These are the original SageMaker artifacts, kept for reference and for the blog post's before/after comparison:
Wine In A Million.ipynb— the original walkthroughsrc/embeddings_script.py— SageMaker job that embedded reviews and wroteembeddings.csv.gzto S3src/nn_script.py— SageMaker job that fit a scikit-learnNearestNeighborsmodel over those embeddings
The 2026 rewrite replaces all of this with a single LanceDB table on object
storage plus a FastAPI service. See ../backend/ and the root
README.md.