@ai-sdk/provider 4.0.0-beta.1 → 4.0.0-beta.11

package/src/skills/v4/skills-v4-upload-skill-result.ts

@@ -0,0 +1,40 @@
+import { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
+import { SharedV4ProviderReference } from '../../shared/v4/shared-v4-provider-reference';
+import { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+
+export interface SkillsV4UploadSkillResult {
+  /**
+   * A provider reference mapping provider names to provider-specific skill identifiers.
+   */
+  providerReference: SharedV4ProviderReference;
+
+  /**
+   * Optional human-readable title for the uploaded skill.
+   */
+  displayTitle?: string;
+
+  /**
+   * Optional name of the uploaded skill.
+   */
+  name?: string;
+
+  /**
+   * Optional description of what the uploaded skill does.
+   */
+  description?: string;
+
+  /**
+   * Optional latest version identifier of the uploaded skill.
+   */
+  latestVersion?: string;
+
+  /**
+   * Additional provider-specific metadata.
+   */
+  providerMetadata?: SharedV4ProviderMetadata;
+
+  /**
+   * Warnings for the call, e.g. unsupported settings.
+   */
+  warnings: SharedV4Warning[];
+}

package/src/skills/v4/skills-v4.ts

@@ -0,0 +1,28 @@
+import { SkillsV4UploadSkillCallOptions } from './skills-v4-upload-skill-call-options';
+import { SkillsV4UploadSkillResult } from './skills-v4-upload-skill-result';
+
+/**
+ * Skills specification version 4.
+ */
+export interface SkillsV4 {
+  /**
+   * The skills implementation must specify which skills interface
+   * version it implements. This will allow us to evolve the skills
+   * 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;
+
+  /**
+   * Uploads a new skill from the given files.
+   */
+  uploadSkill(
+    params: SkillsV4UploadSkillCallOptions,
+  ): PromiseLike<SkillsV4UploadSkillResult>;
+}

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

@@ -1,2 +1,3 @@
 export type { SpeechModelV4 } from './speech-model-v4';
 export type { SpeechModelV4CallOptions } from './speech-model-v4-call-options';
+export type { SpeechModelV4Result } from './speech-model-v4-result';

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

@@ -0,0 +1,64 @@
+import { JSONObject } from '../../json-value';
+import { SharedV2Headers } from '../../shared';
+import { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+
+/**
+ * The result of a speech model doGenerate call.
+ */
+export type SpeechModelV4Result = {
+  /**
+   * Generated audio as an ArrayBuffer.
+   * The audio should be returned without any unnecessary conversion.
+   * If the API returns base64 encoded strings, the audio should be returned
+   * as base64 encoded strings. If the API returns binary data, the audio
+   * should be returned as binary data.
+   */
+  audio: string | Uint8Array;
+
+  /**
+   * Warnings for the call, e.g. unsupported settings.
+   */
+  warnings: Array<SharedV4Warning>;
+
+  /**
+   * Optional request information for telemetry and debugging purposes.
+   */
+  request?: {
+    /**
+     * Response body (available only for providers that use HTTP requests).
+     */
+    body?: unknown;
+  };
+
+  /**
+   * Response information for telemetry and debugging purposes.
+   */
+  response: {
+    /**
+     * Timestamp for the start of the generated response.
+     */
+    timestamp: Date;
+
+    /**
+     * The ID of the response model that was used to generate the response.
+     */
+    modelId: string;
+
+    /**
+     * Response headers.
+     */
+    headers?: SharedV2Headers;
+
+    /**
+     * Response body.
+     */
+    body?: unknown;
+  };
+
+  /**
+   * Additional provider-specific metadata. They are passed through
+   * from the provider to the AI SDK and enable provider-specific
+   * results that can be fully encapsulated in the provider.
+   */
+  providerMetadata?: Record<string, JSONObject>;
+};

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

@@ -1,7 +1,5 @@
-import { JSONObject } from '../../json-value';
-import { SharedV2Headers } from '../../shared';
-import { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
 import { SpeechModelV4CallOptions } from './speech-model-v4-call-options';
+import { SpeechModelV4Result } from './speech-model-v4-result';
 
