@ai-sdk/google 4.0.0-canary.76 → 4.0.0-canary.78

package/dist/internal/index.d.ts

@@ -1,6 +1,7 @@
 import * as _ai_sdk_provider_utils from '@ai-sdk/provider-utils';
 import { WORKFLOW_SERIALIZE, WORKFLOW_DESERIALIZE, Resolvable, FetchFunction, InferSchema } from '@ai-sdk/provider-utils';
-import { LanguageModelV4, JSONObject, LanguageModelV4CallOptions, LanguageModelV4GenerateResult, LanguageModelV4StreamResult } from '@ai-sdk/provider';
+import * as _ai_sdk_provider from '@ai-sdk/provider';
+import { LanguageModelV4, JSONObject, LanguageModelV4CallOptions, LanguageModelV4GenerateResult, LanguageModelV4StreamResult, SpeechModelV4 } from '@ai-sdk/provider';
 import { z } from 'zod/v4';
 
 type GoogleModelId = 'gemini-2.0-flash' | 'gemini-2.0-flash-001' | 'gemini-2.0-flash-lite' | 'gemini-2.0-flash-lite-001' | 'gemini-2.5-pro' | 'gemini-2.5-flash' | 'gemini-2.5-flash-image' | 'gemini-2.5-flash-lite' | 'gemini-2.5-flash-preview-tts' | 'gemini-2.5-pro-preview-tts' | 'gemini-2.5-flash-native-audio-latest' | 'gemini-2.5-flash-native-audio-preview-09-2025' | 'gemini-2.5-flash-native-audio-preview-12-2025' | 'gemini-2.5-computer-use-preview-10-2025' | 'gemini-3-pro-preview' | 'gemini-3-pro-image-preview' | 'gemini-3-flash-preview' | 'gemini-3.1-pro-preview' | 'gemini-3.1-pro-preview-customtools' | 'gemini-3.1-flash-image-preview' | 'gemini-3.1-flash-lite-preview' | 'gemini-3.1-flash-tts-preview' | 'gemini-3.5-flash' | 'gemini-pro-latest' | 'gemini-flash-latest' | 'gemini-flash-lite-latest' | 'deep-research-pro-preview-12-2025' | 'deep-research-max-preview-04-2026' | 'deep-research-preview-04-2026' | 'nano-banana-pro-preview' | 'aqa' | 'gemini-robotics-er-1.5-preview' | 'gemma-3-1b-it' | 'gemma-3-4b-it' | 'gemma-3n-e4b-it' | 'gemma-3n-e2b-it' | 'gemma-3-12b-it' | 'gemma-3-27b-it' | (string & {});
@@ -242,6 +243,35 @@ type SafetyRatingSchema = NonNullable<InferSchema<typeof responseSchema>['candid
 type PromptFeedbackSchema = NonNullable<InferSchema<typeof responseSchema>['promptFeedback']>;
 type UsageMetadataSchema = NonNullable<InferSchema<typeof responseSchema>['usageMetadata']>;
 
+type GoogleSpeechModelId = 'gemini-2.5-flash-preview-tts' | 'gemini-2.5-pro-preview-tts' | 'gemini-3.1-flash-tts-preview' | (string & {});
+
+interface GoogleSpeechModelConfig {
+    provider: string;
+    baseURL: string;
+    headers?: Resolvable<Record<string, string | undefined>>;
+    fetch?: FetchFunction;
+    _internal?: {
+        currentDate?: () => Date;
+    };
+}
+declare class GoogleSpeechModel implements SpeechModelV4 {
+    readonly modelId: GoogleSpeechModelId;
+    private readonly config;
+    readonly specificationVersion = "v4";
+    static [WORKFLOW_SERIALIZE](model: GoogleSpeechModel): {
+        modelId: string;
+        config: _ai_sdk_provider.JSONObject;
+    };
+    static [WORKFLOW_DESERIALIZE](options: {
+        modelId: GoogleSpeechModelId;
+        config: GoogleSpeechModelConfig;
+    }): GoogleSpeechModel;
+    get provider(): string;
+    constructor(modelId: GoogleSpeechModelId, config: GoogleSpeechModelConfig);
+    private getArgs;
+    doGenerate(options: Parameters<SpeechModelV4['doGenerate']>[0]): Promise<Awaited<ReturnType<SpeechModelV4['doGenerate']>>>;
+}
+
 declare const googleTools: {
     /**
      * Creates a Google search tool that gives Google direct access to real-time web content.
@@ -325,4 +355,4 @@ declare const googleTools: {
     }, {}>;
 };
 
