@ai-sdk/provider 4.0.3 → 4.0.5

package/src/image-model/v4/image-model-v4.ts

@@ -6,7 +6,7 @@ type GetMaxImagesPerCallFunction = (options: {
 }) => PromiseLike<number | undefined> | number | undefined;
 
 /**
- * Image generation model specification version 3.
+ * Image generation model specification version 4.
  */
 export type ImageModelV4 = {
   /**

package/src/index.ts

@@ -13,6 +13,7 @@ export * from './reranking-model/index';
 export * from './shared/index';
 export * from './skills/index';
 export * from './speech-model/index';
+export * from './speech-translation-model/index';
 export * from './transcription-model/index';
 export * from './video-model/index';
 

package/src/realtime-model/v4/realtime-model-v4-session-config.ts

@@ -1,3 +1,4 @@
+import type { SharedV4AudioFormat } from '../../shared';
 import type { RealtimeModelV4ToolDefinition } from './realtime-model-v4-tool-definition';
 
 /**
@@ -23,17 +24,7 @@ export type RealtimeModelV4SessionConfig = {
   /**
    * Audio format configuration for input audio.
    */
-  inputAudioFormat?: {
-    /**
-     * Audio format type (e.g. "audio/pcm", "audio/pcmu", "audio/pcma").
-     */
-    type: string;
-
-    /**
-     * Sample rate in Hz. Only applicable for PCM format.
-     */
-    rate?: number;
-  };
+  inputAudioFormat?: SharedV4AudioFormat;
 
   /**
    * Input audio transcription configuration.
@@ -87,17 +78,7 @@ export type RealtimeModelV4SessionConfig = {
   /**
    * Audio format configuration for output audio.
    */
-  outputAudioFormat?: {
-    /**
-     * Audio format type (e.g. "audio/pcm", "audio/pcmu", "audio/pcma").
-     */
-    type: string;
-
-    /**
-     * Sample rate in Hz. Only applicable for PCM format.
-     */
-    rate?: number;
-  };
+  outputAudioFormat?: SharedV4AudioFormat;
 
   /**
    * Voice activity detection configuration.

package/src/reranking-model/v4/reranking-model-v4.ts

@@ -2,7 +2,7 @@ import type { RerankingModelV4CallOptions } from './reranking-model-v4-call-opti
 import type { RerankingModelV4Result } from './reranking-model-v4-result';
 
 /**
- * Specification for a reranking model that implements the reranking model interface version 3.
+ * Specification for a reranking model that implements the reranking model interface version 4.
  */
 export type RerankingModelV4 = {
   /**

package/src/shared/v4/index.ts

@@ -1,3 +1,4 @@
+export * from './shared-v4-audio-format';
 export * from './shared-v4-file-data';
 export * from './shared-v4-headers';
 export * from './shared-v4-provider-metadata';

package/src/shared/v4/shared-v4-audio-format.ts

@@ -0,0 +1,14 @@
+/**
+ * Audio format configuration shared by provider V4 model specifications.
+ */
+export type SharedV4AudioFormat = {
+  /**
+   * Audio format type, e.g. `audio/pcm`, `audio/pcmu`, or `audio/pcma`.
+   */
+  type: string;
+
+  /**
+   * Sample rate in Hz. Only applicable for formats that require a rate.
+   */
+  rate?: number;
+};

package/src/speech-model/v4/speech-model-v4.ts

@@ -2,7 +2,7 @@ import type { SpeechModelV4CallOptions } from './speech-model-v4-call-options';
 import type { SpeechModelV4Result } from './speech-model-v4-result';
 
