goramen-rag-index

A prebuilt Chroma vector store over the Go Ramen blog archive (5,274 chunks, ~512 chars each), built with golden's Golden_Retriever and the thenlper/gte-large embedding model.

This is not tabular data — it's a Chroma collection (chroma.sqlite3 + an HNSW index) plus embedding_settings.json, which records which embedding model to reconstruct on load. There's nothing meaningful for the Hub's dataset viewer to preview here (viewer: false above disables it), and the mention of embedding_settings.json in a file browser is expected, not an error.

Usage

pip install --ignore-requires-python git+https://github.com/johnhalloran321/golden.git

from huggingface_hub import snapshot_download
from golden.golden_retriever import Golden_Retriever

path = snapshot_download(repo_id="johnhalloran/goramen-rag-index", repo_type="dataset")
db = Golden_Retriever.load(path, similarity_fn="cosine")

for doc, score in db.similarity_search_with_score("What did you think of Afuri?", k=4):
    print(score, doc.metadata.get("title"), doc.page_content[:100])

Used by the goramen project's chat.py and colab_demo.ipynb for retrieval-augmented generation over the blog.