@ai-sdk/google 4.0.0-canary.75 → 4.0.0-canary.77

package/dist/index.js

@@ -7,7 +7,7 @@ import {
 } from "@ai-sdk/provider-utils";
 
 // src/version.ts
-var VERSION = true ? "4.0.0-canary.75" : "0.0.0-test";
+var VERSION = true ? "4.0.0-canary.77" : "0.0.0-test";
 
 // src/google-embedding-model.ts
 import {
@@ -3573,12 +3573,11 @@ var googleOperationSchema = z17.object({
   }).nullish()
 });
 
-// src/interactions/google-interactions-language-model.ts
+// src/google-speech-model.ts
 import {
-  combineHeaders as combineHeaders7,
-  createEventSourceResponseHandler as createEventSourceResponseHandler3,
-  createJsonResponseHandler as createJsonResponseHandler7,
-  generateId as defaultGenerateId2,
+  combineHeaders as combineHeaders6,
+  convertBase64ToUint8Array,
+  createJsonResponseHandler as createJsonResponseHandler6,
   parseProviderOptions as parseProviderOptions6,
   postJsonToApi as postJsonToApi5,
   resolve as resolve5,
@@ -3587,6 +3586,266 @@ import {
   WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE4
 } from "@ai-sdk/provider-utils";
 
+// src/google-speech-api.ts
+import { lazySchema as lazySchema16, zodSchema as zodSchema16 } from "@ai-sdk/provider-utils";
+import { z as z18 } from "zod/v4";
+var googleSpeechResponseSchema = lazySchema16(
+  () => zodSchema16(
+    z18.object({
+      candidates: z18.array(
+        z18.object({
+          content: z18.object({
+            parts: z18.array(
+              z18.object({
+                inlineData: z18.object({
+                  mimeType: z18.string().nullish(),
+                  data: z18.string().nullish()
+                }).nullish()
+              })
+            ).nullish()
+          }).nullish()
+        })
+      ).nullish()
+    })
+  )
+);
+
+// src/google-speech-model-options.ts
+import {
+  lazySchema as lazySchema17,
+  zodSchema as zodSchema17
+} from "@ai-sdk/provider-utils";
+import { z as z19 } from "zod/v4";
+var prebuiltVoiceConfigSchema = z19.object({
+  voiceName: z19.string()
+});
+var voiceConfigSchema = z19.object({
+  prebuiltVoiceConfig: prebuiltVoiceConfigSchema
+});
+var googleSpeechProviderOptionsSchema = lazySchema17(
+  () => zodSchema17(
+    z19.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: z19.object({
+        speakerVoiceConfigs: z19.array(
+          z19.object({
+            speaker: z19.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_SERIALIZE4](model) {
+    return serializeModelOptions4({
+      modelId: model.modelId,
+      config: model.config
+    });
+  }
+  static [WORKFLOW_DESERIALIZE4](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 googleOptions = await parseProviderOptions6({
+      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: ${outputFormat}. Using wav instead.`
+      });
+    }
+    const requestBody = {
+      contents: [{ parts: [{ text: promptText }] }],
+      generationConfig: {
+        responseModalities: ["AUDIO"],
+        speechConfig
+      }
+    };
+    return { requestBody, warnings, outputFormat: resolvedOutputFormat };
+  }
+  async doGenerate(options) {
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i;
+    const currentDate = (_c = (_b = (_a = this.config._internal) == null ? void 0 : _a.currentDate) == null ? void 0 : _b.call(_a)) != null ? _c : /* @__PURE__ */ new Date();
+    const { requestBody, warnings, outputFormat } = await this.getArgs(options);
+    const {
+      value: response,
+      responseHeaders,
+      rawValue: rawResponse
+    } = await postJsonToApi5({
+      url: `${this.config.baseURL}/models/${this.modelId}:generateContent`,
+      headers: combineHeaders6(
+        this.config.headers ? await resolve5(this.config.headers) : void 0,
+        options.headers
+      ),
+      body: requestBody,
+      failedResponseHandler: googleFailedResponseHandler,
+      successfulResponseHandler: createJsonResponseHandler6(
+        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/interactions/google-interactions-language-model.ts
+import {
+  combineHeaders as combineHeaders8,
+  createEventSourceResponseHandler as createEventSourceResponseHandler3,
+  createJsonResponseHandler as createJsonResponseHandler8,
+  generateId as defaultGenerateId2,
+  parseProviderOptions as parseProviderOptions7,
+  postJsonToApi as postJsonToApi6,
+  resolve as resolve6,
+  serializeModelOptions as serializeModelOptions5,
+  WORKFLOW_DESERIALIZE as WORKFLOW_DESERIALIZE5,
+  WORKFLOW_SERIALIZE as WORKFLOW_SERIALIZE5
+} from "@ai-sdk/provider-utils";
+
 // src/interactions/convert-google-interactions-usage.ts
 function convertGoogleInteractionsUsage(usage) {
   var _a, _b, _c, _d, _e, _f, _g, _h;
@@ -4686,33 +4945,33 @@ ${block.text}`
 
 // src/interactions/google-interactions-api.ts
 import {
