If you want to identify stuff, you'll be looking for an energy spectrum.
If you're after an actual spectrum and you're thinking DIY then a starting point is a thalium doped sodium iodine cyrstal and electronics
https://alphaspectra.com/scintillation-detector-manufacturin...
https://www.alibaba.com/product-detail/2-inches-NaI-Scintill...
Alibaba is listing a single 50mm round x 50mm thick crystal plus electronics at 3K USD each (less than 10) which seems steep to me given years back we used 42 litre crystal packs (with CSIRO grown crystals and in house PhD electronics engineers, etc - so YMMV).
Now your problem is sampling the output levels several thousand times a second, binning the counts, and having a calibration source and compensating for tempreture drift.
You can buy a box that'll do that for you or you can <cough> DIY that with reference to radiometric survey field guides (there's one from AGSO - probably Bob Minty has his name on it).
A good all in one handheld is probably something like https://www.radiationsolutions.ca/wp-content/uploads/2020/03...
which is one of those "contact us and get a quote" jobs from a Canadian company of good repute with Jens Hovgaard, a Finn, as the Tech CEO | President.
He knows his stuff and has an industry algorithm named after him .. although others did similar work elsewhere about the globe.
A relevant spec sheet to compare against what was outlined above is:
https://www.gammadata.se/assets/Uploads/CsITl-and-Na-data-sh...
If you're in the earth sciences, I linked a data sheet for the crystal: https://www.gammadata.se/assets/Uploads/CsITl-and-Na-data-sh...
There's a reddit community: https://old.reddit.com/r/Radiacode/
and a desktop interface: https://downloads.radiacode.com/EN/RC-101_Windows.pdf
with a few OS choices.
For semi serious use I'd want to continuously download full spectra with a (?) 2 to five second window (subject to understanding how well it counts, subsamples, saturates) and pool those spectra with GPS data to build up NASVD type "typical background" fingerprints to differentiate against to boost outlying signals.
I'd like to see the gadget user groups pooling their data also.
First thoughts on browsing the manual is that spectra export is a UI interface driven one at a time operation so someone would want to get into the guts of that and determine how to continuosly transfer to a server | home PC | nearby phone to retransmit, etc.
This, of course, might be in the reddit forum or Telegram group.
I might get a few of these as Xmas presents for a few folk I know that work with the bigger geophysical survey toys to get some cross over happening.