithReadyThumbnailsExcluding(ctx context.Context, excludeIDs []string, limit int) ([]*Video, error) {
+	return c.randomVideosExcluding(ctx, excludeIDs, limit, true)
+}
+
+func (c *Catalog) randomVideosExcluding(ctx context.Context, excludeIDs []string, limit int, thumbnailReadyOnly bool) ([]*Video, error) {
 	if limit <= 0 {
 		return nil, nil
 	}
@@ -914,6 +948,9 @@ func (c *Catalog) RandomVideosExcluding(ctx context.Context, excludeIDs []string
 	args := make([]any, 0, len(cleaned)+1)
 	whereSQL := `WHERE COALESCE(hidden, 0) = 0
 		           AND ` + uniqueVideoWhereSQL
+	if thumbnailReadyOnly {
+		whereSQL += " AND COALESCE(thumbnail_url, '') != ''"
+	}
 	if len(cleaned) > 0 {
 		placeholders := strings.Repeat("?,", len(cleaned))
 		placeholders = placeholders[:len(placeholders)-1]
@@ -1472,10 +1509,7 @@ func normalizeDriveRootID(kind, rootID string) string {
 		}
 		return rootID
 	case "localstorage", "spider91":
-		if rootID == "" {
-			return "/"
-		}
-		return rootID
+		return "/"
 	default:
 		if rootID == "" {
 			return "0"

backend/internal/catalog/drives_test.go

@@ -61,6 +61,7 @@ func TestUpsertDriveDefaultsRootIDByKind(t *testing.T) {
 		{id: "onedrive", kind: "onedrive", want: "root"},
 		{id: "googledrive", kind: "googledrive", want: "root"},
 		{id: "localstorage", kind: "localstorage", want: "/"},
+		{id: "spider91", kind: "spider91", want: "/"},
 	}
 	for _, tc := range cases {
 		if err := cat.UpsertDrive(ctx, &Drive{
@@ -82,3 +83,44 @@ func TestUpsertDriveDefaultsRootIDByKind(t *testing.T) {
 		}
 	}
 }
+
+func TestUpsertDriveIgnoresRootIDForLocalStorageAndSpider91(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	for _, tc := range []struct {
+		id   string
+		kind string
+	}{
+		{id: "localstorage", kind: "localstorage"},
+		{id: "spider91", kind: "spider91"},
+	} {
+		if err := cat.UpsertDrive(ctx, &Drive{
+			ID:         tc.id,
+			Kind:       tc.kind,
+			Name:       tc.kind,
+			RootID:     "manual-root",
+			ScanRootID: "manual-scan-root",
+		}); err != nil {
+			t.Fatalf("upsert %s: %v", tc.kind, err)
+		}
+		got, err := cat.GetDrive(ctx, tc.id)
+		if err != nil {
+			t.Fatalf("get %s: %v", tc.kind, err)
+		}
+		if got.RootID != "/" {
+			t.Fatalf("%s rootId = %q, want /", tc.kind, got.RootID)
+		}
+		if got.ScanRootID != "/" {
+			t.Fatalf("%s scanRootId = %q, want /", tc.kind, got.ScanRootID)
+		}
+	}
+}

backend/internal/catalog/shorts_test.go

@@ -110,6 +110,62 @@ func TestRandomVideosExcluding(t *testing.T) {
 	}
 }
 
+func TestRandomVideosWithReadyThumbnailsExcluding(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() { _ = cat.Close() })
+
+	now := time.Now()
+	for i := 0; i < 4; i++ {
+		id := "ready-" + string(rune('a'+i))
+		if err := cat.UpsertVideo(ctx, &Video{
+			ID:           id,
+			DriveID:      "drive",
+			FileID:       "f-" + id,
+			Title:        id,
+			ThumbnailURL: "/p/thumb/" + id,
+			PublishedAt:  now,
+			CreatedAt:    now,
+			UpdatedAt:    now,
+		}); err != nil {
+			t.Fatalf("seed %s: %v", id, err)
+		}
+	}
+	for i := 0; i < 4; i++ {
+		id := "pending-" + string(rune('a'+i))
+		if err := cat.UpsertVideo(ctx, &Video{
+			ID:          id,
+			DriveID:     "drive",
+			FileID:      "f-" + id,
+			Title:       id,
+			PublishedAt: now,
+			CreatedAt:   now,
+			UpdatedAt:   now,
+		}); err != nil {
+			t.Fatalf("seed %s: %v", id, err)
+		}
+	}
+
+	got, err := cat.RandomVideosWithReadyThumbnailsExcluding(ctx, []string{"ready-a"}, 10)
+	if err != nil {
+		t.Fatalf("random ready excluding: %v", err)
+	}
+	if len(got) != 3 {
+		t.Fatalf("ready random count = %d, want 3", len(got))
+	}
+	for _, v := range got {
+		if v.ID == "ready-a" {
+			t.Fatal("excluded ready video was returned")
+		}
+		if v.ThumbnailURL == "" {
+			t.Fatalf("pending video %q was returned", v.ID)
+		}
+	}
+}
+
 func TestRandomVideosForPreferredVideoChoosesLeastPopulatedTag(t *testing.T) {
 	ctx := context.Background()
 	cat, err := Open(t.TempDir() + "/catalog.db")

backend/internal/catalog/tags.go

@@ -302,7 +302,15 @@ func (c *Catalog) classifySystemTags(ctx context.Context) error {
 }
 
 func (c *Catalog) backfillVideoTags(ctx context.Context) error {
-	rows, err := c.db.QueryContext(ctx, `SELECT id, COALESCE(tags, '[]') FROM videos`)
+	rows, err := c.db.QueryContext(ctx, `
+SELECT id, COALESCE(tags, '[]')
+FROM videos
+WHERE COALESCE(tags, '') NOT IN ('', '[]', 'null')
+  AND NOT EXISTS (
+	SELECT 1
+	  FROM video_tags vt
+	 WHERE vt.video_id = videos.id
+  )`)
 	if err != nil {
 		return err
 	}
@@ -319,11 +327,14 @@ func (c *Catalog) backfillVideoTags(ctx context.Context) error {
 		if len(labels) == 0 {
 			continue
 		}
-		if err := c.addVideoTags(ctx, videoID, labels, "legacy", true); err != nil {
+		added, err := c.addVideoTags(ctx, videoID, labels, "legacy", true)
+		if err != nil {
 			return err
 		}
-		if err := c.syncVideoTagsJSON(ctx, videoID, false); err != nil {
-			return err
+		if added {
+			if err := c.syncVideoTagsJSON(ctx, videoID, false); err != nil {
+				return err
+			}
 		}
 	}
 	return nil
@@ -392,6 +403,57 @@ func (c *Catalog) CreateTagAndClassify(ctx context.Context, label string, aliase
 	return c.classifyTag(ctx, tag)
 }
 
+func (c *Catalog) EnsureTagForVideoIDPrefix(ctx context.Context, prefix, label string, aliases []string, source string) (int, error) {
+	prefix = strings.TrimSpace(prefix)
+	if prefix == "" {
+		return 0, errors.New("video id prefix is required")
+	}
+	tag, err := c.ensureTag(ctx, label, aliases, source)
+	if err != nil {
+		return 0, err
+	}
+	rows, err := c.db.QueryContext(ctx, `
+SELECT v.id
+  FROM videos v
+ WHERE v.id LIKE ? || '%'
+   AND COALESCE(v.tags_manual, 0) = 0
+   AND NOT EXISTS (
+	 SELECT 1
+	   FROM video_tags vt
+	  WHERE vt.video_id = v.id
+	    AND vt.tag_id = ?
+   )
+ ORDER BY v.id ASC`, prefix, tag.ID)
+	if err != nil {
+		return 0, err
+	}
+	var videoIDs []string
+	for rows.Next() {
+		var videoID string
+		if err := rows.Scan(&videoID); err != nil {
+			rows.Close()
+			return 0, err
+		}
+		videoIDs = append(videoIDs, videoID)
+	}
+	if err := rows.Err(); err != nil {
+		rows.Close()
+		return 0, err
+	}
+	if err := rows.Close(); err != nil {
+		return 0, err
+	}
+	for _, videoID := range videoIDs {
+		if err := c.insertVideoTag(ctx, videoID, tag.ID, "auto"); err != nil {
+			return 0, err
+		}
+		if err := c.syncVideoTagsJSON(ctx, videoID, false); err != nil {
+			return 0, err
+		}
+	}
+	return len(videoIDs), nil
+}
+
 func (c *Catalog) DeleteTag(ctx context.Context, tagID int64) (int, error) {
 	tx, err := c.db.BeginTx(ctx, nil)
 	if err != nil {
@@ -453,10 +515,66 @@ func (c *Catalog) DeleteTag(ctx context.Context, tagID int64) (int, error) {
 
 func (c *Catalog) ListTags(ctx context.Context) ([]Tag, error) {
 	rows, err := c.db.QueryContext(ctx, `
-SELECT t.id, t.label, t.aliases, t.source, COUNT(v.id) AS cnt
+WITH tagged_tags AS (
+	SELECT vt.tag_id,
+	       tagged.id,
+	       COALESCE(tagged.content_hash, '') AS content_hash,
+	       COALESCE(tagged.sampled_sha256, '') AS sampled_sha256,
+	       tagged.size_bytes,
+	       COALESCE(tagged.file_name, '') AS file_name
+	  FROM video_tags vt
+	  JOIN videos tagged ON tagged.id = vt.video_id
+	 WHERE COALESCE(tagged.hidden, 0) = 0
+),
+tag_candidates AS (
+	SELECT tag_id, id AS video_id
+	  FROM tagged_tags
+	UNION ALL
+	SELECT tag_id,
+	       (SELECT canonical.id
+	          FROM videos canonical
+	         WHERE tagged_tags.content_hash != ''
+	           AND canonical.content_hash = tagged_tags.content_hash
+	           AND COALESCE(canonical.content_hash, '') != ''
+	         ORDER BY canonical.created_at ASC, canonical.id ASC
+	         LIMIT 1) AS video_id
+	  FROM tagged_tags
+	 WHERE content_hash != ''
+	UNION ALL
+	SELECT tag_id,
+	       (SELECT canonical.id
+	          FROM videos canonical
+	         WHERE tagged_tags.sampled_sha256 != ''
+	           AND tagged_tags.size_bytes > 0
+	           AND canonical.sampled_sha256 = tagged_tags.sampled_sha256
+	           AND canonical.size_bytes = tagged_tags.size_bytes
+	           AND COALESCE(canonical.sampled_sha256, '') != ''
+	           AND canonical.size_bytes > 0
+	         ORDER BY canonical.created_at ASC, canonical.id ASC
+	         LIMIT 1) AS video_id
+	  FROM tagged_tags
+	 WHERE sampled_sha256 != '' AND size_bytes > 0
+	UNION ALL
+	SELECT tag_id,
+	       (SELECT canonical.id
+	          FROM videos canonical
+	         WHERE tagged_tags.file_name != ''
+	           AND tagged_tags.size_bytes > 0
+	           AND canonical.file_name = tagged_tags.file_name
+	           AND canonical.size_bytes = tagged_tags.size_bytes
+	           AND COALESCE(canonical.file_name, '') != ''
+	           AND canonical.size_bytes > 0
+	         ORDER BY canonical.created_at ASC, canonical.id ASC
+	         LIMIT 1) AS video_id
+	  FROM tagged_tags
+	 WHERE file_name != '' AND size_bytes > 0
+)
+SELECT t.id, t.label, t.aliases, t.source, COUNT(DISTINCT videos.id) AS cnt
 FROM tags t
-LEFT JOIN video_tags vt ON vt.tag_id = t.id
-LEFT JOIN videos v ON v.id = vt.video_id AND COALESCE(v.hidden, 0) = 0
+LEFT JOIN tag_candidates tc ON tc.tag_id = t.id AND tc.video_id IS NOT NULL
+LEFT JOIN videos ON videos.id = tc.video_id
+	AND COALESCE(videos.hidden, 0) = 0
+	AND `+uniqueVideoWhereSQL+`
 GROUP BY t.id, t.label, t.aliases, t.source
 ORDER BY cnt DESC, t.label ASC`)
 	if err != nil {
@@ -474,6 +592,66 @@ ORDER BY cnt DESC, t.label ASC`)
 	return out, nil
 }
 
