@ai-sdk/google-vertex 4.0.146 → 4.0.147

package/dist/index.js

@@ -27,7 +27,7 @@ __export(index_exports, {
 module.exports = __toCommonJS(index_exports);
 
 // src/google-vertex-provider-node.ts
-var import_provider_utils6 = require("@ai-sdk/provider-utils");
+var import_provider_utils7 = require("@ai-sdk/provider-utils");
 
 // src/google-vertex-auth-google-auth-library.ts
 var import_google_auth_library = require("google-auth-library");
@@ -46,10 +46,10 @@ function createAuthTokenGenerator(options) {
 
 // src/google-vertex-provider.ts
 var import_internal3 = require("@ai-sdk/google/internal");
-var import_provider_utils5 = require("@ai-sdk/provider-utils");
+var import_provider_utils6 = require("@ai-sdk/provider-utils");
 
 // src/version.ts
-var VERSION = true ? "4.0.146" : "0.0.0-test";
+var VERSION = true ? "4.0.147" : "0.0.0-test";
 
 // src/google-vertex-embedding-model.ts
 var import_provider = require("@ai-sdk/provider");
@@ -555,10 +555,182 @@ var googleVertexTools = {
   vertexRagStore: import_internal2.googleTools.vertexRagStore
 };
 
-// src/google-vertex-video-model.ts
-var import_provider2 = require("@ai-sdk/provider");
+// src/google-vertex-transcription-model.ts
 var import_provider_utils4 = require("@ai-sdk/provider-utils");
+var import_v46 = require("zod/v4");
+
+// src/google-vertex-transcription-model-options.ts
 var import_v45 = require("zod/v4");
+var googleVertexTranscriptionProviderOptionsSchema = import_v45.z.object({
+  /**
+   * BCP-47 language codes to recognize (e.g. `['en-US']`), or `['auto']` to let
+   * Chirp auto-detect the spoken language. Defaults to `['auto']`. For
+   * `telephony`, pass a supported explicit language code.
+   */
+  languageCodes: import_v45.z.array(import_v45.z.string()).optional(),
+  /**
+   * Whether to add punctuation to the transcript. Defaults to `true`.
+   */
+  enableAutomaticPunctuation: import_v45.z.boolean().optional(),
+  /**
+   * Whether to include word-level timestamps. Defaults to `true` so the
+   * transcription result can include segments.
+   *
+   * Enabling word-level timestamps can reduce transcription quality and speed
+   * for Chirp models.
+   */
+  enableWordTimeOffsets: import_v45.z.boolean().optional(),
+  /**
+   * The Cloud Speech-to-Text region for the request (e.g. `'us'`, `'eu'`,
+   * `'us-central1'`). Defaults to the provider `location`.
+   *
+   * Note: Speech-to-Text regions differ from Vertex AI regions. Chirp is only
+   * available in specific Speech-to-Text regions and is not available in the
+   * `global` location.
+   */
+  region: import_v45.z.string().optional()
+});
+
+// src/google-vertex-transcription-model.ts
+function parseDurationSeconds(value) {
+  if (value == null) {
+    return void 0;
+  }
+  const seconds = Number.parseFloat(value);
+  return Number.isFinite(seconds) ? seconds : void 0;
+}
+function convertBcp47ToIso6391(value) {
+  if (value == null) {
+    return void 0;
+  }
+  try {
+    const language = new Intl.Locale(value).language;
+    return language.length === 2 ? language : void 0;
+  } catch (e) {
+    return void 0;
+  }
+}
+var GoogleVertexTranscriptionModel = class {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v3";
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  async doGenerate(options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const warnings = [];
+    let googleOptions;
+    for (const provider of ["googleVertex", "vertex", "google"]) {
+      googleOptions = await (0, import_provider_utils4.parseProviderOptions)({
+        provider,
+        providerOptions: options.providerOptions,
+        schema: googleVertexTranscriptionProviderOptionsSchema
+      });
+      if (googleOptions != null) {
+        break;
+      }
+    }
+    const region = (_d = googleOptions == null ? void 0 : googleOptions.region) != null ? _d : this.config.location;
+    const languageCodes = (_e = googleOptions == null ? void 0 : googleOptions.languageCodes) != null ? _e : ["auto"];