-export { GoogleLanguageModel, type GoogleModelId, type GroundingMetadataSchema, type PromptFeedbackSchema, type SafetyRatingSchema, type UrlContextMetadataSchema, type UsageMetadataSchema, getGroundingMetadataSchema, getUrlContextMetadataSchema, googleTools };
+export { GoogleLanguageModel, type GoogleModelId, GoogleSpeechModel, type GroundingMetadataSchema, type PromptFeedbackSchema, type SafetyRatingSchema, type UrlContextMetadataSchema, type UsageMetadataSchema, getGroundingMetadataSchema, getUrlContextMetadataSchema, googleTools };

package/dist/internal/index.js

@@ -2421,123 +2421,396 @@ var chunkSchema = lazySchema3(
   )
 );
 
+// src/google-speech-model.ts
+import {
+  combineHeaders as combineHeaders2,
+  convertBase64ToUint8Array,
+  createJsonResponseHandler as createJsonResponseHandler2,
+  parseProviderOptions as parseProviderOptions2,
+  postJsonToApi as postJsonToApi2,
+  resolve as resolve2,
+  serializeModelOptions as serializeModelOptions2,
+  WORKFLOW_DESERIALIZE as WORKFLOW_DESERIALIZE2,
+  WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE2
+} from "@ai-sdk/provider-utils";
+
+// src/google-speech-api.ts
+import { lazySchema as lazySchema4, zodSchema as zodSchema4 } from "@ai-sdk/provider-utils";
+import { z as z4 } from "zod/v4";
+var googleSpeechResponseSchema = lazySchema4(
+  () => zodSchema4(
+    z4.object({
+      candidates: z4.array(
+        z4.object({
+          content: z4.object({
+            parts: z4.array(
+              z4.object({
+                inlineData: z4.object({
+                  mimeType: z4.string().nullish(),
+                  data: z4.string().nullish()
+                }).nullish()
+              })
+            ).nullish()
+          }).nullish()
+        })
+      ).nullish()
+    })
+  )
+);
+
+// src/google-speech-model-options.ts
+import {
+  lazySchema as lazySchema5,
+  zodSchema as zodSchema5
+} from "@ai-sdk/provider-utils";
+import { z as z5 } from "zod/v4";
+var prebuiltVoiceConfigSchema = z5.object({
+  voiceName: z5.string()
+});
+var voiceConfigSchema = z5.object({
+  prebuiltVoiceConfig: prebuiltVoiceConfigSchema
+});
+var googleSpeechProviderOptionsSchema = lazySchema5(
+  () => zodSchema5(
+    z5.object({
+      /**
+       * Multi-speaker configuration for dialogue audio. When provided, this
+       * overrides the top-level `voice`. The Gemini TTS API supports up to two
+       * speakers; each speaker name must match a name used in the input text.
+       *
+       * https://ai.google.dev/gemini-api/docs/speech-generation#multi-speaker
+       */
+      multiSpeakerVoiceConfig: z5.object({
+        speakerVoiceConfigs: z5.array(
+          z5.object({
+            speaker: z5.string(),
+            voiceConfig: voiceConfigSchema
+          })
+        )
+      }).optional()
+    })
+  )
+);
+
+// src/google-speech-model.ts
+var DEFAULT_VOICE = "Kore";
+var DEFAULT_SAMPLE_RATE = 24e3;
+var GoogleSpeechModel = class _GoogleSpeechModel {
+  constructor(modelId, config) {
+    this.modelId = modelId;
+    this.config = config;
+    this.specificationVersion = "v4";
+  }
+  static [WORKFLOW_SERIALIZE2](model) {
+    return serializeModelOptions2({
+      modelId: model.modelId,
+      config: model.config
+    });
+  }
+  static [WORKFLOW_DESERIALIZE2](options) {
+    return new _GoogleSpeechModel(options.modelId, options.config);
+  }
+  get provider() {
+    return this.config.provider;
+  }
+  async getArgs({
+    text,
+    voice = DEFAULT_VOICE,
+    outputFormat,
+    instructions,
+    speed,
+    language,
+    providerOptions
+  }) {
+    const warnings = [];
+    const providerOptionsNames = this.config.provider.includes("vertex") ? ["googleVertex", "vertex"] : ["google"];
+    let googleOptions;
+    for (const name of providerOptionsNames) {
+      googleOptions = await parseProviderOptions2({
+        provider: name,
+        providerOptions,
+        schema: googleSpeechProviderOptionsSchema
+      });
+      if (googleOptions != null) {
+        break;
+      }
+    }
+    if (googleOptions == null && !providerOptionsNames.includes("google")) {
+      googleOptions = await parseProviderOptions2({
+        provider: "google",
+        providerOptions,
+        schema: googleSpeechProviderOptionsSchema
+      });
+    }
+    const multiSpeakerVoiceConfig = googleOptions == null ? void 0 : googleOptions.multiSpeakerVoiceConfig;
+    const speechConfig = multiSpeakerVoiceConfig ? { multiSpeakerVoiceConfig } : { voiceConfig: { prebuiltVoiceConfig: { voiceName: voice } } };
+    let promptText = text;
+    if (instructions != null) {
+      if (multiSpeakerVoiceConfig) {
+        warnings.push({
+          type: "unsupported",
+          feature: "instructions",
+          details: "Google Gemini TTS ignores `instructions` when `multiSpeakerVoiceConfig` is set, because prepending them would break multi-speaker transcript parsing."
+        });
+      } else {
+        promptText = `${instructions}: ${text}`;
+      }
+    }
+    if (speed != null) {
+      warnings.push({
+        type: "unsupported",
+        feature: "speed",
+        details: "Google Gemini TTS models do not support the `speed` option. It was ignored."
+      });
+    }
+    if (language != null) {
+      warnings.push({
+        type: "unsupported",
+        feature: "language",
