@ai-sdk/google 3.0.73 → 3.0.75

package/dist/index.js

@@ -30,7 +30,7 @@ module.exports = __toCommonJS(index_exports);
 var import_provider_utils23 = require("@ai-sdk/provider-utils");
 
 // src/version.ts
-var VERSION = true ? "3.0.73" : "0.0.0-test";
+var VERSION = true ? "3.0.75" : "0.0.0-test";
 
 // src/google-generative-ai-embedding-model.ts
 var import_provider = require("@ai-sdk/provider");
@@ -429,7 +429,7 @@ function convertUrlToolResultPart(url) {
     }
   };
 }
-function appendToolResultParts(parts, toolName, outputValue) {
+function appendToolResultParts(parts, toolName, outputValue, toolCallId) {
   const functionResponseParts = [];
   const responseTextParts = [];
   for (const contentPart of outputValue) {
@@ -468,6 +468,7 @@ function appendToolResultParts(parts, toolName, outputValue) {
   }
   parts.push({
     functionResponse: {
+      ...toolCallId != null ? { id: toolCallId } : {},
       name: toolName,
       response: {
         name: toolName,
@@ -477,12 +478,13 @@ function appendToolResultParts(parts, toolName, outputValue) {
     }
   });
 }
-function appendLegacyToolResultParts(parts, toolName, outputValue) {
+function appendLegacyToolResultParts(parts, toolName, outputValue, toolCallId) {
   for (const contentPart of outputValue) {
     switch (contentPart.type) {
       case "text":
         parts.push({
           functionResponse: {
+            ...toolCallId != null ? { id: toolCallId } : {},
             name: toolName,
             response: {
               name: toolName,
@@ -612,6 +614,7 @@ function convertToGoogleGenerativeAIMessages(prompt, options) {
                 }
                 return {
                   functionCall: {
+                    ...part.toolCallId != null ? { id: part.toolCallId } : {},
                     name: part.toolName,
                     args: part.input
                   },
@@ -668,13 +671,24 @@ function convertToGoogleGenerativeAIMessages(prompt, options) {
           const output = part.output;
           if (output.type === "content") {
             if (supportsFunctionResponseParts) {
-              appendToolResultParts(parts, part.toolName, output.value);
+              appendToolResultParts(
+                parts,
+                part.toolName,
+                output.value,
+                part.toolCallId
+              );
             } else {
-              appendLegacyToolResultParts(parts, part.toolName, output.value);
+              appendLegacyToolResultParts(
+                parts,
+                part.toolName,
+                output.value,
+                part.toolCallId
+              );
             }
           } else {
             parts.push({
               functionResponse: {
+                ...part.toolCallId != null ? { id: part.toolCallId } : {},
                 name: part.toolName,
                 response: {
                   name: part.toolName,
@@ -1494,7 +1508,7 @@ var GoogleGenerativeAILanguageModel = class {
     };
   }
   async doGenerate(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q;
+    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r;
     const { args, warnings, providerOptionsName } = await this.getArgs(options);
     const mergedHeaders = (0, import_provider_utils6.combineHeaders)(
       await (0, import_provider_utils6.resolve)(this.config.headers),
@@ -1565,9 +1579,9 @@ var GoogleGenerativeAILanguageModel = class {
       } else if ("functionCall" in part && part.functionCall.name != null) {
         content.push({
           type: "tool-call",
-          toolCallId: this.config.generateId(),
+          toolCallId: (_e = part.functionCall.id) != null ? _e : this.config.generateId(),
           toolName: part.functionCall.name,
-          input: JSON.stringify((_e = part.functionCall.args) != null ? _e : {}),
+          input: JSON.stringify((_f = part.functionCall.args) != null ? _f : {}),
           providerMetadata: part.thoughtSignature ? {
             [providerOptionsName]: {
               thoughtSignature: part.thoughtSignature
@@ -1589,13 +1603,13 @@ var GoogleGenerativeAILanguageModel = class {
           } : void 0