+func videoMatchesTagLabelSQL(videoAlias string) string {
+	return fmt.Sprintf(`%s.id IN (
+			WITH tagged_videos AS (
+				SELECT tagged.id,
+				       COALESCE(tagged.content_hash, '') AS content_hash,
+				       COALESCE(tagged.sampled_sha256, '') AS sampled_sha256,
+				       tagged.size_bytes,
+				       COALESCE(tagged.file_name, '') AS file_name
+				  FROM video_tags vt
+				  JOIN tags tag_filter ON tag_filter.id = vt.tag_id
+				  JOIN videos tagged ON tagged.id = vt.video_id
+				 WHERE tag_filter.label = ? COLLATE NOCASE
+				   AND COALESCE(tagged.hidden, 0) = 0
+			),
+			tag_candidates AS (
+				SELECT id AS video_id
+				  FROM tagged_videos
+				UNION ALL
+				SELECT (SELECT canonical.id
+				          FROM videos canonical
+				         WHERE tagged_videos.content_hash != ''
+				           AND canonical.content_hash = tagged_videos.content_hash
+				           AND COALESCE(canonical.content_hash, '') != ''
+				         ORDER BY canonical.created_at ASC, canonical.id ASC
+				         LIMIT 1) AS video_id
+				  FROM tagged_videos
+				 WHERE content_hash != ''
+				UNION ALL
+				SELECT (SELECT canonical.id
+				          FROM videos canonical
+				         WHERE tagged_videos.sampled_sha256 != ''
+				           AND tagged_videos.size_bytes > 0
+				           AND canonical.sampled_sha256 = tagged_videos.sampled_sha256
+				           AND canonical.size_bytes = tagged_videos.size_bytes
+				           AND COALESCE(canonical.sampled_sha256, '') != ''
+				           AND canonical.size_bytes > 0
+				         ORDER BY canonical.created_at ASC, canonical.id ASC
+				         LIMIT 1) AS video_id
+				  FROM tagged_videos
+				 WHERE sampled_sha256 != '' AND size_bytes > 0
+				UNION ALL
+				SELECT (SELECT canonical.id
+				          FROM videos canonical
+				         WHERE tagged_videos.file_name != ''
+				           AND tagged_videos.size_bytes > 0
+				           AND canonical.file_name = tagged_videos.file_name
+				           AND canonical.size_bytes = tagged_videos.size_bytes
+				           AND COALESCE(canonical.file_name, '') != ''
+				           AND canonical.size_bytes > 0
+				         ORDER BY canonical.created_at ASC, canonical.id ASC
+				         LIMIT 1) AS video_id
+				  FROM tagged_videos
+				 WHERE file_name != '' AND size_bytes > 0
+			)
+			SELECT video_id
+			  FROM tag_candidates
+			 WHERE video_id IS NOT NULL
+		)`, videoAlias)
+}
+
 func (c *Catalog) SetManualVideoTags(ctx context.Context, videoID string, labels []string) error {
 	if _, err := c.GetVideo(ctx, videoID); err != nil {
 		return err
@@ -592,6 +770,10 @@ func (c *Catalog) getTagByLabel(ctx context.Context, label string) (Tag, error)
 }
 
 func (c *Catalog) classifyTag(ctx context.Context, tag Tag) (int, error) {
+	existingIDs, err := c.videoIDSetForTagID(ctx, tag.ID)
+	if err != nil {
+		return 0, err
+	}
 	rows, err := c.db.QueryContext(ctx, `
 SELECT id, title, COALESCE(author, ''), COALESCE(category, ''), COALESCE(tags_manual, 0)
 FROM videos`)
@@ -623,13 +805,14 @@ FROM videos`)
 				continue
 			}
 		}
-		added, err := c.addVideoTag(ctx, videoID, tag.Label, "auto", false)
-		if err != nil {
+		if existingIDs[videoID] {
+			continue
+		}
+		if err := c.insertVideoTag(ctx, videoID, tag.ID, "auto"); err != nil {
 			return 0, err
 		}
-		if added {
-			classified++
-		}
+		existingIDs[videoID] = true
+		classified++
 		if err := c.syncVideoTagsJSON(ctx, videoID, false); err != nil {
 			return 0, err
 		}
@@ -689,17 +872,22 @@ func (c *Catalog) replaceVideoTags(ctx context.Context, videoID string, labels [
 	return c.syncVideoTagsJSON(ctx, videoID, manual)
 }
 
-func (c *Catalog) addVideoTags(ctx context.Context, videoID string, labels []string, source string, createMissing bool) error {
+func (c *Catalog) addVideoTags(ctx context.Context, videoID string, labels []string, source string, createMissing bool) (bool, error) {
 	labels = uniqueStrings(cleanLabels(labels))
 	if source != "manual" {
 		labels = c.filterDeletedTagLabels(ctx, labels)
 	}
+	changed := false
 	for _, label := range labels {
-		if _, err := c.addVideoTag(ctx, videoID, label, source, createMissing); err != nil {
-			return err
+		added, err := c.addVideoTag(ctx, videoID, label, source, createMissing)
+		if err != nil {
+			return false, err
+		}
+		if added {
+			changed = true
 		}
 	}
-	return nil
+	return changed, nil
 }
 
 func (c *Catalog) addVideoTag(ctx context.Context, videoID, label, source string, createMissing bool) (bool, error) {
@@ -729,12 +917,33 @@ func (c *Catalog) addVideoTag(ctx context.Context, videoID, label, source string
 	return n > 0, nil
 }
 
+func (c *Catalog) insertVideoTag(ctx context.Context, videoID string, tagID int64, source string) error {
+	_, err := c.db.ExecContext(ctx,
+		`INSERT OR IGNORE INTO video_tags (video_id, tag_id, source, created_at) VALUES (?, ?, ?, ?)`,
+		videoID, tagID, source, time.Now().UnixMilli())
+	return err
+}
+
 func (c *Catalog) addCollectionTagToVideos(ctx context.Context, category string) error {
 	return c.addTagToVideosByCategory(ctx, category, category, "auto")
 }
 
 func (c *Catalog) addTagToVideosByCategory(ctx context.Context, category, label, source string) error {
-	rows, err := c.db.QueryContext(ctx, `SELECT id FROM videos WHERE category = ? AND COALESCE(tags_manual, 0) = 0`, category)
+	tag, err := c.getTagByLabel(ctx, label)
+	if err != nil {
+		return err
+	}
+	rows, err := c.db.QueryContext(ctx, `
+SELECT v.id
+  FROM videos v
+ WHERE v.category = ?
+   AND COALESCE(v.tags_manual, 0) = 0
+   AND NOT EXISTS (
+	 SELECT 1
+	   FROM video_tags vt
+	  WHERE vt.video_id = v.id
+	    AND vt.tag_id = ?
+   )`, category, tag.ID)
 	if err != nil {
 		return err
 	}
@@ -753,7 +962,7 @@ func (c *Catalog) addTagToVideosByCategory(ctx context.Context, category, label,
 		return err
 	}
 	for _, videoID := range videoIDs {
-		if _, err := c.addVideoTag(ctx, videoID, label, source, false); err != nil {
+		if err := c.insertVideoTag(ctx, videoID, tag.ID, source); err != nil {
 			return err
 		}
 		if err := c.syncVideoTagsJSON(ctx, videoID, false); err != nil {
@@ -837,6 +1046,23 @@ func (c *Catalog) videoIDsForTagID(ctx context.Context, tagID int64) ([]string,
 	return videoIDs, rows.Err()
 }
 
+func (c *Catalog) videoIDSetForTagID(ctx context.Context, tagID int64) (map[string]bool, error) {
+	rows, err := c.db.QueryContext(ctx, `SELECT video_id FROM video_tags WHERE tag_id = ?`, tagID)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	out := map[string]bool{}
+	for rows.Next() {
+		var videoID string
+		if err := rows.Scan(&videoID); err != nil {
+			return nil, err
+		}
+		out[videoID] = true
+	}
+	return out, rows.Err()
+}
+
 func (c *Catalog) validateTagsExist(ctx context.Context, labels []string) error {
 	for _, label := range labels {
 		if _, err := c.getTagByLabel(ctx, label); err != nil {

backend/internal/catalog/tags_test.go

@@ -326,6 +326,75 @@ func TestCreateTagAndClassifyRestoresDeletedTag(t *testing.T) {
 	}
 }
 
+func TestEnsureTagForVideoIDPrefixBackfillsSourceTag(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	now := time.Now()
+	for _, seed := range []struct {
+		id     string
+		manual bool
+	}{
+		{id: "spider91-91-spider-1200001"},
+		{id: "spider91-91-spider-1200002", manual: true},
+		{id: "spider91-other-1200003"},
+	} {
+		if err := cat.UpsertVideo(ctx, &Video{
+			ID:          seed.id,
+			DriveID:     "91-spider",
+			FileID:      seed.id + ".mp4",
+			Title:       "legacy title without source text",
+			PublishedAt: now,
+			CreatedAt:   now,
+			UpdatedAt:   now,
+		}); err != nil {
+			t.Fatalf("seed %s: %v", seed.id, err)
+		}
+		if seed.manual {
+			if err := cat.SetManualVideoTags(ctx, seed.id, nil); err != nil {
+				t.Fatalf("mark %s manual: %v", seed.id, err)
+			}
+		}
+	}
+
+	added, err := cat.EnsureTagForVideoIDPrefix(ctx, "spider91-91-spider-", "91porn", nil, "system")
+	if err != nil {
+		t.Fatalf("ensure prefix tag: %v", err)
+	}
+	if added != 1 {
+		t.Fatalf("added = %d, want 1", added)
+	}
+	got, err := cat.GetVideo(ctx, "spider91-91-spider-1200001")
+	if err != nil {
+		t.Fatalf("get tagged video: %v", err)
+	}
+	if !sameStrings(got.Tags, []string{"91porn"}) {
+		t.Fatalf("tagged video tags = %#v, want 91porn", got.Tags)
+	}
+	manual, err := cat.GetVideo(ctx, "spider91-91-spider-1200002")
+	if err != nil {
+		t.Fatalf("get manual video: %v", err)
+	}
+	if len(manual.Tags) != 0 {
+		t.Fatalf("manual video tags = %#v, want unchanged", manual.Tags)
+	}
+	other, err := cat.GetVideo(ctx, "spider91-other-1200003")
+	if err != nil {
+		t.Fatalf("get other prefix video: %v", err)
+	}
+	if len(other.Tags) != 0 {
+		t.Fatalf("other prefix video tags = %#v, want unchanged", other.Tags)
+	}
+}
+
 func TestDeleteTagRejectsSystemTags(t *testing.T) {
 	ctx := context.Background()
 	cat, err := Open(t.TempDir() + "/catalog.db")
@@ -398,6 +467,84 @@ VALUES
 	}
 }
 