 /**
- * Speech model specification version 3.
+ * Speech model specification version 4.
  */
 export type SpeechModelV4 = {
   /**

package/src/speech-translation-model/index.ts

@@ -0,0 +1 @@
+export * from './v4/index';

package/src/speech-translation-model/v4/index.ts

@@ -0,0 +1,5 @@
+export type { SpeechTranslationModelV4 as Experimental_SpeechTranslationModelV4 } from './speech-translation-model-v4';
+export type { SpeechTranslationModelV4StreamOptions as Experimental_SpeechTranslationModelV4StreamOptions } from './speech-translation-model-v4-stream-options';
+export type { SpeechTranslationModelV4StreamPart as Experimental_SpeechTranslationModelV4StreamPart } from './speech-translation-model-v4-stream-part';
+export type { SpeechTranslationModelV4StreamResult as Experimental_SpeechTranslationModelV4StreamResult } from './speech-translation-model-v4-stream-result';
+export type { SpeechTranslationModelV4Usage as Experimental_SpeechTranslationModelV4Usage } from './speech-translation-model-v4-usage';

package/src/speech-translation-model/v4/speech-translation-model-v4-stream-options.ts

@@ -0,0 +1,65 @@
+import type { JSONObject } from '../../json-value/json-value';
+import type { SharedV4AudioFormat } from '../../shared';
+
+type SpeechTranslationModelV4ProviderOptions = Record<string, JSONObject>;
+
+/**
+ * Options for a speech translation model stream call.
+ */
+export type SpeechTranslationModelV4StreamOptions = {
+  /**
+   * Source audio chunks to transform.
+   *
+   * `Uint8Array` chunks contain raw audio bytes. `string` chunks contain
+   * base64-encoded raw audio bytes.
+   */
+  audio: ReadableStream<Uint8Array | string>;
+
+  /**
+   * The input audio format for the raw audio chunks.
+   */
+  inputAudioFormat: SharedV4AudioFormat;
+
+  /**
+   * The language to produce output audio and text in, as a BCP-47-style
+   * language tag (e.g. `en`, `es`, `fr-CA`). Supported values are
+   * provider-specific and validated by the provider.
+   */
+  targetLanguage: string;
+
+  /**
+   * The language of the source audio, as a BCP-47-style language tag.
+   * When absent, providers should auto-detect the source language.
+   */
+  sourceLanguage?: string;
+
+  /**
+   * The desired audio format for output audio chunks.
+   * When absent, the provider default output format is used.
+   */
+  outputAudioFormat?: SharedV4AudioFormat;
+
+  /**
+   * Additional provider-specific options that are passed through to the provider.
+   *
+   * The outer record is keyed by the provider name, and the inner record is keyed
+   * by provider-specific option names.
+   */
+  providerOptions?: SpeechTranslationModelV4ProviderOptions;
+
+  /**
+   * Abort signal for cancelling the operation.
+   */
+  abortSignal?: AbortSignal;
+
+  /**
+   * Additional HTTP headers to be sent with the request.
+   * Only applicable for HTTP/WebSocket-based providers that support headers.
+   */
+  headers?: Record<string, string | undefined>;
+
+  /**
+   * When true, providers should include raw provider chunks in the stream.
+   */
+  includeRawChunks?: boolean;
+};

package/src/speech-translation-model/v4/speech-translation-model-v4-stream-part.ts

