@ai-sdk/provider 0.0.0-01d6317c-20260129172110

package/src/speech-model/v3/speech-model-v3.ts

@@ -0,0 +1,89 @@
+import { JSONObject } from '../../json-value';
+import { SharedV2Headers } from '../../shared';
+import { SharedV3Warning } from '../../shared/v3/shared-v3-warning';
+import { SpeechModelV3CallOptions } from './speech-model-v3-call-options';
+
+/**
+ * Speech model specification version 3.
+ */
+export type SpeechModelV3 = {
+  /**
+   * The speech model must specify which speech model interface
+   * version it implements. This will allow us to evolve the speech
+   * model interface and retain backwards compatibility. The different
+   * implementation versions can be handled as a discriminated union
+   * on our side.
+   */
+  readonly specificationVersion: 'v3';
+
+  /**
+   * Name of the provider for logging purposes.
+   */
+  readonly provider: string;
+
+  /**
+   * Provider-specific model ID for logging purposes.
+   */
+  readonly modelId: string;
+
+  /**
+   * Generates speech audio from text.
+   */
+  doGenerate(options: SpeechModelV3CallOptions): 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<SharedV3Warning>;
+
+    /**
+     * 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/transcription-model/index.ts

@@ -0,0 +1,2 @@
+export * from './v2/index';
+export * from './v3/index';

package/src/transcription-model/v2/index.ts

@@ -0,0 +1,3 @@
+export type { TranscriptionModelV2 } from './transcription-model-v2';
+export type { TranscriptionModelV2CallOptions } from './transcription-model-v2-call-options';
+export type { TranscriptionModelV2CallWarning } from './transcription-model-v2-call-warning';

package/src/transcription-model/v2/transcription-model-v2-call-options.ts

@@ -0,0 +1,48 @@
+import { JSONValue } from '../../json-value/json-value';
+
+type TranscriptionModelV2ProviderOptions = Record<
+  string,
+  Record<string, JSONValue>
+>;
+
+export type TranscriptionModelV2CallOptions = {
+  /**
+   * Audio data to transcribe.
+   * Accepts a `Uint8Array` or `string`, where `string` is a base64 encoded audio file.
+   */
+  audio: Uint8Array | string;
+
+  /**
+   * The IANA media type of the audio data.
+   *
+   * @see https://www.iana.org/assignments/media-types/media-types.xhtml
+   */
+  mediaType: string;
+
+  /**
+   * Additional provider-specific options that are passed through to the provider
+   * as body parameters.
+   *
+   * The outer record is keyed by the provider name, and the inner
+   * record is keyed by the provider-specific metadata key.
+   * ```ts
+   * {
+   * "openai": {
+   * "timestampGranularities": ["word"]
+   * }
+   * }
+   * ```
+   */
+  providerOptions?: TranscriptionModelV2ProviderOptions;
+
+  /**
+   * Abort signal for cancelling the operation.
+   */
+  abortSignal?: AbortSignal;
+
+  /**
+   * Additional HTTP headers to be sent with the request.
+   * Only applicable for HTTP-based providers.
+   */
+  headers?: Record<string, string | undefined>;
+};

package/src/transcription-model/v2/transcription-model-v2-call-warning.ts

@@ -0,0 +1,16 @@
+import { TranscriptionModelV2CallOptions } from './transcription-model-v2-call-options';
+
+/**
+ * Warning from the model provider for this call. The call will proceed, but e.g.
+ * some settings might not be supported, which can lead to suboptimal results.
+ */
+export type TranscriptionModelV2CallWarning =
+  | {
+      type: 'unsupported-setting';
+      setting: keyof TranscriptionModelV2CallOptions;
+      details?: string;
+    }
+  | {
+      type: 'other';
+      message: string;
+    };