-  lazySchema as lazySchema16,
-  zodSchema as zodSchema16
+  lazySchema as lazySchema18,
+  zodSchema as zodSchema18
 } from "@ai-sdk/provider-utils";
-import { z as z18 } from "zod/v4";
-var tokenByModalitySchema = () => z18.object({
-  modality: z18.string().nullish(),
-  tokens: z18.number().nullish()
+import { z as z20 } from "zod/v4";
+var tokenByModalitySchema = () => z20.object({
+  modality: z20.string().nullish(),
+  tokens: z20.number().nullish()
 }).loose();
-var usageSchema2 = () => z18.object({
-  total_input_tokens: z18.number().nullish(),
-  total_output_tokens: z18.number().nullish(),
-  total_thought_tokens: z18.number().nullish(),
-  total_cached_tokens: z18.number().nullish(),
-  total_tool_use_tokens: z18.number().nullish(),
-  total_tokens: z18.number().nullish(),
-  input_tokens_by_modality: z18.array(tokenByModalitySchema()).nullish(),
-  output_tokens_by_modality: z18.array(tokenByModalitySchema()).nullish(),
-  cached_tokens_by_modality: z18.array(tokenByModalitySchema()).nullish(),
-  tool_use_tokens_by_modality: z18.array(tokenByModalitySchema()).nullish(),
-  grounding_tool_count: z18.array(
-    z18.object({
-      type: z18.string().nullish(),
-      count: z18.number().nullish()
+var usageSchema2 = () => z20.object({
+  total_input_tokens: z20.number().nullish(),
+  total_output_tokens: z20.number().nullish(),
+  total_thought_tokens: z20.number().nullish(),
+  total_cached_tokens: z20.number().nullish(),
+  total_tool_use_tokens: z20.number().nullish(),
+  total_tokens: z20.number().nullish(),
+  input_tokens_by_modality: z20.array(tokenByModalitySchema()).nullish(),
+  output_tokens_by_modality: z20.array(tokenByModalitySchema()).nullish(),
+  cached_tokens_by_modality: z20.array(tokenByModalitySchema()).nullish(),
+  tool_use_tokens_by_modality: z20.array(tokenByModalitySchema()).nullish(),
+  grounding_tool_count: z20.array(
+    z20.object({
+      type: z20.string().nullish(),
+      count: z20.number().nullish()
     }).loose()
   ).nullish()
 }).loose();
-var interactionStatusSchema = () => z18.enum([
+var interactionStatusSchema = () => z20.enum([
   "in_progress",
   "requires_action",
   "completed",
@@ -4721,62 +4980,62 @@ var interactionStatusSchema = () => z18.enum([
   "incomplete"
 ]);
 var annotationSchema = () => {
-  const urlCitation = z18.object({
-    type: z18.literal("url_citation"),
-    url: z18.string().nullish(),
-    title: z18.string().nullish(),
-    start_index: z18.number().nullish(),
-    end_index: z18.number().nullish()
+  const urlCitation = z20.object({
+    type: z20.literal("url_citation"),
+    url: z20.string().nullish(),
+    title: z20.string().nullish(),
+    start_index: z20.number().nullish(),
+    end_index: z20.number().nullish()
   }).loose();
-  const fileCitation = z18.object({
-    type: z18.literal("file_citation"),
-    file_name: z18.string().nullish(),
-    document_uri: z18.string().nullish(),
-    url: z18.string().nullish(),
-    page_number: z18.number().nullish(),
-    media_id: z18.string().nullish(),
-    start_index: z18.number().nullish(),
-    end_index: z18.number().nullish(),
-    custom_metadata: z18.record(z18.string(), z18.unknown()).nullish()
+  const fileCitation = z20.object({
+    type: z20.literal("file_citation"),
+    file_name: z20.string().nullish(),
+    document_uri: z20.string().nullish(),
+    url: z20.string().nullish(),
+    page_number: z20.number().nullish(),
+    media_id: z20.string().nullish(),
+    start_index: z20.number().nullish(),
+    end_index: z20.number().nullish(),
+    custom_metadata: z20.record(z20.string(), z20.unknown()).nullish()
   }).loose();
-  const placeCitation = z18.object({
-    type: z18.literal("place_citation"),
-    name: z18.string().nullish(),
-    url: z18.string().nullish(),
-    place_id: z18.string().nullish(),
-    start_index: z18.number().nullish(),
-    end_index: z18.number().nullish()
+  const placeCitation = z20.object({
+    type: z20.literal("place_citation"),
+    name: z20.string().nullish(),
+    url: z20.string().nullish(),
+    place_id: z20.string().nullish(),
+    start_index: z20.number().nullish(),
+    end_index: z20.number().nullish()
   }).loose();
-  return z18.union([
+  return z20.union([
     urlCitation,
     fileCitation,
     placeCitation,
-    z18.object({ type: z18.string() }).loose()
+    z20.object({ type: z20.string() }).loose()
   ]);
 };
-var thoughtSummaryItemSchema = () => z18.object({
-  type: z18.string(),
-  text: z18.string().nullish(),
-  data: z18.string().nullish(),
-  mime_type: z18.string().nullish()
+var thoughtSummaryItemSchema = () => z20.object({
+  type: z20.string(),
+  text: z20.string().nullish(),
+  data: z20.string().nullish(),
+  mime_type: z20.string().nullish()
 }).loose();
