Skip to main content

Table of Contents

spoon_ai.rag.embeddings
- HashEmbeddingClient
- get_embedding_client

Module `spoon_ai.rag.embeddings`

`HashEmbeddingClient` Objects

class HashEmbeddingClient(EmbeddingClient)

Deterministic offline embedding via hashing.

Produces fixed-length vectors in [0,1] normalized range. Not semantically meaningful but stable for tests and offline demos.

`get_embedding_client`

def get_embedding_client(
        provider: Optional[str],
        *,
        openai_api_key: Optional[str] = None,
        openai_model: str = "text-embedding-3-small") -> EmbeddingClient

Create an embedding client.

Provider selection rules:

provider is None/"auto": pick the first configured embeddings provider using a dedicated priority order (OpenAI > OpenRouter > Gemini).
provider is "openai" / "openrouter" / "gemini" / "ollama": force that provider (uses core env config when applicable).
provider is "openai_compatible": use OpenAI-compatible embeddings via RAG_EMBEDDINGS_* env vars.
otherwise: deterministic hash embeddings (offline).

HashEmbeddingClient Objects