diff --git a/chunk.go b/chunk.go
new file mode 100644
index 0000000..65f2dba
--- /dev/null
+++ b/chunk.go
@@ -0,0 +1,366 @@
+package main
+
+import (
+	"encoding/json"
+	"io"
+	"log"
+	"net/http"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strconv"
+	"sync"
+	"time"
+
+	"github.com/gin-gonic/gin"
+)
+
+const (
+	chunkDir       = "./data/chunks"
+	chunkThreshold = 5 << 30 // >5GB 才走分片
+	chunkSize      = 1 << 30 // 每片 1GB
+)
+
+var uploadIDRe = regexp.MustCompile(`^[a-zA-Z0-9_-]{8,64}$`)
+
+type chunkMeta struct {
+	UploadID    string       `json:"upload_id"`
+	Filename    string       `json:"filename"`
+	TotalSize   int64        `json:"total_size"`
+	TotalChunks int          `json:"total_chunks"`
+	Received    map[int]bool `json:"received"`
+	CreatedAt   time.Time    `json:"created_at"`
+	Reserved    int64        `json:"reserved"`
+}
+
+// 全局 map 存 in-flight 上传; 持锁只做元数据操作, 1GB 分片 io.Copy 在锁外进行.
+// ponytail: 全局锁, 若并发大文件上传成为瓶颈再换 per-upload 锁.
+var (
+	chunkMu      sync.RWMutex
+	chunkMetaMap = map[string]*chunkMeta{}
+)
+
+func chunkDirPath(id string) string { return filepath.Join(chunkDir, id) }
+func chunkFilePath(id string, idx int) string {
+	return filepath.Join(chunkDir, id, strconv.Itoa(idx))
+}
+
+func metaPath(id string) string { return filepath.Join(chunkDir, id, "meta.json") }
+
+// saveMeta 原子写 meta.json (.tmp + rename).
+func saveMeta(m *chunkMeta) error {
+	b, err := json.Marshal(m)
+	if err != nil {
+		return err
+	}
+	tmp := metaPath(m.UploadID) + ".tmp"
+	if err := os.WriteFile(tmp, b, 0o644); err != nil {
+		return err
+	}
+	return os.Rename(tmp, metaPath(m.UploadID))
+}
+
+func loadMeta(id string) *chunkMeta {
+	b, err := os.ReadFile(metaPath(id))
+	if err != nil {
+		return nil
+	}
+	var m chunkMeta
+	if err := json.Unmarshal(b, &m); err != nil {
+		return nil
+	}
+	return &m
+}
+
+// expectedChunks totalSize 按 chunkSize 分片应有的片数
+func expectedChunks(totalSize int64) int {
+	return int((totalSize + chunkSize - 1) / chunkSize)
+}
+
+// expectedChunkSize 第 i 片应有的字节数 (最后一片是余数)
+func expectedChunkSize(totalSize int64, i int) int64 {
+	if n := totalSize - int64(i)*chunkSize; n < chunkSize {
+		return n
+	}
+	return chunkSize
+}
+
+func uploadChunkHandler(c *gin.Context) {
+	uploadID := c.GetHeader("X-Upload-Id")
+	if !uploadIDRe.MatchString(uploadID) {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid upload id"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "invalid_upload_id"})
+		return
+	}
+	idx, err := strconv.Atoi(c.GetHeader("X-Chunk-Index"))
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid chunk index"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "invalid_chunk_index"})
+		return
+	}
+	totalChunks, err := strconv.Atoi(c.GetHeader("X-Total-Chunks"))
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid total chunks"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "invalid_total_chunks"})
+		return
+	}
+	filename := c.GetHeader("X-Filename")
+	if !checkFilenameSafe(filename) {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid filename"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "invalid_filename"})
+		return
+	}
+	totalSize, err := strconv.ParseInt(c.GetHeader("X-Total-Size"), 10, 64)
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "invalid total size"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "invalid_total_size"})
+		return
+	}
+	if totalSize <= chunkThreshold {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "chunked upload is for files > 5GB"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "below_threshold"})
+		return
+	}
+	if totalChunks < 1 || idx < 0 || idx >= totalChunks {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "chunk index out of range"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "total": totalChunks, "result": "index_out_of_range"})
