Kushalkhemka's picture
Upload Chroma index
a87541a verified
metadata
dataset_name: cve-mixedbread-index
language:
  - en
license: mit
tags:
  - cve
  - rag
  - vector-database
  - chroma
task_categories:
  - text-retrieval
pretty_name: CVE MixedBread Chroma Index
dataset_summary: >-
  Serialized ChromaDB index built from the CVE/NVD corpus using MixedBread
  embeddings.

CVE MixedBread Chroma Index

This directory contains the ready-to-use ChromaDB index that powers the /rag_mixedbread query service. It was built over the Kushalkhemka/embedding-cve-nvd-dataset corpus using MixedBread embeddings.

Contents

  • chroma.sqlite3 (~6.5 GB): sqlite metadata DB with documents, embeddings, and collection info.
  • */ uuid folders: Chroma's persisted embeddings/uuid partitions.

Usage

Download the repo and point Chroma to the folder:

from chromadb import PersistentClient
client = PersistentClient(path="/path/to/index")
collection = client.get_collection("cve-mixedbread")

You can now query with new embeddings created via MixedBread to retrieve CVE passages.

Notes

  • Requires ChromaDB >=0.5.
  • Embeddings are stored as float32; total size ~6.5 GB.

License

MIT, but respect the upstream CVE/NVD terms.