Class RealtimeModel

Google Realtime Model for real-time voice conversations with Gemini models

Hierarchy (view full)

llm.RealtimeModel
- RealtimeModel

Index

Constructors

constructor

new RealtimeModel(options?): beta.realtime.RealtimeModel
Parameters
- options: {
      apiKey?: string;
      apiVersion?: string;
      candidateCount?: number;
      connOptions?: APIConnectOptions;
      contextWindowCompression?: ContextWindowCompressionConfig;
      enableAffectiveDialog?: boolean;
      frequencyPenalty?: number;
      geminiTools?: LLMTools;
      httpOptions?: HttpOptions;
      imageEncodeOptions?: {
          format: "JPEG";
          quality: number;
          resizeOptions: {
              height: number;
              strategy: "scale_aspect_fit";
              width: number;
          };
      };
      inputAudioTranscription?: null | AudioTranscriptionConfig;
      instructions?: string;
      language?: string;
      location?: string;
      maxOutputTokens?: number;
      modalities?: Modality[];
      model?: string;
      outputAudioTranscription?: null | AudioTranscriptionConfig;
      presencePenalty?: number;
      proactivity?: boolean;
      project?: string;
      realtimeInputConfig?: RealtimeInputConfig;
      temperature?: number;
      topK?: number;
      topP?: number;
      vertexai?: boolean;
      voice?: string;
  } = {}
  - Optional apiKey?: string
    Google Gemini API key. If not provided, will attempt to read from GOOGLE_API_KEY environment variable
  - Optional apiVersion?: string
    API version to use
  - Optional candidateCount?: number
    The number of candidate responses to generate
  - Optional connOptions?: APIConnectOptions
    The configuration for the API connection
  - Optional contextWindowCompression?: ContextWindowCompressionConfig
    The configuration for context window compression
  - Optional enableAffectiveDialog?: boolean
    Whether to enable affective dialog
  - Optional frequencyPenalty?: number
    The frequency penalty for response generation
  - Optional geminiTools?: LLMTools
    Gemini-specific tools to use for the session
  - Optional httpOptions?: HttpOptions
    HTTP options for API requests
  - Optional imageEncodeOptions?: {
        format: "JPEG";
        quality: number;
        resizeOptions: {
            height: number;
            strategy: "scale_aspect_fit";
            width: number;
        };
    }
    The configuration for image encoding
    - format: "JPEG"
    - quality: number
    - resizeOptions: {
          height: number;
          strategy: "scale_aspect_fit";
          width: number;
      }
      height: number
      strategy: "scale_aspect_fit"
      width: number
  - Optional inputAudioTranscription?: null | AudioTranscriptionConfig
    The configuration for input audio transcription
  - Optional instructions?: string
    Initial system instructions for the model
  - Optional language?: string
    The language (BCP-47 Code) to use for the API See https://ai.google.dev/gemini-api/docs/live#supported-languages
  - Optional location?: string
    The location to use for the API (for VertexAI)
  - Optional maxOutputTokens?: number
    Maximum number of tokens in the response
  - Optional modalities?: Modality[]
    Modalities to use, such as [Modality.TEXT, Modality.AUDIO]
  - Optional model?: string
    The name of the model to use
  - Optional outputAudioTranscription?: null | AudioTranscriptionConfig
    The configuration for output audio transcription
  - Optional presencePenalty?: number
    The presence penalty for response generation
  - Optional proactivity?: boolean
    Whether to enable proactive audio
  - Optional project?: string
    The project ID to use for the API (for VertexAI)
  - Optional realtimeInputConfig?: RealtimeInputConfig
    The configuration for realtime input
  - Optional temperature?: number
    Sampling temperature for response generation
  - Optional topK?: number
    The top-k value for response generation
  - Optional topP?: number
    The top-p value for response generation
  - Optional vertexai?: boolean
    Whether to use VertexAI for the API
  - Optional voice?: string
    Voice setting for audio outputs
Returns beta.realtime.RealtimeModel
Overrides RealtimeModel.constructor
- Defined in plugins/google/src/beta/realtime/realtime_api.ts:139