 /**
  * Speech model specification version 3.
@@ -29,61 +27,7 @@ export type SpeechModelV4 = {
   /**
    * Generates speech audio from text.
    */
-  doGenerate(options: SpeechModelV4CallOptions): PromiseLike<{
-    /**
-     * Generated audio as an ArrayBuffer.
-     * The audio should be returned without any unnecessary conversion.
-     * If the API returns base64 encoded strings, the audio should be returned
-     * as base64 encoded strings. If the API returns binary data, the audio
-     * should be returned as binary data.
-     */
-    audio: string | Uint8Array;
-
-    /**
-     * Warnings for the call, e.g. unsupported settings.
-     */
-    warnings: Array<SharedV4Warning>;
-
-    /**
-     * Optional request information for telemetry and debugging purposes.
-     */
-    request?: {
-      /**
-       * Response body (available only for providers that use HTTP requests).
-       */
-      body?: unknown;
-    };
-
-    /**
-     * Response information for telemetry and debugging purposes.
-     */
-    response: {
-      /**
-       * Timestamp for the start of the generated response.
-       */
-      timestamp: Date;
-
-      /**
-       * The ID of the response model that was used to generate the response.
-       */
-      modelId: string;
-
-      /**
-       * Response headers.
-       */
-      headers?: SharedV2Headers;
-
-      /**
-       * Response body.
-       */
-      body?: unknown;
-    };
-
-    /**
-     * Additional provider-specific metadata. They are passed through
-     * from the provider to the AI SDK and enable provider-specific
-     * results that can be fully encapsulated in the provider.
-     */
-    providerMetadata?: Record<string, JSONObject>;
-  }>;
+  doGenerate(
+    options: SpeechModelV4CallOptions,
+  ): PromiseLike<SpeechModelV4Result>;
 };

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

@@ -1,2 +1,3 @@
 export type { TranscriptionModelV4 } from './transcription-model-v4';
 export type { TranscriptionModelV4CallOptions } from './transcription-model-v4-call-options';
+export type { TranscriptionModelV4Result } from './transcription-model-v4-result';

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

@@ -0,0 +1,92 @@
+import { JSONObject } from '../../json-value';
+import { SharedV4Headers } from '../../shared';
+import { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+
+/**
+ * The result of a transcription model doGenerate call.
+ */
+export type TranscriptionModelV4Result = {
+  /**
+   * The complete transcribed text from the audio.
+   */
+  text: string;
+
+  /**
+   * Array of transcript segments with timing information.
+   * Each segment represents a portion of the transcribed text with start and end times.
+   */
+  segments: Array<{
+    /**
+     * The text content of this segment.
+     */
+    text: string;
+    /**
+     * The start time of this segment in seconds.
+     */
+    startSecond: number;
+    /**
+     * The end time of this segment in seconds.
+     */
+    endSecond: number;
+  }>;
+
+  /**
+   * The detected language of the audio content, as an ISO-639-1 code (e.g., 'en' for English).
+   * May be undefined if the language couldn't be detected.
+   */
+  language: string | undefined;
+
+  /**
+   * The total duration of the audio file in seconds.
+   * May be undefined if the duration couldn't be determined.
+   */
+  durationInSeconds: number | undefined;
+
+  /**
+   * Warnings for the call, e.g. unsupported settings.
+   */
+  warnings: Array<SharedV4Warning>;
+
+  /**
+   * Optional request information for telemetry and debugging purposes.
+   */
+  request?: {
+    /**
+     * Raw request HTTP body that was sent to the provider API as a string (JSON should be stringified).
+     * Non-HTTP(s) providers should not set this.
+     */
+    body?: string;
+  };
+
+  /**
+   * Response information for telemetry and debugging purposes.
+   */
+  response: {
+    /**
+     * Timestamp for the start of the generated 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;
+  };
+
+  /**
+   * Additional provider-specific metadata. They are passed through
+   * from the provider to the AI SDK and enable provider-specific
+   * results that can be fully encapsulated in the provider.
+   */
+  providerMetadata?: Record<string, JSONObject>;
+};