+        details: "Google Gemini TTS models do not support the `language` option. Language is detected automatically from the input text."
+      });
+    }
+    let resolvedOutputFormat = "wav";
+    if (outputFormat === "pcm") {
+      resolvedOutputFormat = "pcm";
+    } else if (outputFormat != null && outputFormat !== "wav") {
+      warnings.push({
+        type: "unsupported",
+        feature: "outputFormat",
+        details: `Unsupported output format: ${outputFormat}. Using wav instead.`
+      });
+    }
+    const requestBody = {
+      contents: [{ role: "user", parts: [{ text: promptText }] }],
+      generationConfig: {
+        responseModalities: ["AUDIO"],
+        speechConfig
+      }
+    };
+    return { requestBody, warnings, outputFormat: resolvedOutputFormat };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const { requestBody, warnings, outputFormat } = await this.getArgs(options);
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi2({
+      url: `${this.config.baseURL}/models/${this.modelId}:generateContent`,
+      headers: combineHeaders2(
+        this.config.headers ? await resolve2(this.config.headers) : void 0,
+        options.headers
+      ),
+      body: requestBody,
+      failedResponseHandler: googleFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler2(
+        googleSpeechResponseSchema
+      ),
+      abortSignal: options.abortSignal,
+      fetch: this.config.fetch
+    });
+    let base64Audio;
+    let mimeType;
+    for (const candidate of (_d = response.candidates) != null ? _d : []) {
+      for (const part of (_f = (_e = candidate.content) == null ? void 0 : _e.parts) != null ? _f : []) {
+        if ((_g = part.inlineData) == null ? void 0 : _g.data) {
+          base64Audio = part.inlineData.data;
+          mimeType = (_h = part.inlineData.mimeType) != null ? _h : void 0;
+          break;
+        }
+      }
+      if (base64Audio != null) {
+        break;
+      }
+    }
+    const sampleRate = (_i = parseSampleRate(mimeType)) != null ? _i : DEFAULT_SAMPLE_RATE;
+    const pcm = base64Audio != null ? convertBase64ToUint8Array(base64Audio) : new Uint8Array(0);
+    const audio = outputFormat === "pcm" || pcm.length === 0 ? pcm : addWavHeader(pcm, sampleRate);
+    if (outputFormat === "pcm" && pcm.length > 0) {
+      warnings.push({
+        type: "unsupported",
+        feature: "outputFormat",
+        details: `Returning raw PCM audio (signed 16-bit little-endian, mono, ${sampleRate} Hz). These bytes have no container header and are not directly playable; see providerMetadata.google for the sample rate and mime type.`
+      });
+    }
+    return {
+      audio,
+      warnings,
+      request: {
+        body: JSON.stringify(requestBody)
+      },
+      response: {
+        timestamp: currentDate,
+        modelId: this.modelId,
+        headers: responseHeaders,
+        body: rawResponse
+      },
+      providerMetadata: {
+        google: {
+          sampleRate,
+          mimeType: mimeType != null ? mimeType : null
+        }
+      }
+    };
+  }
+};
+function parseSampleRate(mimeType) {
+  if (mimeType == null) {
+    return void 0;
+  }
+  const match = /rate=(\d+)/.exec(mimeType);
+  return match ? Number.parseInt(match[1], 10) : void 0;
+}
+function addWavHeader(pcm, sampleRate) {
+  const numChannels = 1;
+  const bitsPerSample = 16;
+  const blockAlign = numChannels * bitsPerSample / 8;
+  const byteRate = sampleRate * blockAlign;
+  const dataSize = pcm.length;
+  const buffer = new ArrayBuffer(44 + dataSize);
+  const view = new DataView(buffer);
+  writeAscii(view, 0, "RIFF");
+  view.setUint32(4, 36 + dataSize, true);
+  writeAscii(view, 8, "WAVE");
+  writeAscii(view, 12, "fmt ");
+  view.setUint32(16, 16, true);
+  view.setUint16(20, 1, true);
+  view.setUint16(22, numChannels, true);
+  view.setUint32(24, sampleRate, true);
+  view.setUint32(28, byteRate, true);
+  view.setUint16(32, blockAlign, true);
+  view.setUint16(34, bitsPerSample, true);
+  writeAscii(view, 36, "data");
+  view.setUint32(40, dataSize, true);
+  const out = new Uint8Array(buffer);
+  out.set(pcm, 44);
+  return out;
+}
+function writeAscii(view, offset, text) {
+  for (let i = 0; i < text.length; i++) {
+    view.setUint8(offset + i, text.charCodeAt(i));
+  }
+}
+
 // src/tool/code-execution.ts
 import { createProviderExecutedToolFactory } from "@ai-sdk/provider-utils";
-import { z as z4 } from "zod/v4";
+import { z as z6 } from "zod/v4";
 var codeExecution = createProviderExecutedToolFactory({
   id: "google.code_execution",
-  inputSchema: z4.object({
-    language: z4.string().describe("The programming language of the code."),
-    code: z4.string().describe("The code to be executed.")
+  inputSchema: z6.object({
+    language: z6.string().describe("The programming language of the code."),
+    code: z6.string().describe("The code to be executed.")
   }),
-  outputSchema: z4.object({
-    outcome: z4.string().describe('The outcome of the execution (e.g., "OUTCOME_OK").'),
-    output: z4.string().describe("The output from the code execution.")
+  outputSchema: z6.object({
+    outcome: z6.string().describe('The outcome of the execution (e.g., "OUTCOME_OK").'),
+    output: z6.string().describe("The output from the code execution.")
   })
 });
 
