Use Google Magika for file type detection instead of Apache Tika

[vlsi]

Use case

Currently, JMeter uses Tika to detect file type

Possible solution

We could replace Tika with https://google.github.io/magika/

On the other hand, Magika depends on Tensorflow, which might be a non-trivial dependency

Possible workarounds

No response

JMeter Version

5.6.3

Java Version

No response

OS Version

No response

[FSchumacher]

Do we have a real problem with using Tika?
I read the repo of magika, that it is a python (and javascript?) solution. Is it easy to add to our dependencies? Is it working locally (without internet access)?

[vlsi]

I thought tika consumed significant space dependency-wise.
Magika model is ~1MiB. However, Magika requires tensorflow, so it would probably involve a lot of deps :(
Yes, it works without Internet access