+func TestMigrateDoesNotRewriteAlreadySyncedVideoTags(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	now := time.Now()
+	for _, id := range []string{"video-1", "video-2", "video-3"} {
+		if err := cat.UpsertVideo(ctx, &Video{
+			ID:          id,
+			DriveID:     "drive",
+			FileID:      id,
+			Title:       "巨乳后入合集",
+			Category:    "Better Call Saul S03",
+			PublishedAt: now,
+			CreatedAt:   now,
+			UpdatedAt:   now,
+		}); err != nil {
+			t.Fatalf("seed %s: %v", id, err)
+		}
+	}
+	if err := cat.migrate(ctx); err != nil {
+		t.Fatalf("first migrate: %v", err)
+	}
+
+	before := videoUpdatedAtByID(t, ctx, cat, "video-1", "video-2", "video-3")
+	time.Sleep(5 * time.Millisecond)
+	if err := cat.migrate(ctx); err != nil {
+		t.Fatalf("second migrate: %v", err)
+	}
+	after := videoUpdatedAtByID(t, ctx, cat, "video-1", "video-2", "video-3")
+	for id, want := range before {
+		if got := after[id]; got != want {
+			t.Fatalf("%s updated_at changed on no-op migrate: got %d, want %d", id, got, want)
+		}
+	}
+}
+
+func TestMigrateBackfillsLegacyTagsWithoutRelations(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	now := time.Now().UnixMilli()
+	if _, err := cat.db.ExecContext(ctx, `
+INSERT INTO videos (id, drive_id, file_id, title, tags, tags_manual, published_at, created_at, updated_at)
+VALUES ('legacy-video', 'drive', 'file-legacy', 'legacy title', '["legacy-tag"]', 0, ?, ?, ?)`,
+		now, now, now); err != nil {
+		t.Fatalf("seed legacy video: %v", err)
+	}
+	if err := cat.migrate(ctx); err != nil {
+		t.Fatalf("migrate: %v", err)
+	}
+
+	tag := mustTagByLabel(t, ctx, cat, "legacy-tag")
+	var count int
+	if err := cat.db.QueryRowContext(ctx,
+		`SELECT COUNT(*) FROM video_tags WHERE video_id = 'legacy-video' AND tag_id = ?`, tag.ID).Scan(&count); err != nil {
+		t.Fatalf("count video tag: %v", err)
+	}
+	if count != 1 {
+		t.Fatalf("legacy video tag relation count = %d, want 1", count)
+	}
+}
+
 func TestOpenMigratesLegacyVideosWithoutFileName(t *testing.T) {
 	path := t.TempDir() + "/catalog.db"
 	db, err := sql.Open("sqlite", path)
@@ -859,6 +1006,98 @@ func TestListVideosHidesDuplicateContentHashes(t *testing.T) {
 	}
 }
 
+func TestTagFilterMatchesCanonicalDuplicateVideo(t *testing.T) {
+	ctx := context.Background()
+	cat, err := Open(t.TempDir() + "/catalog.db")
+	if err != nil {
+		t.Fatalf("open catalog: %v", err)
+	}
+	t.Cleanup(func() {
+		if err := cat.Close(); err != nil {
+			t.Fatalf("close catalog: %v", err)
+		}
+	})
+
+	now := time.Now()
+	for _, v := range []*Video{
+		{
+			ID:          "pikpak-canonical",
+			DriveID:     "pikpak",
+			FileID:      "canonical.mp4",
+			Title:       "Canonical",
+			Size:        1024,
+			PublishedAt: now,
+			CreatedAt:   now,
+			UpdatedAt:   now,
+		},
+		{
+			ID:          "spider91-dup-1",
+			DriveID:     "91-spider",
+			FileID:      "dup-1.mp4",
+			Title:       "Spider duplicate 1",
+			Tags:        []string{"91porn"},
+			Size:        1024,
+			PublishedAt: now.Add(time.Second),
+			CreatedAt:   now.Add(time.Second),
+			UpdatedAt:   now.Add(time.Second),
+		},
+		{
+			ID:          "spider91-dup-2",
+			DriveID:     "91-spider",
+			FileID:      "dup-2.mp4",
+			Title:       "Spider duplicate 2",
+			Tags:        []string{"91porn"},
+			Size:        1024,
+			PublishedAt: now.Add(2 * time.Second),
+			CreatedAt:   now.Add(2 * time.Second),
+			UpdatedAt:   now.Add(2 * time.Second),
+		},
+		{
+			ID:          "spider91-visible",
+			DriveID:     "91-spider",
+			FileID:      "visible.mp4",
+			Title:       "Spider visible",
+			Tags:        []string{"91porn"},
+			Size:        2048,
+			PublishedAt: now.Add(3 * time.Second),
+			CreatedAt:   now.Add(3 * time.Second),
+			UpdatedAt:   now.Add(3 * time.Second),
+		},
+	} {
+		if err := cat.UpsertVideo(ctx, v); err != nil {
+			t.Fatalf("seed %s: %v", v.ID, err)
+		}
+	}
+	for _, id := range []string{"pikpak-canonical", "spider91-dup-1", "spider91-dup-2"} {
+		if err := cat.UpdateVideoFingerprint(ctx, id, "same-sampled-sha256", "ready", ""); err != nil {
+			t.Fatalf("fingerprint %s: %v", id, err)
+		}
+	}
+	if err := cat.UpdateVideoFingerprint(ctx, "spider91-visible", "unique-sampled-sha256", "ready", ""); err != nil {
+		t.Fatalf("fingerprint visible: %v", err)
+	}
+
+	items, total, err := cat.ListVideos(ctx, ListParams{Tag: "91porn", Page: 1, PageSize: 10})
+	if err != nil {
+		t.Fatalf("list videos by tag: %v", err)
+	}
+	if total != 2 || len(items) != 2 {
+		t.Fatalf("tagged videos total=%d len=%d, want 2", total, len(items))
+	}
+	gotIDs := map[string]bool{}
+	for _, item := range items {
+		gotIDs[item.ID] = true
+	}
+	for _, want := range []string{"pikpak-canonical", "spider91-visible"} {
+		if !gotIDs[want] {
+			t.Fatalf("tagged video ids = %#v, want %s", gotIDs, want)
+		}
+	}
+	if got := mustTagByLabel(t, ctx, cat, "91porn").Count; got != 2 {
+		t.Fatalf("91porn count = %d, want 2 visible canonical videos", got)
+	}
+}
+
 func TestListVideosCanFilterReadyThumbnails(t *testing.T) {
 	ctx := context.Background()
 	cat, err := Open(t.TempDir() + "/catalog.db")
@@ -944,6 +1183,19 @@ func mustTagByLabel(t *testing.T, ctx context.Context, cat *Catalog, label strin
 	return Tag{}
 }
 
+func videoUpdatedAtByID(t *testing.T, ctx context.Context, cat *Catalog, ids ...string) map[string]int64 {
+	t.Helper()
+	out := make(map[string]int64, len(ids))
+	for _, id := range ids {
+		var updatedAt int64
+		if err := cat.db.QueryRowContext(ctx, `SELECT updated_at FROM videos WHERE id = ?`, id).Scan(&updatedAt); err != nil {
+			t.Fatalf("read updated_at for %s: %v", id, err)
+		}
+		out[id] = updatedAt
+	}
+	return out
+}
+
 // 删除 collection 标签的最后一个引用视频后，标签应当自动从 tags 表里消失。
 // user/system 标签不受影响：用户/系统标签的语义由人维护，孤儿状态保留。
 func TestDeleteVideoPrunesOrphanCollectionTag(t *testing.T) {

backend/internal/drives/spider91/crawler.go

@@ -8,6 +8,7 @@ import (
 	"fmt"
 	"io"
 	"log"
+	"net"
 	"net/http"
 	"net/url"
 	"os"
@@ -20,6 +21,7 @@ import (
 	"time"
 
 	"github.com/video-site/backend/internal/catalog"
+	"golang.org/x/net/proxy"
 )
 
 // 默认 author/tag 标签，便于在前端筛选 spider91 来源的视频。
@@ -79,29 +81,123 @@ func NewCrawler(cfg CrawlerConfig) *Crawler {
 		cfg.DownloadTimeout = 30 * time.Minute
 	}
 	if cfg.HTTPClient == nil {
-		// 选 proxy 函数：显式 ProxyURL > 环境变量 > 直连
-		proxyFn := http.ProxyFromEnvironment
-		if strings.TrimSpace(cfg.ProxyURL) != "" {
-			if u, err := url.Parse(cfg.ProxyURL); err == nil {
-				proxyFn = http.ProxyURL(u)
-			} else {
-				log.Printf("[spider91] invalid proxy URL %q, falling back to env: %v", cfg.ProxyURL, err)
-			}
+		transport := &http.Transport{
+			Proxy:                 http.ProxyFromEnvironment,
+			ResponseHeaderTimeout: 60 * time.Second,
+			MaxIdleConns:          10,
+			IdleConnTimeout:       90 * time.Second,
+		}
+		if err := configureExplicitProxy(transport, cfg.ProxyURL); err != nil {
+			log.Printf("[spider91] invalid configured proxy URL, falling back to env: %v", err)
 		}
 		cfg.HTTPClient = &http.Client{
 			// 不限制总下载时长，靠 ctx 控制；只挡 dial / handshake / header
-			Timeout: 0,
-			Transport: &http.Transport{
-				Proxy:                 proxyFn,
-				ResponseHeaderTimeout: 60 * time.Second,
-				MaxIdleConns:          10,
-				IdleConnTimeout:       90 * time.Second,
-			},
+			Timeout:   0,
+			Transport: transport,
 		}
 	}
 	return &Crawler{cfg: cfg}
 }
 
