Class GeminiEmbedCapabilityImpl

Adapts Gemini embeddings into ProviderPlaneAI's normalized embedding artifact surface.

Supports scalar and batched embedding inputs, forwards Gemini-specific task options, and normalizes returned vectors into NormalizedEmbedding[].

Implements

EmbedCapability<ClientEmbeddingRequest, NormalizedEmbedding[]>

Index

Constructors

Methods

Constructors

constructor

new GeminiEmbedCapabilityImpl(
provider: BaseProvider,
client: GoogleGenAI,
): GeminiEmbedCapabilityImpl
Creates a new Gemini embedding capability adapter.
Parameters
- provider: BaseProvider
  Owning provider instance used for initialization checks and merged config access.
- client: GoogleGenAI
  Initialized Google GenAI SDK client.
Returns GeminiEmbedCapabilityImpl
- Defined in src/providers/gemini/capabilities/GeminiEmbedCapabilityImpl.ts:37

Methods

embed

embed(
    request: AIRequest<ClientEmbeddingRequest>,
    _executionContext?: MultiModalExecutionContext,
    signal?: AbortSignal,
): Promise<AIResponse<NormalizedEmbedding[]>>
Executes a Gemini embeddings request.

Responsibilities:
- validate embedding input
- resolve merged model/runtime options
- normalize scalar input into Gemini's batched request shape
- execute models.embedContent
- attach provider/model/usage metadata to normalized embeddings
Parameters
- request: AIRequest<ClientEmbeddingRequest>
  Unified embedding request envelope.
- Optional_executionContext: MultiModalExecutionContext
  Optional execution context. Unused directly in this adapter.
- Optionalsignal: AbortSignal
  Optional cancellation signal.
Returns Promise<AIResponse<NormalizedEmbedding[]>>
Provider-normalized embedding artifacts.

Throws
When input is invalid or Gemini returns no embeddings.
Implementation of EmbedCapability.embed
- Defined in src/providers/gemini/capabilities/GeminiEmbedCapabilityImpl.ts:58