Are you an LLM? You can read better optimized documentation at /api/@nhtio/adk/batteries/embeddings/openai/types/interfaces/BaseEmbeddingsAdapterOptions.md for this page in Markdown format
Interface: BaseEmbeddingsAdapterOptions
Defined in: src/batteries/embeddings/openai/types.ts:61
Options shared by every embeddings battery, regardless of engine.
Remarks
Engine-specific batteries extend this with their own transport/engine fields. model is required with no default — naming the embedding model is the caller's responsibility, never the battery's.
Extended by
WebLLMEmbeddingsAdapterOptionsTransformersJsEmbeddingsAdapterOptionsOpenAIEmbeddingsAdapterOptionsOllamaEmbeddingsAdapterOptions
Properties
| Property | Type | Description | Defined in |
|---|---|---|---|
dimensions? | number | Declared output dimensionality, surfaced via adapter.dimensions. Optional metadata for callers that pre-allocate buffers; the battery does not enforce it against responses. | src/batteries/embeddings/openai/types.ts:84 |
documentPrefix? | string | Instruction prefix prepended to inputs embedded with kind: 'document'. Unset → no prefix. Most asymmetric models (including Arctic) use no document prefix; exposed for symmetry and the models that do. | src/batteries/embeddings/openai/types.ts:79 |
model | string | Model id to embed with. Required, no default. For OpenAI this is e.g. text-embedding-3-small; for WebLLM it is an MLC embedding model id (e.g. snowflake-arctic-embed-m-q0f32-MLC). | src/batteries/embeddings/openai/types.ts:67 |
queryPrefix? | string | Instruction prefix prepended to inputs embedded with kind: 'query'. Unset → no prefix. For Snowflake Arctic Embed, the documented value is 'Represent this sentence for searching relevant passages: '. | src/batteries/embeddings/openai/types.ts:73 |