 var contentBlockSchema = () => {
-  const textContent = z18.object({
-    type: z18.literal("text"),
-    text: z18.string(),
-    annotations: z18.array(annotationSchema()).nullish()
+  const textContent = z20.object({
+    type: z20.literal("text"),
+    text: z20.string(),
+    annotations: z20.array(annotationSchema()).nullish()
   }).loose();
-  const imageContent = z18.object({
-    type: z18.literal("image"),
-    data: z18.string().nullish(),
-    mime_type: z18.string().nullish(),
-    resolution: z18.enum(["low", "medium", "high", "ultra_high"]).nullish(),
-    uri: z18.string().nullish()
+  const imageContent = z20.object({
+    type: z20.literal("image"),
+    data: z20.string().nullish(),
+    mime_type: z20.string().nullish(),
+    resolution: z20.enum(["low", "medium", "high", "ultra_high"]).nullish(),
+    uri: z20.string().nullish()
   }).loose();
-  return z18.union([
+  return z20.union([
     textContent,
     imageContent,
-    z18.object({ type: z18.string() }).loose()
+    z20.object({ type: z20.string() }).loose()
   ]);
 };
 var BUILTIN_TOOL_CALL_STEP_TYPES = [
@@ -4796,152 +5055,152 @@ var BUILTIN_TOOL_RESULT_STEP_TYPES = [
   "mcp_server_tool_result"
 ];
 var stepSchema = () => {
-  const userInputStep = z18.object({
-    type: z18.literal("user_input"),
-    content: z18.array(contentBlockSchema()).nullish()
+  const userInputStep = z20.object({
+    type: z20.literal("user_input"),
+    content: z20.array(contentBlockSchema()).nullish()
   }).loose();
-  const modelOutputStep = z18.object({
-    type: z18.literal("model_output"),
-    content: z18.array(contentBlockSchema()).nullish()
+  const modelOutputStep = z20.object({
+    type: z20.literal("model_output"),
+    content: z20.array(contentBlockSchema()).nullish()
   }).loose();
-  const functionCallStep = z18.object({
-    type: z18.literal("function_call"),
-    id: z18.string(),
-    name: z18.string(),
-    arguments: z18.record(z18.string(), z18.unknown()).nullish(),
-    signature: z18.string().nullish()
+  const functionCallStep = z20.object({
+    type: z20.literal("function_call"),
+    id: z20.string(),
+    name: z20.string(),
+    arguments: z20.record(z20.string(), z20.unknown()).nullish(),
+    signature: z20.string().nullish()
   }).loose();
-  const thoughtStep = z18.object({
-    type: z18.literal("thought"),
-    signature: z18.string().nullish(),
-    summary: z18.array(thoughtSummaryItemSchema()).nullish()
+  const thoughtStep = z20.object({
+    type: z20.literal("thought"),
+    signature: z20.string().nullish(),
+    summary: z20.array(thoughtSummaryItemSchema()).nullish()
   }).loose();
-  const builtinToolCallStep = z18.object({
-    type: z18.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
-    id: z18.string(),
-    arguments: z18.record(z18.string(), z18.unknown()).nullish(),
-    name: z18.string().nullish(),
-    server_name: z18.string().nullish(),
-    search_type: z18.string().nullish(),
-    signature: z18.string().nullish()
+  const builtinToolCallStep = z20.object({
+    type: z20.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
+    id: z20.string(),
+    arguments: z20.record(z20.string(), z20.unknown()).nullish(),
+    name: z20.string().nullish(),
+    server_name: z20.string().nullish(),
+    search_type: z20.string().nullish(),
+    signature: z20.string().nullish()
   }).loose();
-  const builtinToolResultStep = z18.object({
-    type: z18.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
-    call_id: z18.string(),
-    result: z18.unknown().nullish(),
-    is_error: z18.boolean().nullish(),
-    name: z18.string().nullish(),
-    server_name: z18.string().nullish(),
-    signature: z18.string().nullish()
+  const builtinToolResultStep = z20.object({
+    type: z20.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
+    call_id: z20.string(),
+    result: z20.unknown().nullish(),
+    is_error: z20.boolean().nullish(),
+    name: z20.string().nullish(),
+    server_name: z20.string().nullish(),
+    signature: z20.string().nullish()
   }).loose();
-  return z18.union([
+  return z20.union([
     userInputStep,
     modelOutputStep,
     functionCallStep,
     thoughtStep,
     builtinToolCallStep,
     builtinToolResultStep,
-    z18.object({ type: z18.string() }).loose()
+    z20.object({ type: z20.string() }).loose()
   ]);
 };
-var googleInteractionsResponseSchema = lazySchema16(
-  () => zodSchema16(
-    z18.object({
+var googleInteractionsResponseSchema = lazySchema18(
+  () => zodSchema18(
+    z20.object({
       /*
        * `id` is omitted from the response body when `store: false` (fully
        * stateless mode) — there is no server-side interaction record for the
        * client to reference. `nullish` lets the schema accept that shape.