@@ -0,0 +1,141 @@
+import type { JSONObject } from '../../json-value/json-value';
+import type { SharedV4Headers } from '../../shared';
+import type { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
+import type { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+import type { SpeechTranslationModelV4Usage } from './speech-translation-model-v4-usage';
+
+export type SpeechTranslationModelV4StreamPart =
+  | {
+      /**
+       * Stream start event with warnings for the call, e.g. unsupported settings.
+       */
+      type: 'stream-start';
+      warnings: Array<SharedV4Warning>;
+    }
+  | {
+      /**
+       * Output audio chunk.
+       *
+       * `Uint8Array` chunks contain raw audio bytes. `string` chunks contain
+       * base64-encoded raw audio bytes.
+       */
+      type: 'audio';
+      id?: string;
+      audio: Uint8Array | string;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Append-only output text delta.
+       *
+       * Output text is append-only: providers stream `output-text-delta`
+       * parts and finalize per-utterance with `output-text-final`. There is
+       * no partial/revision part for output text by design for now.
+       */
+      type: 'output-text-delta';
+      id?: string;
+      delta: string;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Final output text for a provider-defined segment or utterance.
+       *
+       * Output text is append-only: providers stream `output-text-delta`
+       * parts and finalize per-utterance with `output-text-final`. There is
+       * no partial/revision part for output text by design for now.
+       */
+      type: 'output-text-final';
+      id?: string;
+      text: string;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Append-only source transcript delta.
+       */
+      type: 'source-transcript-delta';
+      id?: string;
+      delta: string;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Non-final source transcript text. The text may be revised by later parts.
+       */
+      type: 'source-transcript-partial';
+      id?: string;
+      text: string;
+      startSecond?: number;
+      endSecond?: number;
+      channelIndex?: number;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Final source transcript text for a provider-defined segment or utterance.
+       */
+      type: 'source-transcript-final';
+      id?: string;
+      text: string;
+      startSecond?: number;
+      endSecond?: number;
+      channelIndex?: number;
+      providerMetadata?: SharedV4ProviderMetadata;
+    }
+  | {
+      /**
+       * Metadata for the response, emitted once available.
+       */
+      type: 'response-metadata';
+      timestamp?: Date;
+      modelId?: string;
+      headers?: SharedV4Headers;
+      body?: unknown;
+    }
+  | {
+      /**
+       * Metadata that is available after the stream is finished.
+       */
+      type: 'finish';
+
+      /**
+       * The final source-language transcript of the input audio.
+       */
+      sourceText: string;
+
+      /**
+       * The final output text. May be an empty string for providers that
+       * produce only audio output.
+       */
+      outputText: string;
+
+      /**
+       * The duration of the source audio in seconds, if available.
+       */
+      durationInSeconds?: number;
+
+      /**
+       * Usage information for the call, if reported by the provider.
+       */
+      usage?: SpeechTranslationModelV4Usage;
+
+      /**
+       * Additional provider-specific metadata.
+       */
+      providerMetadata?: Record<string, JSONObject>;
+    }
+  | {
+      /**
+       * Raw provider chunks if enabled.
+       */
+      type: 'raw';
+      rawValue: unknown;
+    }
+  | {
+      /**
+       * Error parts are streamed, allowing for multiple errors.
+       */
+      type: 'error';
+      error: unknown;
+    };

package/src/speech-translation-model/v4/speech-translation-model-v4-stream-result.ts

@@ -0,0 +1,47 @@
+import type { SharedV4Headers } from '../../shared';
+import type { SpeechTranslationModelV4StreamPart } from './speech-translation-model-v4-stream-part';
+
+/**
+ * The result of a speech translation model doStream call.
+ */
+export type SpeechTranslationModelV4StreamResult = {
+  /**
+   * The stream.
+   */
+  stream: ReadableStream<SpeechTranslationModelV4StreamPart>;
+
+  /**
+   * Optional request information for telemetry and debugging purposes.
+   */
+  request?: {
+    /**
+     * Request body or setup payload that was sent to the provider API.
+     */
+    body?: unknown;
+  };
+
+  /**
+   * Optional response data.
+   */
+  response?: {
+    /**
+     * Timestamp for the start of the streamed response.
+     */
+    timestamp?: Date;
+
+    /**
+     * The ID of the response model that was used to generate the response.
+     */
+    modelId?: string;
+
+    /**
+     * Response headers.
+     */
+    headers?: SharedV4Headers;
+
+    /**
+     * Response body.
+     */
+    body?: unknown;
+  };
+};

