In many cases, the tokenizer-s3 library can significantly enhance the speed of S3 file detection.

I am not in favor of relying solely on matching the ppt/ path, as this approach feels more like an educated guess rather than a reliable method for identifying a pptx file.

A more robust solution might involve extracting and analyzing the docProps/app.xml [Content_Types].xml file, which offers a much more specialized detection mechanism. However, this approach requires decompression and XML parsing, which adds complexity. Consequently, it might be better to delegate this functionality to an add-on for Office or zipped formats.

Yes, I do use the tokenizer/s3 library.

My code is more or less:

import { fileTypeFromTokenizer } from "file-type";
import { makeTokenizer } from "@tokenizer/s3";

const tokenizer = await makeTokenizer(s3Client, {
			Bucket: location.bucket,
			Key: location.key
		});

const type = await fileTypeFromTokenizer(tokenizer);
await tokenizer.close();