+func configureExplicitProxy(transport *http.Transport, raw string) error {
+	proxyURL := strings.TrimSpace(raw)
+	if proxyURL == "" {
+		return nil
+	}
+	u, err := url.Parse(proxyURL)
+	if err != nil || u.Scheme == "" || u.Host == "" {
+		return fmt.Errorf("invalid proxy URL")
+	}
+	switch strings.ToLower(u.Scheme) {
+	case "http", "https":
+		transport.Proxy = http.ProxyURL(u)
+		transport.DialContext = nil
+		return nil
+	case "socks5", "socks5h":
+		dialContext, err := socksProxyDialContext(u)
+		if err != nil {
+			return err
+		}
+		transport.Proxy = nil
+		transport.DialContext = dialContext
+		return nil
+	default:
+		return fmt.Errorf("unsupported proxy scheme %q", u.Scheme)
+	}
+}
+
+func socksProxyDialContext(proxyURL *url.URL) (func(context.Context, string, string) (net.Conn, error), error) {
+	var auth *proxy.Auth
+	if proxyURL.User != nil {
+		username := proxyURL.User.Username()
+		password, _ := proxyURL.User.Password()
+		auth = &proxy.Auth{User: username, Password: password}
+	}
+	dialer, err := proxy.SOCKS5("tcp", proxyURL.Host, auth, &net.Dialer{Timeout: 60 * time.Second})
+	if err != nil {
+		return nil, err
+	}
+	remoteDNS := strings.EqualFold(proxyURL.Scheme, "socks5h")
+	return func(ctx context.Context, network, addr string) (net.Conn, error) {
+		target := addr
+		if !remoteDNS {
+			resolved, err := resolveSocksTarget(ctx, addr)
+			if err != nil {
+				return nil, err
+			}
+			target = resolved
+		}
+		if ctxDialer, ok := dialer.(proxy.ContextDialer); ok {
+			return ctxDialer.DialContext(ctx, network, target)
+		}
+		type result struct {
+			conn net.Conn
+			err  error
+		}
+		ch := make(chan result, 1)
+		go func() {
+			conn, err := dialer.Dial(network, target)
+			ch <- result{conn: conn, err: err}
+		}()
+		select {
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		case res := <-ch:
+			return res.conn, res.err
+		}
+	}, nil
+}
+
+func resolveSocksTarget(ctx context.Context, addr string) (string, error) {
+	host, port, err := net.SplitHostPort(addr)
+	if err != nil || net.ParseIP(host) != nil {
+		return addr, nil
+	}
+	ips, err := net.DefaultResolver.LookupIPAddr(ctx, host)
+	if err != nil {
+		return "", err
+	}
+	ip := selectSocksTargetIP(ips)
+	if ip == nil {
+		return "", fmt.Errorf("resolve %s: no address", host)
+	}
+	return net.JoinHostPort(ip.String(), port), nil
+}
+
+func selectSocksTargetIP(ips []net.IPAddr) net.IP {
+	for _, addr := range ips {
+		if ip4 := addr.IP.To4(); ip4 != nil {
+			return ip4
+		}
+	}
+	for _, addr := range ips {
+		if addr.IP != nil {
+			return addr.IP
+		}
+	}
+	return nil
+}
+
 // CrawlResult 汇总一次 RunOnce 的结果。
 type CrawlResult struct {
 	// TargetNew 是本次 RunOnce 的目标新增数（来自 drive.Credentials.target_new）。
@@ -324,6 +420,16 @@ func (c *Crawler) startSpiderTargetNew(ctx context.Context, targetNew int, seenP
 	if c.cfg.WorkDir != "" {
 		cmd.Dir = c.cfg.WorkDir
 	}
+	if proxyURL := strings.TrimSpace(c.cfg.ProxyURL); proxyURL != "" {
+		cmd.Env = append(os.Environ(),
+			"HTTP_PROXY="+proxyURL,
+			"HTTPS_PROXY="+proxyURL,
+			"http_proxy="+proxyURL,
+			"https_proxy="+proxyURL,
+			"NO_PROXY=",
+			"no_proxy=",
+		)
+	}
 	stdout, err := cmd.StdoutPipe()
 	if err != nil {
 		return nil, nil, fmt.Errorf("stdout pipe: %w", err)
@@ -427,6 +533,17 @@ func (c *Crawler) processOne(ctx context.Context, videoID string, item spiderVid
 		}
 	}
 
+	title := strings.TrimSpace(item.Title)
+	if title == "" {
+		title = sourceID
+	}
+	tags := []string{DefaultTag}
+	if matched, err := c.cfg.Catalog.MatchTags(ctx, title+" "+DefaultAuthor); err == nil {
+		tags = mergeCatalogTags(tags, matched)
+	} else {
+		log.Printf("[spider91] drive=%s viewkey=%s source_id=%s match tags: %v", c.cfg.Driver.ID(), viewkey, sourceID, err)
+	}
+
 	// 入库
 	now := time.Now()
 	v := &catalog.Video{
@@ -434,9 +551,9 @@ func (c *Crawler) processOne(ctx context.Context, videoID string, item spiderVid
 		DriveID:       c.cfg.Driver.ID(),
 		FileID:        videoFile,
 		FileName:      videoFile,
-		Title:         strings.TrimSpace(item.Title),
+		Title:         title,
 		Author:        DefaultAuthor,
-		Tags:          []string{DefaultTag},
+		Tags:          tags,
 		Ext:           strings.TrimPrefix(videoExt, "."),
 		Quality:       "HD",
 		Size:          videoSize,
@@ -445,9 +562,6 @@ func (c *Crawler) processOne(ctx context.Context, videoID string, item spiderVid
 		CreatedAt:     now,
 		UpdatedAt:     now,
 	}
-	if v.Title == "" {
-		v.Title = sourceID
-	}
 	if thumbReady {
 		// 设了 ThumbnailURL 后 thumb worker 会跳过这条视频，
 		// 不再尝试用 ffmpeg 抽帧（封面已经是网站原图）。
@@ -888,6 +1002,26 @@ func copyFileAtomic(src, dst string) error {
 	return os.Rename(tmp, dst)
 }
 
+func mergeCatalogTags(lists ...[]string) []string {
+	out := []string{}
+	seen := map[string]bool{}
+	for _, list := range lists {
+		for _, tag := range list {
+			tag = strings.TrimSpace(tag)
+			if tag == "" {
+				continue
+			}
+			key := strings.ToLower(tag)
+			if seen[key] {
+				continue
+			}
+			seen[key] = true
+			out = append(out, tag)
+		}
+	}
+	return out
+}
+
 // BuildVideoID 给定 driveID + 91 源视频 ID，按统一规则生成 catalog 中 videos.id。
 // 与 scanner 用法一致：<kind>-<driveID>-<fileID>。
 func BuildVideoID(driveID, sourceID string) string {

backend/internal/drives/spider91/crawler_test.go

@@ -3,6 +3,8 @@ package spider91
 import (
 	"context"
 	"encoding/json"
+	"io"
+	"net"
 	"net/http"
 	"net/http/httptest"
 	"net/url"
@@ -53,7 +55,7 @@ func TestCrawlerRunOnceFullFlow(t *testing.T) {
 	//    同时仍写 --output 文件作归档。
 	videoEntries := []map[string]string{
 		{
-			"title":      "Video One",
+			"title":      "Video One 口交",
 			"thumb_url":  srv.URL + "/thumb/not-120001.jpg",
 			"video_url":  srv.URL + "/videos/120001.mp4",
 			"viewkey":    "vk-001",
@@ -94,6 +96,9 @@ func TestCrawlerRunOnceFullFlow(t *testing.T) {
 	}); err != nil {
 		t.Fatalf("upsert drive: %v", err)
 	}
+	if _, err := cat.CreateTagAndClassify(context.Background(), "Video One", nil, "user"); err != nil {
+		t.Fatalf("create user tag: %v", err)
+	}
 
 	var newVideos []*catalog.Video
 	c := NewCrawler(CrawlerConfig{
@@ -188,6 +193,17 @@ func TestCrawlerRunOnceFullFlow(t *testing.T) {
 		if !hasDefaultTag {
 			t.Fatalf("video %s tags = %v, want contain %q", videoID, v.Tags, DefaultTag)
 		}
+		if sourceID == "120001" {
+			if !containsString(v.Tags, "口交") {
+				t.Fatalf("video %s tags = %v, want contain built-in tag 口交", videoID, v.Tags)
+			}
+			if !containsString(v.Tags, "Video One") {
+				t.Fatalf("video %s tags = %v, want contain user tag Video One", videoID, v.Tags)
+			}
+		}
+		if sourceID == "120002" && (containsString(v.Tags, "口交") || containsString(v.Tags, "Video One")) {
+			t.Fatalf("video %s tags = %v, should not inherit tags from other spider91 videos", videoID, v.Tags)
+		}
 	}
 
 	// 7. 第二次 RunOnce：源视频 ID 已存在 → 全部 skipped，无新文件下载
@@ -233,6 +249,108 @@ func TestCrawlerRunOnceMissingScript(t *testing.T) {
 	}
 }
 
+func TestCrawlerPassesProxyToSpiderProcess(t *testing.T) {
+	if runtime.GOOS == "windows" {
+		t.Skip("shell-based fake script only on unix")
+	}
+
+	tmp := t.TempDir()
+	scriptPath := filepath.Join(tmp, "print_proxy_env.sh")
+	script := `#!/bin/sh
+printf 'HTTP_PROXY=%s\n' "$HTTP_PROXY"
+printf 'HTTPS_PROXY=%s\n' "$HTTPS_PROXY"
+printf 'http_proxy=%s\n' "$http_proxy"
+printf 'https_proxy=%s\n' "$https_proxy"
+printf 'NO_PROXY=%s\n' "$NO_PROXY"
+printf 'no_proxy=%s\n' "$no_proxy"
+`
+	if err := os.WriteFile(scriptPath, []byte(script), 0o755); err != nil {
+		t.Fatalf("write script: %v", err)
+	}
+
+	proxyURL := "socks5h://proxy.local:1080"
+	drv := New(Config{ID: "proxy-drive", RootDir: filepath.Join(tmp, "proxy-drive")})
+	c := NewCrawler(CrawlerConfig{
+		Driver:     drv,
+		PythonPath: "sh",
+		ScriptPath: scriptPath,
+		ProxyURL:   proxyURL,
+	})
+	cmd, stdout, err := c.startSpiderTargetNew(
+		context.Background(),
+		1,
+		filepath.Join(tmp, "seen.txt"),
+		filepath.Join(tmp, "out.json"),
+	)
+	if err != nil {
+		t.Fatalf("startSpiderTargetNew: %v", err)
+	}
+	raw, err := io.ReadAll(stdout)
+	if err != nil {
+		t.Fatalf("read stdout: %v", err)
+	}
+	if err := cmd.Wait(); err != nil {
+		t.Fatalf("wait: %v", err)
+	}
+
+	want := strings.Join([]string{
+		"HTTP_PROXY=" + proxyURL,
+		"HTTPS_PROXY=" + proxyURL,
+		"http_proxy=" + proxyURL,
+		"https_proxy=" + proxyURL,
+		"NO_PROXY=",
+		"no_proxy=",
+	}, "\n") + "\n"
+	if string(raw) != want {
+		t.Fatalf("proxy env = %q, want %q", string(raw), want)
+	}
+}
+
+func TestConfigureExplicitProxySupportsSocksSchemes(t *testing.T) {
+	for _, raw := range []string{
+		"socks5://127.0.0.1:1080",
+		"socks5h://proxy-user:proxy-pass@127.0.0.1:1080",
+	} {
+		t.Run(raw, func(t *testing.T) {
+			transport := &http.Transport{Proxy: http.ProxyFromEnvironment}
+			if err := configureExplicitProxy(transport, raw); err != nil {
+				t.Fatalf("configureExplicitProxy: %v", err)
+			}
+			if transport.Proxy != nil {
+				t.Fatalf("Transport.Proxy should be nil for SOCKS proxy")
+			}
+			if transport.DialContext == nil {
+				t.Fatalf("Transport.DialContext should be set for SOCKS proxy")
+			}
+		})
+	}
+
+	transport := &http.Transport{Proxy: http.ProxyFromEnvironment}
+	if err := configureExplicitProxy(transport, "http://127.0.0.1:7890"); err != nil {
+		t.Fatalf("configureExplicitProxy http: %v", err)
+	}
+	if transport.Proxy == nil {
+		t.Fatalf("Transport.Proxy should be set for HTTP proxy")
+	}
+	if transport.DialContext != nil {
+		t.Fatalf("Transport.DialContext should not be set for HTTP proxy")
+	}
+
+	if err := configureExplicitProxy(&http.Transport{}, "ftp://127.0.0.1:21"); err == nil {
+		t.Fatalf("expected unsupported proxy scheme error")
+	}
+}
+
+func TestSelectSocksTargetIPPrefersIPv4(t *testing.T) {
+	got := selectSocksTargetIP([]net.IPAddr{
+		{IP: net.ParseIP("2606:4700:20::681a:229")},
+		{IP: net.ParseIP("104.26.3.41")},
+	})
+	if got == nil || got.String() != "104.26.3.41" {
+		t.Fatalf("selectSocksTargetIP = %v, want IPv4 104.26.3.41", got)
+	}
+}
+
 // TestCrawlerThumbDownloadFailureMarksStatusFailed 验证：网站封面下载失败时
 // crawler 把 thumbnail_status 显式标 'failed'，避免后续封面补队列一直重复
 // 捞到这条 spider91 视频。
@@ -658,3 +776,12 @@ func buildFakeSpiderScript(entries []map[string]string) string {
 	sb.WriteString("fi\n")
 	return sb.String()
 }
+
+func containsString(values []string, want string) bool {
+	for _, value := range values {
+		if value == want {
+			return true
+		}
+	}
+	return false
+}

