手写 SSE vs Vercel AI SDK

手写版 · 关键路径约 20 行

// 手写版:app/api/ai/stream/route.ts(节选关键路径,约 20 行)
export async function GET(req: NextRequest) {
    const encoder = new TextEncoder();
    let cancelled = false;

    const stream = new ReadableStream<Uint8Array>({
        async start(controller) {
            for (const token of TOKENS) {
                if (cancelled || req.signal.aborted) { break; }   // 取消检查 ×2
                await sleep(nextDelay());
                if (cancelled || req.signal.aborted) { break; }
                controller.enqueue(                                  // 自己拼帧
                    encoder.encode(`data: ${JSON.stringify({ token })}\n\n`),
                );
            }
            controller.enqueue(encoder.encode("data: [DONE]\n\n")); // 终止哨兵
            controller.close();
        },
        cancel() { cancelled = true; },                              // 取消回调
    });

    return new Response(stream, {
        headers: { "Content-Type": "text/event-stream; charset=utf-8",
                   "Cache-Control": "no-cache, no-transform",
                   "X-Accel-Buffering": "no" },                       // 防代理缓冲
    });
}

SDK 版 · 核心 5 行

// SDK 版:app/api/ai/sdk-chat/route.ts(核心就这几行)
export async function POST(req: Request) {
    const { messages }: { messages: UIMessage[] } = await req.json();
    const result = streamText({
        model: mockModel,  // 生产换成 openai("gpt-5"),其余不动
        messages: await convertToModelMessages(messages),
    });
    return createUIMessageStreamResponse({
        stream: toUIMessageStream({ stream: result.stream }),
    });
}
// 帧格式、取消传播、防缓冲响应头、错误帧 —— 全部由 SDK 维护