+		return
+	}
+	if e := expectedChunks(totalSize); totalChunks < e-1 || totalChunks > e+1 {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "total chunks mismatch"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "total": totalChunks, "expected": e, "result": "chunk_count_mismatch"})
+		return
+	}
+	if totalSize > quotaRemaining() {
+		c.JSON(http.StatusInsufficientStorage, gin.H{"error": "quota exceeded"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "size": totalSize, "result": "quota_exceeded"})
+		return
+	}
+
+	chunk, err := c.FormFile("chunk")
+	if err != nil {
+		c.JSON(http.StatusBadRequest, gin.H{"error": "missing form field 'chunk'"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "missing_form_field"})
+		return
+	}
+
+	dir := chunkDirPath(uploadID)
+	chunkMu.RLock()
+	meta, inMap := chunkMetaMap[uploadID]
+	chunkMu.RUnlock()
+	if !inMap {
+		meta = loadMeta(uploadID) // 进程重启后从磁盘恢复
+		if meta != nil {
+			chunkMu.Lock()
+			chunkMetaMap[uploadID] = meta
+			chunkMu.Unlock()
+		}
+	}
+	if meta == nil {
+		if idx != 0 {
+			c.JSON(http.StatusNotFound, gin.H{"error": "upload not found"})
+			audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "upload_not_found"})
+			return
+		}
+		if err := os.MkdirAll(dir, 0o755); err != nil {
+			c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+			audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "mkdir_failed"})
+			return
+		}
+		if !quotaReserve(totalSize) {
+			_ = os.RemoveAll(dir)
+			c.JSON(http.StatusInsufficientStorage, gin.H{"error": "quota exceeded"})
+			audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "size": totalSize, "result": "quota_exceeded"})
+			return
+		}
+		meta = &chunkMeta{
+			UploadID: uploadID, Filename: filename, TotalSize: totalSize,
+			TotalChunks: totalChunks, Received: map[int]bool{},
+			CreatedAt: time.Now(), Reserved: totalSize,
+		}
+		chunkMu.Lock()
+		chunkMetaMap[uploadID] = meta
+		err = saveMeta(meta)
+		chunkMu.Unlock()
+		if err != nil {
+			abortChunkUpload(uploadID, meta)
+			c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+			audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "meta_save_failed"})
+			return
+		}
+	} else if idx > 0 {
+		if _, err := os.Stat(dir); err != nil {
+			c.JSON(http.StatusNotFound, gin.H{"error": "upload not found"})
+			audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "result": "upload_not_found"})
+			return
+		}
+	}
+
+	// 写分片: multipart → 流式 io.Copy 落盘, 不进内存
+	dst := chunkFilePath(uploadID, idx)
+	in, err := chunk.Open()
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "result": "open_failed"})
+		return
+	}
+	out, err := os.Create(dst)
+	if err != nil {
+		in.Close()
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "result": "create_failed"})
+		return
+	}
+	_, cpErr := io.Copy(out, in)
+	in.Close()
+	if cerr := out.Close(); cerr != nil && cpErr == nil {
+		cpErr = cerr
+	}
+	if cpErr != nil {
+		_ = os.Remove(dst)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "result": "write_failed"})
+		return
+	}
+
+	// 更新 Received + 持久化 (持写锁, 防并发分片时 meta.json 丢进度)
+	chunkMu.Lock()
+	meta.Received[idx] = true
+	err = saveMeta(meta)
+	complete := len(meta.Received) == meta.TotalChunks
+	chunkMu.Unlock()
+	if err != nil {
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "result": "meta_save_failed"})
+		return
+	}
+
+	if !complete {
+		c.JSON(http.StatusAccepted, gin.H{"received": len(meta.Received), "total": meta.TotalChunks})
+		audit(c, "upload_chunk", gin.H{"upload_id": uploadID, "index": idx, "received": len(meta.Received), "total": meta.TotalChunks, "result": "ok"})
+		return
+	}
+	assembleChunks(c, uploadID, meta)
+}
+
+// assembleChunks 所有分片到位后顺序拼装落盘并清理.