backend/vendor/golang.org/x/net/internal/socks/client.go

@@ -0,0 +1,168 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package socks
+
+import (
+	"context"
+	"errors"
+	"io"
+	"net"
+	"strconv"
+	"time"
+)
+
+var (
+	noDeadline   = time.Time{}
+	aLongTimeAgo = time.Unix(1, 0)
+)
+
+func (d *Dialer) connect(ctx context.Context, c net.Conn, address string) (_ net.Addr, ctxErr error) {
+	host, port, err := splitHostPort(address)
+	if err != nil {
+		return nil, err
+	}
+	if deadline, ok := ctx.Deadline(); ok && !deadline.IsZero() {
+		c.SetDeadline(deadline)
+		defer c.SetDeadline(noDeadline)
+	}
+	if ctx != context.Background() {
+		errCh := make(chan error, 1)
+		done := make(chan struct{})
+		defer func() {
+			close(done)
+			if ctxErr == nil {
+				ctxErr = <-errCh
+			}
+		}()
+		go func() {
+			select {
+			case <-ctx.Done():
+				c.SetDeadline(aLongTimeAgo)
+				errCh <- ctx.Err()
+			case <-done:
+				errCh <- nil
+			}
+		}()
+	}
+
+	b := make([]byte, 0, 6+len(host)) // the size here is just an estimate
+	b = append(b, Version5)
+	if len(d.AuthMethods) == 0 || d.Authenticate == nil {
+		b = append(b, 1, byte(AuthMethodNotRequired))
+	} else {
+		ams := d.AuthMethods
+		if len(ams) > 255 {
+			return nil, errors.New("too many authentication methods")
+		}
+		b = append(b, byte(len(ams)))
+		for _, am := range ams {
+			b = append(b, byte(am))
+		}
+	}
+	if _, ctxErr = c.Write(b); ctxErr != nil {
+		return
+	}
+
+	if _, ctxErr = io.ReadFull(c, b[:2]); ctxErr != nil {
+		return
+	}
+	if b[0] != Version5 {
+		return nil, errors.New("unexpected protocol version " + strconv.Itoa(int(b[0])))
+	}
+	am := AuthMethod(b[1])
+	if am == AuthMethodNoAcceptableMethods {
+		return nil, errors.New("no acceptable authentication methods")
+	}
+	if d.Authenticate != nil {
+		if ctxErr = d.Authenticate(ctx, c, am); ctxErr != nil {
+			return
+		}
+	}
+
+	b = b[:0]
+	b = append(b, Version5, byte(d.cmd), 0)
+	if ip := net.ParseIP(host); ip != nil {
+		if ip4 := ip.To4(); ip4 != nil {
+			b = append(b, AddrTypeIPv4)
+			b = append(b, ip4...)
+		} else if ip6 := ip.To16(); ip6 != nil {
+			b = append(b, AddrTypeIPv6)
+			b = append(b, ip6...)
+		} else {
+			return nil, errors.New("unknown address type")
+		}
+	} else {
+		if len(host) > 255 {
+			return nil, errors.New("FQDN too long")
+		}
+		b = append(b, AddrTypeFQDN)
+		b = append(b, byte(len(host)))
+		b = append(b, host...)
+	}
+	b = append(b, byte(port>>8), byte(port))
+	if _, ctxErr = c.Write(b); ctxErr != nil {
+		return
+	}
+
+	if _, ctxErr = io.ReadFull(c, b[:4]); ctxErr != nil {
+		return
+	}
+	if b[0] != Version5 {
+		return nil, errors.New("unexpected protocol version " + strconv.Itoa(int(b[0])))
+	}
+	if cmdErr := Reply(b[1]); cmdErr != StatusSucceeded {
+		return nil, errors.New("unknown error " + cmdErr.String())
+	}
+	if b[2] != 0 {
+		return nil, errors.New("non-zero reserved field")
+	}
+	l := 2
+	var a Addr
+	switch b[3] {
+	case AddrTypeIPv4:
+		l += net.IPv4len
+		a.IP = make(net.IP, net.IPv4len)
+	case AddrTypeIPv6:
+		l += net.IPv6len
+		a.IP = make(net.IP, net.IPv6len)
+	case AddrTypeFQDN:
+		if _, err := io.ReadFull(c, b[:1]); err != nil {
+			return nil, err
+		}
+		l += int(b[0])
+	default:
+		return nil, errors.New("unknown address type " + strconv.Itoa(int(b[3])))
+	}
+	if cap(b) < l {
+		b = make([]byte, l)
+	} else {
+		b = b[:l]
+	}
+	if _, ctxErr = io.ReadFull(c, b); ctxErr != nil {
+		return
+	}
+	if a.IP != nil {
+		copy(a.IP, b)
+	} else {
+		a.Name = string(b[:len(b)-2])
+	}
+	a.Port = int(b[len(b)-2])<<8 | int(b[len(b)-1])
+	return &a, nil
+}
+
+func splitHostPort(address string) (string, int, error) {
+	host, port, err := net.SplitHostPort(address)
+	if err != nil {
+		return "", 0, err
+	}
+	portnum, err := strconv.Atoi(port)
+	if err != nil {
+		return "", 0, err
+	}
+	if 1 > portnum || portnum > 0xffff {
+		return "", 0, errors.New("port number out of range " + port)
+	}
+	return host, portnum, nil
+}

