luplo.core.embedding
====================

.. py:module:: luplo.core.embedding

.. autoapi-nested-parse::

   Embedding backend abstraction.

   Default is ``NullEmbedding`` — no ML dependencies, vector reranking disabled.
   Install ``luplo[vector-local]`` for sentence-transformers support.



Submodules
----------

.. toctree::
   :maxdepth: 1

   /api/luplo/core/embedding/local/index
   /api/luplo/core/embedding/null/index
   /api/luplo/core/embedding/protocol/index


Classes
-------

.. autoapisummary::

   luplo.core.embedding.NullEmbedding
   luplo.core.embedding.EmbeddingBackend


Package Contents
----------------

.. py:class:: NullEmbedding

   Embedding backend that produces no vectors.


   .. py:method:: embed(texts: list[str]) -> list[list[float] | None]
      :async:


      Return ``None`` for every text — embeddings disabled.



.. py:class:: EmbeddingBackend

   Bases: :py:obj:`Protocol`


   Abstraction over text embedding models.

   Implementations produce fixed-dimension float vectors from text.
   The null backend returns ``None`` for every input, which causes the
   search pipeline to skip vector reranking entirely.


   .. py:method:: embed(texts: list[str]) -> list[list[float] | None]
      :async:


      Embed a batch of texts.

      :param texts: Strings to embed.

      :returns: One vector per input text, or ``None`` if embedding is
                unavailable for that text.