@ai-sdk/gladia 3.0.0-beta.4 → 3.0.0-beta.50

package/src/gladia-transcription-model.ts

@@ -1,7 +1,7 @@
 import {
   AISDKError,
-  TranscriptionModelV4,
-  SharedV4Warning,
+  type TranscriptionModelV4,
+  type SharedV4Warning,
 } from '@ai-sdk/provider';
 import {
   combineHeaders,
@@ -10,322 +10,19 @@ import {
   mediaTypeToExtension,
   delay,
   getFromApi,
+  isSameOrigin,
   parseProviderOptions,
   postFormDataToApi,
   postJsonToApi,
+  serializeModelOptions,
+  WORKFLOW_SERIALIZE,
+  WORKFLOW_DESERIALIZE,
 } from '@ai-sdk/provider-utils';
 import { z } from 'zod/v4';
-import { GladiaConfig } from './gladia-config';
+import type { GladiaConfig } from './gladia-config';
 import { gladiaFailedResponseHandler } from './gladia-error';
-import { GladiaTranscriptionInitiateAPITypes } from './gladia-api-types';
-
-// https://docs.gladia.io/api-reference/v2/pre-recorded/init
-const gladiaTranscriptionModelOptionsSchema = z.object({
-  /**
-   * Optional context prompt to guide the transcription.
-   */
-  contextPrompt: z.string().nullish(),
-
-  /**
-   * Custom vocabulary to improve transcription accuracy.
-   * Can be a boolean or an array of custom terms.
-   */
-  customVocabulary: z.union([z.boolean(), z.array(z.any())]).nullish(),
-
-  /**
-   * Configuration for custom vocabulary.
-   */
-  customVocabularyConfig: z
-    .object({
-      /**
-       * Array of vocabulary terms or objects with pronunciation details.
-       */
-      vocabulary: z.array(
-        z.union([
-          z.string(),
-          z.object({
-            /**
-             * The vocabulary term.
-             */
-            value: z.string(),
-            /**
-             * Intensity of the term in recognition (optional).
-             */
-            intensity: z.number().nullish(),
-            /**
-             * Alternative pronunciations for the term (optional).
-             */
-            pronunciations: z.array(z.string()).nullish(),
-            /**
-             * Language of the term (optional).
-             */
-            language: z.string().nullish(),
-          }),
-        ]),
-      ),
-      /**
-       * Default intensity for all vocabulary terms.
-       */
-      defaultIntensity: z.number().nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to automatically detect the language of the audio.
-   */
-  detectLanguage: z.boolean().nullish(),
-
-  /**
-   * Whether to enable code switching (multiple languages in the same audio).
-   */
-  enableCodeSwitching: z.boolean().nullish(),
-
-  /**
-   * Configuration for code switching.
-   */
-  codeSwitchingConfig: z
-    .object({
-      /**
-       * Languages to consider for code switching.
-       */
-      languages: z.array(z.string()).nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Specific language for transcription.
-   */
-  language: z.string().nullish(),
-
-  /**
-   * Whether to enable callback when transcription is complete.
-   */
-  callback: z.boolean().nullish(),
-
-  /**
-   * Configuration for callback.
-   */
-  callbackConfig: z
-    .object({
-      /**
-       * URL to send the callback to.
-       */
-      url: z.string(),
-      /**
-       * HTTP method for the callback.
-       */
-      method: z.enum(['POST', 'PUT']).nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to generate subtitles.
-   */
-  subtitles: z.boolean().nullish(),
-
-  /**
-   * Configuration for subtitles generation.
-   */
-  subtitlesConfig: z
-    .object({
-      /**
-       * Subtitle file formats to generate.
-       */
-      formats: z.array(z.enum(['srt', 'vtt'])).nullish(),
-      /**
-       * Minimum duration for subtitle segments.
-       */
-      minimumDuration: z.number().nullish(),
-      /**
-       * Maximum duration for subtitle segments.
-       */
-      maximumDuration: z.number().nullish(),
-      /**
-       * Maximum characters per row in subtitles.
-       */
-      maximumCharactersPerRow: z.number().nullish(),
-      /**
-       * Maximum rows per caption in subtitles.
-       */
-      maximumRowsPerCaption: z.number().nullish(),
-      /**
-       * Style of subtitles.