backend/vendor/golang.org/x/net/internal/socks/socks.go

@@ -0,0 +1,317 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package socks provides a SOCKS version 5 client implementation.
+//
+// SOCKS protocol version 5 is defined in RFC 1928.
+// Username/Password authentication for SOCKS version 5 is defined in
+// RFC 1929.
+package socks
+
+import (
+	"context"
+	"errors"
+	"io"
+	"net"
+	"strconv"
+)
+
+// A Command represents a SOCKS command.
+type Command int
+
+func (cmd Command) String() string {
+	switch cmd {
+	case CmdConnect:
+		return "socks connect"
+	case cmdBind:
+		return "socks bind"
+	default:
+		return "socks " + strconv.Itoa(int(cmd))
+	}
+}
+
+// An AuthMethod represents a SOCKS authentication method.
+type AuthMethod int
+
+// A Reply represents a SOCKS command reply code.
+type Reply int
+
+func (code Reply) String() string {
+	switch code {
+	case StatusSucceeded:
+		return "succeeded"
+	case 0x01:
+		return "general SOCKS server failure"
+	case 0x02:
+		return "connection not allowed by ruleset"
+	case 0x03:
+		return "network unreachable"
+	case 0x04:
+		return "host unreachable"
+	case 0x05:
+		return "connection refused"
+	case 0x06:
+		return "TTL expired"
+	case 0x07:
+		return "command not supported"
+	case 0x08:
+		return "address type not supported"
+	default:
+		return "unknown code: " + strconv.Itoa(int(code))
+	}
+}
+
+// Wire protocol constants.
+const (
+	Version5 = 0x05
+
+	AddrTypeIPv4 = 0x01
+	AddrTypeFQDN = 0x03
+	AddrTypeIPv6 = 0x04
+
+	CmdConnect Command = 0x01 // establishes an active-open forward proxy connection
+	cmdBind    Command = 0x02 // establishes a passive-open forward proxy connection
+
+	AuthMethodNotRequired         AuthMethod = 0x00 // no authentication required
+	AuthMethodUsernamePassword    AuthMethod = 0x02 // use username/password
+	AuthMethodNoAcceptableMethods AuthMethod = 0xff // no acceptable authentication methods
+
+	StatusSucceeded Reply = 0x00
+)
+
+// An Addr represents a SOCKS-specific address.
+// Either Name or IP is used exclusively.
+type Addr struct {
+	Name string // fully-qualified domain name
+	IP   net.IP
+	Port int
+}
+
+func (a *Addr) Network() string { return "socks" }
+
+func (a *Addr) String() string {
+	if a == nil {
+		return "<nil>"
+	}
+	port := strconv.Itoa(a.Port)
+	if a.IP == nil {
+		return net.JoinHostPort(a.Name, port)
+	}
+	return net.JoinHostPort(a.IP.String(), port)
+}
+
+// A Conn represents a forward proxy connection.
+type Conn struct {
+	net.Conn
+
+	boundAddr net.Addr
+}
+
+// BoundAddr returns the address assigned by the proxy server for
+// connecting to the command target address from the proxy server.
+func (c *Conn) BoundAddr() net.Addr {
+	if c == nil {
+		return nil
+	}
+	return c.boundAddr
+}
+
+// A Dialer holds SOCKS-specific options.
+type Dialer struct {
+	cmd          Command // either CmdConnect or cmdBind
+	proxyNetwork string  // network between a proxy server and a client
+	proxyAddress string  // proxy server address
+
+	// ProxyDial specifies the optional dial function for
+	// establishing the transport connection.
+	ProxyDial func(context.Context, string, string) (net.Conn, error)
+
+	// AuthMethods specifies the list of request authentication
+	// methods.
+	// If empty, SOCKS client requests only AuthMethodNotRequired.
+	AuthMethods []AuthMethod
+
+	// Authenticate specifies the optional authentication
+	// function. It must be non-nil when AuthMethods is not empty.
+	// It must return an error when the authentication is failed.
+	Authenticate func(context.Context, io.ReadWriter, AuthMethod) error
+}
+
+// DialContext connects to the provided address on the provided
+// network.
+//
+// The returned error value may be a net.OpError. When the Op field of
+// net.OpError contains "socks", the Source field contains a proxy
+// server address and the Addr field contains a command target
+// address.
+//
+// See func Dial of the net package of standard library for a
+// description of the network and address parameters.
+func (d *Dialer) DialContext(ctx context.Context, network, address string) (net.Conn, error) {
+	if err := d.validateTarget(network, address); err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	if ctx == nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: errors.New("nil context")}
+	}
+	var err error
+	var c net.Conn
+	if d.ProxyDial != nil {
+		c, err = d.ProxyDial(ctx, d.proxyNetwork, d.proxyAddress)
+	} else {
+		var dd net.Dialer
+		c, err = dd.DialContext(ctx, d.proxyNetwork, d.proxyAddress)
+	}
+	if err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	a, err := d.connect(ctx, c, address)
+	if err != nil {
+		c.Close()
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	return &Conn{Conn: c, boundAddr: a}, nil
+}
+
+// DialWithConn initiates a connection from SOCKS server to the target
+// network and address using the connection c that is already
+// connected to the SOCKS server.
+//
+// It returns the connection's local address assigned by the SOCKS
+// server.
+func (d *Dialer) DialWithConn(ctx context.Context, c net.Conn, network, address string) (net.Addr, error) {
+	if err := d.validateTarget(network, address); err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	if ctx == nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: errors.New("nil context")}
+	}
+	a, err := d.connect(ctx, c, address)
+	if err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	return a, nil
+}
+
+// Dial connects to the provided address on the provided network.
+//
+// Unlike DialContext, it returns a raw transport connection instead
+// of a forward proxy connection.
+//
+// Deprecated: Use DialContext or DialWithConn instead.
+func (d *Dialer) Dial(network, address string) (net.Conn, error) {
+	if err := d.validateTarget(network, address); err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	var err error
+	var c net.Conn
+	if d.ProxyDial != nil {
+		c, err = d.ProxyDial(context.Background(), d.proxyNetwork, d.proxyAddress)
+	} else {
+		c, err = net.Dial(d.proxyNetwork, d.proxyAddress)
+	}
+	if err != nil {
+		proxy, dst, _ := d.pathAddrs(address)
+		return nil, &net.OpError{Op: d.cmd.String(), Net: network, Source: proxy, Addr: dst, Err: err}
+	}
+	if _, err := d.DialWithConn(context.Background(), c, network, address); err != nil {
+		c.Close()
+		return nil, err
+	}
+	return c, nil
+}
+
+func (d *Dialer) validateTarget(network, address string) error {
+	switch network {
+	case "tcp", "tcp6", "tcp4":
+	default:
+		return errors.New("network not implemented")
+	}
+	switch d.cmd {
+	case CmdConnect, cmdBind:
+	default:
+		return errors.New("command not implemented")
+	}
+	return nil
+}
+
+func (d *Dialer) pathAddrs(address string) (proxy, dst net.Addr, err error) {
+	for i, s := range []string{d.proxyAddress, address} {
+		host, port, err := splitHostPort(s)
+		if err != nil {
+			return nil, nil, err
+		}
+		a := &Addr{Port: port}
+		a.IP = net.ParseIP(host)
+		if a.IP == nil {
+			a.Name = host
+		}
+		if i == 0 {
+			proxy = a
+		} else {
+			dst = a
+		}
+	}
+	return
+}
+
+// NewDialer returns a new Dialer that dials through the provided
+// proxy server's network and address.
+func NewDialer(network, address string) *Dialer {
+	return &Dialer{proxyNetwork: network, proxyAddress: address, cmd: CmdConnect}
+}
+
+const (
+	authUsernamePasswordVersion = 0x01
+	authStatusSucceeded         = 0x00
+)
+
+// UsernamePassword are the credentials for the username/password
+// authentication method.
+type UsernamePassword struct {
+	Username string
+	Password string
+}
+
+// Authenticate authenticates a pair of username and password with the
+// proxy server.
+func (up *UsernamePassword) Authenticate(ctx context.Context, rw io.ReadWriter, auth AuthMethod) error {
+	switch auth {
+	case AuthMethodNotRequired:
+		return nil
+	case AuthMethodUsernamePassword:
+		if len(up.Username) == 0 || len(up.Username) > 255 || len(up.Password) > 255 {
+			return errors.New("invalid username/password")
+		}
+		b := []byte{authUsernamePasswordVersion}
+		b = append(b, byte(len(up.Username)))
+		b = append(b, up.Username...)
+		b = append(b, byte(len(up.Password)))
+		b = append(b, up.Password...)
+		// TODO(mikio): handle IO deadlines and cancelation if
+		// necessary
+		if _, err := rw.Write(b); err != nil {
+			return err
+		}
+		if _, err := io.ReadFull(rw, b[:2]); err != nil {
+			return err
+		}
+		if b[0] != authUsernamePasswordVersion {
+			return errors.New("invalid username/password version")
+		}
+		if b[1] != authStatusSucceeded {
+			return errors.New("username/password authentication failed")
+		}
+		return nil
+	}
+	return errors.New("unsupported authentication method " + strconv.Itoa(int(auth)))
+}

backend/vendor/golang.org/x/net/proxy/dial.go

@@ -0,0 +1,54 @@
+// Copyright 2019 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package proxy
+
+import (
+	"context"
+	"net"
+)
+
+// A ContextDialer dials using a context.
+type ContextDialer interface {
+	DialContext(ctx context.Context, network, address string) (net.Conn, error)
+}
+
+// Dial works like DialContext on net.Dialer but using a dialer returned by FromEnvironment.
+//
+// The passed ctx is only used for returning the Conn, not the lifetime of the Conn.
+//
+// Custom dialers (registered via RegisterDialerType) that do not implement ContextDialer
+// can leak a goroutine for as long as it takes the underlying Dialer implementation to timeout.
+//
+// A Conn returned from a successful Dial after the context has been cancelled will be immediately closed.
+func Dial(ctx context.Context, network, address string) (net.Conn, error) {
+	d := FromEnvironment()
+	if xd, ok := d.(ContextDialer); ok {
+		return xd.DialContext(ctx, network, address)
+	}
+	return dialContext(ctx, d, network, address)
+}
+
+// WARNING: this can leak a goroutine for as long as the underlying Dialer implementation takes to timeout
+// A Conn returned from a successful Dial after the context has been cancelled will be immediately closed.
+func dialContext(ctx context.Context, d Dialer, network, address string) (net.Conn, error) {
+	var (
+		conn net.Conn
+		done = make(chan struct{}, 1)
+		err  error
+	)
+	go func() {
+		conn, err = d.Dial(network, address)
+		close(done)
+		if conn != nil && ctx.Err() != nil {
+			conn.Close()
+		}
+	}()
+	select {
+	case <-ctx.Done():
+		err = ctx.Err()
+	case <-done:
+	}
+	return conn, err
+}

backend/vendor/golang.org/x/net/proxy/direct.go

@@ -0,0 +1,31 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package proxy
+
+import (
+	"context"
+	"net"
+)
+
+type direct struct{}
+
+// Direct implements Dialer by making network connections directly using net.Dial or net.DialContext.
+var Direct = direct{}
+
+var (
+	_ Dialer        = Direct
+	_ ContextDialer = Direct
+)
+
+// Dial directly invokes net.Dial with the supplied parameters.
+func (direct) Dial(network, addr string) (net.Conn, error) {
+	return net.Dial(network, addr)
+}
+
+// DialContext instantiates a net.Dialer and invokes its DialContext receiver with the supplied parameters.
+func (direct) DialContext(ctx context.Context, network, addr string) (net.Conn, error) {
+	var d net.Dialer
+	return d.DialContext(ctx, network, addr)
+}

backend/vendor/golang.org/x/net/proxy/per_host.go

@@ -0,0 +1,151 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package proxy
+
+import (
+	"context"
+	"net"
+	"strings"
+)
+
+// A PerHost directs connections to a default Dialer unless the host name
+// requested matches one of a number of exceptions.
+type PerHost struct {
+	def, bypass Dialer
+
+	bypassNetworks []*net.IPNet
+	bypassIPs      []net.IP
+	bypassZones    []string
+	bypassHosts    []string
+}
+
+// NewPerHost returns a PerHost Dialer that directs connections to either
+// defaultDialer or bypass, depending on whether the connection matches one of
+// the configured rules.
+func NewPerHost(defaultDialer, bypass Dialer) *PerHost {
+	return &PerHost{
+		def:    defaultDialer,
+		bypass: bypass,
+	}
+}
+
+// Dial connects to the address addr on the given network through either
+// defaultDialer or bypass.
+func (p *PerHost) Dial(network, addr string) (c net.Conn, err error) {
+	host, _, err := net.SplitHostPort(addr)
+	if err != nil {
+		return nil, err
+	}
+
+	return p.dialerForRequest(host).Dial(network, addr)
+}
+
+// DialContext connects to the address addr on the given network through either
+// defaultDialer or bypass.
+func (p *PerHost) DialContext(ctx context.Context, network, addr string) (c net.Conn, err error) {
+	host, _, err := net.SplitHostPort(addr)
+	if err != nil {
+		return nil, err
+	}
+	d := p.dialerForRequest(host)
+	if x, ok := d.(ContextDialer); ok {
+		return x.DialContext(ctx, network, addr)
+	}
+	return dialContext(ctx, d, network, addr)
+}
+
+func (p *PerHost) dialerForRequest(host string) Dialer {
+	if ip := net.ParseIP(host); ip != nil {
+		for _, net := range p.bypassNetworks {
+			if net.Contains(ip) {
+				return p.bypass
+			}
+		}
+		for _, bypassIP := range p.bypassIPs {
+			if bypassIP.Equal(ip) {
+				return p.bypass
+			}
+		}
+		return p.def
+	}
+
+	for _, zone := range p.bypassZones {
+		if strings.HasSuffix(host, zone) {
+			return p.bypass
+		}
+		if host == zone[1:] {
+			// For a zone ".example.com", we match "example.com"
+			// too.
+			return p.bypass
+		}
+	}
+	for _, bypassHost := range p.bypassHosts {
+		if bypassHost == host {
+			return p.bypass
+		}
+	}
+	return p.def
+}
+
+// AddFromString parses a string that contains comma-separated values
+// specifying hosts that should use the bypass proxy. Each value is either an
+// IP address, a CIDR range, a zone (*.example.com) or a host name
+// (localhost). A best effort is made to parse the string and errors are
+// ignored.
+func (p *PerHost) AddFromString(s string) {
+	hosts := strings.Split(s, ",")
+	for _, host := range hosts {
+		host = strings.TrimSpace(host)
+		if len(host) == 0 {
+			continue
+		}
+		if strings.Contains(host, "/") {
+			// We assume that it's a CIDR address like 127.0.0.0/8
+			if _, net, err := net.ParseCIDR(host); err == nil {
+				p.AddNetwork(net)
+			}
+			continue
+		}
+		if ip := net.ParseIP(host); ip != nil {
+			p.AddIP(ip)
+			continue
+		}
+		if strings.HasPrefix(host, "*.") {
+			p.AddZone(host[1:])
+			continue
+		}
+		p.AddHost(host)
+	}
+}
+
+// AddIP specifies an IP address that will use the bypass proxy. Note that
+// this will only take effect if a literal IP address is dialed. A connection
+// to a named host will never match an IP.
+func (p *PerHost) AddIP(ip net.IP) {
+	p.bypassIPs = append(p.bypassIPs, ip)
+}
+
+// AddNetwork specifies an IP range that will use the bypass proxy. Note that
+// this will only take effect if a literal IP address is dialed. A connection
+// to a named host will never match.
+func (p *PerHost) AddNetwork(net *net.IPNet) {
+	p.bypassNetworks = append(p.bypassNetworks, net)
+}
+
+// AddZone specifies a DNS suffix that will use the bypass proxy. A zone of
+// "example.com" matches "example.com" and all of its subdomains.
+func (p *PerHost) AddZone(zone string) {
+	zone = strings.TrimSuffix(zone, ".")
+	if !strings.HasPrefix(zone, ".") {
+		zone = "." + zone
+	}
+	p.bypassZones = append(p.bypassZones, zone)
+}
+
+// AddHost specifies a host name that will use the bypass proxy.
+func (p *PerHost) AddHost(host string) {
+	host = strings.TrimSuffix(host, ".")
+	p.bypassHosts = append(p.bypassHosts, host)
+}