        */
-      id: z18.string().nullish(),
-      created: z18.string().nullish(),
-      updated: z18.string().nullish(),
+      id: z20.string().nullish(),
+      created: z20.string().nullish(),
+      updated: z20.string().nullish(),
       status: interactionStatusSchema(),
-      model: z18.string().nullish(),
-      agent: z18.string().nullish(),
-      steps: z18.array(stepSchema()).nullish(),
+      model: z20.string().nullish(),
+      agent: z20.string().nullish(),
+      steps: z20.array(stepSchema()).nullish(),
       usage: usageSchema2().nullish(),
-      service_tier: z18.string().nullish(),
-      previous_interaction_id: z18.string().nullish(),
-      response_modalities: z18.array(z18.string()).nullish()
+      service_tier: z20.string().nullish(),
+      previous_interaction_id: z20.string().nullish(),
+      response_modalities: z20.array(z20.string()).nullish()
     }).loose()
   )
 );
-var googleInteractionsEventSchema = lazySchema16(
-  () => zodSchema16(
+var googleInteractionsEventSchema = lazySchema18(
+  () => zodSchema18(
     (() => {
       const status = interactionStatusSchema();
       const annotation = annotationSchema();
       const thoughtSummaryItem = thoughtSummaryItemSchema();
-      const interactionCreatedEvent = z18.object({
-        event_type: z18.literal("interaction.created"),
-        event_id: z18.string().nullish(),
-        interaction: z18.object({
+      const interactionCreatedEvent = z20.object({
+        event_type: z20.literal("interaction.created"),
+        event_id: z20.string().nullish(),
+        interaction: z20.object({
           /*
            * `id` is omitted when `store: false` (fully stateless mode);
            * see the matching note on `googleInteractionsResponseSchema.id`.
            */
-          id: z18.string().nullish(),
-          created: z18.string().nullish(),
-          model: z18.string().nullish(),
-          agent: z18.string().nullish(),
+          id: z20.string().nullish(),
+          created: z20.string().nullish(),
+          model: z20.string().nullish(),
+          agent: z20.string().nullish(),
           status: status.nullish()
         }).loose()
       }).loose();
-      const stepStartEvent = z18.object({
-        event_type: z18.literal("step.start"),
-        event_id: z18.string().nullish(),
-        index: z18.number(),
+      const stepStartEvent = z20.object({
+        event_type: z20.literal("step.start"),
+        event_id: z20.string().nullish(),
+        index: z20.number(),
         step: stepSchema()
       }).loose();
-      const stepDeltaText = z18.object({
-        type: z18.literal("text"),
-        text: z18.string()
+      const stepDeltaText = z20.object({
+        type: z20.literal("text"),
+        text: z20.string()
       }).loose();
-      const stepDeltaThoughtSummary = z18.object({
-        type: z18.literal("thought_summary"),
+      const stepDeltaThoughtSummary = z20.object({
+        type: z20.literal("thought_summary"),
         content: thoughtSummaryItem.nullish()
       }).loose();
-      const stepDeltaThoughtSignature = z18.object({
-        type: z18.literal("thought_signature"),
-        signature: z18.string().nullish()
+      const stepDeltaThoughtSignature = z20.object({
+        type: z20.literal("thought_signature"),
+        signature: z20.string().nullish()
       }).loose();
-      const stepDeltaArgumentsDelta = z18.object({
-        type: z18.literal("arguments_delta"),
-        arguments: z18.string().nullish(),
-        id: z18.string().nullish(),
-        signature: z18.string().nullish()
+      const stepDeltaArgumentsDelta = z20.object({
+        type: z20.literal("arguments_delta"),
+        arguments: z20.string().nullish(),
+        id: z20.string().nullish(),
+        signature: z20.string().nullish()
       }).loose();
-      const stepDeltaTextAnnotation = z18.object({
-        type: z18.enum(["text_annotation_delta", "text_annotation"]),
-        annotations: z18.array(annotation).nullish()
+      const stepDeltaTextAnnotation = z20.object({
+        type: z20.enum(["text_annotation_delta", "text_annotation"]),
+        annotations: z20.array(annotation).nullish()
       }).loose();
-      const stepDeltaImage = z18.object({
-        type: z18.literal("image"),
-        data: z18.string().nullish(),
-        mime_type: z18.string().nullish(),
-        resolution: z18.enum(["low", "medium", "high", "ultra_high"]).nullish(),
-        uri: z18.string().nullish()
+      const stepDeltaImage = z20.object({
+        type: z20.literal("image"),
+        data: z20.string().nullish(),
+        mime_type: z20.string().nullish(),
+        resolution: z20.enum(["low", "medium", "high", "ultra_high"]).nullish(),
+        uri: z20.string().nullish()
       }).loose();
