@ai-sdk/provider - npm Package Compare versions

Comparing version 4.0.3 to 4.0.4

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

+/**
+ * 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;
+};

src/speech-translation-model/index.ts

+export * from './v4/index';

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

+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';

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

+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;
+};

src/speech-transla...ech-translation-model-v4-stream-part.ts

+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;
+    };

src/speech-transla...h-translation-model-v4-stream-result.ts

+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;
+  };
+};

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

+/**
+ * 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;
+};

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

+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>;
+};

CHANGELOG.md

 # @ai-sdk/provider
 
+## 4.0.4
+
+### Patch Changes
+
+- 1e2f324: feat: add experimental speech translation model specification (`Experimental_SpeechTranslationModelV4`) and `experimental_streamTranslate` for streaming speech-to-speech translation
+
 ## 4.0.3
@@ -4,0 +10,0 @@

package.json

 {
   "name": "@ai-sdk/provider",
-  "version": "4.0.3",
+  "version": "4.0.4",
   "type": "module",
@@ -5,0 +5,0 @@ "license": "Apache-2.0",

src/embedding-model/v4/embedding-model-v4.ts

@@ -6,3 +6,3 @@ import type { EmbeddingModelV4CallOptions } from './embedding-model-v4-call-options';
  * Specification for an embedding model that implements the embedding model
- * interface version 3.
+ * interface version 4.
  *
@@ -9,0 +9,0 @@ * It is specific to text embeddings.

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

@@ -9,3 +9,3 @@ import type { ImageModelV4CallOptions } from './image-model-v4-call-options';
 /**
- * Image generation model specification version 3.
+ * Image generation model specification version 4.
  */
@@ -12,0 +12,0 @@ export type ImageModelV4 = {

src/index.ts

@@ -16,2 +16,3 @@ export * from './embedding-model/index';
 export * from './speech-model/index';
+export * from './speech-translation-model/index';
 export * from './transcription-model/index';
@@ -18,0 +19,0 @@ export * from './video-model/index';

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

@@ -0,1 +1,2 @@
+import type { SharedV4AudioFormat } from '../../shared';
 import type { RealtimeModelV4ToolDefinition } from './realtime-model-v4-tool-definition';
@@ -26,14 +27,4 @@
    */
-  inputAudioFormat?: {
-    /**
-     * Audio format type (e.g. "audio/pcm", "audio/pcmu", "audio/pcma").
-     */
-    type: string;
+  inputAudioFormat?: SharedV4AudioFormat;
 
-    /**
-     * Sample rate in Hz. Only applicable for PCM format.
-     */
-    rate?: number;
-  };
-
   /**
@@ -91,14 +82,4 @@ * Input audio transcription configuration.
    */
-  outputAudioFormat?: {
-    /**
-     * Audio format type (e.g. "audio/pcm", "audio/pcmu", "audio/pcma").
-     */
-    type: string;
+  outputAudioFormat?: SharedV4AudioFormat;
 
-    /**
-     * Sample rate in Hz. Only applicable for PCM format.
-     */
-    rate?: number;
-  };
-
   /**
@@ -105,0 +86,0 @@ * Voice activity detection configuration.

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

@@ -5,3 +5,3 @@ import type { RerankingModelV4CallOptions } from './reranking-model-v4-call-options';
 /**
- * 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.
  */
@@ -8,0 +8,0 @@ export type RerankingModelV4 = {

src/shared/v4/index.ts

@@ -0,1 +1,2 @@
+export * from './shared-v4-audio-format';
 export * from './shared-v4-file-data';
@@ -2,0 +3,0 @@ export * from './shared-v4-headers';

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

@@ -5,3 +5,3 @@ import type { SpeechModelV4CallOptions } from './speech-model-v4-call-options';
 /**
- * Speech model specification version 3.
+ * Speech model specification version 4.
  */
@@ -8,0 +8,0 @@ export type SpeechModelV4 = {

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

 import type { JSONObject } from '../../json-value/json-value';
+import type { SharedV4AudioFormat } from '../../shared';
 
@@ -17,14 +18,4 @@ type TranscriptionModelV4ProviderOptions = Record<string, JSONObject>;
    */
-  inputAudioFormat: {
-    /**
-     * Audio format type, e.g. `audio/pcm`, `audio/pcmu`, or `audio/pcma`.
-     */
-    type: string;
+  inputAudioFormat: SharedV4AudioFormat;
 
-    /**
-     * Sample rate in Hz. Only applicable for formats that require a rate.
-     */
-    rate?: number;
-  };
-
   /**
@@ -31,0 +22,0 @@ * Additional provider-specific options that are passed through to the provider.

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

@@ -7,3 +7,3 @@ import type { TranscriptionModelV4CallOptions } from './transcription-model-v4-call-options';
 /**
- * Transcription model specification version 3.
+ * Transcription model specification version 4.
  */
@@ -10,0 +10,0 @@ export type TranscriptionModelV4 = {

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

@@ -9,3 +9,3 @@ import type { VideoModelV4CallOptions } from './video-model-v4-call-options';
 /**
- * Video generation model specification version 3.
+ * Video generation model specification version 4.
  */
@@ -12,0 +12,0 @@ export type VideoModelV4 = {

No alert changes

Improved metrics

Total package byte prevSize

598488

3.11%

Number of package files

240

3.45%

Lines of code

16628

3.41%

No dependency changes