backend/vendor/golang.org/x/net/proxy/proxy.go

@@ -0,0 +1,149 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package proxy provides support for a variety of protocols to proxy network
+// data.
+package proxy // import "golang.org/x/net/proxy"
+
+import (
+	"errors"
+	"net"
+	"net/url"
+	"os"
+	"sync"
+)
+
+// A Dialer is a means to establish a connection.
+// Custom dialers should also implement ContextDialer.
+type Dialer interface {
+	// Dial connects to the given address via the proxy.
+	Dial(network, addr string) (c net.Conn, err error)
+}
+
+// Auth contains authentication parameters that specific Dialers may require.
+type Auth struct {
+	User, Password string
+}
+
+// FromEnvironment returns the dialer specified by the proxy-related
+// variables in the environment and makes underlying connections
+// directly.
+func FromEnvironment() Dialer {
+	return FromEnvironmentUsing(Direct)
+}
+
+// FromEnvironmentUsing returns the dialer specify by the proxy-related
+// variables in the environment and makes underlying connections
+// using the provided forwarding Dialer (for instance, a *net.Dialer
+// with desired configuration).
+func FromEnvironmentUsing(forward Dialer) Dialer {
+	allProxy := allProxyEnv.Get()
+	if len(allProxy) == 0 {
+		return forward
+	}
+
+	proxyURL, err := url.Parse(allProxy)
+	if err != nil {
+		return forward
+	}
+	proxy, err := FromURL(proxyURL, forward)
+	if err != nil {
+		return forward
+	}
+
+	noProxy := noProxyEnv.Get()
+	if len(noProxy) == 0 {
+		return proxy
+	}
+
+	perHost := NewPerHost(proxy, forward)
+	perHost.AddFromString(noProxy)
+	return perHost
+}
+
+// proxySchemes is a map from URL schemes to a function that creates a Dialer
+// from a URL with such a scheme.
+var proxySchemes map[string]func(*url.URL, Dialer) (Dialer, error)
+
+// RegisterDialerType takes a URL scheme and a function to generate Dialers from
+// a URL with that scheme and a forwarding Dialer. Registered schemes are used
+// by FromURL.
+func RegisterDialerType(scheme string, f func(*url.URL, Dialer) (Dialer, error)) {
+	if proxySchemes == nil {
+		proxySchemes = make(map[string]func(*url.URL, Dialer) (Dialer, error))
+	}
+	proxySchemes[scheme] = f
+}
+
+// FromURL returns a Dialer given a URL specification and an underlying
+// Dialer for it to make network requests.
+func FromURL(u *url.URL, forward Dialer) (Dialer, error) {
+	var auth *Auth
+	if u.User != nil {
+		auth = new(Auth)
+		auth.User = u.User.Username()
+		if p, ok := u.User.Password(); ok {
+			auth.Password = p
+		}
+	}
+
+	switch u.Scheme {
+	case "socks5", "socks5h":
+		addr := u.Hostname()
+		port := u.Port()
+		if port == "" {
+			port = "1080"
+		}
+		return SOCKS5("tcp", net.JoinHostPort(addr, port), auth, forward)
+	}
+
+	// If the scheme doesn't match any of the built-in schemes, see if it
+	// was registered by another package.
+	if proxySchemes != nil {
+		if f, ok := proxySchemes[u.Scheme]; ok {
+			return f(u, forward)
+		}
+	}
+
+	return nil, errors.New("proxy: unknown scheme: " + u.Scheme)
+}
+
+var (
+	allProxyEnv = &envOnce{
+		names: []string{"ALL_PROXY", "all_proxy"},
+	}
+	noProxyEnv = &envOnce{
+		names: []string{"NO_PROXY", "no_proxy"},
+	}
+)
+
+// envOnce looks up an environment variable (optionally by multiple
+// names) once. It mitigates expensive lookups on some platforms
+// (e.g. Windows).
+// (Borrowed from net/http/transport.go)
+type envOnce struct {
+	names []string
+	once  sync.Once
+	val   string
+}
+
+func (e *envOnce) Get() string {
+	e.once.Do(e.init)
+	return e.val
+}
+
+func (e *envOnce) init() {
+	for _, n := range e.names {
+		e.val = os.Getenv(n)
+		if e.val != "" {
+			return
+		}
+	}
+}
+
+// reset is used by tests
+func (e *envOnce) reset() {
+	e.once = sync.Once{}
+	e.val = ""
+}

backend/vendor/golang.org/x/net/proxy/socks5.go

@@ -0,0 +1,42 @@
+// Copyright 2011 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package proxy
+
+import (
+	"context"
+	"net"
+
+	"golang.org/x/net/internal/socks"
+)
+
+// SOCKS5 returns a Dialer that makes SOCKSv5 connections to the given
+// address with an optional username and password.
+// See RFC 1928 and RFC 1929.
+func SOCKS5(network, address string, auth *Auth, forward Dialer) (Dialer, error) {
+	d := socks.NewDialer(network, address)
+	if forward != nil {
+		if f, ok := forward.(ContextDialer); ok {
+			d.ProxyDial = func(ctx context.Context, network string, address string) (net.Conn, error) {
+				return f.DialContext(ctx, network, address)
+			}
+		} else {
+			d.ProxyDial = func(ctx context.Context, network string, address string) (net.Conn, error) {
+				return dialContext(ctx, forward, network, address)
+			}
+		}
+	}
+	if auth != nil {
+		up := socks.UsernamePassword{
+			Username: auth.User,
+			Password: auth.Password,
+		}
+		d.AuthMethods = []socks.AuthMethod{
+			socks.AuthMethodNotRequired,
+			socks.AuthMethodUsernamePassword,
+		}
+		d.Authenticate = up.Authenticate
+	}
+	return d, nil
+}

backend/vendor/modules.txt

@@ -67,6 +67,8 @@ github.com/skip2/go-qrcode/reedsolomon
 golang.org/x/crypto/curve25519
 # golang.org/x/net v0.27.0
 ## explicit; go 1.18
+golang.org/x/net/internal/socks
+golang.org/x/net/proxy
 golang.org/x/net/publicsuffix
 # golang.org/x/sys v0.30.0
 ## explicit; go 1.18

src/admin/DrivesPage.tsx

@@ -93,9 +93,9 @@ export function DrivesPage() {
   const [form, setForm] = useState<FormState>(emptyForm);
   const [saving, setSaving] = useState(false);
   const [regenFailedId, setRegenFailedId] = useState("");
-  // 与 regenFailedId 并列：失败封面重新入队按钮的 disable 状态。两套独立按钮 →
-  // 两个 state 互不阻塞，避免操作 teaser 时锁住封面那条按钮（反之亦然）。
+  // 失败重试按钮各自维护 pending 状态，避免操作 teaser / 封面 / 指纹时互相锁住。
   const [regenFailedThumbId, setRegenFailedThumbId] = useState("");
+  const [regenFailedFingerprintId, setRegenFailedFingerprintId] = useState("");
   // togglingTeaserId 在请求未返回前禁用按钮，避免连点导致两次切换互相覆盖。
   const [togglingTeaserId, setTogglingTeaserId] = useState("");
   const [scanningAll, setScanningAll] = useState(false);
@@ -169,7 +169,7 @@ export function DrivesPage() {
       kind: d.kind,
       name: d.name,
       rootId: d.rootId,
-      creds: {},
+      creds: d.kind === "spider91" ? { proxy: d.spider91Proxy ?? "" } : {},
       spider91UploadDriveId: settings?.spider91UploadDriveId ?? "",
     });
     setModalOpen(true);
@@ -185,7 +185,9 @@ export function DrivesPage() {
     const driveID = existing
       ? form.id
       : makeUniqueDriveId(form.kind, name, list);
-    const rootId = form.rootId.trim() || defaultRootId(form.kind);
+    const rootId = usesRootDirectoryID(form.kind)
+      ? form.rootId.trim() || defaultRootId(form.kind)
+      : defaultRootId(form.kind);
     // 若编辑且没有提供凭证，提示一下但仍允许保存（不改凭证）
     setSaving(true);
     try {
@@ -310,6 +312,19 @@ export function DrivesPage() {
     }
   }
 
+  async function handleRegenFailedFingerprints(d: api.AdminDrive) {
+    setRegenFailedFingerprintId(d.id);
+    try {
+      await api.regenFailedFingerprints(d.id);
+      show("已触发失败指纹重新生成", "success");
+      refresh();
+    } catch (e) {
+      show(e instanceof Error ? e.message : "触发失败", "error");
+    } finally {
+      setRegenFailedFingerprintId("");
+    }
+  }
+
   async function handleToggleTeaser(d: api.AdminDrive) {
     const next = !d.teaserEnabled;
     setTogglingTeaserId(d.id);
@@ -395,7 +410,7 @@ export function DrivesPage() {
                   <span className="admin-detail-label">网盘 ID</span>
                   <span className="admin-detail-value admin-mono-cell">{d.id}</span>
                 </div>
-                {d.kind !== "spider91" && (
+                {usesRootDirectoryID(d.kind) && (
                   <>
                     <div className="admin-detail-row">
                       <span className="admin-detail-label">根目录 ID</span>
@@ -479,14 +494,14 @@ export function DrivesPage() {
                     style={{ padding: "4px 10px", fontSize: "11px" }}
                   >
                     {d.teaserEnabled ? <Power size={11} /> : <PowerOff size={11} />}
-                    <span>{d.teaserEnabled ? "Teaser: 开" : "Teaser: 关"}</span>
+                    <span>{d.teaserEnabled ? "预览视频生成：开" : "预览视频生成：关"}</span>
                   </button>
                 </div>
               </header>
 
               <div className="admin-detail-grid">
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">封面状态</span>
+                  <span className="admin-detail-label">封面生成状态</span>
                   <div className="admin-detail-value">
                     <GenerationStatusLine label="封面" status={d.thumbnailGenerationStatus} />
                   </div>
@@ -503,13 +518,13 @@ export function DrivesPage() {
                   </div>
                 </div>
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">Teaser 状态</span>
+                  <span className="admin-detail-label">预览视频生成状态</span>
                   <div className="admin-detail-value">
                     <GenerationStatusLine label="预览" status={d.previewGenerationStatus} />
                   </div>
                 </div>
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">Teaser 数量</span>
+                  <span className="admin-detail-label">预览视频数量</span>
                   <div className="admin-detail-value">
                     <GenerationCounts
                       ready={d.teaserReadyCount}
@@ -519,13 +534,13 @@ export function DrivesPage() {
                   </div>
                 </div>
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">指纹状态</span>
+                  <span className="admin-detail-label">视频指纹生成状态</span>
                   <div className="admin-detail-value">
                     <GenerationStatusLine label="指纹" status={d.fingerprintGenerationStatus} />
                   </div>
                 </div>
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">指纹数量</span>
+                  <span className="admin-detail-label">视频指纹数量</span>
                   <div className="admin-detail-value">
                     <GenerationCounts
                       ready={d.fingerprintReadyCount}
@@ -553,6 +568,17 @@ export function DrivesPage() {
                   <RotateCcw size={13} />
                   <span>重试失败封面</span>
                 </button>
+                <button
+                  className="admin-btn"
+                  disabled={
+                    (d.fingerprintFailedCount ?? 0) <= 0 ||
+                    regenFailedFingerprintId === d.id
+                  }
+                  onClick={() => handleRegenFailedFingerprints(d)}
+                >
+                  <RotateCcw size={13} />
+                  <span>重试失败指纹</span>
+                </button>
               </div>
             </div>
 