         });
       } else if ("toolCall" in part && part.toolCall) {
-        const toolCallId = (_f = part.toolCall.id) != null ? _f : this.config.generateId();
+        const toolCallId = (_g = part.toolCall.id) != null ? _g : this.config.generateId();
         lastServerToolCallId = toolCallId;
         content.push({
           type: "tool-call",
           toolCallId,
           toolName: `server:${part.toolCall.toolType}`,
-          input: JSON.stringify((_g = part.toolCall.args) != null ? _g : {}),
+          input: JSON.stringify((_h = part.toolCall.args) != null ? _h : {}),
           providerExecuted: true,
           dynamic: true,
           providerMetadata: part.thoughtSignature ? {
@@ -1612,12 +1626,12 @@ var GoogleGenerativeAILanguageModel = class {
           }
         });
       } else if ("toolResponse" in part && part.toolResponse) {
-        const responseToolCallId = (_h = lastServerToolCallId != null ? lastServerToolCallId : part.toolResponse.id) != null ? _h : this.config.generateId();
+        const responseToolCallId = (_i = lastServerToolCallId != null ? lastServerToolCallId : part.toolResponse.id) != null ? _i : this.config.generateId();
         content.push({
           type: "tool-result",
           toolCallId: responseToolCallId,
           toolName: `server:${part.toolResponse.toolType}`,
-          result: (_i = part.toolResponse.response) != null ? _i : {},
+          result: (_j = part.toolResponse.response) != null ? _j : {},
           providerMetadata: part.thoughtSignature ? {
             [providerOptionsName]: {
               thoughtSignature: part.thoughtSignature,
@@ -1634,10 +1648,10 @@ var GoogleGenerativeAILanguageModel = class {
         lastServerToolCallId = void 0;
       }
     }
-    const sources = (_j = extractSources({
+    const sources = (_k = extractSources({
       groundingMetadata: candidate.groundingMetadata,
       generateId: this.config.generateId
-    })) != null ? _j : [];
+    })) != null ? _k : [];
     for (const source of sources) {
       content.push(source);
     }
@@ -1651,19 +1665,19 @@ var GoogleGenerativeAILanguageModel = class {
             (part) => part.type === "tool-call" && !part.providerExecuted
           )
         }),
-        raw: (_k = candidate.finishReason) != null ? _k : void 0
+        raw: (_l = candidate.finishReason) != null ? _l : void 0
       },
       usage: convertGoogleGenerativeAIUsage(usageMetadata),
       warnings,
       providerMetadata: {
         [providerOptionsName]: {
-          promptFeedback: (_l = response.promptFeedback) != null ? _l : null,
-          groundingMetadata: (_m = candidate.groundingMetadata) != null ? _m : null,
-          urlContextMetadata: (_n = candidate.urlContextMetadata) != null ? _n : null,
-          safetyRatings: (_o = candidate.safetyRatings) != null ? _o : null,
+          promptFeedback: (_m = response.promptFeedback) != null ? _m : null,
+          groundingMetadata: (_n = candidate.groundingMetadata) != null ? _n : null,
+          urlContextMetadata: (_o = candidate.urlContextMetadata) != null ? _o : null,
+          safetyRatings: (_p = candidate.safetyRatings) != null ? _p : null,
           usageMetadata: usageMetadata != null ? usageMetadata : null,
-          finishMessage: (_p = candidate.finishMessage) != null ? _p : null,
-          serviceTier: (_q = response.serviceTier) != null ? _q : null
+          finishMessage: (_q = candidate.finishMessage) != null ? _q : null,
+          serviceTier: (_r = response.serviceTier) != null ? _r : null
         }
       },
       request: { body: args },
@@ -1719,7 +1733,7 @@ var GoogleGenerativeAILanguageModel = class {
             controller.enqueue({ type: "stream-start", warnings });
           },
           transform(chunk, controller) {
-            var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l;
+            var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o;
             if (options.includeRawChunks) {
               controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
             }
@@ -1925,7 +1939,7 @@ var GoogleGenerativeAILanguageModel = class {
