feat: 拆分引入本地视频上传支持

合并维护者拆分分支，保留 PR #68 中低风险且边界清楚的改动：本地视频上传、OpenAI Base URL 规范化、Whisper 缓存校验失败重试、ffmpeg/ffprobe 路径配置，以及对应测试。

Author: liuzhao1225

.env.example

@@ -7,6 +7,13 @@ CORS_ALLOW_ORIGIN_REGEX=
 # Device: cuda, cuda:0, cpu, or auto
 DEVICE=cuda
 
+# Optional explicit media binary paths when they are not on PATH.
+# FFMPEG_PATH=C:\path\to\ffmpeg.exe
+# FFPROBE_PATH=C:\path\to\ffprobe.exe
+
+# Maximum local video upload size in bytes. Default: 4 GiB.
+LOCAL_UPLOAD_MAX_BYTES=4294967296
+
 # OpenAI-compatible translator
 OPENAI_BASE_URL=https://api.openai.com/v1
 OPENAI_API_KEY=

apps/web/src/app/page.tsx

@@ -2,13 +2,15 @@
 
 import Link from "next/link"
 import { useRouter } from "next/navigation"
-import { FormEvent, useEffect, useState } from "react"
-import { ChevronRight, Play } from "lucide-react"
+import { ChangeEvent, FormEvent, useEffect, useRef, useState } from "react"
+import { ChevronRight, Play, Upload } from "lucide-react"
 
 import {
   TaskSummary,
+  LocalDirection,
   createTask,
   listTasks,
+  uploadLocalTask,
 } from "@/lib/api"
 import { useI18n } from "@/lib/i18n"
 import { statusBadgeClass } from "@/lib/status"
@@ -24,6 +26,13 @@ import {
 import { Input } from "@/components/ui/input"
 import { Label } from "@/components/ui/label"
 import { ScrollArea } from "@/components/ui/scroll-area"
+import {
+  Select,
+  SelectContent,
+  SelectItem,
+  SelectTrigger,
+  SelectValue,
+} from "@/components/ui/select"
 
 function isActive(status: string) {
   return status === "queued" || status === "running"
@@ -47,8 +56,11 @@ function activeCount(tasks: TaskSummary[]) {
 export default function Home() {
   const router = useRouter()
   const { activeTasksText, stageLabel, statusLabel, t } = useI18n()
+  const fileInputRef = useRef<HTMLInputElement>(null)
   const [youtubeUrl, setYoutubeUrl] = useState("")
   const [bilibiliUrl, setBilibiliUrl] = useState("")
+  const [localFile, setLocalFile] = useState<File | null>(null)
+  const [localDirection, setLocalDirection] = useState<LocalDirection>("en-zh")
   const [tasks, setTasks] = useState<TaskSummary[]>([])
   const [error, setError] = useState("")
   const [submitting, setSubmitting] = useState(false)
@@ -78,16 +90,27 @@ export default function Home() {
     }
   }, [t.home.loadError])
 
+  function selectLocalFile(event: ChangeEvent<HTMLInputElement>) {
+    setError("")
+    setLocalFile(event.target.files?.[0] || null)
+  }
+
   async function submitTask(event: FormEvent<HTMLFormElement>) {
     event.preventDefault()
     setError("")
     const submittedUrl = youtubeUrl.trim() || bilibiliUrl.trim()
-    if (!submittedUrl) return
+    if (!submittedUrl && !localFile) return
     setSubmitting(true)
     try {
-      const created = await createTask(submittedUrl)
+      const created = localFile
+        ? await uploadLocalTask(localFile, localDirection)
+        : await createTask(submittedUrl)
       setYoutubeUrl("")
       setBilibiliUrl("")
+      setLocalFile(null)
+      if (fileInputRef.current) {
+        fileInputRef.current.value = ""
+      }
       refreshTasks().catch(() => undefined)
       router.push(`/tasks/${created.id}`)
     } catch (err) {
@@ -98,7 +121,9 @@ export default function Home() {
   }
 
