Class Gemini

constructor

• new Gemini(params: GeminiParams): Gemini

  Parameters

  • params: GeminiParams

    The parameters for creating a Gemini instance.

  Returns Gemini

Readonly[GEMINI_MODEL_SYMBOL]

ReadonlyuseInteractionsApi

Static ReadonlysupportedModels

apiBackend

• get apiBackend(): GoogleLLMVariant

  Returns GoogleLLMVariant

apiClient

• get apiClient(): GoogleGenAI

  Returns GoogleGenAI

liveApiClient

• get liveApiClient(): GoogleGenAI

  Returns GoogleGenAI

liveApiVersion

• get liveApiVersion(): string

  Returns string

connect

• connect(llmRequest: LlmRequest): Promise<BaseLlmConnection>

  Connects to the Gemini model and returns an llm connection.

  Parameters

  • llmRequest: LlmRequest

    LlmRequest, the request to send to the Gemini model.

  Returns Promise<BaseLlmConnection>

  BaseLlmConnection, the connection to the Gemini model.

generateContentAsync

• generateContentAsync(
      llmRequest: LlmRequest,
      stream?: boolean,
      abortSignal?: AbortSignal,
  ): AsyncGenerator<LlmResponse, void>

  Sends a request to the Gemini model.

  Parameters

  • llmRequest: LlmRequest

    The request to send to the Gemini model.

  • stream: boolean = false

    Whether to make a streaming call. Defaults to false.

  • OptionalabortSignal: AbortSignal

  Returns AsyncGenerator<LlmResponse, void>

  Yields

  Each LlmResponse produced by the model. A non-streaming call yields exactly one; a streaming call yields the partial responses as they arrive, then a final aggregated response whenever the stream carried either content or usage/grounding/citation metadata. A metadata-only stream therefore still ends with one, holding the metadata and no content; only a stream with neither ends silently.