-      const stepDeltaBuiltinToolCall = z18.object({
-        type: z18.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
-        id: z18.string().nullish(),
-        arguments: z18.record(z18.string(), z18.unknown()).nullish(),
-        name: z18.string().nullish(),
-        server_name: z18.string().nullish(),
-        search_type: z18.string().nullish(),
-        signature: z18.string().nullish()
+      const stepDeltaBuiltinToolCall = z20.object({
+        type: z20.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
+        id: z20.string().nullish(),
+        arguments: z20.record(z20.string(), z20.unknown()).nullish(),
+        name: z20.string().nullish(),
+        server_name: z20.string().nullish(),
+        search_type: z20.string().nullish(),
+        signature: z20.string().nullish()
       }).loose();
-      const stepDeltaBuiltinToolResult = z18.object({
-        type: z18.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
-        call_id: z18.string().nullish(),
-        result: z18.unknown().nullish(),
-        is_error: z18.boolean().nullish(),
-        name: z18.string().nullish(),
-        server_name: z18.string().nullish(),
-        signature: z18.string().nullish()
+      const stepDeltaBuiltinToolResult = z20.object({
+        type: z20.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
+        call_id: z20.string().nullish(),
+        result: z20.unknown().nullish(),
+        is_error: z20.boolean().nullish(),
+        name: z20.string().nullish(),
+        server_name: z20.string().nullish(),
+        signature: z20.string().nullish()
       }).loose();
-      const stepDeltaUnknown = z18.object({ type: z18.string() }).loose();
-      const stepDeltaUnion = z18.union([
+      const stepDeltaUnknown = z20.object({ type: z20.string() }).loose();
+      const stepDeltaUnion = z20.union([
         stepDeltaText,
         stepDeltaImage,
         stepDeltaThoughtSummary,
@@ -4952,55 +5211,55 @@ var googleInteractionsEventSchema = lazySchema16(
         stepDeltaBuiltinToolResult,
         stepDeltaUnknown
       ]);
-      const stepDeltaEvent = z18.object({
-        event_type: z18.literal("step.delta"),
-        event_id: z18.string().nullish(),
-        index: z18.number(),
+      const stepDeltaEvent = z20.object({
+        event_type: z20.literal("step.delta"),
+        event_id: z20.string().nullish(),
+        index: z20.number(),
         delta: stepDeltaUnion
       }).loose();
-      const stepStopEvent = z18.object({
-        event_type: z18.literal("step.stop"),
-        event_id: z18.string().nullish(),
-        index: z18.number()
+      const stepStopEvent = z20.object({
+        event_type: z20.literal("step.stop"),
+        event_id: z20.string().nullish(),
+        index: z20.number()
       }).loose();
-      const interactionStatusUpdateEvent = z18.object({
-        event_type: z18.literal("interaction.status_update"),
-        event_id: z18.string().nullish(),
-        interaction_id: z18.string().nullish(),
+      const interactionStatusUpdateEvent = z20.object({
+        event_type: z20.literal("interaction.status_update"),
+        event_id: z20.string().nullish(),
+        interaction_id: z20.string().nullish(),
         status: status.nullish()
       }).loose();
-      const interactionInProgressEvent = z18.object({
-        event_type: z18.literal("interaction.in_progress"),
-        event_id: z18.string().nullish(),
-        interaction_id: z18.string().nullish(),
+      const interactionInProgressEvent = z20.object({
+        event_type: z20.literal("interaction.in_progress"),
+        event_id: z20.string().nullish(),
+        interaction_id: z20.string().nullish(),
         status: status.nullish()
       }).loose();
-      const interactionRequiresActionEvent = z18.object({
-        event_type: z18.literal("interaction.requires_action"),
-        event_id: z18.string().nullish(),
-        interaction_id: z18.string().nullish(),
+      const interactionRequiresActionEvent = z20.object({
+        event_type: z20.literal("interaction.requires_action"),
+        event_id: z20.string().nullish(),
+        interaction_id: z20.string().nullish(),
         status: status.nullish()
       }).loose();
-      const interactionCompletedEvent = z18.object({
-        event_type: z18.literal("interaction.completed"),
-        event_id: z18.string().nullish(),
-        interaction: z18.object({
-          id: z18.string().nullish(),
+      const interactionCompletedEvent = z20.object({
+        event_type: z20.literal("interaction.completed"),
+        event_id: z20.string().nullish(),
+        interaction: z20.object({
+          id: z20.string().nullish(),
           status: status.nullish(),
           usage: usageSchema2().nullish(),