+// ponytail: 末片重传/并发组装不额外去重, 前端单分片串行重试不会触发.
+func assembleChunks(c *gin.Context, uploadID string, meta *chunkMeta) {
+	dir := chunkDirPath(uploadID)
+	for i := 0; i < meta.TotalChunks; i++ {
+		info, err := os.Stat(chunkFilePath(uploadID, i))
+		if err != nil || info.Size() != expectedChunkSize(meta.TotalSize, i) {
+			abortChunkUpload(uploadID, meta)
+			c.JSON(http.StatusBadRequest, gin.H{"error": "chunk size mismatch"})
+			audit(c, "upload_assembled", gin.H{"upload_id": uploadID, "index": i, "result": "chunk_size_mismatch"})
+			return
+		}
+	}
+
+	name := uniqueFilename(*uploadDir, meta.Filename)
+	dst := filepath.Join(*uploadDir, name)
+	out, err := os.Create(dst)
+	if err != nil {
+		abortChunkUpload(uploadID, meta)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_assembled", gin.H{"upload_id": uploadID, "result": "create_failed"})
+		return
+	}
+	var realSize int64
+	for i := 0; i < meta.TotalChunks; i++ {
+		in, err := os.Open(chunkFilePath(uploadID, i))
+		if err != nil {
+			out.Close()
+			_ = os.Remove(dst)
+			abortChunkUpload(uploadID, meta)
+			c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+			audit(c, "upload_assembled", gin.H{"upload_id": uploadID, "index": i, "result": "open_failed"})
+			return
+		}
+		n, cErr := io.Copy(out, in)
+		in.Close()
+		if cErr != nil {
+			out.Close()
+			_ = os.Remove(dst)
+			abortChunkUpload(uploadID, meta)
+			c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+			audit(c, "upload_assembled", gin.H{"upload_id": uploadID, "index": i, "result": "copy_failed"})
+			return
+		}
+		realSize += n
+	}
+	if err := out.Close(); err != nil {
+		_ = os.Remove(dst)
+		abortChunkUpload(uploadID, meta)
+		c.JSON(http.StatusInternalServerError, gin.H{"error": "internal error"})
+		audit(c, "upload_assembled", gin.H{"upload_id": uploadID, "result": "close_failed"})
+		return
+	}
+
+	// 配额: 释放预留, 按真实大小重新占用
+	quotaSub(meta.Reserved)
+	quotaReserve(realSize)
+	chunkMu.Lock()
+	delete(chunkMetaMap, uploadID)
+	chunkMu.Unlock()
+	_ = os.RemoveAll(dir)
+
+	now := time.Now()
+	_ = os.Chtimes(dst, now, now)
+	c.JSON(http.StatusOK, fileResponse{
+		Filename: name, URL: "/download/" + name, Size: realSize,
+		UploadedAt: now, ExpiresAt: now.Add(*fileTTL),
+	})
+	audit(c, "upload_assembled", gin.H{
+		"upload_id": uploadID, "file": name, "size": realSize, "result": "ok",
+	})
+}
+
+// abortChunkUpload 失败时释放预留配额并清掉分片目录与内存记录
+func abortChunkUpload(uploadID string, meta *chunkMeta) {
+	quotaSub(meta.Reserved)
+	chunkMu.Lock()
+	delete(chunkMetaMap, uploadID)
+	chunkMu.Unlock()
+	_ = os.RemoveAll(chunkDirPath(uploadID))
+}
+
+// cleanupChunks 清理超过 maxAge 的孤儿分片目录并释放预留配额
+func cleanupChunks(maxAge time.Duration) {
+	entries, err := os.ReadDir(chunkDir)
+	if err != nil {
+		if !os.IsNotExist(err) {
+			log.Printf("[cleaner] read chunk dir failed: %v", err)
+		}
+		return
+	}
+	now := time.Now()
+	for _, e := range entries {
+		if !e.IsDir() {
+			continue
+		}
+		id := e.Name()
+		dir := chunkDirPath(id)
+		if !uploadIDRe.MatchString(id) {
+			_ = os.RemoveAll(dir) // 非法目录名 (非本服务生成), 直接清
+			continue
+		}
