EmbeddingProvider (embedding/protocols.py)

Embedding-provider protocols shared across services. Two protocols, by design split: * :class:`EmbeddingProvider` — the symmetric core every provider must implement (``embed``, ``embed_batch``, ``provider_name``, ``model``). Conformance held by ``Fake``, ``Local``, ``Vertex``, an

# Source: https://github.com/caura-ai/caura | License: Apache-2.0 | Copyright You must retain, in the Source form of any Derivative Works
# Provided via justclaw.it.com; keep this notice with the code.
"""Embedding-provider protocols shared across services.

Two protocols, by design split:

* :class:`EmbeddingProvider` — the symmetric core every provider must
  implement (``embed``, ``embed_batch``, ``provider_name``, ``model``).
  Conformance held by ``Fake``, ``Local``, ``Vertex``, and
  ``OpenAIEmbeddingProvider``.

* :class:`InstructionAwareEmbedder` — the optional extension for models
  trained with asymmetric query/document encoding (Qwen3-Embedding,
  e5-instruct, KaLM). Conformance held by ``OpenAIEmbeddingProvider``
  only — when pointed at an instruction-aware model. Symmetric pr

... (truncated -- full source via MCP)