vectorstores 0.1.2.dev50

Project for uploading vector embeddings directly to weaviate.

vectorstores

Utilities for pushing vectors and metadata into a Weaviate instance. The package ships a WeaviateVectorStore implementation plus a simple VectorStore interface you can build on.

Installation

pip install vectorstores

Quickstart

import weaviate
from vectorstores import WeaviateVectorStore

# create a client (configure URL/auth to match your deployment)
client = weaviate.connect_to_local()  # or weaviate.connect_to_wcs(...)

store = WeaviateVectorStore(
    client=client,
    index_name="MyIndex",
    text_key="text",
)

# add data (stores text plus optional metadata; vectors can be provided explicitly)
ids = store.add_vector(
    vectors=["Hello world", "Another sample"],
    metadatas=[{"topic": "greeting"}, {"topic": "example"}],
)
print(ids)

# delete by id
store.delete(ids=[ids[0]])

Notes

• If you want multi-tenancy, pass use_multi_tenancy=True when constructing the store and provide tenant=... to write/delete calls.
• If you already have embeddings and want to bypass internal embedding, pass them via vectors and supply matching metadatas/ids.

Development

Create a virtual environment and install dependencies with tests:

python3 -m venv .venv
source .venv/bin/activate
pip install -e .[tests]

Run checks:

ruff check .
mypy .
pytest

Building and uploading to PyPI

1. Clean old artifacts (optional but recommended):

rm -rf dist build *.egg-info

2. Build the wheel and sdist:

python -m build

3. Upload to PyPI with Twine:

twine upload dist/*

Make sure your PyPI credentials are configured (e.g., via ~/.pypirc or environment variables) before running the upload.