+		meta := loadMeta(id)
+		if meta == nil {
+			// 无 meta 的孤儿目录: 无法确定 reserved, 删目录即可 (quota 由 cleaner 的 rescanQuota 校正)
+			_ = os.RemoveAll(dir)
+			chunkMu.Lock()
+			delete(chunkMetaMap, id)
+			chunkMu.Unlock()
+			continue
+		}
+		if now.Sub(meta.CreatedAt) > maxAge {
+			quotaSub(meta.Reserved)
+			chunkMu.Lock()
+			delete(chunkMetaMap, id)
+			chunkMu.Unlock()
+			_ = os.RemoveAll(dir)
+			log.Printf("[cleaner] removed orphan chunks: %s (age=%v)", id, now.Sub(meta.CreatedAt))
+		}
+	}
+}
diff --git a/chunk_test.go b/chunk_test.go
new file mode 100644
index 0000000..ca5e8cd
--- /dev/null
+++ b/chunk_test.go
@@ -0,0 +1,84 @@
+package main
+
+import (
+	"bytes"
+	"mime/multipart"
+	"net/http"
+	"net/http/httptest"
+	"strconv"
+	"testing"
+
+	"github.com/gin-gonic/gin"
+)
+
+// 分片数学: 片数 ceil 与每片期望大小
+func TestChunkMath(t *testing.T) {
+	if got := expectedChunks(5<<30 + 1); got != 6 {
+		t.Fatalf("expectedChunks(5GB+1) = %d, want 6", got)
+	}
+	if got := expectedChunks(5 << 30); got != 5 {
+		t.Fatalf("expectedChunks(5GB) = %d, want 5", got)
+	}
+	if got := expectedChunkSize(5<<30+1, 0); got != chunkSize {
+		t.Fatalf("chunk0 size = %d, want 1GB", got)
+	}
+	if got := expectedChunkSize(5<<30+1, 5); got != 1 {
+		t.Fatalf("last chunk size = %d, want 1", got)
+	}
+}
+
+func chunkRequest(t *testing.T, uploadID string, idx, totalChunks int, filename string, totalSize int64, content []byte) *gin.Context {
+	t.Helper()
+	var buf bytes.Buffer
+	w := multipart.NewWriter(&buf)
+	fw, err := w.CreateFormFile("chunk", "chunk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	fw.Write(content)
+	w.Close()
+	req := httptest.NewRequest("POST", "/upload-chunk", &buf)
+	req.Header.Set("Content-Type", w.FormDataContentType())
+	req.Header.Set("X-Upload-Id", uploadID)
+	req.Header.Set("X-Chunk-Index", strconv.Itoa(idx))
+	req.Header.Set("X-Total-Chunks", strconv.Itoa(totalChunks))
+	req.Header.Set("X-Filename", filename)
+	req.Header.Set("X-Total-Size", strconv.FormatInt(totalSize, 10))
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	c.Request = req
+	return c
+}
+
+// 校验拒绝路径: 非法 id / 小于阈值 / 越界 / 片数不匹配 / 超配额, 均不落盘
+func TestChunkHandlerRejections(t *testing.T) {
+	*auditDir = "" // 测试不写审计日志
+	oldQuota := *quota
+	*quota = 1 << 30
+	defer func() { *quota = oldQuota }()
+
+	cases := []struct {
+		name   string
+		id     string
+		idx    int
+		total  int
+		fn     string
+		size   int64
+		status int
+	}{
+		{"invalid upload id", "bad id!", 0, 6, "x.bin", chunkThreshold + 1, http.StatusBadRequest},
+		{"below threshold", "uploadid123456", 0, 1, "x.bin", chunkThreshold, http.StatusBadRequest},
+		{"index out of range", "uploadid123456", 6, 6, "x.bin", chunkThreshold + 1, http.StatusBadRequest},
+		{"chunk count mismatch", "uploadid123456", 0, 2, "x.bin", chunkThreshold + 1, http.StatusBadRequest},
+		{"quota exceeded", "uploadid123456", 0, 6, "x.bin", chunkThreshold + 1, http.StatusInsufficientStorage},
+	}
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			c := chunkRequest(t, tc.id, tc.idx, tc.total, tc.fn, tc.size, []byte("data"))
+			uploadChunkHandler(c)
