Turbopack Jest worker 在 LLM 生成大内容时崩溃

[emmett001]

Bug Description

当 LLM Agent 生成较大响应时（例如根据多章节课程大纲生成完整课堂），Next.js 16 Turbopack 内建的 Jest worker进程会因内存不足崩溃，导致整个 dev server 挂掉。

Steps to Reproduce

1. npm run dev
2. 将一份约 7 章的大学课程大纲（如《信号与系统》：傅里叶、拉普拉斯、z 变换、抽样、调制等，约 1200
   字结构化文本）粘贴到聊天输入框
3. 要求 Agent 根据此内容生成课堂
4. 等待 LLM 生成完毕
5. 观察到崩溃

Expected Behavior

课堂正常生成并在界面中渲染。

Actual Behavior

Runtime Error
Jest worker encountered 2 child process exceptions, exceeding retry limit

Dev server 无响应，必须重启。

Deployment Method

Local development (npm run dev / pnpm dev / yarn dev)

Browser

148.0.7778.179（64 位稳定版）

Operating System

Windows11专业版

Relevant Logs / Screenshots

Additional Context

No response

[wyuc]

Thanks for the detailed repro, this was enough to trace it.

Short version: this is a dev-only limitation and not a bug in OpenMAIC's own code.

What happens: in dev mode, Next 16 forks an internal worker process for any App Router dynamic route (here /classroom/[id]), to evaluate generateStaticParams. This happens even when the route doesn't define one. When you generate a large classroom, the dev process is holding a big in-memory object right as that worker is forked, so under memory pressure the worker child gets OOM-killed, retried once, and killed again. That surfaces as Jest worker encountered 2 child process exceptions, exceeding retry limit and takes the dev server down with it. The "Jest worker" name is just Next's internal process pool and has nothing to do with testing.

There's no route-level config that disables that worker, so this is effectively an upstream Next dev-server robustness limitation rather than something OpenMAIC can fix directly.

Workaround for generating large content locally, raise the dev heap limit, sized to your machine's RAM (don't over-set it or you trade a V8 OOM for a system-level one):

# macOS / Linux
NODE_OPTIONS=--max-old-space-size=4096 npm run dev

# Windows (PowerShell)
$env:NODE_OPTIONS="--max-old-space-size=4096"; npm run dev

Production builds are unaffected (they're prebuilt and never fork this worker). Marking as dev-only / low priority.

[mvanhorn]

Implemented in #808 - documents your dev-only NODE_OPTIONS heap workaround in the Getting Started troubleshooting docs.