-       */
-      style: z.enum(['default', 'compliance']).nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to enable speaker diarization (speaker identification).
-   */
-  diarization: z.boolean().nullish(),
-
-  /**
-   * Configuration for diarization.
-   */
-  diarizationConfig: z
-    .object({
-      /**
-       * Exact number of speakers to identify.
-       */
-      numberOfSpeakers: z.number().nullish(),
-      /**
-       * Minimum number of speakers to identify.
-       */
-      minSpeakers: z.number().nullish(),
-      /**
-       * Maximum number of speakers to identify.
-       */
-      maxSpeakers: z.number().nullish(),
-      /**
-       * Whether to use enhanced diarization.
-       */
-      enhanced: z.boolean().nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to translate the transcription.
-   */
-  translation: z.boolean().nullish(),
-
-  /**
-   * Configuration for translation.
-   */
-  translationConfig: z
-    .object({
-      /**
-       * Target languages for translation.
-       */
-      targetLanguages: z.array(z.string()),
-      /**
-       * Translation model to use.
-       */
-      model: z.enum(['base', 'enhanced']).nullish(),
-      /**
-       * Whether to match original utterances in translation.
-       */
-      matchOriginalUtterances: z.boolean().nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to generate a summary of the transcription.
-   */
-  summarization: z.boolean().nullish(),
-
-  /**
-   * Configuration for summarization.
-   */
-  summarizationConfig: z
-    .object({
-      /**
-       * Type of summary to generate.
-       */
-      type: z.enum(['general', 'bullet_points', 'concise']).nullish(),
-    })
-    .nullish(),
-
-  /**
-   * Whether to enable content moderation.
-   */
-  moderation: z.boolean().nullish(),
-
-  /**
-   * Whether to enable named entity recognition.
-   */
-  namedEntityRecognition: z.boolean().nullish(),
-
-  /**
-   * Whether to enable automatic chapter creation.
-   */
-  chapterization: z.boolean().nullish(),
-
-  /**
-   * Whether to ensure consistent naming of entities.
-   */
-  nameConsistency: z.boolean().nullish(),
-
-  /**
-   * Whether to enable custom spelling.
-   */
-  customSpelling: z.boolean().nullish(),
-
-  /**
-   * Configuration for custom spelling.
-   */
-  customSpellingConfig: z
-    .object({
-      /**
-       * Dictionary of custom spellings.
-       */
-      spellingDictionary: z.record(z.string(), z.array(z.string())),
-    })
-    .nullish(),
-
-  /**
-   * Whether to extract structured data from the transcription.
-   */
-  structuredDataExtraction: z.boolean().nullish(),
-
-  /**
-   * Configuration for structured data extraction.
-   */
-  structuredDataExtractionConfig: z
-    .object({
-      /**
-       * Classes of data to extract.
-       */
-      classes: z.array(z.string()),
-    })
-    .nullish(),
-
-  /**
-   * Whether to perform sentiment analysis on the transcription.
-   */
-  sentimentAnalysis: z.boolean().nullish(),
-
-  /**
-   * Whether to send audio to a language model for processing.
-   */
-  audioToLlm: z.boolean().nullish(),
-
-  /**
-   * Configuration for audio to language model processing.
-   */
-  audioToLlmConfig: z
-    .object({
-      /**
-       * Prompts to send to the language model.
-       */
-      prompts: z.array(z.string()),
-    })
-    .nullish(),
-
-  /**
-   * Custom metadata to include with the transcription.
-   */
-  customMetadata: z.record(z.string(), z.any()).nullish(),
-
-  /**
-   * Whether to include sentence-level segmentation.
-   */
-  sentences: z.boolean().nullish(),
-
-  /**
-   * Whether to enable display mode.
-   */
-  displayMode: z.boolean().nullish(),
-
-  /**
-   * Whether to enhance punctuation in the transcription.
-   */
-  punctuationEnhanced: z.boolean().nullish(),
-});
-
-export type GladiaTranscriptionModelOptions = z.infer<
-  typeof gladiaTranscriptionModelOptionsSchema
->;
+import { gladiaTranscriptionModelOptionsSchema } from './gladia-transcription-model-options';
+import type { GladiaTranscriptionInitiateAPITypes } from './gladia-api-types';
 