 // src/tool/enterprise-web-search.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory2,
-  lazySchema as lazySchema4,
-  zodSchema as zodSchema4
+  lazySchema as lazySchema6,
+  zodSchema as zodSchema6
 } from "@ai-sdk/provider-utils";
-import { z as z5 } from "zod/v4";
+import { z as z7 } from "zod/v4";
 var enterpriseWebSearch = createProviderExecutedToolFactory2({
   id: "google.enterprise_web_search",
-  inputSchema: lazySchema4(() => zodSchema4(z5.object({}))),
-  outputSchema: lazySchema4(() => zodSchema4(z5.object({})))
+  inputSchema: lazySchema6(() => zodSchema6(z7.object({}))),
+  outputSchema: lazySchema6(() => zodSchema6(z7.object({})))
 });
 
 // src/tool/file-search.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory3,
-  lazySchema as lazySchema5,
-  zodSchema as zodSchema5
+  lazySchema as lazySchema7,
+  zodSchema as zodSchema7
 } from "@ai-sdk/provider-utils";
-import { z as z6 } from "zod/v4";
-var fileSearchArgsBaseSchema = z6.object({
+import { z as z8 } from "zod/v4";
+var fileSearchArgsBaseSchema = z8.object({
   /** The names of the file_search_stores to retrieve from.
    *  Example: `fileSearchStores/my-file-search-store-123`
    */
-  fileSearchStoreNames: z6.array(z6.string()).describe(
+  fileSearchStoreNames: z8.array(z8.string()).describe(
     "The names of the file_search_stores to retrieve from. Example: `fileSearchStores/my-file-search-store-123`"
   ),
   /** The number of file search retrieval chunks to retrieve. */
-  topK: z6.number().int().positive().describe("The number of file search retrieval chunks to retrieve.").optional(),
+  topK: z8.number().int().positive().describe("The number of file search retrieval chunks to retrieve.").optional(),
   /** Metadata filter to apply to the file search retrieval documents.
    *  See https://google.aip.dev/160 for the syntax of the filter expression.
    */
-  metadataFilter: z6.string().describe(
+  metadataFilter: z8.string().describe(
     "Metadata filter to apply to the file search retrieval documents. See https://google.aip.dev/160 for the syntax of the filter expression."
   ).optional()
 }).passthrough();
 var fileSearch = createProviderExecutedToolFactory3({
   id: "google.file_search",
-  inputSchema: lazySchema5(() => zodSchema5(z6.object({}))),
-  outputSchema: lazySchema5(() => zodSchema5(z6.object({})))
+  inputSchema: lazySchema7(() => zodSchema7(z8.object({}))),
+  outputSchema: lazySchema7(() => zodSchema7(z8.object({})))
 });
 