   const queued = activeCount(tasks)
-  const canSubmit = Boolean((youtubeUrl.trim() || bilibiliUrl.trim()) && !submitting)
+  const hasUrl = Boolean(youtubeUrl.trim() || bilibiliUrl.trim())
+  const hasLocalFile = Boolean(localFile)
+  const canSubmit = Boolean((hasUrl || hasLocalFile) && !submitting)
 
   return (
     <main className="min-h-screen bg-[linear-gradient(135deg,#fff5f5_0%,#f2fbff_48%,#fff4fa_100%)] text-foreground">
@@ -118,7 +143,7 @@ export default function Home() {
                   value={youtubeUrl}
                   onChange={(event) => setYoutubeUrl(event.target.value)}
                   placeholder="https://www.youtube.com/watch?v=..."
-                  disabled={Boolean(bilibiliUrl.trim())}
+                  disabled={Boolean(bilibiliUrl.trim()) || hasLocalFile}
                 />
               </div>
               <div className="space-y-2">
@@ -128,9 +153,38 @@ export default function Home() {
                   value={bilibiliUrl}
                   onChange={(event) => setBilibiliUrl(event.target.value)}
                   placeholder="https://www.bilibili.com/video/BV..."
-                  disabled={Boolean(youtubeUrl.trim())}
+                  disabled={Boolean(youtubeUrl.trim()) || hasLocalFile}
                 />
               </div>
+              <div className="grid gap-3 sm:grid-cols-[1fr_180px]">
+                <div className="space-y-2">
+                  <Label htmlFor="local-video">{t.home.localVideoLabel}</Label>
+                  <Input
+                    ref={fileInputRef}
+                    id="local-video"
+                    type="file"
+                    accept="video/*,.mp4,.mov,.m4v,.mkv,.webm,.avi,.flv,.wmv"
+                    onChange={selectLocalFile}
+                    disabled={hasUrl}
+                  />
+                </div>
+                <div className="space-y-2">
+                  <Label htmlFor="local-direction">{t.home.localDirectionLabel}</Label>
+                  <Select
+                    value={localDirection}
+                    onValueChange={(value) => setLocalDirection(value as LocalDirection)}
+                    disabled={hasUrl}
+                  >
+                    <SelectTrigger id="local-direction" className="h-10">
+                      <SelectValue />
+                    </SelectTrigger>
+                    <SelectContent>
+                      <SelectItem value="en-zh">{t.home.localEnZh}</SelectItem>
+                      <SelectItem value="zh-en">{t.home.localZhEn}</SelectItem>
+                    </SelectContent>
+                  </Select>
+                </div>
+              </div>
               <div className="flex items-center justify-between gap-3">
                 {queued > 0 ? (
                   <p className="text-xs text-muted-foreground">
@@ -140,7 +194,7 @@ export default function Home() {
                   <span />
                 )}
                 <Button type="submit" disabled={!canSubmit}>
-                  <Play className="size-4" />
+                  {hasLocalFile ? <Upload className="size-4" /> : <Play className="size-4" />}
                   {submitting ? t.home.submitting : t.home.createTask}
                 </Button>
               </div>

apps/web/src/lib/api.ts

@@ -60,6 +60,8 @@ export type YtdlpSettings = {
   proxy_port: string
 }
 
+export type LocalDirection = "en-zh" | "zh-en"
+
 async function request<T>(path: string, options?: RequestInit): Promise<T> {
   const response = await fetch(`${API_BASE}${path}`, {
     ...options,
@@ -131,6 +133,23 @@ export function createTask(url: string) {
   })
 }
 
