I'm trying to download a zip file, untar it, pull out two files and send them to another service.

This route fails to deploy when I call tar.extract() so I'm guessing vercel blocks this (doesn't throw a helpful error message, just fails silently).

My guess is that I'll need to stream this zip file into memory instead of disk (but these files are 200-300MB so that could get fun)

Am I right that Vercel doesn't allow this level of file i/o? Is that documented anywhere? Or is it maybe just an issue with this 3rd party tar package?

Curious if others are able to do this and what the best practices are around it.

    const pipeline = promisify(stream.pipeline);
    const uniqueId = uuidv4();
    const isDev = process.env.NODE_ENV === 'development';
    console.log('Is dev: ', isDev)
    const tempDir = isDev? join(tmpdir(), uniqueId) : '/tmp' ;
    if(isDev) {
      await mkdirSync(tempDir, { recursive: true });
    }
    const tempFilePath = join(tempDir, `trained_model.tar`);

    const response = await axios({
      method: 'get',
      url: url,
      responseType: 'stream',
    });

    console.log('Fetched tar file, starting stream');

    await pipeline(response.data, createWriteStream(tempFilePath));

    await tar.extract({ cwd: tempDir, file: tempFilePath });