-          service_tier: z18.string().nullish()
+          service_tier: z20.string().nullish()
         }).loose()
       }).loose();
-      const errorEvent = z18.object({
-        event_type: z18.literal("error"),
-        event_id: z18.string().nullish(),
-        error: z18.object({
-          code: z18.string().nullish(),
-          message: z18.string().nullish()
+      const errorEvent = z20.object({
+        event_type: z20.literal("error"),
+        event_id: z20.string().nullish(),
+        error: z20.object({
+          code: z20.string().nullish(),
+          message: z20.string().nullish()
         }).loose().nullish()
       }).loose();
-      const unknownEvent = z18.object({ event_type: z18.string() }).loose();
-      return z18.union([
+      const unknownEvent = z20.object({ event_type: z20.string() }).loose();
+      return z20.union([
         interactionCreatedEvent,
         stepStartEvent,
         stepDeltaEvent,
@@ -5018,29 +5277,29 @@ var googleInteractionsEventSchema = lazySchema16(
 
 // src/interactions/google-interactions-language-model-options.ts
 import {
-  lazySchema as lazySchema17,
-  zodSchema as zodSchema17
+  lazySchema as lazySchema19,
+  zodSchema as zodSchema19
 } from "@ai-sdk/provider-utils";
-import { z as z19 } from "zod/v4";
-var googleInteractionsLanguageModelOptions = lazySchema17(
-  () => zodSchema17(
-    z19.object({
-      previousInteractionId: z19.string().nullish(),
-      store: z19.boolean().nullish(),
-      agent: z19.string().nullish(),
-      agentConfig: z19.union([
-        z19.object({
-          type: z19.literal("dynamic")
+import { z as z21 } from "zod/v4";
+var googleInteractionsLanguageModelOptions = lazySchema19(
+  () => zodSchema19(
+    z21.object({
+      previousInteractionId: z21.string().nullish(),
+      store: z21.boolean().nullish(),
+      agent: z21.string().nullish(),
+      agentConfig: z21.union([
+        z21.object({
+          type: z21.literal("dynamic")
         }).loose(),
-        z19.object({
-          type: z19.literal("deep-research"),
-          thinkingSummaries: z19.enum(["auto", "none"]).nullish(),
-          visualization: z19.enum(["off", "auto"]).nullish(),
-          collaborativePlanning: z19.boolean().nullish()
+        z21.object({
+          type: z21.literal("deep-research"),
+          thinkingSummaries: z21.enum(["auto", "none"]).nullish(),
+          visualization: z21.enum(["off", "auto"]).nullish(),
+          collaborativePlanning: z21.boolean().nullish()
         })
       ]).nullish(),
-      thinkingLevel: z19.enum(["minimal", "low", "medium", "high"]).nullish(),
-      thinkingSummaries: z19.enum(["auto", "none"]).nullish(),
+      thinkingLevel: z21.enum(["minimal", "low", "medium", "high"]).nullish(),
+      thinkingSummaries: z21.enum(["auto", "none"]).nullish(),
       /**
        * Output-format entries that map directly to the API's `response_format`
        * array. Use this to request image, audio, or non-JSON text outputs
@@ -5050,17 +5309,17 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
        * type: 'json', schema }` still drives JSON-mode and adds a matching
        * text entry automatically; entries listed here are appended.
        */
-      responseFormat: z19.array(
-        z19.union([
-          z19.object({
-            type: z19.literal("text"),
-            mimeType: z19.string().nullish(),
-            schema: z19.unknown().nullish()
+      responseFormat: z21.array(
+        z21.union([
+          z21.object({
+            type: z21.literal("text"),
+            mimeType: z21.string().nullish(),
+            schema: z21.unknown().nullish()
           }).loose(),
-          z19.object({
-            type: z19.literal("image"),
-            mimeType: z19.string().nullish(),
-            aspectRatio: z19.enum([
+          z21.object({
+            type: z21.literal("image"),
+            mimeType: z21.string().nullish(),
+            aspectRatio: z21.enum([
               "1:1",
               "2:3",
               "3:2",
@@ -5076,11 +5335,11 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
               "1:4",
               "4:1"
             ]).nullish(),
-            imageSize: z19.enum(["1K", "2K", "4K", "512"]).nullish()
+            imageSize: z21.enum(["1K", "2K", "4K", "512"]).nullish()
           }).loose(),
-          z19.object({
-            type: z19.literal("audio"),
-            mimeType: z19.string().nullish()
+          z21.object({
+            type: z21.literal("audio"),
+            mimeType: z21.string().nullish()
           }).loose()
         ])
       ).nullish(),
@@ -5090,8 +5349,8 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
        * translates it into a matching `response_format` image entry and
        * emits a warning when set.