+export async function uploadLocalTask(file: File, direction: LocalDirection) {
+  const form = new FormData()
+  form.append("direction", direction)
+  form.append("file", file)
+
+  const response = await fetch(`${API_BASE}/api/tasks/upload`, {
+    method: "POST",
+    body: form,
+    cache: "no-store",
+  })
+  if (!response.ok) {
+    const body = await response.json().catch(() => ({}))
+    throw new Error(body.detail || `Request failed: ${response.status}`)
+  }
+  return response.json() as Promise<Task>
+}
+
 export function getCookieInfo() {
   return request<CookieInfo>("/api/cookies/youtube")
 }

apps/web/src/lib/i18n.tsx

@@ -41,10 +41,14 @@ const messages: Record<UiLanguage, Messages> = {
       createTitle: "Create new task",
       youtubeLabel: "YouTube URL (English -> Chinese)",
       bilibiliLabel: "Bilibili URL (Chinese -> English)",
+      localVideoLabel: "Local video file",
+      localDirectionLabel: "Translation direction",
+      localEnZh: "English -> Chinese",
+      localZhEn: "Chinese -> English",
       submitting: "Submitting",
       createTask: "Create task",
       taskHistory: "Task history",
-      empty: "No tasks yet. Submit a YouTube or Bilibili URL above to start.",
+      empty: "No tasks yet. Submit a URL or upload a local video above to start.",
       loadError: "Failed to load tasks",
       createError: "Failed to create task",
     },
@@ -147,10 +151,14 @@ const messages: Record<UiLanguage, Messages> = {
       createTitle: "新建任务",
       youtubeLabel: "YouTube 链接（英文 -> 中文）",
       bilibiliLabel: "Bilibili 链接（中文 -> 英文）",
+      localVideoLabel: "本地视频文件",
+      localDirectionLabel: "翻译方向",
+      localEnZh: "英文 -> 中文",
+      localZhEn: "中文 -> 英文",
       submitting: "提交中",
       createTask: "创建任务",
       taskHistory: "任务历史",
-      empty: "暂无任务。输入 YouTube 或 Bilibili 链接后即可开始。",
+      empty: "暂无任务。输入链接或上传本地视频后即可开始。",
       loadError: "加载任务失败",
       createError: "创建任务失败",
     },

backend/app/adapters/ffmpeg.py

@@ -5,6 +5,8 @@
 import subprocess
 from pathlib import Path
 
+from ..config import ffmpeg_binary, ffprobe_binary
+
 SUBTITLE_PUNCTUATION = {"，", ",", "；", ";", "：", ":", "。", "?", "？", "!", "！", "、"}
 SUBTITLE_PROTECTED_PAIRS = {"《": "》", "（": "）", "【": "】", "「": "」", "『": "』"}
 SUBTITLE_CLOSING_QUOTES = {'"', "'", "」", "』", "》", "）", "】", "\u201d", "\u2019", "]"}
@@ -187,7 +189,7 @@ def write_srt(translation_file: Path, session: Path) -> Path:
 def probe_video_size(video_file: Path) -> tuple[int, int] | None:
     result = subprocess.run(
         [
-            "ffprobe",
+            ffprobe_binary(),
             "-v",
             "error",
             "-select_streams",
@@ -251,7 +253,7 @@ def merge_video(video_file: Path, dubbing_file: Path, bgm_file: Path, timings_fi
     mixed_audio = tmp_dir / "audio_mixed.m4a"
     subprocess.run(
         [
-            "ffmpeg",
+            ffmpeg_binary(),
             "-y",
             "-i",
             str(dubbing_file),
@@ -269,7 +271,7 @@ def merge_video(video_file: Path, dubbing_file: Path, bgm_file: Path, timings_fi
     )
     subprocess.run(
         [
-            "ffmpeg",
+            ffmpeg_binary(),
             "-y",
             "-i",
             str(video_file),