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

@@ -1,7 +1,5 @@
-import { JSONObject } from '../../json-value';
-import { SharedV4Headers } from '../../shared';
-import { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
 import { TranscriptionModelV4CallOptions } from './transcription-model-v4-call-options';
+import { TranscriptionModelV4Result } from './transcription-model-v4-result';
 
 /**
  * Transcription model specification version 3.
@@ -29,89 +27,7 @@ export type TranscriptionModelV4 = {
   /**
    * Generates a transcript.
    */
-  doGenerate(options: TranscriptionModelV4CallOptions): PromiseLike<{
-    /**
-     * The complete transcribed text from the audio.
-     */
-    text: string;
-
-    /**
-     * Array of transcript segments with timing information.
-     * Each segment represents a portion of the transcribed text with start and end times.
-     */
-    segments: Array<{
-      /**
-       * The text content of this segment.
-       */
-      text: string;
-      /**
-       * The start time of this segment in seconds.
-       */
-      startSecond: number;
-      /**
-       * The end time of this segment in seconds.
-       */
-      endSecond: number;
-    }>;
-
-    /**
-     * The detected language of the audio content, as an ISO-639-1 code (e.g., 'en' for English).
-     * May be undefined if the language couldn't be detected.
-     */
-    language: string | undefined;
-
-    /**
-     * The total duration of the audio file in seconds.
-     * May be undefined if the duration couldn't be determined.
-     */
-    durationInSeconds: number | undefined;
-
-    /**
-     * Warnings for the call, e.g. unsupported settings.
-     */
-    warnings: Array<SharedV4Warning>;
-
-    /**
-     * Optional request information for telemetry and debugging purposes.
-     */
-    request?: {
-      /**
-       * Raw request HTTP body that was sent to the provider API as a string (JSON should be stringified).
-       * Non-HTTP(s) providers should not set this.
-       */
-      body?: string;
-    };
-
-    /**
-     * Response information for telemetry and debugging purposes.
-     */
-    response: {
-      /**
-       * Timestamp for the start of the generated 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;
-    };
-
-    /**
-     * Additional provider-specific metadata. They are passed through
-     * from the provider to the AI SDK and enable provider-specific
-     * results that can be fully encapsulated in the provider.
-     */
-    providerMetadata?: Record<string, JSONObject>;
-  }>;
+  doGenerate(
+    options: TranscriptionModelV4CallOptions,
+  ): PromiseLike<TranscriptionModelV4Result>;
 };

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

@@ -1,6 +1,7 @@
+export type { VideoModelV4 as Experimental_VideoModelV4 } from './video-model-v4';
+export type { VideoModelV4CallOptions as Experimental_VideoModelV4CallOptions } from './video-model-v4-call-options';
 export type {
-  VideoModelV4 as Experimental_VideoModelV4,
+  VideoModelV4Result as Experimental_VideoModelV4Result,
   VideoModelV4VideoData as Experimental_VideoModelV4VideoData,
-} from './video-model-v4';
-export type { VideoModelV4CallOptions as Experimental_VideoModelV4CallOptions } from './video-model-v4-call-options';
+} from './video-model-v4-result';
 export type { VideoModelV4File as Experimental_VideoModelV4File } from './video-model-v4-file';