Accessors

capabilities

get capabilities(): RealtimeCapabilities
Returns RealtimeCapabilities
Inherited from llm.RealtimeModel.capabilities
- Defined in agents/dist/llm/realtime.d.ts:45

Methods

close

close(): Promise<void>
Close the model and cleanup resources

Returns Promise<void>
Overrides RealtimeModel.close
- Defined in plugins/google/src/beta/realtime/realtime_api.ts:359

session

session(): RealtimeSession
Create a new realtime session

Returns RealtimeSession
Overrides RealtimeModel.session
- Defined in plugins/google/src/beta/realtime/realtime_api.ts:338

updateOptions

updateOptions(options): void
Update model options
Parameters
- options: {
  temperature?: number;
  voice?: string;
  }
  - Optional temperature?: number
  - Optional voice?: string
Returns void
- Defined in plugins/google/src/beta/realtime/realtime_api.ts:345

Class RealtimeModel

Hierarchy (view full)

Index

Constructors

Accessors

Methods

Constructors

constructor

Parameters

Optional apiKey?: string

Optional apiVersion?: string

Optional candidateCount?: number

Optional connOptions?: APIConnectOptions

Optional contextWindowCompression?: ContextWindowCompressionConfig

Optional enableAffectiveDialog?: boolean

Optional frequencyPenalty?: number

Optional geminiTools?: LLMTools

Optional httpOptions?: HttpOptions

Optional imageEncodeOptions?: { format: "JPEG"; quality: number; resizeOptions: { height: number; strategy: "scale_aspect_fit"; width: number; }; }

format: "JPEG"

quality: number

resizeOptions: { height: number; strategy: "scale_aspect_fit"; width: number; }

height: number

strategy: "scale_aspect_fit"

width: number

Optional inputAudioTranscription?: null | AudioTranscriptionConfig

Optional instructions?: string

Optional language?: string

Optional location?: string

Optional maxOutputTokens?: number

Optional modalities?: Modality[]

Optional model?: string

Optional outputAudioTranscription?: null | AudioTranscriptionConfig

Optional presencePenalty?: number

Optional proactivity?: boolean

Optional project?: string

Optional realtimeInputConfig?: RealtimeInputConfig

Optional temperature?: number

Optional topK?: number

Optional topP?: number

Optional vertexai?: boolean

Optional voice?: string

Returns beta.realtime.RealtimeModel

Accessors

capabilities

Returns RealtimeCapabilities

Methods

close

Returns Promise<void>

session

Returns RealtimeSession

updateOptions

Parameters

Optional temperature?: number

Optional voice?: string

Returns void

Settings

Member Visibility

Theme

On This Page

`Optional` apiKey?: string

`Optional` apiVersion?: string

`Optional` candidateCount?: number

`Optional` connOptions?: APIConnectOptions

`Optional` contextWindowCompression?: ContextWindowCompressionConfig

`Optional` enableAffectiveDialog?: boolean

`Optional` frequencyPenalty?: number

`Optional` geminiTools?: LLMTools

`Optional` httpOptions?: HttpOptions

`Optional` imageEncodeOptions?: {
format: "JPEG";
quality: number;
resizeOptions: {
height: number;
strategy: "scale_aspect_fit";
width: number;
};
}

resizeOptions: {
height: number;
strategy: "scale_aspect_fit";
width: number;
}

`Optional` inputAudioTranscription?: null | AudioTranscriptionConfig

`Optional` instructions?: string

`Optional` language?: string

`Optional` location?: string

`Optional` maxOutputTokens?: number

`Optional` modalities?: Modality[]

`Optional` model?: string

`Optional` outputAudioTranscription?: null | AudioTranscriptionConfig

`Optional` presencePenalty?: number

`Optional` proactivity?: boolean

`Optional` project?: string

`Optional` realtimeInputConfig?: RealtimeInputConfig

`Optional` temperature?: number

`Optional` topK?: number

`Optional` topP?: number

`Optional` vertexai?: boolean

`Optional` voice?: string

`Optional` temperature?: number

`Optional` voice?: string