 // src/tool/google-maps.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory4,
-  lazySchema as lazySchema6,
-  zodSchema as zodSchema6
+  lazySchema as lazySchema8,
+  zodSchema as zodSchema8
 } from "@ai-sdk/provider-utils";
-import { z as z7 } from "zod/v4";
+import { z as z9 } from "zod/v4";
 var googleMaps = createProviderExecutedToolFactory4({
   id: "google.google_maps",
-  inputSchema: lazySchema6(() => zodSchema6(z7.object({}))),
-  outputSchema: lazySchema6(() => zodSchema6(z7.object({})))
+  inputSchema: lazySchema8(() => zodSchema8(z9.object({}))),
+  outputSchema: lazySchema8(() => zodSchema8(z9.object({})))
 });
 
 // src/tool/google-search.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory5,
-  lazySchema as lazySchema7,
-  zodSchema as zodSchema7
+  lazySchema as lazySchema9,
+  zodSchema as zodSchema9
 } from "@ai-sdk/provider-utils";
-import { z as z8 } from "zod/v4";
-var googleSearchToolArgsBaseSchema = z8.object({
-  searchTypes: z8.object({
-    webSearch: z8.object({}).optional(),
-    imageSearch: z8.object({}).optional()
+import { z as z10 } from "zod/v4";
+var googleSearchToolArgsBaseSchema = z10.object({
+  searchTypes: z10.object({
+    webSearch: z10.object({}).optional(),
+    imageSearch: z10.object({}).optional()
   }).optional(),
-  timeRangeFilter: z8.object({
-    startTime: z8.string(),
-    endTime: z8.string()
+  timeRangeFilter: z10.object({
+    startTime: z10.string(),
+    endTime: z10.string()
   }).optional()
 }).passthrough();
 var googleSearch = createProviderExecutedToolFactory5({
   id: "google.google_search",
-  inputSchema: lazySchema7(() => zodSchema7(z8.object({}))),
-  outputSchema: lazySchema7(() => zodSchema7(z8.object({})))
+  inputSchema: lazySchema9(() => zodSchema9(z10.object({}))),
+  outputSchema: lazySchema9(() => zodSchema9(z10.object({})))
 });
 
 // src/tool/url-context.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory6,
-  lazySchema as lazySchema8,
-  zodSchema as zodSchema8
+  lazySchema as lazySchema10,
+  zodSchema as zodSchema10
 } from "@ai-sdk/provider-utils";
-import { z as z9 } from "zod/v4";
+import { z as z11 } from "zod/v4";
 var urlContext = createProviderExecutedToolFactory6({
   id: "google.url_context",
-  inputSchema: lazySchema8(() => zodSchema8(z9.object({}))),
-  outputSchema: lazySchema8(() => zodSchema8(z9.object({})))
+  inputSchema: lazySchema10(() => zodSchema10(z11.object({}))),
+  outputSchema: lazySchema10(() => zodSchema10(z11.object({})))
 });
 
 // src/tool/vertex-rag-store.ts
 import {
   createProviderExecutedToolFactory as createProviderExecutedToolFactory7,
-  lazySchema as lazySchema9,
-  zodSchema as zodSchema9
+  lazySchema as lazySchema11,
+  zodSchema as zodSchema11
 } from "@ai-sdk/provider-utils";
-import { z as z10 } from "zod/v4";
+import { z as z12 } from "zod/v4";
 var vertexRagStore = createProviderExecutedToolFactory7({
   id: "google.vertex_rag_store",
-  inputSchema: lazySchema9(() => zodSchema9(z10.object({}))),
-  outputSchema: lazySchema9(() => zodSchema9(z10.object({})))
+  inputSchema: lazySchema11(() => zodSchema11(z12.object({}))),
+  outputSchema: lazySchema11(() => zodSchema11(z12.object({})))
 });
 
 // src/google-tools.ts
@@ -2601,6 +2874,7 @@ var googleTools = {
 };
 export {
   GoogleLanguageModel,
+  GoogleSpeechModel,
   getGroundingMetadataSchema,
   getUrlContextMetadataSchema,
   googleTools