package/src/speech-translation-model/v4/speech-translation-model-v4-usage.ts

@@ -0,0 +1,32 @@
+/**
+ * Usage information for a speech translation call.
+ *
+ * All fields are optional because providers report usage with different
+ * granularity (seconds of audio, audio tokens, and/or text tokens).
+ */
+export type SpeechTranslationModelV4Usage = {
+  /**
+   * Seconds of input audio that were processed, if reported.
+   */
+  inputAudioSeconds?: number;
+
+  /**
+   * Number of input audio tokens, if reported.
+   */
+  inputAudioTokens?: number;
+
+  /**
+   * Number of output audio tokens, if reported.
+   */
+  outputAudioTokens?: number;
+
+  /**
+   * Number of input text tokens, if reported.
+   */
+  inputTextTokens?: number;
+
+  /**
+   * Number of output text tokens, if reported.
+   */
+  outputTextTokens?: number;
+};

package/src/speech-translation-model/v4/speech-translation-model-v4.ts

@@ -0,0 +1,40 @@
+import type { SpeechTranslationModelV4StreamOptions } from './speech-translation-model-v4-stream-options';
+import type { SpeechTranslationModelV4StreamResult } from './speech-translation-model-v4-stream-result';
+
+/**
+ * Speech translation model specification version 4.
+ *
+ * Speech translation is a streaming-only modality: models translate live
+ * source audio into target-language audio and text.
+ *
+ * Experimental: the speech translation model contract may change in patch
+ * releases while the functions built on it are experimental. All types of
+ * this modality are exported with `Experimental_` prefixes for this reason.
+ */
+export type SpeechTranslationModelV4 = {
+  /**
+   * The speech translation model must specify which speech translation model
+   * interface version it implements. This will allow us to evolve the
+   * speech translation model interface and retain backwards compatibility.
+   * The different implementation versions can be handled as a discriminated
+   * union on our side.
+   */
+  readonly specificationVersion: 'v4';
+
+  /**
+   * Name of the provider for logging purposes.
+   */
+  readonly provider: string;
+
+  /**
+   * Provider-specific model ID for logging purposes.
+   */
+  readonly modelId: string;
+
+  /**
+   * Streams a speech translation for live audio.
+   */
+  doStream(
+    options: SpeechTranslationModelV4StreamOptions,
+  ): PromiseLike<SpeechTranslationModelV4StreamResult>;
+};

package/src/transcription-model/v4/transcription-model-v4-stream-options.ts

@@ -1,4 +1,5 @@
 import type { JSONObject } from '../../json-value/json-value';
+import type { SharedV4AudioFormat } from '../../shared';
 
 type TranscriptionModelV4ProviderOptions = Record<string, JSONObject>;
 
@@ -14,17 +15,7 @@ export type TranscriptionModelV4StreamOptions = {
   /**
    * The input audio format for the raw audio chunks.
    */
-  inputAudioFormat: {
-    /**
-     * Audio format type, e.g. `audio/pcm`, `audio/pcmu`, or `audio/pcma`.
-     */
-    type: string;
-
-    /**
-     * Sample rate in Hz. Only applicable for formats that require a rate.
-     */
-    rate?: number;
-  };
+  inputAudioFormat: SharedV4AudioFormat;
 
   /**
    * Additional provider-specific options that are passed through to the provider.

package/src/transcription-model/v4/transcription-model-v4.ts

@@ -4,7 +4,7 @@ import type { TranscriptionModelV4StreamOptions } from './transcription-model-v4
 import type { TranscriptionModelV4StreamResult } from './transcription-model-v4-stream-result';
 
 /**
- * Transcription model specification version 3.
+ * Transcription model specification version 4.
  */
 export type TranscriptionModelV4 = {
   /**

package/src/video-model/v4/index.ts

@@ -5,6 +5,9 @@ export type {
   VideoModelV4VideoData as Experimental_VideoModelV4VideoData,
 } from './video-model-v4-result';
 export type { VideoModelV4File as Experimental_VideoModelV4File } from './video-model-v4-file';
+export type { VideoModelV4OperationStartResult as Experimental_VideoModelV4OperationStartResult } from './video-model-v4-operation-start-result';
+export type { VideoModelV4OperationStatusResult as Experimental_VideoModelV4OperationStatusResult } from './video-model-v4-operation-status-result';
+export type { VideoModelV4OperationWebhook as Experimental_VideoModelV4OperationWebhook } from './video-model-v4-operation-webhook';
 export type {
   VideoModelV4FrameImage as Experimental_VideoModelV4FrameImage,
   VideoModelV4FrameType as Experimental_VideoModelV4FrameType,

