Installation¶

denselinkage requires Python 3.10+. The core depends only on NumPy and pandas; every heavy ML backend is an optional extra, so the base install stays light.

pip install denselinkage                        # core: numpy + pandas

Optional extras¶

Install an extra only when you reach past the dependency-free default stack:

Extra	Adds	Enables
`faiss`	`faiss-cpu`	`FaissFlatIndex` — scalable vector search
`sentence-transformers`	`sentence-transformers`	`SentenceTransformerEmbedder` — semantic embeddings
`langchain`	`langchain-core`, `langchain-openai`	`LangChainMatcher` — LLM matching
`all`	all of the above	—

pip install "denselinkage[faiss]"
pip install "denselinkage[sentence-transformers]"
pip install "denselinkage[langchain]"
pip install "denselinkage[all]"

The adapters import their backend lazily, so a missing extra surfaces only if you actually construct that adapter — never at import denselinkage.

Development¶

The project uses uv:

uv sync --dev
uv run ruff check . && uv run ruff format --check . && uv run mypy && uv run pytest

To build these docs locally, install the docs extra and run the Sphinx build (it treats warnings as errors, matching CI):

pip install -e ".[docs]"
cd docs && make html          # or: make.bat html   (Windows)

The rendered site lands in docs/_build/html/. See CONTRIBUTING.md for the full contributor guide.