+    const content = typeof options.audio === "string" ? options.audio : (0, import_provider_utils4.convertUint8ArrayToBase64)(options.audio);
+    const requestBody = {
+      config: {
+        model: this.modelId,
+        languageCodes,
+        // Let Speech-to-Text auto-detect the audio encoding (wav/mp3/flac/…).
+        autoDecodingConfig: {},
+        features: {
+          // Word timing populates `segments`.
+          enableWordTimeOffsets: (_f = googleOptions == null ? void 0 : googleOptions.enableWordTimeOffsets) != null ? _f : true,
+          enableAutomaticPunctuation: (_g = googleOptions == null ? void 0 : googleOptions.enableAutomaticPunctuation) != null ? _g : true
+        }
+      },
+      content
+    };
+    const host = region === "global" ? "speech.googleapis.com" : `${region}-speech.googleapis.com`;
+    const url = `https://${host}/v2/projects/${this.config.project}/locations/${region}/recognizers/_:recognize`;
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await (0, import_provider_utils4.postJsonToApi)({
+      url,
+      headers: (0, import_provider_utils4.combineHeaders)(
+        this.config.headers ? await (0, import_provider_utils4.resolve)(this.config.headers) : void 0,
+        options.headers
+      ),
+      body: requestBody,
+      failedResponseHandler: googleVertexFailedResponseHandler,
+      successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
+        googleVertexTranscriptionResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    const results = (_h = response.results) != null ? _h : [];
+    const text = results.map((result) => {
+      var _a2, _b2, _c2;
+      return (_c2 = (_b2 = (_a2 = result.alternatives) == null ? void 0 : _a2[0]) == null ? void 0 : _b2.transcript) != null ? _c2 : "";
+    }).join(" ").trim();
+    const segments = results.flatMap(
+      (result) => {
+        var _a2, _b2, _c2, _d2;
+        return (_d2 = (_c2 = (_b2 = (_a2 = result.alternatives) == null ? void 0 : _a2[0]) == null ? void 0 : _b2.words) == null ? void 0 : _c2.flatMap((word) => {
+          const startSecond = parseDurationSeconds(word.startOffset);
+          const endSecond = parseDurationSeconds(word.endOffset);
+          return word.word == null || startSecond == null || endSecond == null ? [] : [{ text: word.word, startSecond, endSecond }];
+        })) != null ? _d2 : [];
+      }
+    );
+    const language = convertBcp47ToIso6391((_i = results[0]) == null ? void 0 : _i.languageCode);
+    return {
+      text,
+      segments,
+      language,
+      durationInSeconds: parseDurationSeconds(
+        (_j = response.metadata) == null ? void 0 : _j.totalBilledDuration
+      ),
+      warnings,
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      }
+    };
+  }
+};
+var googleVertexTranscriptionResponseSchema = import_v46.z.object({
+  results: import_v46.z.array(
+    import_v46.z.object({
+      alternatives: import_v46.z.array(
+        import_v46.z.object({
+          transcript: import_v46.z.string().nullish(),
+          words: import_v46.z.array(
+            import_v46.z.object({
+              word: import_v46.z.string().nullish(),
+              startOffset: import_v46.z.string().nullish(),
+              endOffset: import_v46.z.string().nullish()
+            })
+          ).nullish()
+        })
+      ).nullish(),
+      languageCode: import_v46.z.string().nullish()
+    })
+  ).nullish(),
+  metadata: import_v46.z.object({
+    totalBilledDuration: import_v46.z.string().nullish()
+  }).nullish()
+});
+
+// src/google-vertex-video-model.ts
+var import_provider2 = require("@ai-sdk/provider");
+var import_provider_utils5 = require("@ai-sdk/provider-utils");
+var import_v47 = require("zod/v4");
 var GoogleVertexVideoModel = class {
   constructor(modelId, config) {
     this.modelId = modelId;
@@ -575,7 +747,7 @@ var GoogleVertexVideoModel = class {
     var _a, _b, _c, _d, _e, _f;