+			if c.Writer.Status() != tc.status {
+				t.Fatalf("status = %d, want %d", c.Writer.Status(), tc.status)
+			}
+		})
+	}
+}
diff --git a/cleaner.go b/cleaner.go
index b205674..695c1ba 100644
--- a/cleaner.go
+++ b/cleaner.go
@@ -51,6 +51,15 @@ func startCleaner() {
 			log.Fatalf("add cron job failed: %v", err)
 		}
 		c.Start()
+		// 分片目录独立清理: 复用同一周期, 清掉孤儿分片并释放预留配额
+		go func() {
+			cleanupChunks(*fileTTL) // 启动先清一轮, 处理上次进程遗留
+			ticker := time.NewTicker(*scanEvery)
+			defer ticker.Stop()
+			for range ticker.C {
+				cleanupChunks(*fileTTL)
+			}
+		}()
 		log.Printf("[cleaner] started, schedule=@every %s dir=%s ttl=%s",
 			scanEvery, *uploadDir, *fileTTL)
 	})
diff --git a/handlers.go b/handlers.go
index 87ff228..16da746 100644
--- a/handlers.go
+++ b/handlers.go
@@ -24,6 +24,12 @@ type fileResponse struct {
 	ExpiresAt  time.Time `json:"expires_at"`
 }
 
+// checkFilenameSafe 校验文件名: 不含 NUL 字节, 且不超过 200 字节 (Linux NAME_MAX=255 的余量).
+// 供 uploadHandler 与 uploadChunkHandler 复用 (NUL/200 校验只写这一份).
+func checkFilenameSafe(name string) bool {
+	return !strings.ContainsRune(name, 0) && len(name) <= 200
+}
+
 // validateFilename 从 URL/JSON 参数提取并校验 filename.
 // 返回 (sanitized, true) 表示安全可访问, 否则返回 ("", false).
 // 规则与 downloadHandler 原内联逻辑一致: filepath.Base + 拒绝 "" / "." / ".."
@@ -44,14 +50,14 @@ func uploadHandler(c *gin.Context) {
 		audit(c, "upload", gin.H{"file": "", "result": "missing_form_field"})
 		return
 	}
-	// 早 reject: 防止 NUL 字节 + 超长文件名打爆 Stat/Save
-	if strings.ContainsRune(f.Filename, 0) {
-		c.JSON(http.StatusBadRequest, gin.H{"error": "filename contains null byte"})
-		audit(c, "upload", gin.H{"file": f.Filename, "result": "null_byte"})
-		return
-	}
-	// Linux/macOS 路径分量 NAME_MAX = 255 字节
-	if len(f.Filename) > 200 {
+	// 早 reject: 防止 NUL 字节 + 超长文件名打爆 Stat/Save (规则与分片上传共用)
+	if !checkFilenameSafe(f.Filename) {
+		if strings.ContainsRune(f.Filename, 0) {
+			c.JSON(http.StatusBadRequest, gin.H{"error": "filename contains null byte"})
+			audit(c, "upload", gin.H{"file": f.Filename, "result": "null_byte"})
+			return
+		}
+		// Linux/macOS 路径分量 NAME_MAX = 255 字节
 		c.JSON(http.StatusBadRequest, gin.H{"error": "filename too long (max 200 bytes)"})
 		audit(c, "upload", gin.H{"file": f.Filename, "result": "filename_too_long"})
 		return
diff --git a/main.go b/main.go
index f0bcbd3..d9ccf58 100644
--- a/main.go
+++ b/main.go
@@ -58,6 +58,7 @@ func main() {
 
 	// 业务路由
 	r.POST("/upload", uploadHandler)
+	r.POST("/upload-chunk", uploadChunkHandler)
 	r.GET("/download/:filename", downloadHandler)
 	r.POST("/download-zip", batchDownloadHandler)
 	r.POST("/files-delete", batchDeleteHandler)
diff --git a/static/app.js b/static/app.js
index ab782d7..91763ff 100644
--- a/static/app.js
+++ b/static/app.js
@@ -90,6 +90,8 @@ function addFiles(fileList) {
       id: nextId++,
       file: f,
       xhr: null,
+      chunkXhrs: new Set(),
+      chunkProgress: null,
       status: 'pending',
       progress: 0,
       loaded: 0,
@@ -117,9 +119,9 @@ function renderPending() {
       item.status === 'error' ? '✗' :
       item.status === 'cancelled' ? '⊘' : '';
 
