streaming

API reference for @fal-ai/client streaming

Classes & Interfaces

FalStream

interface FalStream

The class representing a streaming response. With t

Properties

Name	Type	Description
config	RequiredConfig	-
endpointId	string	-
url	string	-
options	StreamOptions<Input>	-
listeners	Map<FalStreamEventType, EventHandler[]>	-
buffer	Output[]	-
currentData	Output | undefined	-
lastEventTimestamp	any	-
streamClosed	any	-
_requestId	string | null	-
donePromise	Promise<Output>	-
abortController	any	-
start	any	-
handleResponse	any	-
handleError	any	-
on	any	-
emit	any	-
done	any	Gets a reference to the Promise that indicates whether the streaming is done or not. Developers should always call this in their apps to ensure the request is over. An alternative to this, is to use on('done') in case your application architecture works best with event listeners.
abort	any	Aborts the streaming request. Note: This method is noop in case the request is already done.

Methods

[Symbol.asyncIterator]

async [Symbol.asyncIterator](): any

Returns: any

StreamingClient

interface StreamingClient

The streaming client interface.

Methods

stream

stream(endpointId: Id, options: StreamOptions<InputType<Id>>): Promise<FalStream<InputType<Id>, OutputType<Id>>>

Calls a fal app that supports streaming and provides a streaming-capable object as a result, that can be used to get partial results through either AsyncIterator or through an event listener.

Parameter	Type	Description
endpointId	Id	the endpoint id, e.g. fal-ai/llavav15-13b.
options	StreamOptions<InputType<Id>>	the request options, including the input payload.

Returns: Promise<FalStream<InputType<Id>, OutputType<Id>>>

---

Functions

createStreamingClient

function createStreamingClient({ config, storage, }: StreamingClientDependencies): StreamingClient

Parameter	Type	Description
{ config, storage, }	StreamingClientDependencies	-

Returns: StreamingClient

---

Types

StreamingConnectionMode

type StreamingConnectionMode = "client" | "server"

StreamOptions

type StreamOptions = {
  /**
   * The endpoint URL. If not provided, it will be generated from the
   * `endpointId` and the `queryParams`.
   */
  readonly url?: string;

  /**
   * The API input payload.
   */
  readonly input?: Input;

  /**
   * The query parameters to be sent with the request.
   */
  readonly queryParams?: Record<string, string>;

  /**
   * The maximum time interval in milliseconds between stream chunks. Defaults to 15s.
   */
  readonly timeout?: number;

  /**
   * Whether it should auto-upload File-like types to fal's storage
   * or not.
   */
  readonly autoUpload?: boolean;

  /**
   * The HTTP method, defaults to `post`;
   */
  readonly method?: "get" | "post" | "put" | "delete" | string;

  /**
   * The content type the client accepts as response.
   * By default this is set to `text/event-stream`.
   */
  readonly accept?: string;

  /**
   * The streaming connection mode. This is used to determine
   * whether the streaming will be done from the browser itself (client)
   * or through your own server, either when running on NodeJS or when
   * using a proxy that supports streaming.
   *
   * It defaults to `server`. Set to `client` if your server proxy doesn't
   * support streaming.
   */
  readonly connectionMode?: StreamingConnectionMode;

  /**
   * The signal to abort the request.
   */
  readonly signal?: AbortSignal;

  /**
   * A custom token provider function. Only used when `connectionMode` is `"client"`.
   * When provided, this function will be used to fetch authentication tokens
   * instead of the default internal token fetching mechanism.
   */
  readonly tokenProvider?: TokenProvider;
}

The stream API options. It requires the API input and also offers configuration options.