Prompt Finder uses Stanford GloVe word vectors (Wikipedia 2014 + Gigaword 5,
6 billion tokens, 50 dimensions). Pretrained data is made available under
the Open Data Commons Public Domain Dedication and License (PDDL) v1.0:
https://opendatacommons.org/licenses/pddl/1-0/
https://nlp.stanford.edu/projects/glove/

Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014.
GloVe: Global Vectors for Word Representation. EMNLP.

Source-format conversion: gensim-data glove-wiki-gigaword-50 release.
Our Python preparation retains the first 20,000 alphabetic lowercase words,
normalizes their vectors, and quantizes to signed integers. We did not train
GloVe. This feature averages those word vectors; it is not a sentence model,
an LLM, or retrieval-augmented generation. The original 40 game prompts are
the retrieval corpus. No song recordings or lyrics are included.