        */
-      imageConfig: z19.object({
-        aspectRatio: z19.enum([
+      imageConfig: z21.object({
+        aspectRatio: z21.enum([
           "1:1",
           "2:3",
           "3:2",
@@ -5107,35 +5366,35 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
           "1:4",
           "4:1"
         ]).nullish(),
-        imageSize: z19.enum(["1K", "2K", "4K", "512"]).nullish()
+        imageSize: z21.enum(["1K", "2K", "4K", "512"]).nullish()
       }).nullish(),
-      mediaResolution: z19.enum(["low", "medium", "high", "ultra_high"]).nullish(),
-      responseModalities: z19.array(z19.enum(["text", "image", "audio", "video", "document"])).nullish(),
-      serviceTier: z19.enum(["flex", "standard", "priority"]).nullish(),
+      mediaResolution: z21.enum(["low", "medium", "high", "ultra_high"]).nullish(),
+      responseModalities: z21.array(z21.enum(["text", "image", "audio", "video", "document"])).nullish(),
+      serviceTier: z21.enum(["flex", "standard", "priority"]).nullish(),
       /**
        * Alternative to AI SDK `system` message. If both are set, the AI SDK
        * `system` message wins and a warning is emitted.
        */
-      systemInstruction: z19.string().nullish(),
+      systemInstruction: z21.string().nullish(),
       /**
        * Per-block signature for round-tripping `thought.signature` and
        * `function_call.signature` blocks. Set by the SDK on output reasoning /
        * tool-call parts; passed back unchanged on input parts so the API
        * accepts the prior turn.
        */
-      signature: z19.string().nullish(),
+      signature: z21.string().nullish(),
       /**
        * Set by the SDK on output assistant messages. The converter uses it to
        * decide which messages to drop when compacting under
        * `previousInteractionId`.
        */
-      interactionId: z19.string().nullish(),
+      interactionId: z21.string().nullish(),
       /**
        * Maximum time, in milliseconds, to poll a background interaction (agent
        * call) before giving up. Defaults to 30 minutes. Long-running agents
        * such as deep research can take tens of minutes — increase if needed.
        */
-      pollingTimeoutMs: z19.number().int().positive().nullish(),
+      pollingTimeoutMs: z21.number().int().positive().nullish(),
       /**
        * Run the interaction in the background. Required for agents whose
        * server-side workflow cannot complete within a single request/response.
@@ -5144,7 +5403,7 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
        * reject `true`; see the agent's documentation for which mode it
        * requires.
        */
-      background: z19.boolean().nullish(),
+      background: z21.boolean().nullish(),
       /**
        * Environment configuration for the agent sandbox. Only applies to agent
        * calls (`google.interactions({ agent })`); ignored on model-id calls.
@@ -5154,36 +5413,36 @@ var googleInteractionsLanguageModelOptions = lazySchema17(
        *   - object: provision a fresh sandbox and optionally preload `sources`
        *     and/or constrain outbound traffic via `network`.
        */
-      environment: z19.union([
-        z19.string(),
-        z19.object({
-          type: z19.literal("remote"),
-          sources: z19.array(
-            z19.union([
-              z19.object({
-                type: z19.literal("gcs"),
-                source: z19.string(),
-                target: z19.string().nullish()
+      environment: z21.union([
+        z21.string(),
+        z21.object({
+          type: z21.literal("remote"),
+          sources: z21.array(
+            z21.union([
+              z21.object({
+                type: z21.literal("gcs"),
+                source: z21.string(),
+                target: z21.string().nullish()
               }),
-              z19.object({
-                type: z19.literal("repository"),
-                source: z19.string(),
-                target: z19.string().nullish()
+              z21.object({
+                type: z21.literal("repository"),
+                source: z21.string(),
+                target: z21.string().nullish()
               }),
-              z19.object({
-                type: z19.literal("inline"),
-                content: z19.string(),
-                target: z19.string()
+              z21.object({
+                type: z21.literal("inline"),
+                content: z21.string(),
+                target: z21.string()
               })
             ])
           ).nullish(),
-          network: z19.union([
-            z19.literal("disabled"),
-            z19.object({
-              allowlist: z19.array(
-                z19.object({
-                  domain: z19.string(),
-                  transform: z19.array(z19.record(z19.string(), z19.string())).nullish()
+          network: z21.union([
+            z21.literal("disabled"),
+            z21.object({
+              allowlist: z21.array(
+                z21.object({
+                  domain: z21.string(),
+                  transform: z21.array(z21.record(z21.string(), z21.string())).nullish()
                 })
               )
             })
@@ -5356,7 +5615,7 @@ function parseGoogleInteractionsOutputs({
 
 // src/interactions/poll-google-interactions.ts
 import {
-  createJsonResponseHandler as createJsonResponseHandler6,
+  createJsonResponseHandler as createJsonResponseHandler7,
   delay as delay3,
