I did actually try this a back in 2021, but apparently didn't post my notes

Anyway, here they are
It identifies the arm and xtensa firmware. It unsurprisingly did not ID the jpcore or mecha blobs I tried.
Basic usage is just install per the readme and run
./cpu_rec.py PRIMARY.BIN
You can use it directly in Ghidra python like
>>> import sys
>>> sys.path.append('c:\\path\to\cpu_rec')
>>> from cpu_rec import which_arch
>>> which_arch(str(bytearray(getBytes(toAddr('fc020000'),0x1000))))
'ARMhf'
The prompt re-appears before it finishes executing, which can take a long time for large address ranges.
Since Ghidra python doesn't include lzma, you need to unzip the corpus files first, like
xz -d *.xz
The command line variant tries sliding windows of various sizes over the whole file, which IDed part of the jpcore blob as 6502 (virtually certain to be wrong).