A package for visualising vector embedding collections as part of the Chroma vector database.

Uses Flask, Vite, and react-three-fiber to host a live 3D view of the data in a web browser, should perform well up to 10k documents. Dimensional reduction is performed using PCA for colors down to 50 dimensions, followed by tSNE down to 3.

pip install chromaviz or pip install git https://github.com/mtybadger/chromaviz/. After installing from pip, simply call visualize_collection with a valid ChromaDB collection, and chromaviz will do the rest.

from chromaviz import visualize_collection
visualize_collection(chromadb.Collection)

It also works with Langchain Chroma, as in:

from langchain.vectorstores import Chroma
vectordb = Chroma.from_documents(data, embeddings, ids)

from chromaviz import visualize_collection
visualize_collection(vectordb._collection)

• More dimensional reduction options and flexibility
• Refactor extremely shoddy React code
• Improve UX