Module embeddings 

Explicit embedding split/batch calls over one selected upstream target.

Structs

EmbeddingBatchError

Failure never presents completed partial vectors as a successful batch. completed_calls counts 2xx calls whose JSON was decoded, including one whose typed data proved invalid. A transport/body failure can have an unknown upstream outcome; this counter is not a retry authorization. completed_items counts only vectors validated before the failed call.

EmbeddingBatchOptions

EmbeddingBatchResult

Enums

EmbeddingFailure

Functions

gemini_batch_to_openai

Execute a Gemini embedding batch through the selected OpenAI model. Varying output dimensions are split into separate calls and returned in input order.

gemini_single_to_openai

One Gemini input through OpenAI embeddings. The selected target request carries its explicit model and declared dimensionality.

openai_to_gemini_batch

Convert an OpenAI embedding batch and call Gemini batchEmbedContents in source order. template is the selected target HTTP endpoint, never the original client request; its body is replaced and authentication is host-owned.

openai_to_gemini_single

One OpenAI input through Gemini embedContent, retaining requested encoding and validating the returned dimension and factual usage.