latent.rag.embeddings¶

Embedding providers using raw SDKs (no LangChain).

Classes¶

BedrockEmbeddings¶

BedrockEmbeddings(model: str = 'amazon.titan-embed-text-v1', region_name: str = 'us-east-1', kwargs: Any = {})

AWS Bedrock embeddings via boto3.

HuggingFaceEmbeddings¶

HuggingFaceEmbeddings(model: str = 'BAAI/bge-m3', kwargs: Any = {})

Local embeddings via sentence-transformers.

LiteLLMEmbeddings¶

LiteLLMEmbeddings(model: str, kwargs: Any = {})

Embeddings via litellm — supports any LiteLLM model string.

OpenAIEmbeddings¶

OpenAIEmbeddings(model: str = 'text-embedding-3-small', kwargs: Any = {})

OpenAI embeddings via the openai SDK.

VoyageEmbeddings¶

VoyageEmbeddings(model: str = 'voyage-3', kwargs: Any = {})

Voyage AI embeddings via the voyageai SDK.

Functions¶

create_embedding_provider¶

create_embedding_provider(provider: str | None = None, model: str | None = None, kwargs: Any = {}) -> EmbeddingProvider

Factory: create an embedding provider by name.

Parameters¶

provider: One of "openai", "voyage", "bedrock", "litellm". Required — there is no default. model: Model identifier passed to the provider constructor. **kwargs: Extra keyword arguments forwarded to the provider.

Methods¶

BedrockEmbeddings.embed_documents¶

embed_documents(texts: list[str]) -> list[list[float]]

BedrockEmbeddings.embed_query¶

embed_query(text: str) -> list[float]

HuggingFaceEmbeddings.embed_documents¶

embed_documents(texts: list[str]) -> list[list[float]]

HuggingFaceEmbeddings.embed_query¶

embed_query(text: str) -> list[float]

LiteLLMEmbeddings.embed_documents¶

embed_documents(texts: list[str]) -> list[list[float]]

LiteLLMEmbeddings.embed_query¶

embed_query(text: str) -> list[float]

OpenAIEmbeddings.embed_documents¶

embed_documents(texts: list[str]) -> list[list[float]]

OpenAIEmbeddings.embed_query¶

embed_query(text: str) -> list[float]

VoyageEmbeddings.embed_documents¶

embed_documents(texts: list[str]) -> list[list[float]]

VoyageEmbeddings.embed_query¶

embed_query(text: str) -> list[float]