     const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
     const warnings = [];
-    const vertexOptions = await (0, import_provider_utils4.parseProviderOptions)({
+    const vertexOptions = await (0, import_provider_utils5.parseProviderOptions)({
       provider: "vertex",
       providerOptions: options.providerOptions,
       schema: googleVertexVideoModelOptionsSchema
@@ -593,7 +765,7 @@ var GoogleVertexVideoModel = class {
           details: "Vertex AI video models require base64-encoded images or GCS URIs. URL will be ignored."
         });
       } else {
-        const base64Data = typeof options.image.data === "string" ? options.image.data : (0, import_provider_utils4.convertUint8ArrayToBase64)(options.image.data);
+        const base64Data = typeof options.image.data === "string" ? options.image.data : (0, import_provider_utils5.convertUint8ArrayToBase64)(options.image.data);
         instance.image = {
           bytesBase64Encoded: base64Data,
           mimeType: options.image.mediaType
@@ -651,17 +823,17 @@ var GoogleVertexVideoModel = class {
         }
       }
     }
-    const { value: operation } = await (0, import_provider_utils4.postJsonToApi)({
+    const { value: operation } = await (0, import_provider_utils5.postJsonToApi)({
       url: `${this.config.baseURL}/models/${this.modelId}:predictLongRunning`,
-      headers: (0, import_provider_utils4.combineHeaders)(
-        await (0, import_provider_utils4.resolve)(this.config.headers),
+      headers: (0, import_provider_utils5.combineHeaders)(
+        await (0, import_provider_utils5.resolve)(this.config.headers),
         options.headers
       ),
       body: {
         instances,
         parameters
       },
-      successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
+      successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
         vertexOperationSchema
       ),
       failedResponseHandler: googleVertexFailedResponseHandler,
@@ -687,23 +859,23 @@ var GoogleVertexVideoModel = class {
           message: `Video generation timed out after ${pollTimeoutMs}ms`
         });
       }
-      await (0, import_provider_utils4.delay)(pollIntervalMs);
+      await (0, import_provider_utils5.delay)(pollIntervalMs);
       if ((_f = options.abortSignal) == null ? void 0 : _f.aborted) {
         throw new import_provider2.AISDKError({
           name: "VERTEX_VIDEO_GENERATION_ABORTED",
           message: "Video generation request was aborted"
         });
       }
-      const { value: statusOperation, responseHeaders: pollHeaders } = await (0, import_provider_utils4.postJsonToApi)({
+      const { value: statusOperation, responseHeaders: pollHeaders } = await (0, import_provider_utils5.postJsonToApi)({
         url: `${this.config.baseURL}/models/${this.modelId}:fetchPredictOperation`,
-        headers: (0, import_provider_utils4.combineHeaders)(
-          await (0, import_provider_utils4.resolve)(this.config.headers),
+        headers: (0, import_provider_utils5.combineHeaders)(
+          await (0, import_provider_utils5.resolve)(this.config.headers),
           options.headers
         ),
         body: {
           operationName
         },
-        successfulResponseHandler: (0, import_provider_utils4.createJsonResponseHandler)(
+        successfulResponseHandler: (0, import_provider_utils5.createJsonResponseHandler)(
           vertexOperationSchema
         ),
         failedResponseHandler: googleVertexFailedResponseHandler,
@@ -772,38 +944,38 @@ var GoogleVertexVideoModel = class {
     };
   }
 };
-var vertexOperationSchema = import_v45.z.object({
-  name: import_v45.z.string().nullish(),
-  done: import_v45.z.boolean().nullish(),
-  error: import_v45.z.object({
-    code: import_v45.z.number().nullish(),
-    message: import_v45.z.string(),
-    status: import_v45.z.string().nullish()
+var vertexOperationSchema = import_v47.z.object({
+  name: import_v47.z.string().nullish(),
