@ai-sdk/openai 4.0.110 → 4.0.113

package/src/transcription/openai-transcription-model-options.ts

@@ -18,91 +18,92 @@ export type OpenAITranscriptionModelId =
   | (string & {});
 
 // https://platform.openai.com/docs/api-reference/audio/createTranscription
-export const openAITranscriptionModelOptions = lazySchema(() =>
-  zodSchema(
-    z.object({
-      /**
-       * Additional information to include in the transcription response.
-       */
+/**
+ * Shape of the OpenAI transcription model options. Exported so that routers
+ * that also call the OpenAI transcription API (e.g. Azure) can accept these
+ * options under their own providerOptions key.
+ */
+export const openAITranscriptionModelOptionsShape = () => ({
+  /**
+   * Additional information to include in the transcription response.
+   */
 
-      include: z.array(z.string()).optional(),
+  include: z.array(z.string()).optional(),
 
-      /**
-       * The language of the input audio in ISO-639-1 format.
-       */
-      language: z.string().optional(),
+  /**
+   * The language of the input audio in ISO-639-1 format.
+   */
+  language: z.string().optional(),
 
-      /**
-       * Expected languages for REST transcription, e.g. with gpt-transcribe.
-       * Use this instead of language. The API validates model support.
-       */
-      languages: z.array(z.string()).optional(),
+  /**
+   * Expected languages for REST transcription, e.g. with gpt-transcribe.
+   * Use this instead of language. The API validates model support.
+   */
+  languages: z.array(z.string()).optional(),
 
-      /**
-       * Vocabulary hints for REST transcription, e.g. names or technical terms.
-       * The API validates model support.
-       */
-      keywords: z.array(z.string()).optional(),
+  /**
+   * Vocabulary hints for REST transcription, e.g. names or technical terms.
+   * The API validates model support.
+   */
+  keywords: z.array(z.string()).optional(),
 
-      /**
-       * An optional text to guide the model's style or continue a previous audio segment.
-       */
-      prompt: z.string().optional(),
+  /**
+   * An optional text to guide the model's style or continue a previous audio segment.
+   */
+  prompt: z.string().optional(),
 
-      /**
-       * The sampling temperature, between 0 and 1.
-       * @default 0
-       */
-      temperature: z.number().min(0).max(1).default(0).optional(),
+  /**
+   * The sampling temperature, between 0 and 1.
+   * @default 0
+   */
+  temperature: z.number().min(0).max(1).default(0).optional(),
 
-      /**
-       * The timestamp granularities to populate for this transcription.
-       * Defaults to ['segment'] when the response format is verbose_json.
-       */
-      timestampGranularities: z.array(z.enum(['word', 'segment'])).optional(),
+  /**
+   * The timestamp granularities to populate for this transcription.
+   * Defaults to ['segment'] when the response format is verbose_json.
+   */
+  timestampGranularities: z.array(z.enum(['word', 'segment'])).optional(),
 
-      /**
-       * The format of the transcription response.
-       */
-      responseFormat: z
-        .enum(['json', 'verbose_json', 'diarized_json'])
-        .optional(),
+  /**
+   * The format of the transcription response.
+   */
+  responseFormat: z.enum(['json', 'verbose_json', 'diarized_json']).optional(),
+
+  /**
+   * Controls how the audio is split into chunks before transcription.
+   */
+  chunkingStrategy: z
+    .union([
+      z.literal('auto'),
+      z.object({
+        type: z.literal('server_vad'),
+        threshold: z.number().min(0).max(1).optional(),
+        prefixPaddingMs: z.number().int().min(0).optional(),
+        silenceDurationMs: z.number().int().min(0).optional(),
+      }),
+    ])
+    .optional(),
 
+  /**
+   * Options for streaming transcription models such as `gpt-live-transcribe`.
+   */
+  streaming: z
+    .object({
       /**
-       * Controls how the audio is split into chunks before transcription.
+       * Latency/accuracy tradeoff for realtime transcription.
        */
-      chunkingStrategy: z
-        .union([
-          z.literal('auto'),
-          z.object({
-            type: z.literal('server_vad'),
-            threshold: z.number().min(0).max(1).optional(),
-            prefixPaddingMs: z.number().int().min(0).optional(),
-            silenceDurationMs: z.number().int().min(0).optional(),
-          }),
-        ])
-        .optional(),
+      delay: z.enum(['minimal', 'low', 'medium', 'high', 'xhigh']).optional(),
 
       /**
-       * Options for streaming transcription models such as `gpt-live-transcribe`.
+       * Additional fields to include in realtime transcription events.
        */
-      streaming: z
-        .object({
-          /**
-           * Latency/accuracy tradeoff for realtime transcription.
-           */
-          delay: z
-            .enum(['minimal', 'low', 'medium', 'high', 'xhigh'])
-            .optional(),
+      include: z.array(z.string()).optional(),
+    })
+    .optional(),
+});
 
-          /**
-           * Additional fields to include in realtime transcription events.
-           */
-          include: z.array(z.string()).optional(),
-        })
-        .optional(),
-    }),
-  ),
+export const openAITranscriptionModelOptions = lazySchema(() =>
+  zodSchema(z.object(openAITranscriptionModelOptionsShape())),
 );
 
 export type OpenAITranscriptionModelOptions = InferSchema<