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

@@ -0,0 +1,92 @@
+import type { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
+import type { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+
+/**
+ * Generated video data. Can be a URL, base64-encoded string, or binary data.
+ */
+export type VideoModelV4VideoData =
+  | {
+      /**
+       * Video available as a URL (most common for video providers).
+       */
+      type: 'url';
+      url: string;
+      mediaType: string;
+    }
+  | {
+      /**
+       * Video as base64-encoded string.
+       */
+      type: 'base64';
+      data: string;
+      mediaType: string;
+    }
+  | {
+      /**
+       * Video as binary data.
+       */
+      type: 'binary';
+      data: Uint8Array;
+      mediaType: string;
+    };
+
+/**
+ * The result of a video model doGenerate call.
+ */
+export type VideoModelV4Result = {
+  /**
+   * Generated videos as URLs, base64 strings, or binary data.
+   *
+   * Most providers return URLs to video files (MP4, WebM) due to large file sizes.
+   * Use the discriminated union to indicate the type of video data being returned.
+   */
+  videos: Array<VideoModelV4VideoData>;
+
+  /**
+   * Warnings for the call, e.g. unsupported features.
+   */
+  warnings: Array<SharedV4Warning>;
+
+  /**
+   * Additional provider-specific metadata. They are passed through
+   * from the provider to the AI SDK and enable provider-specific
+   * results that can be fully encapsulated in the provider.
+   *
+   * The outer record is keyed by the provider name, and the inner
+   * record is provider-specific metadata.
+   *
+   * ```ts
+   * {
+   *   "fal": {
+   *     "videos": [{
+   *       "duration": 5.0,
+   *       "fps": 24,
+   *       "width": 1280,
+   *       "height": 720
+   *     }]
+   *   }
+   * }
+   * ```
+   */
+  providerMetadata?: SharedV4ProviderMetadata;
+
+  /**
+   * Response information for telemetry and debugging purposes.
+   */
+  response: {
+    /**
+     * Timestamp for the start of the generated response.
+     */
+    timestamp: Date;
+
+    /**
+     * The ID of the response model that was used to generate the response.
+     */
+    modelId: string;
+
+    /**
+     * Response headers.
+     */
+    headers: Record<string, string> | undefined;
+  };
+};

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

@@ -1,40 +1,10 @@
 import type { VideoModelV4CallOptions } from './video-model-v4-call-options';
-import type { SharedV4ProviderMetadata } from '../../shared/v4/shared-v4-provider-metadata';
-import type { SharedV4Warning } from '../../shared/v4/shared-v4-warning';
+import type { VideoModelV4Result } from './video-model-v4-result';
 
 type GetMaxVideosPerCallFunction = (options: {
   modelId: string;
 }) => PromiseLike<number | undefined> | number | undefined;
 
-/**
- * Generated video data. Can be a URL, base64-encoded string, or binary data.
- */
-export type VideoModelV4VideoData =
-  | {
-      /**
-       * Video available as a URL (most common for video providers).
-       */
-      type: 'url';
-      url: string;
-      mediaType: string;
-    }
-  | {
-      /**
-       * Video as base64-encoded string.
-       */
-      type: 'base64';
-      data: string;
-      mediaType: string;
-    }
-  | {
-      /**
-       * Video as binary data.
-       */
-      type: 'binary';
-      data: Uint8Array;
-      mediaType: string;
-    };
-
 /**
  * Video generation model specification version 3.
  */
@@ -72,61 +42,5 @@ export type VideoModelV4 = {
   /**
    * Generates an array of videos.
    */
-  doGenerate(options: VideoModelV4CallOptions): PromiseLike<{
-    /**
-     * Generated videos as URLs, base64 strings, or binary data.
-     *
-     * Most providers return URLs to video files (MP4, WebM) due to large file sizes.
-     * Use the discriminated union to indicate the type of video data being returned.
-     */
-    videos: Array<VideoModelV4VideoData>;
-
-    /**
-     * Warnings for the call, e.g. unsupported features.
-     */
-    warnings: Array<SharedV4Warning>;
-
-    /**
-     * Additional provider-specific metadata. They are passed through
-     * from the provider to the AI SDK and enable provider-specific
-     * results that can be fully encapsulated in the provider.
-     *
-     * The outer record is keyed by the provider name, and the inner
-     * record is provider-specific metadata.
-     *
-     * ```ts
-     * {
-     *   "fal": {
-     *     "videos": [{
-     *       "duration": 5.0,
-     *       "fps": 24,
-     *       "width": 1280,
-     *       "height": 720
-     *     }]
-     *   }
-     * }
-     * ```
-     */
-    providerMetadata?: SharedV4ProviderMetadata;
-
-    /**
-     * Response information for telemetry and debugging purposes.
-     */
-    response: {
-      /**
-       * Timestamp for the start of the generated response.
-       */
-      timestamp: Date;
-
-      /**
-       * The ID of the response model that was used to generate the response.
-       */
-      modelId: string;
-
-      /**
-       * Response headers.
-       */
-      headers: Record<string, string> | undefined;
-    };
-  }>;
+  doGenerate(options: VideoModelV4CallOptions): PromiseLike<VideoModelV4Result>;
 };