+  done: import_v47.z.boolean().nullish(),
+  error: import_v47.z.object({
+    code: import_v47.z.number().nullish(),
+    message: import_v47.z.string(),
+    status: import_v47.z.string().nullish()
   }).nullish(),
-  response: import_v45.z.object({
-    videos: import_v45.z.array(
-      import_v45.z.object({
-        bytesBase64Encoded: import_v45.z.string().nullish(),
-        gcsUri: import_v45.z.string().nullish(),
-        mimeType: import_v45.z.string().nullish()
+  response: import_v47.z.object({
+    videos: import_v47.z.array(
+      import_v47.z.object({
+        bytesBase64Encoded: import_v47.z.string().nullish(),
+        gcsUri: import_v47.z.string().nullish(),
+        mimeType: import_v47.z.string().nullish()
       })
     ).nullish(),
-    raiMediaFilteredCount: import_v45.z.number().nullish()
+    raiMediaFilteredCount: import_v47.z.number().nullish()
   }).nullish()
 });
-var googleVertexVideoModelOptionsSchema = (0, import_provider_utils4.lazySchema)(
-  () => (0, import_provider_utils4.zodSchema)(
-    import_v45.z.object({
-      pollIntervalMs: import_v45.z.number().positive().nullish(),
-      pollTimeoutMs: import_v45.z.number().positive().nullish(),
-      personGeneration: import_v45.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
-      negativePrompt: import_v45.z.string().nullish(),
-      generateAudio: import_v45.z.boolean().nullish(),
-      gcsOutputDirectory: import_v45.z.string().nullish(),
-      referenceImages: import_v45.z.array(
-        import_v45.z.object({
-          bytesBase64Encoded: import_v45.z.string().nullish(),
-          gcsUri: import_v45.z.string().nullish()
+var googleVertexVideoModelOptionsSchema = (0, import_provider_utils5.lazySchema)(
+  () => (0, import_provider_utils5.zodSchema)(
+    import_v47.z.object({
+      pollIntervalMs: import_v47.z.number().positive().nullish(),
+      pollTimeoutMs: import_v47.z.number().positive().nullish(),
+      personGeneration: import_v47.z.enum(["dont_allow", "allow_adult", "allow_all"]).nullish(),
+      negativePrompt: import_v47.z.string().nullish(),
+      generateAudio: import_v47.z.boolean().nullish(),
+      gcsOutputDirectory: import_v47.z.string().nullish(),
+      referenceImages: import_v47.z.array(
+        import_v47.z.object({
+          bytesBase64Encoded: import_v47.z.string().nullish(),
+          gcsUri: import_v47.z.string().nullish()
         })
       ).nullish()
     }).passthrough()
@@ -817,7 +989,7 @@ function createExpressModeFetch(apiKey, customFetch) {
     const modifiedInit = {
       ...init,
       headers: {
-        ...(init == null ? void 0 : init.headers) ? (0, import_provider_utils5.normalizeHeaders)(init.headers) : {},
+        ...(init == null ? void 0 : init.headers) ? (0, import_provider_utils6.normalizeHeaders)(init.headers) : {},
         "x-goog-api-key": apiKey
       }
     };
@@ -825,17 +997,17 @@ function createExpressModeFetch(apiKey, customFetch) {
   };
 }
 function createVertex(options = {}) {
-  const apiKey = (0, import_provider_utils5.loadOptionalSetting)({
+  const apiKey = (0, import_provider_utils6.loadOptionalSetting)({
     settingValue: options.apiKey,
     environmentVariableName: "GOOGLE_VERTEX_API_KEY"
   });
-  const loadVertexProject = () => (0, import_provider_utils5.loadSetting)({
+  const loadVertexProject = () => (0, import_provider_utils6.loadSetting)({
     settingValue: options.project,
     settingName: "project",
     environmentVariableName: "GOOGLE_VERTEX_PROJECT",
     description: "Google Vertex project"
   });
-  const loadVertexLocation = () => (0, import_provider_utils5.loadSetting)({
+  const loadVertexLocation = () => (0, import_provider_utils6.loadSetting)({
     settingValue: options.location,
     settingName: "location",
     environmentVariableName: "GOOGLE_VERTEX_LOCATION",
@@ -844,7 +1016,7 @@ function createVertex(options = {}) {
   const loadBaseURL = () => {
     var _a, _b;
     if (apiKey) {