-    const sizeInfo = item.file.size > 0
+    const sizeInfo = (item.file.size > 0
       ? `${formatSize(item.loaded)} / ${formatSize(item.file.size)}`
-      : formatSize(item.file.size);
+      : formatSize(item.file.size)) + (item.file.size > CHUNK_THRESHOLD ? ' (分片上传)' : '');
 
     div.innerHTML = `
       <div class="progress-item-head">
@@ -179,6 +181,8 @@ function renderPending() {
 function cancelItem(id) {
   const item = pending.find(p => p.id === id);
   if (!item) return;
+  item.chunkXhrs.forEach(x => x.abort());
+  item.chunkXhrs.clear();
   if (item.xhr) {
     item.xhr.abort();
     item.status = 'cancelled';
@@ -191,8 +195,10 @@ function cancelItem(id) {
 
 function removeItem(id) {
   const item = pending.find(p => p.id === id);
-  if (item && item.xhr && item.xhr.readyState !== XMLHttpRequest.DONE) {
-    item.xhr.abort();
+  if (item) {
+    item.chunkXhrs.forEach(x => x.abort());
+    item.chunkXhrs.clear();
+    if (item.xhr && item.xhr.readyState !== XMLHttpRequest.DONE) item.xhr.abort();
   }
   pending = pending.filter(p => p.id !== id);
   renderPending();
@@ -201,6 +207,8 @@ function removeItem(id) {
 clearBtn.addEventListener('click', () => {
   // 取消所有进行中的上传
   pending.forEach(p => {
+    p.chunkXhrs.forEach(x => x.abort());
+    p.chunkXhrs.clear();
     if (p.xhr && p.xhr.readyState !== XMLHttpRequest.DONE) p.xhr.abort();
   });
   pending = [];
@@ -238,7 +246,7 @@ uploadBtn.addEventListener('click', async () => {
     item.error = null;
     renderPending();
     try {
-      await uploadOne(item);
+      await (item.file.size > CHUNK_THRESHOLD ? uploadChunked(item) : uploadOne(item));
       item.status = 'done';
       item.progress = 100;
     } catch (err) {
@@ -308,6 +316,86 @@ function uploadOne(item) {
   });
 }
 
+// ---------------------------------------------------------------------------
+// 大文件分片上传 (>5GB, 1GB/片)
+// ---------------------------------------------------------------------------
+async function uploadChunked(item) {
+  const uploadId = uuid();
+  const totalChunks = Math.ceil(item.file.size / CHUNK_SIZE);
+  item.chunkProgress = new Array(totalChunks).fill(0);
+  let result = null;
+  const tasks = Array.from({ length: totalChunks }, (_, i) => async () => {
+    if (item.status !== 'uploading') return; // 已取消则不再发起新分片
+    const r = await uploadOneChunk(item, uploadId, i, totalChunks);
+    if (i === totalChunks - 1) result = r; // 末片响应即整个上传结果
+  });
+  await runWithLimit(tasks, 3);
+  return result;
+}
+
+// 单分片上传: 失败最多重试 3 次, 退避 500/1000/2000ms
+function uploadOneChunk(item, uploadId, idx, total) {
+  return new Promise((resolve, reject) => {
+    const delays = [500, 1000, 2000];
+    const attempt = n => {
+      if (item.status === 'cancelled') {
+        const e = new Error('aborted'); e.name = 'AbortError'; reject(e);
+        return;
+      }
+      const xhr = new XMLHttpRequest();
+      item.chunkXhrs.add(xhr);
+      const fd = new FormData();
+      const start = idx * CHUNK_SIZE;
+      const end = Math.min(start + CHUNK_SIZE, item.file.size);