backend/app/adapters/local_video.py

@@ -0,0 +1,110 @@
+from __future__ import annotations
+
+import json
+import shutil
+import subprocess
+from pathlib import Path
+from urllib.parse import parse_qs, urlparse
+
+from ..config import ffmpeg_binary
+from ..sanitize import sanitize_text
+from ..sources import SourceConfig
+from ..youtube import local_upload_task_id
+
+
+def upload_dir(workfolder: Path, task_id: str) -> Path:
+    return workfolder / "_uploads" / task_id
+
+
+def remove_upload(workfolder: Path, task_id: str) -> None:
+    target = upload_dir(workfolder, task_id)
+    if target.exists():
+        shutil.rmtree(target)
+
+
+def _uploaded_video_file(workfolder: Path, task_id: str) -> Path:
+    root = upload_dir(workfolder, task_id)
+    if not root.exists():
+        raise FileNotFoundError(f"Local upload is missing for task {task_id}.")
+
+    files = sorted(path for path in root.iterdir() if path.is_file())
+    if not files:
+        raise FileNotFoundError(f"Local upload is missing for task {task_id}.")
+    if len(files) > 1:
+        raise RuntimeError(f"Local upload has multiple files for task {task_id}.")
+    return files[0]
+
+
+def _title_from_url(url: str, source_file: Path) -> str:
+    query = parse_qs(urlparse(url.strip()).query)
+    filename = (query.get("filename") or [""])[0].strip()
+    if filename:
+        return Path(filename).stem or source_file.stem
+    return source_file.stem
+
+
+def _session_path(workfolder: Path, task_id: str, title: str) -> Path:
+    safe_title = sanitize_text(title) or "local-video"
+    return workfolder / "local" / f"{safe_title}__{task_id}"
+
+
+def _transcode_to_mp4(source_file: Path, video_file: Path) -> None:
+    subprocess.run(
+        [
+            ffmpeg_binary(),
+            "-y",
+            "-i",
+            str(source_file),
+            "-map",
+            "0:v:0",
+            "-map",
+            "0:a:0?",
+            "-c:v",
+            "libx264",
+            "-preset",
+            "fast",
+            "-crf",
+            "23",
+            "-c:a",
+            "aac",
+            "-movflags",
+            "+faststart",
+            str(video_file),
+        ],
+        check=True,
+    )
+
+
+def import_local_video(url: str, workfolder: Path, source: SourceConfig) -> tuple[Path, dict]:
+    task_id = local_upload_task_id(url)
+    if not task_id:
+        raise ValueError("Invalid local upload URL.")
+
+    source_file = _uploaded_video_file(workfolder, task_id)
+    title = _title_from_url(url, source_file)
+    session = _session_path(workfolder, task_id, title)
+    media_dir = session / "media"
+    metadata_dir = session / "metadata"
+    media_dir.mkdir(parents=True, exist_ok=True)
+    metadata_dir.mkdir(parents=True, exist_ok=True)
+
+    video_file = media_dir / "video_source.mp4"
+    info = {
+        "id": task_id,
+        "title": title,
+        "source": "local",
+        "webpage_url": url,
+        "original_path": str(source_file),
+        "asr_language": source.asr_language,
+        "target_language": source.target_language,
+    }
+    metadata_file = metadata_dir / "local_info.json"
+    metadata_file.write_text(json.dumps(info, ensure_ascii=False, indent=2), encoding="utf-8")
+
+    if video_file.exists() and video_file.stat().st_size > 0:
+        return session, info
+
+    _transcode_to_mp4(source_file, video_file)
+    if not video_file.exists() or video_file.stat().st_size == 0:
+        raise RuntimeError("ffmpeg finished without producing media/video_source.mp4")
+    return session, info

backend/app/adapters/openai_client.py

@@ -0,0 +1,11 @@
+from __future__ import annotations
+
+
+def normalize_openai_base_url(base_url: str) -> str:
+    url = base_url.strip().rstrip("/")
+    lowered = url.lower()
+    for suffix in ("/chat/completions", "/completions"):
+        if lowered.endswith(suffix):
+            url = url[: -len(suffix)].rstrip("/")
+            lowered = url.lower()
+    return url or "https://api.openai.com/v1"