@@ -570,7 +596,7 @@ export function DrivesPage() {
                   <span className="admin-detail-value">{formatBytes(driveStorage?.thumbnailBytes ?? 0)}</span>
                 </div>
                 <div className="admin-detail-row">
-                  <span className="admin-detail-label">Teaser 占用</span>
+                  <span className="admin-detail-label">预览视频占用</span>
                   <span className="admin-detail-value">{formatBytes(driveStorage?.teaserBytes ?? 0)}</span>
                 </div>
                 <div className="admin-detail-row">
@@ -739,7 +765,7 @@ function StorageSummary({ storage }: { storage: api.AdminDriveStorage }) {
         <strong>{formatBytes(storage.thumbnailBytes)}</strong>
       </div>
       <div className="admin-storage-summary__metric">
-        <span>Teaser 占用</span>
+        <span>预览视频占用</span>
         <strong>{formatBytes(storage.teaserBytes)}</strong>
       </div>
       <div className="admin-storage-summary__metric">
@@ -950,17 +976,19 @@ function DriveForm({
           <option value="wopan">联通沃盘</option>
         </select>
       </div>
-      <div className="admin-form__row">
-        <label>根目录 ID</label>
-        <input
-          value={form.rootId}
-          onChange={(e) => set("rootId", e.target.value)}
-          placeholder={rootIdPlaceholder(form.kind)}
-        />
-        <div className="admin-form__help">
-          留空时使用该网盘类型的默认根目录，具体目录ID获取方式请参考OpenList文档
+      {usesRootDirectoryID(form.kind) && (
+        <div className="admin-form__row">
+          <label>根目录 ID</label>
+          <input
+            value={form.rootId}
+            onChange={(e) => set("rootId", e.target.value)}
+            placeholder={rootIdPlaceholder(form.kind)}
+          />
+          <div className="admin-form__help">
+            留空时使用该网盘类型的默认根目录，具体目录ID获取方式请参考OpenList文档
+          </div>
         </div>
-      </div>
+      )}
 
       {(help || fields.length > 0) && (
         <>
@@ -1063,7 +1091,7 @@ function credentialHelp(kind: Kind, isEdit: boolean): string {
     case "localstorage":
       return `把服务器上的一个已有目录作为视频来源扫描。填写绝对路径，例如 /mnt/videos；系统会读取该目录及子目录中的视频，并生成封面、Teaser 和指纹。${note}`;
     case "spider91":
-      return "91 爬虫会把定时抓取到的视频和封面先保存到本机，并作为一个视频来源接入站点；它不是外部网盘，不需要填写 Cookie 或目录 ID。后续流水线会把较早的视频上传到你选择的 115 / PikPak / OneDrive 目标盘。";
+      return "91 爬虫会把定时抓取到的视频和封面先保存到本机，并作为一个视频来源接入站点；可按服务器网络情况单独配置代理。后续流水线会把较早的视频上传到你选择的 115 / PikPak / OneDrive 目标盘。";
     default:
       return "";
   }
@@ -1164,7 +1192,14 @@ function credentialFields(kind: Kind): Array<{
         },
       ];
     case "spider91":
-      return [];
+      return [
+        {
+          key: "proxy",
+          label: "代理地址（可选）",
+          placeholder: "http://127.0.0.1:7890",
+          help: "仅用于 91Spider 的列表/详情请求和视频、封面下载；留空则使用服务器环境变量 HTTP_PROXY / HTTPS_PROXY 或直连。支持 http://、https://、socks5:// 或 socks5h://。",
+        },
+      ];
   }
 }
 
@@ -1177,6 +1212,10 @@ function defaultRootId(kind: Kind): string {
   return "0";
 }
 
+function usesRootDirectoryID(kind: Kind): boolean {
+  return kind !== "localstorage" && kind !== "spider91";
+}
+
 function rootIdPlaceholder(kind: Kind): string {
   const rootId = defaultRootId(kind);
   return rootId ? `默认：${rootId}` : "留空表示根目录";

src/admin/api.ts

@@ -93,6 +93,8 @@ export type AdminDrive = {
   skipDirIds: string[];
   // spider91 上次成功爬取时间（unix 秒）；其它 kind 留空。
   lastCrawlAt?: number;
+  // spider91 专用代理地址；仅后台管理接口返回，用于编辑表单回显。
+  spider91Proxy?: string;
   thumbnailGenerationStatus?: DriveGenerationStatus;
   previewGenerationStatus?: DriveGenerationStatus;
   fingerprintGenerationStatus?: DriveGenerationStatus;
@@ -243,6 +245,13 @@ export function regenFailedThumbnails(id: string) {
   );
 }
 
+export function regenFailedFingerprints(id: string) {
+  return request<{ ok: boolean }>(
+    `/drives/${encodeURIComponent(id)}/fingerprints/failed/regenerate`,
+    { method: "POST" }
+  );
+}
+
 // ---------- Videos ----------
 
 export type AdminVideo = {

src/data/videos.ts

@@ -1,8 +1,13 @@
 import type { VideoDetail, VideoItem } from "@/types";
 
 // 真实后端接口调用。未配置网盘时，各接口返回空数据。
-export function fetchHomeVideos(): Promise<VideoItem[]> {
-  return apiGet<VideoItem[]>("/api/home").catch(() => []);
+export function fetchHomeVideos(excludeIds?: string[]): Promise<VideoItem[]> {
+  const qs = new URLSearchParams();
+  for (const id of excludeIds ?? []) {
+    if (id.trim()) qs.append("exclude", id.trim());
+  }
+  const suffix = qs.toString() ? `?${qs.toString()}` : "";
+  return apiGet<VideoItem[]>(`/api/home${suffix}`).catch(() => []);
 }
 
 export function fetchListing(

src/pages/HomePage.tsx

@@ -10,6 +10,8 @@ import type { VideoItem } from "@/types";
 
 const DESKTOP_COUNT = 12;
 const MOBILE_COUNT = 8;
+const HOME_RECENT_KEY = "home.random.recentVideoIds";
+const HOME_RECENT_LIMIT = 72;
 
 function useIsMobile() {
   const [mobile, setMobile] = useState(window.innerWidth <= 640);
@@ -26,6 +28,35 @@ function useIsMobile() {
 let cachedRanking: VideoItem[] | null = null;
 let cachedLatest: VideoItem[] | null = null;
 
+function loadRecentHomeVideoIds(): string[] {
+  try {
+    const raw = window.localStorage.getItem(HOME_RECENT_KEY);
+    const parsed = raw ? JSON.parse(raw) : [];
+    return Array.isArray(parsed)
+      ? parsed.filter((id): id is string => typeof id === "string" && id.length > 0)
+      : [];
+  } catch {
+    return [];
+  }
+}
+
+function rememberHomeVideos(items: VideoItem[]) {
+  const merged = [...items.map((item) => item.id), ...loadRecentHomeVideoIds()];
+  const seen = new Set<string>();
+  const recent: string[] = [];
+  for (const id of merged) {
+    if (!id || seen.has(id)) continue;
+    seen.add(id);
+    recent.push(id);
+    if (recent.length >= HOME_RECENT_LIMIT) break;
+  }
+  try {
+    window.localStorage.setItem(HOME_RECENT_KEY, JSON.stringify(recent));
+  } catch {
+    // localStorage 不可用时只影响连续刷新去重，不影响首页展示。
+  }
+}
+
 export default function HomePage() {
   const [rankingVideos, setRankingVideos] = useState<VideoItem[]>(cachedRanking ?? []);
   const [latestVideos, setLatestVideos] = useState<VideoItem[]>(cachedLatest ?? []);
@@ -40,11 +71,13 @@ export default function HomePage() {
 
     let active = true;
     setLoading(true);
+    const excludeIds = loadRecentHomeVideoIds();
     Promise.all([
-      fetchHomeVideos(),
+      fetchHomeVideos(excludeIds),
       fetchListing(1, DESKTOP_COUNT, { sort: "latest" }),
     ]).then(([rankingItems, latestResult]) => {
       if (!active) return;
+      rememberHomeVideos(rankingItems);
       cachedRanking = rankingItems;
       cachedLatest = latestResult.items;
       setRankingVideos(rankingItems);

tests/adminDriveForm.test.ts

@@ -8,6 +8,9 @@ const drivesPageSource = readFileSync(
 );
 
 test("spider91 drive form does not expose advanced crawler credentials", () => {
+  assert.match(drivesPageSource, /key: "proxy"/);
+  assert.match(drivesPageSource, /label: "代理地址（可选）"/);
+  assert.match(drivesPageSource, /支持 http:\/\/、https:\/\/、socks5:\/\/ 或 socks5h:\/\//);
   assert.doesNotMatch(drivesPageSource, /target_new/);
   assert.doesNotMatch(drivesPageSource, /crawl_hour/);
   assert.doesNotMatch(drivesPageSource, /python_path/);
@@ -24,8 +27,14 @@ test("spider91 upload target uses explicit local-save option instead of auto tar
   assert.doesNotMatch(drivesPageSource, /自动模式/);
 });
 
-test("drive form shows a root directory id field for all drive kinds", () => {
+test("drive form hides root directory id for localstorage and spider91", () => {
   assert.match(drivesPageSource, /<label>根目录 ID<\/label>/);
+  assert.match(
+    drivesPageSource,
+    /function usesRootDirectoryID\(kind: Kind\): boolean \{\s*return kind !== "localstorage" && kind !== "spider91";\s*\}/
+  );
+  assert.match(drivesPageSource, /\{usesRootDirectoryID\(form\.kind\) && \(/);
+  assert.match(drivesPageSource, /\{usesRootDirectoryID\(d\.kind\) && \(/);
   assert.match(drivesPageSource, /placeholder=\{rootIdPlaceholder\(form\.kind\)\}/);
   assert.doesNotMatch(drivesPageSource, /扫描起点目录 ID/);
   assert.doesNotMatch(drivesPageSource, /set\("scanRootId"/);
@@ -94,6 +103,7 @@ test("localstorage drive form asks for a server directory path", () => {
   assert.match(fields, /key: "path"/);
   assert.match(fields, /label: "本地目录路径"/);
   assert.match(drivesPageSource, /if \(kind === "localstorage"\) return "\/"/);
+  assert.match(drivesPageSource, /kind !== "localstorage" && kind !== "spider91"/);
 });
 
 test("drive type selector keeps primary source order", () => {