+      fd.append('chunk', item.file.slice(start, end));
+      xhr.open('POST', '/upload-chunk');
+      xhr.setRequestHeader('X-Upload-Id', uploadId);
+      xhr.setRequestHeader('X-Chunk-Index', String(idx));
+      xhr.setRequestHeader('X-Total-Chunks', String(total));
+      xhr.setRequestHeader('X-Filename', item.file.name);
+      xhr.setRequestHeader('X-Total-Size', String(item.file.size));
+      xhr.upload.onprogress = e => {
+        if (e.lengthComputable) {
+          item.chunkProgress[idx] = e.loaded;
+          item.loaded = item.chunkProgress.reduce((s, v) => s + v, 0);
+          item.progress = item.file.size ? Math.round(item.loaded / item.file.size * 100) : 0;
+          renderPending();
+        }
+      };
+      const cleanup = () => item.chunkXhrs.delete(xhr);
+      xhr.onload = () => {
+        cleanup();
+        if (xhr.status >= 200 && xhr.status < 300) {
+          if (idx === total - 1) {
+            let res = {};
+            try { res = JSON.parse(xhr.responseText); } catch {}
+            resolve(res);
+          } else {
+            resolve(undefined);
+          }
+        } else if (n < 4) {
+          setTimeout(() => attempt(n + 1), delays[n - 1]);
+        } else {
+          let msg = 'HTTP ' + xhr.status;
+          try { msg = JSON.parse(xhr.responseText).error || msg; } catch {}
+          reject(new Error(msg));
+        }
+      };
+      xhr.onerror = () => {
+        cleanup();
+        if (n < 4) setTimeout(() => attempt(n + 1), delays[n - 1]);
+        else reject(new Error('network error'));
+      };
+      xhr.onabort = () => {
+        cleanup();
+        const e = new Error('aborted'); e.name = 'AbortError'; reject(e);
+      };
+      xhr.send(fd);
+    };
+    attempt(1);
+  });
+}
+
 // ---------------------------------------------------------------------------
 // File list
 // ---------------------------------------------------------------------------
diff --git a/static/helpers.js b/static/helpers.js
index 9993069..0e16757 100644
--- a/static/helpers.js
+++ b/static/helpers.js
@@ -1,5 +1,22 @@
 'use strict';
 
+// ---------------------------------------------------------------------------
+// 大文件分片上传常量 (见 app.js uploadChunked/uploadOneChunk)
+// ---------------------------------------------------------------------------
+const CHUNK_THRESHOLD = 5 * 1024 * 1024 * 1024; // >5GB 走分片
+const CHUNK_SIZE = 1 * 1024 * 1024 * 1024;      // 每片 1GB
+
+function uuid() {
+  // crypto.randomUUID 仅在 secure context (HTTPS / localhost) 可用, HTTP 非 localhost 会抛.
+  // 用 crypto.getRandomValues (所有现代浏览器均支持, 不要求 secure context) 手动拼一个 v4.
+  const b = new Uint8Array(16);
+  crypto.getRandomValues(b);
+  b[6] = (b[6] & 0x0f) | 0x40; // version 4
+  b[8] = (b[8] & 0x3f) | 0x80; // variant RFC 4122
+  const h = [...b].map(x => x.toString(16).padStart(2, '0')).join('');
+  return `${h.slice(0,8)}-${h.slice(8,12)}-${h.slice(12,16)}-${h.slice(16,20)}-${h.slice(20)}`;
+}
+
 // ---------------------------------------------------------------------------
 // 纯函数 + 倒计时 interval
 // 跨文件共享 (classic <script> 加载顺序: app.js → helpers.js → batch.js).