-      return (_a = (0, import_provider_utils5.withoutTrailingSlash)(options.baseURL)) != null ? _a : EXPRESS_MODE_BASE_URL;
+      return (_a = (0, import_provider_utils6.withoutTrailingSlash)(options.baseURL)) != null ? _a : EXPRESS_MODE_BASE_URL;
     }
     const region = loadVertexLocation();
     const project = loadVertexProject();
@@ -857,13 +1029,13 @@ function createVertex(options = {}) {
         return `${region}-aiplatform.googleapis.com`;
       }
     };
-    return (_b = (0, import_provider_utils5.withoutTrailingSlash)(options.baseURL)) != null ? _b : `https://${getHost()}/v1beta1/projects/${project}/locations/${region}/publishers/google`;
+    return (_b = (0, import_provider_utils6.withoutTrailingSlash)(options.baseURL)) != null ? _b : `https://${getHost()}/v1beta1/projects/${project}/locations/${region}/publishers/google`;
   };
   const createConfig = (name) => {
     const getHeaders = async () => {
       var _a;
-      const originalHeaders = await (0, import_provider_utils5.resolve)((_a = options.headers) != null ? _a : {});
-      return (0, import_provider_utils5.withUserAgentSuffix)(
+      const originalHeaders = await (0, import_provider_utils6.resolve)((_a = options.headers) != null ? _a : {});
+      return (0, import_provider_utils6.withUserAgentSuffix)(
         originalHeaders,
         `ai-sdk/google-vertex/${VERSION}`
       );
@@ -879,7 +1051,7 @@ function createVertex(options = {}) {
     var _a;
     return new import_internal3.GoogleGenerativeAILanguageModel(modelId, {
       ...createConfig("chat"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils5.generateId,
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId,
       supportedUrls: () => ({
         "*": [
           // HTTP URLs:
@@ -895,14 +1067,29 @@ function createVertex(options = {}) {
     var _a;
     return new GoogleVertexImageModel(modelId, {
       ...createConfig("image"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils5.generateId
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId
     });
   };
   const createVideoModel = (modelId) => {
     var _a;
     return new GoogleVertexVideoModel(modelId, {
       ...createConfig("video"),
-      generateId: (_a = options.generateId) != null ? _a : import_provider_utils5.generateId
+      generateId: (_a = options.generateId) != null ? _a : import_provider_utils6.generateId
+    });
+  };
+  const createTranscriptionModel = (modelId) => {
+    if (apiKey) {
+      throw new Error(
+        "Google Vertex transcription models do not support Express Mode API keys. Use standard Google Cloud credentials instead."
+      );
+    }
+    const config = createConfig("transcription");
+    return new GoogleVertexTranscriptionModel(modelId, {
+      provider: config.provider,
+      headers: config.headers,
+      fetch: config.fetch,
+      project: loadVertexProject(),
+      location: loadVertexLocation()
     });
   };
   const provider = function(modelId) {
@@ -921,13 +1108,15 @@ function createVertex(options = {}) {
   provider.imageModel = createImageModel;
   provider.video = createVideoModel;
   provider.videoModel = createVideoModel;
+  provider.transcription = createTranscriptionModel;
+  provider.transcriptionModel = createTranscriptionModel;
   provider.tools = googleVertexTools;
   return provider;
 }
 
 // src/google-vertex-provider-node.ts
 function createVertex2(options = {}) {
-  const apiKey = (0, import_provider_utils6.loadOptionalSetting)({
+  const apiKey = (0, import_provider_utils7.loadOptionalSetting)({
     settingValue: options.apiKey,
     environmentVariableName: "GOOGLE_VERTEX_API_KEY"
   });
@@ -943,7 +1132,7 @@ function createVertex2(options = {}) {
     ...options,
     headers: async () => ({
       Authorization: `Bearer ${await generateAuthToken()}`,
-      ...await (0, import_provider_utils6.resolve)(options.headers)
+      ...await (0, import_provider_utils7.resolve)(options.headers)
     })
   });
 }