 interface GladiaTranscriptionModelConfig extends GladiaConfig {
   _internal?: {
@@ -340,8 +37,22 @@ export class GladiaTranscriptionModel implements TranscriptionModelV4 {
     return this.config.provider;
   }
 
+  static [WORKFLOW_SERIALIZE](model: GladiaTranscriptionModel) {
+    return serializeModelOptions({
+      modelId: model.modelId,
+      config: model.config,
+    });
+  }
+
+  static [WORKFLOW_DESERIALIZE](options: {
+    modelId: 'default';
+    config: GladiaTranscriptionModelConfig;
+  }) {
+    return new GladiaTranscriptionModel(options.modelId, options.config);
+  }
+
   constructor(
-    readonly modelId: 'default',
+    readonly modelId: string,
     private readonly config: GladiaTranscriptionModelConfig,
   ) {}
 
@@ -510,7 +221,7 @@ export class GladiaTranscriptionModel implements TranscriptionModelV4 {
         path: '/v2/upload',
         modelId: 'default',
       }),
-      headers: combineHeaders(this.config.headers(), options.headers),
+      headers: combineHeaders(this.config.headers?.(), options.headers),
       formData,
       failedResponseHandler: gladiaFailedResponseHandler,
       successfulResponseHandler: createJsonResponseHandler(
@@ -527,7 +238,7 @@ export class GladiaTranscriptionModel implements TranscriptionModelV4 {
         path: '/v2/pre-recorded',
         modelId: 'default',
       }),
-      headers: combineHeaders(this.config.headers(), options.headers),
+      headers: combineHeaders(this.config.headers?.(), options.headers),
       body: {
         ...body,
         audio_url: uploadResponse.audio_url,
@@ -540,8 +251,11 @@ export class GladiaTranscriptionModel implements TranscriptionModelV4 {
       fetch: this.config.fetch,
     });
 
-    // Poll the result URL until the transcription is done or an error occurs
+    // Poll the result URL until the transcription is done or an error occurs.
+    // The result URL comes from the provider response; only send credentials
+    // when it stays on the provider's own origin.
     const resultUrl = transcriptionInitResponse.result_url;
+    const apiOrigin = this.config.url({ modelId: 'default', path: '' });
     let transcriptionResult;
     let transcriptionResultHeaders;
     const timeoutMs = 60 * 1000; // 60 seconds timeout
@@ -560,7 +274,9 @@ export class GladiaTranscriptionModel implements TranscriptionModelV4 {
 
       const response = await getFromApi({
         url: resultUrl,
-        headers: combineHeaders(this.config.headers(), options.headers),
+        headers: isSameOrigin(resultUrl, apiOrigin)
+          ? combineHeaders(this.config.headers?.(), options.headers)
+          : undefined,
         failedResponseHandler: gladiaFailedResponseHandler,
         successfulResponseHandler: createJsonResponseHandler(
           gladiaTranscriptionResultResponseSchema,

package/src/index.ts

@@ -1,4 +1,4 @@
 export { createGladia, gladia } from './gladia-provider';
 export type { GladiaProvider, GladiaProviderSettings } from './gladia-provider';
-export type { GladiaTranscriptionModelOptions } from './gladia-transcription-model';
+export type { GladiaTranscriptionModelOptions } from './gladia-transcription-model-options';
 export { VERSION } from './version';

