Fix three import errors that break modules on a fresh checkout

[egeboy35]

Three modules currently fail to import on a fresh checkout. Each fix is small and verifiable without SDR hardware.

Problem and fix

• sdradi/myofdm.py imports tfmodules at module level. That file lives in sdradi/pluto_test/ and requires TensorFlow, so import myofdm fails from sdradi/ even though only MyDemapper uses those names. → The import is now optional; MyDemapper raises an ImportError explaining how to enable it.
• WebApp/FastAPIbackend/radar/waveform.py runs pip install plotly as a side effect of a failed import. → It now raises a clear ImportError with the install command instead. It also adds a __file__-based sys.path entry so its AIRadar.AIRadarLib imports resolve when the backend is run from WebApp/FastAPIbackend/ as the README describes.
• AIRadar/AIradar_model_comparison.py imports RadarTransformerNet from AIradar_transformer, which is not in the repository. → It now imports from AIRadarLib.modeling_transformer, matching the sibling script AIradar_transformer_train.py; the constructor arguments already line up.

Verification

In a clean venv without TensorFlow:

cd sdradi && python -c "import myofdm"                          # succeeds
cd WebApp/FastAPIbackend && python -c "import radar.waveform"    # succeeds

MyDemapper now raises the new explanatory error when tfmodules is unavailable, and RadarTransformerNet resolves from AIRadarLib.

Not included

While tracing these I noticed a few more import problems I did not want to guess at: PacketFramer is imported from sdr_video_comm by four scripts but does not appear to be defined anywhere in the repository, AIradar_fmcw_dataset is missing, and AIradar_model_comparison.py also imports RadarNet / RadarTimeToFreqNet from AImodels_joint, where those names are not defined. I left all of those alone — happy to follow up if you can point me at the intended sources. The WebApp/FastAPIbackend/main.py startup path also needs work, but part of it only runs with an SDR attached, so I have left it for when it can actually be exercised.

🤖 Generated with Claude Code