Add v2 pipeline model (prithivMLmods), route by model param

Replace corrupted LFS pointer with actual model weights

Fix backbone weight loading: strip wav2vec2 prefix, add sanity check, remove unused /predict endpoint

Fix predict_b64 to handle both JSON and form content types

Fix predict_b64 to accept both JSON body and form data

Real Wav2Vec2 SER model inference from model.safetensors

Upload model.safetensors

Update with 7-class SER model scaffold and correct label mapping

Add speech emotion recognition API scaffolding

initial commit