ai-chatbot/app/api/canvas/audio-separate/route.ts

import { createHash, randomUUID } from "node:crypto";
import { mkdtemp, readFile, rm, stat, writeFile } from "node:fs/promises";
import { tmpdir } from "node:os";
import path from "node:path";
import { NextResponse } from "next/server";
import { auth } from "@/app/(auth)/auth";
import { uploadFile } from "@/lib/storage/service";
import { ffmpeg, mediaSafety, probeMedia } from "@/lib/video-canvas/smart-edit/local";
import {
  remoteAudio,
  remoteAudioConfig,
} from "@/lib/video-canvas/smart-edit/remote-audio";

export const runtime = "nodejs";
export const maxDuration = 600;

const running = new Set<string>();
const allowedStems = ["vocals", "background"] as const;

export async function POST(request: Request) {
  const session = await auth();
  const userId = session?.user?.id;
  if (!userId)
    return NextResponse.json({ success: false, error: "未登录" }, { status: 401 });
  const body = (await request.json().catch(() => null)) as {
    url?: unknown;
    stem?: unknown;
  } | null;
  const url = typeof body?.url === "string" ? body.url.trim() : "";
  const stem = typeof body?.stem === "string" ? body.stem : "";
  if (!/^https?:\/\//i.test(url) || !allowedStems.includes(stem as (typeof allowedStems)[number]))
    return NextResponse.json({ success: false, error: "参数不完整" }, { status: 400 });
  if (!remoteAudioConfig())
    return NextResponse.json(
      { success: false, error: "音频分离服务未配置，请联系管理员" },
      { status: 503 }
    );
  const lockKey = `${userId}:${url}:${stem}`;
  if (running.has(lockKey) || running.size >= 3)
    return NextResponse.json(
      { success: false, error: "已有分离任务在进行，请稍候" },
      { status: 429 }
    );
  running.add(lockKey);
  let dir = "";
  try {
    dir = await mkdtemp(path.join(tmpdir(), "canvas-separate-"));
    const input = path.join(dir, "input");
    const response = await fetch(url, {
      signal: AbortSignal.timeout(180000),
      cache: "no-store",
    });
    if (!response.ok) throw new Error("素材下载失败，请确认素材仍可访问");
    const bytes = Buffer.from(await response.arrayBuffer());
    if (!bytes.length || bytes.length > 150 * 1024 * 1024)
      throw new Error("素材必须非空且不超过 125 MiB");
    await writeFile(input, bytes);
    const info = await probeMedia(input, 600);
    if (!info.hasAudio) throw new Error("当前视频没有音轨");
    const inputAudio = path.join(dir, "separation-input.wav");
    await ffmpeg(
      [
        ...mediaSafety,
        "-i",
        input,
        "-vn",
        "-af",
        "apad",
        "-ac",
        "2",
        "-ar",
        "44100",
        "-c:a",
        "pcm_f32le",
        inputAudio,
      ],
      undefined,
      240000
    );
    const requestKey = createHash("sha256")
      .update(`${url}:${stem}:${(await stat(inputAudio)).size}`)
      .digest("hex");
    await remoteAudio("separate", inputAudio, userId, requestKey, dir);
    const stemFile = path.join(dir, `${stem}.wav`);
    const wav = await readFile(stemFile);
    if (wav.length < 1024) throw new Error("分离结果异常，请稍后重试");
    const uploaded = await uploadFile(
      `canvas-separate-${randomUUID()}.wav`,
      await new Blob([wav]).arrayBuffer()
    );
    return NextResponse.json({
      success: true,
      data: {
        url: uploaded.url,
        duration: Math.round(info.duration / 1e6),
      },
    });
  } catch (error) {
    return NextResponse.json(
      {
        success: false,
        error:
          error instanceof Error ? error.message : "人声分离失败，原音轨已保留",
      },
      { status: 500 }
    );
  } finally {
    running.delete(lockKey);
    if (dir) await rm(dir, { recursive: true, force: true });
  }
}