   getFromApi as getFromApi3,
   isAbortError
@@ -5364,7 +5623,7 @@ import {
 
 // src/interactions/cancel-google-interaction.ts
 import {
-  combineHeaders as combineHeaders6,
+  combineHeaders as combineHeaders7,
   getRuntimeEnvironmentUserAgent,
   withUserAgentSuffix
 } from "@ai-sdk/provider-utils";
@@ -5383,7 +5642,7 @@ async function cancelGoogleInteraction({
     const response = await fetch(url, {
       method: "POST",
       headers: withUserAgentSuffix(
-        combineHeaders6({ "Content-Type": "application/json" }, headers),
+        combineHeaders7({ "Content-Type": "application/json" }, headers),
         getRuntimeEnvironmentUserAgent()
       ),
       body: "{}"
@@ -5443,7 +5702,7 @@ async function pollGoogleInteractionUntilTerminal({
         url,
         headers,
         failedResponseHandler: googleFailedResponseHandler,
-        successfulResponseHandler: createJsonResponseHandler6(
+        successfulResponseHandler: createJsonResponseHandler7(
           googleInteractionsResponseSchema
         ),
         abortSignal,
@@ -5935,16 +6194,16 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
     }
     this.config = config;
   }
-  static [WORKFLOW_SERIALIZE4](model) {
+  static [WORKFLOW_SERIALIZE5](model) {
     return {
-      ...serializeModelOptions4({
+      ...serializeModelOptions5({
         modelId: model.modelId,
         config: model.config
       }),
       agent: model.agent
     };
   }
-  static [WORKFLOW_DESERIALIZE4](options) {
+  static [WORKFLOW_DESERIALIZE5](options) {
     return new _GoogleInteractionsLanguageModel(
       options.agent != null ? { agent: options.agent } : options.modelId,
       options.config
@@ -5971,7 +6230,7 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
   async getArgs(options) {
     var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t, _u, _v, _w, _x, _y, _z;
     const warnings = [];
-    const googleOptions = await parseProviderOptions6({
+    const googleOptions = await parseProviderOptions7({
       provider: "google",
       providerOptions: options.providerOptions,
       schema: googleInteractionsLanguageModelOptions
@@ -6203,17 +6462,17 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
     var _a, _b, _c, _d, _e, _f;
     const { args, warnings, isAgent, pollingTimeoutMs } = await this.getArgs(options);
     const url = `${this.config.baseURL}/interactions`;
-    const mergedHeaders = combineHeaders7(
+    const mergedHeaders = combineHeaders8(
       INTERACTIONS_API_REVISION_HEADER,
-      this.config.headers ? await resolve5(this.config.headers) : void 0,
+      this.config.headers ? await resolve6(this.config.headers) : void 0,
       options.headers
     );
-    const postResult = await postJsonToApi5({
+    const postResult = await postJsonToApi6({
       url,
       headers: mergedHeaders,
       body: args,
       failedResponseHandler: googleFailedResponseHandler,
-      successfulResponseHandler: createJsonResponseHandler7(
+      successfulResponseHandler: createJsonResponseHandler8(
         googleInteractionsResponseSchema
       ),
       abortSignal: options.abortSignal,
@@ -6284,9 +6543,9 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
     var _a;
     const { args, warnings, isBackground, pollingTimeoutMs } = await this.getArgs(options);
     const url = `${this.config.baseURL}/interactions`;
-    const mergedHeaders = combineHeaders7(
+    const mergedHeaders = combineHeaders8(
       INTERACTIONS_API_REVISION_HEADER,
-      this.config.headers ? await resolve5(this.config.headers) : void 0,
+      this.config.headers ? await resolve6(this.config.headers) : void 0,
       options.headers
     );
     if (isBackground) {
@@ -6300,7 +6559,7 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
       });
     }
     const body = { ...args, stream: true };
-    const { responseHeaders, value: response } = await postJsonToApi5({
+    const { responseHeaders, value: response } = await postJsonToApi6({
       url,
       headers: mergedHeaders,
       body,
@@ -6351,12 +6610,12 @@ var GoogleInteractionsLanguageModel = class _GoogleInteractionsLanguageModel {
     pollingTimeoutMs
   }) {
     var _a, _b;
-    const postResult = await postJsonToApi5({
+    const postResult = await postJsonToApi6({
       url,
       headers: mergedHeaders,
       body: args,
       failedResponseHandler: googleFailedResponseHandler,
-      successfulResponseHandler: createJsonResponseHandler7(
+      successfulResponseHandler: createJsonResponseHandler8(
         googleInteractionsResponseSchema
       ),
       abortSignal: options.abortSignal,
@@ -6483,6 +6742,12 @@ function createGoogle(options = {}) {
       generateId: (_a2 = options.generateId) != null ? _a2 : generateId2
     });
   };
+  const createSpeechModel = (modelId) => new GoogleSpeechModel(modelId, {
+    provider: `${providerName}.speech`,
+    baseURL,
+    headers: getHeaders,
+    fetch: options.fetch
+  });
   const createInteractionsModel = (modelIdOrAgent) => {
     var _a2;
     return new GoogleInteractionsLanguageModel(
@@ -6517,6 +6782,8 @@ function createGoogle(options = {}) {
   provider.video = createVideoModel;
   provider.videoModel = createVideoModel;
   provider.files = createFiles;
+  provider.speech = createSpeechModel;
+  provider.speechModel = createSpeechModel;
   provider.interactions = createInteractionsModel;
   provider.tools = googleTools;
   return provider;