package/src/video-model/v4/video-model-v4-operation-start-result.ts

@@ -0,0 +1,44 @@
+import type { JSONValue } from '../../json-value/json-value';
+import type { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
+import type { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+
+/**
+ * Result returned by `doStart` when initiating an asynchronous video generation.
+ */
+export type VideoModelV4OperationStartResult = {
+  /**
+   * JSON-serializable opaque reference passed to `doStatus` to check the
+   * status of the generation (e.g., a task ID or prediction URL).
+   */
+  operation: JSONValue;
+
+  /**
+   * Warnings for the call, e.g. unsupported features.
+   */
+  warnings: Array<SharedV4Warning>;
+
+  /**
+   * Additional provider-specific metadata.
+   */
+  providerMetadata?: SharedV4ProviderMetadata;
+
+  /**
+   * Response information for telemetry and debugging purposes.
+   */
+  response: {
+    /**
+     * Timestamp for the start of the response.
+     */
+    timestamp: Date;
+
+    /**
+     * The ID of the response model that was used.
+     */
+    modelId: string;
+
+    /**
+     * Response headers.
+     */
+    headers: Record<string, string> | undefined;
+  };
+};

package/src/video-model/v4/video-model-v4-operation-status-result.ts

@@ -0,0 +1,89 @@
+import type { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
+import type { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+import type { VideoModelV4VideoData } from './video-model-v4-result';
+
+/**
+ * Result returned by `doStatus` when checking the status of an
+ * asynchronous video generation.
+ */
+export type VideoModelV4OperationStatusResult =
+  | {
+      /**
+       * The video generation is still in progress.
+       */
+      status: 'pending';
+
+      /**
+       * Warnings for the call.
+       */
+      warnings?: Array<SharedV4Warning>;
+
+      /**
+       * Additional provider-specific metadata.
+       */
+      providerMetadata?: SharedV4ProviderMetadata;
+
+      /**
+       * Response information for telemetry and debugging purposes.
+       */
+      response: {
+        timestamp: Date;
+        modelId: string;
+        headers: Record<string, string> | undefined;
+      };
+    }
+  | {
+      /**
+       * The video generation is complete.
+       */
+      status: 'completed';
+
+      /**
+       * Generated videos.
+       */
+      videos: Array<VideoModelV4VideoData>;
+
+      /**
+       * Warnings for the call.
+       */
+      warnings: Array<SharedV4Warning>;
+
+      /**
+       * Additional provider-specific metadata.
+       */
+      providerMetadata?: SharedV4ProviderMetadata;
+
+      /**
+       * Response information for telemetry and debugging purposes.
+       */
+      response: {
+        timestamp: Date;
+        modelId: string;
+        headers: Record<string, string> | undefined;
+      };
+    }
+  | {
+      /**
+       * The video generation failed.
+       */
+      status: 'error';
+
+      /**
+       * A human-readable error message describing why the generation failed.
+       */
+      error: string;
+
+      /**
+       * Additional provider-specific metadata.
+       */
+      providerMetadata?: SharedV4ProviderMetadata;
+
+      /**
+       * Response information for telemetry and debugging purposes.
+       */
+      response: {
+        timestamp: Date;
+        modelId: string;
+        headers: Record<string, string> | undefined;
+      };
+    };

package/src/video-model/v4/video-model-v4-operation-webhook.ts

@@ -0,0 +1,11 @@
+import type { JSONValue } from '../../json-value/json-value';
+
+/**
+ * Data received from a webhook notification during asynchronous video
+ * generation. Generic over the body type so providers/consumers can
+ * narrow it to a specific shape.
+ */
+export type VideoModelV4OperationWebhook<TBody = JSONValue> = {
+  headers: Record<string, string>;
+  body: TBody;
+};