ethanvillalovoz/knowledge-graph-rag-retrieval-artifacts

Knowledge Graph RAG Retrieval Artifacts

Pinned vector-retrieval artifacts for the Knowledge Graph RAG Assistant, a Washington State University capstone project combining knowledge-graph and dense-vector retrieval.

This repository is a project-owned, documented mirror of the two binary artifacts used by the maintained application. The files are byte-identical to the current artifacts originally hosted in miverson9/acme10-he-ragapp-embeddings at revision b88b9c93be2943f05485874914af00c47b82fc18. The mirror preserves that origin and removes the maintained runtime's dependency on an individual teammate's namespace.

Contents

File	Format	Size	SHA-256	Role
text_embeddings.npy	NumPy, float32, shape (769770, 384)	1,182,366,848 bytes	98592d86c93dbf474decba8b79426cd3c57c73c607b650692ce2df0398fbad74	Dense passage embeddings loaded by the retrieval backend
index.faiss	FAISS binary index	1,182,366,765 bytes	1e87e64080acfce1cbc3ecad2b3a8ae80900dde935135042dc2481675d340b1a	Vector index paired with the embedding matrix

The two files form one versioned artifact set and should be used together. This is a binary reproducibility-artifact repository, not a row-oriented training dataset, so the Hugging Face Dataset Viewer is not expected to render it.

Download

Download the tagged artifact set with the Hugging Face CLI:

hf download ethanvillalovoz/knowledge-graph-rag-retrieval-artifacts \
  index.faiss text_embeddings.npy checksums.sha256 \
  --repo-type dataset \
  --revision v1.0.0 \
  --local-dir ./retrieval-artifacts

Verify the files after download:

cd retrieval-artifacts
shasum -a 256 -c checksums.sha256

The application repository provides the expected destination paths and an automated checksum-verifying downloader in scripts/download_retrieval_artifacts.py.

Provenance

The artifacts were produced during the original capstone from a passage corpus derived from Simple English Wikipedia. The original project record does not preserve the exact Wikipedia snapshot date, complete passage-construction environment, or all model/runtime controls used to create the files. This mirror does not attempt to reconstruct that missing provenance or imply that the artifacts constitute a fully reconstructible benchmark.

For traceability:

• Original artifact repository: miverson9/acme10-he-ragapp-embeddings
• Original pinned revision: b88b9c93be2943f05485874914af00c47b82fc18
• Maintained application: ethanvillalovoz/knowledge-graph-rag-assistant
• Capstone implementation snapshot: 1ad5cc0
• Technical report: Knowledge Graph RAG Assistant

Project team and stewardship

The capstone project was created by Molly Iverson, Ethan Villalovoz, Chandler Juego, Adam Shtrikman, Vikas Aditya, and Parteek Kumar. The artifacts were originally published from Molly Iverson's Hugging Face account. This documented mirror is maintained by Ethan Villalovoz for continuity of the public research artifact and does not replace or obscure the original contribution.

Affiliations recorded by the project are Washington State University and HackerEarth. See the application repository's CITATION.cff for the citation metadata maintained with the software release.

License and reuse

The repository's source-code MIT license does not automatically apply to these data artifacts. The underlying Simple English Wikipedia material is available under the Creative Commons Attribution-ShareAlike 4.0 International license and, where applicable, the GNU Free Documentation License. Reuse must preserve the attribution and share-alike obligations that apply to the source material. See DATA_LICENSE.md for the complete notice used by this mirror.

Limitations and appropriate use

• These files support reproduction of the project's historical retrieval prototype; they are not presented as a general-purpose embedding benchmark.
• The missing source-snapshot date and construction environment limit exact end-to-end reconstruction.
• The artifacts do not include the source passage text. They must be paired with the separately documented corpus and compatible application code.
• Pin the repository revision and verify the hashes before reporting results.
• Do not infer new scientific claims from the existence or packaging of these artifacts.

Version

v1.0.0 is the first project-owned archival mirror of the exact runtime files from the original pinned Hugging Face revision. File-level provenance and integrity metadata are also available in artifact-manifest.json.