package/src/transcription-model/v2/transcription-model-v2.ts

@@ -0,0 +1,117 @@
+import { JSONValue } from '../../json-value';
+import { SharedV2Headers } from '../../shared';
+import { TranscriptionModelV2CallOptions } from './transcription-model-v2-call-options';
+import { TranscriptionModelV2CallWarning } from './transcription-model-v2-call-warning';
+
+/**
+ * Transcription model specification version 2.
+ */
+export type TranscriptionModelV2 = {
+  /**
+   * The transcription model must specify which transcription model interface
+   * version it implements. This will allow us to evolve the transcription
+   * model interface and retain backwards compatibility. The different
+   * implementation versions can be handled as a discriminated union
+   * on our side.
+   */
+  readonly specificationVersion: 'v2';
+
+  /**
+   * Name of the provider for logging purposes.
+   */
+  readonly provider: string;
+
+  /**
+   * Provider-specific model ID for logging purposes.
+   */
+  readonly modelId: string;
+
+  /**
+   * Generates a transcript.
+   */
+  doGenerate(options: TranscriptionModelV2CallOptions): 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<TranscriptionModelV2CallWarning>;
+
+    /**
+     * 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?: 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, Record<string, JSONValue>>;
+  }>;
+};

package/src/transcription-model/v3/index.ts

@@ -0,0 +1,2 @@
+export type { TranscriptionModelV3 } from './transcription-model-v3';
+export type { TranscriptionModelV3CallOptions } from './transcription-model-v3-call-options';

package/src/transcription-model/v3/transcription-model-v3-call-options.ts

@@ -0,0 +1,45 @@
+import { JSONObject } from '../../json-value/json-value';
+
+type TranscriptionModelV3ProviderOptions = Record<string, JSONObject>;
+
+export type TranscriptionModelV3CallOptions = {
+  /**
+   * Audio data to transcribe.
+   * Accepts a `Uint8Array` or `string`, where `string` is a base64 encoded audio file.
+   */
+  audio: Uint8Array | string;
+
+  /**
+   * The IANA media type of the audio data.
+   *
+   * @see https://www.iana.org/assignments/media-types/media-types.xhtml
+   */
+  mediaType: string;
+
+  /**
+   * Additional provider-specific options that are passed through to the provider
+   * as body parameters.
+   *
+   * The outer record is keyed by the provider name, and the inner
+   * record is keyed by the provider-specific metadata key.
+   * ```ts
+   * {
+   * "openai": {
+   * "timestampGranularities": ["word"]
+   * }
+   * }
+   * ```
+   */
+  providerOptions?: TranscriptionModelV3ProviderOptions;
+
+  /**
+   * Abort signal for cancelling the operation.
+   */
+  abortSignal?: AbortSignal;
+
+  /**
+   * Additional HTTP headers to be sent with the request.
+   * Only applicable for HTTP-based providers.
+   */
+  headers?: Record<string, string | undefined>;
+};

package/src/transcription-model/v3/transcription-model-v3.ts

@@ -0,0 +1,117 @@
+import { JSONObject } from '../../json-value';
+import { SharedV3Headers } from '../../shared';
+import { SharedV3Warning } from '../../shared/v3/shared-v3-warning';
+import { TranscriptionModelV3CallOptions } from './transcription-model-v3-call-options';
+
+/**
+ * Transcription model specification version 3.
+ */
+export type TranscriptionModelV3 = {
+  /**
+   * The transcription model must specify which transcription model interface
+   * version it implements. This will allow us to evolve the transcription
+   * model interface and retain backwards compatibility. The different
+   * implementation versions can be handled as a discriminated union
+   * on our side.
+   */
+  readonly specificationVersion: 'v3';
+
+  /**
+   * Name of the provider for logging purposes.
+   */
+  readonly provider: string;
+
+  /**
+   * Provider-specific model ID for logging purposes.
+   */
+  readonly modelId: string;
+
+  /**
+   * Generates a transcript.
+   */
+  doGenerate(options: TranscriptionModelV3CallOptions): 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<SharedV3Warning>;
+
+    /**
+     * 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?: SharedV3Headers;
+
+      /**
+       * 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>;
+  }>;
+};