package/dist/index.d.mts

@@ -1,158 +0,0 @@
-import { TranscriptionModelV4, ProviderV4 } from '@ai-sdk/provider';
-import { FetchFunction } from '@ai-sdk/provider-utils';
-import { z } from 'zod/v4';
-
-type GladiaConfig = {
-    provider: string;
-    url: (options: {
-        modelId: string;
-        path: string;
-    }) => string;
-    headers: () => Record<string, string | undefined>;
-    fetch?: FetchFunction;
-    generateId?: () => string;
-};
-
-declare const gladiaTranscriptionModelOptionsSchema: z.ZodObject<{
-    contextPrompt: z.ZodOptional<z.ZodNullable<z.ZodString>>;
-    customVocabulary: z.ZodOptional<z.ZodNullable<z.ZodUnion<readonly [z.ZodBoolean, z.ZodArray<z.ZodAny>]>>>;
-    customVocabularyConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        vocabulary: z.ZodArray<z.ZodUnion<readonly [z.ZodString, z.ZodObject<{
-            value: z.ZodString;
-            intensity: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-            pronunciations: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodString>>>;
-            language: z.ZodOptional<z.ZodNullable<z.ZodString>>;
-        }, z.core.$strip>]>>;
-        defaultIntensity: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-    }, z.core.$strip>>>;
-    detectLanguage: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    enableCodeSwitching: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    codeSwitchingConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        languages: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodString>>>;
-    }, z.core.$strip>>>;
-    language: z.ZodOptional<z.ZodNullable<z.ZodString>>;
-    callback: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    callbackConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        url: z.ZodString;
-        method: z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-            POST: "POST";
-            PUT: "PUT";
-        }>>>;
-    }, z.core.$strip>>>;
-    subtitles: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    subtitlesConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        formats: z.ZodOptional<z.ZodNullable<z.ZodArray<z.ZodEnum<{
-            srt: "srt";
-            vtt: "vtt";
-        }>>>>;
-        minimumDuration: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        maximumDuration: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        maximumCharactersPerRow: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        maximumRowsPerCaption: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        style: z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-            default: "default";
-            compliance: "compliance";
-        }>>>;
-    }, z.core.$strip>>>;
-    diarization: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    diarizationConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        numberOfSpeakers: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        minSpeakers: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        maxSpeakers: z.ZodOptional<z.ZodNullable<z.ZodNumber>>;
-        enhanced: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    }, z.core.$strip>>>;
-    translation: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    translationConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        targetLanguages: z.ZodArray<z.ZodString>;
-        model: z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-            base: "base";
-            enhanced: "enhanced";
-        }>>>;
-        matchOriginalUtterances: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    }, z.core.$strip>>>;
-    summarization: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    summarizationConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        type: z.ZodOptional<z.ZodNullable<z.ZodEnum<{
-            general: "general";
-            bullet_points: "bullet_points";
-            concise: "concise";
-        }>>>;
-    }, z.core.$strip>>>;
-    moderation: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    namedEntityRecognition: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    chapterization: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    nameConsistency: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    customSpelling: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    customSpellingConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        spellingDictionary: z.ZodRecord<z.ZodString, z.ZodArray<z.ZodString>>;
-    }, z.core.$strip>>>;
-    structuredDataExtraction: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    structuredDataExtractionConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        classes: z.ZodArray<z.ZodString>;
-    }, z.core.$strip>>>;
-    sentimentAnalysis: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    audioToLlm: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    audioToLlmConfig: z.ZodOptional<z.ZodNullable<z.ZodObject<{
-        prompts: z.ZodArray<z.ZodString>;
-    }, z.core.$strip>>>;
-    customMetadata: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodAny>>>;
-    sentences: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    displayMode: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-    punctuationEnhanced: z.ZodOptional<z.ZodNullable<z.ZodBoolean>>;
-}, z.core.$strip>;
-type GladiaTranscriptionModelOptions = z.infer<typeof gladiaTranscriptionModelOptionsSchema>;
-interface GladiaTranscriptionModelConfig extends GladiaConfig {
-    _internal?: {
-        currentDate?: () => Date;
-    };
-}
-declare class GladiaTranscriptionModel implements TranscriptionModelV4 {
-    readonly modelId: 'default';
-    private readonly config;
-    readonly specificationVersion = "v4";
-    get provider(): string;
-    constructor(modelId: 'default', config: GladiaTranscriptionModelConfig);
-    private getArgs;
-    doGenerate(options: Parameters<TranscriptionModelV4['doGenerate']>[0]): Promise<Awaited<ReturnType<TranscriptionModelV4['doGenerate']>>>;
-}
-
-interface GladiaProvider extends ProviderV4 {
-    (): {
-        transcription: GladiaTranscriptionModel;
-    };
-    /**
-     * Creates a model for transcription.
-     */
-    transcription(): TranscriptionModelV4;
-    /**
-     * @deprecated Use `embeddingModel` instead.
-     */
-    textEmbeddingModel(modelId: string): never;
-}
-interface GladiaProviderSettings {
-    /**
-     * API key for authenticating requests.
-     */
-    apiKey?: string;
-    /**
-     * Custom headers to include in the requests.
-     */
-    headers?: Record<string, string>;
-    /**
-     * Custom fetch implementation. You can use it as a middleware to intercept requests,
-     * or to provide a custom fetch implementation for e.g. testing.
-     */
-    fetch?: FetchFunction;
-}
-/**
- * Create a Gladia provider instance.
- */
-declare function createGladia(options?: GladiaProviderSettings): GladiaProvider;
-/**
- * Default Gladia provider instance.
- */
-declare const gladia: GladiaProvider;
-
-declare const VERSION: string;
-
-export { type GladiaProvider, type GladiaProviderSettings, type GladiaTranscriptionModelOptions, VERSION, createGladia, gladia };