Skip to main content

hackagent.models.embeddings

Embedding-only provider requests shared by retrieval and preflight.

normalize_embedding_endpoint​

def normalize_embedding_endpoint(endpoint: str,
                                 *,
                                 ollama: bool = False) -> str

Return an OpenAI-compatible API base, not a full embeddings URL.

Ollama's native endpoint spellings select its compatible /v1 API. Preserve reverse-proxy prefixes and custom OpenAI-compatible API paths.

embedding_request_kwargs​

def embedding_request_kwargs(config: Dict[str, Any]) -> Dict[str, Any]

Resolve provider, endpoint and credentials without consulting storage.

validate_embedding_vector​

def validate_embedding_vector(value: Any,
                              *,
                              dimension: Optional[int] = None) -> np.ndarray

Require a nonempty, finite, one-dimensional float32 numeric vector.

extract_embedding_vector​

def extract_embedding_vector(response: Any) -> np.ndarray

Extract and validate the single requested vector from a LiteLLM response.

request_embedding​

def request_embedding(config: Dict[str, Any], text: str) -> np.ndarray

Request a real embedding, never chat completions or textual signatures.