add models.json: machine-readable index with backend support, input contract and checksums

Update SHA256SUMS for DFT-truncated embedding backbone

Add DFT-truncated fp32 embedding backbone (bit-exact, ~2x faster; keeps both outputs, drop-in, new filename)

checksums: add clean-named variants

Add birdnet-v2.4-embeddings-fp32.onnx (clean canonical filename, add-only)

docs: add explicit BirdNET attribution (Powered by BirdNET)

Upload folder using huggingface_hub

Add BirdNET v2.4 embeddings model (shared backbone for all classifiers)

Add SHA256 checksums for all model and label files

Add regional translations for bat labels

Initial upload: ONNX-converted BattyBirdNET regional bat classifiers

initial commit