This gist illustrates how to store vectors of your documents in chroma without providing your actual text documents. Useful if your docs contain sensitive info or you are mindful of the storage.

chroma_no_docs.py

import uuid

from chromadb.utils import embedding_functions
import chromadb
ef = embedding_functions.DefaultEmbeddingFunction()

docs = ["Article by john", "Article by Jack", "Article by Jill"]
client = chromadb.Client()
embeddings = ef(docs)
collection = client.get_or_create_collection("test-where-list")
collection.upsert(documents=["" for _ in range(len(docs))], embeddings=embeddings, metadatas=[{"source": "blogger.com","author":"John"}, {"source": "medium","author":"Jack"}, {"source": "notion","author":"Jill"}],ids=[str(uuid.uuid4()) for _ in range(len(docs))])

#collection.get(include=["embeddings","metadatas"])

qr = collection.query(query_texts=["All articles by John"], include=["metadatas","distances"])
print(qr)