                 const isNoArgsCompleteCall = part.functionCall.name != null && part.functionCall.args == null && part.functionCall.partialArgs == null && part.functionCall.willContinue !== true;
                 if (isStreamingChunk) {
                   if (part.functionCall.name != null && part.functionCall.willContinue === true) {
-                    const toolCallId = generateId3();
+                    const toolCallId = (_i = part.functionCall.id) != null ? _i : generateId3();
                     const accumulator = new GoogleJSONAccumulator();
                     activeStreamingToolCalls.push({
                       toolCallId,
@@ -1991,9 +2005,9 @@ var GoogleGenerativeAILanguageModel = class {
                   });
                   hasToolCalls = true;
                 } else if (isCompleteCall) {
-                  const toolCallId = generateId3();
+                  const toolCallId = (_j = part.functionCall.id) != null ? _j : generateId3();
                   const toolName = part.functionCall.name;
-                  const args2 = typeof part.functionCall.args === "string" ? part.functionCall.args : JSON.stringify((_i = part.functionCall.args) != null ? _i : {});
+                  const args2 = typeof part.functionCall.args === "string" ? part.functionCall.args : JSON.stringify((_k = part.functionCall.args) != null ? _k : {});
                   controller.enqueue({
                     type: "tool-input-start",
                     id: toolCallId,
@@ -2020,7 +2034,7 @@ var GoogleGenerativeAILanguageModel = class {
                   });
                   hasToolCalls = true;
                 } else if (isNoArgsCompleteCall) {
-                  const toolCallId = generateId3();
+                  const toolCallId = (_l = part.functionCall.id) != null ? _l : generateId3();
                   const toolName = part.functionCall.name;
                   controller.enqueue({
                     type: "tool-input-start",
@@ -2054,12 +2068,12 @@ var GoogleGenerativeAILanguageModel = class {
               };
               providerMetadata = {
                 [providerOptionsName]: {
-                  promptFeedback: (_j = value.promptFeedback) != null ? _j : null,
+                  promptFeedback: (_m = value.promptFeedback) != null ? _m : null,
                   groundingMetadata: lastGroundingMetadata,
                   urlContextMetadata: lastUrlContextMetadata,
-                  safetyRatings: (_k = candidate.safetyRatings) != null ? _k : null,
+                  safetyRatings: (_n = candidate.safetyRatings) != null ? _n : null,
                   usageMetadata: usageMetadata != null ? usageMetadata : null,
-                  finishMessage: (_l = candidate.finishMessage) != null ? _l : null,
+                  finishMessage: (_o = candidate.finishMessage) != null ? _o : null,
                   serviceTier
                 }
               };
@@ -2249,6 +2263,7 @@ var getContentSchema = () => import_v45.z.object({
       // note: order matters since text can be fully empty
       import_v45.z.object({
         functionCall: import_v45.z.object({
+          id: import_v45.z.string().nullish(),
           name: import_v45.z.string().nullish(),
           args: import_v45.z.unknown().nullish(),
           partialArgs: import_v45.z.array(partialArgSchema).nullish(),
@@ -3110,7 +3125,7 @@ function annotationToSource({
     }
     case "file_citation": {
       const a = annotation;
-      const uri = (_b = (_a = a.document_uri) != null ? _a : a.source) != null ? _b : a.file_name;
+      const uri = (_b = (_a = a.url) != null ? _a : a.document_uri) != null ? _b : a.file_name;
       if (uri == null || uri.length === 0) return void 0;
       if (uri.startsWith("http://") || uri.startsWith("https://")) {
         return {
@@ -3204,7 +3219,7 @@ function builtinToolResultToSources({
       for (const raw of result) {
         if (raw == null || typeof raw !== "object") continue;
         const entry = raw;
-        const uri = (_g = (_f = entry.document_uri) != null ? _f : entry.source) != null ? _g : entry.file_name;
+        const uri = (_g = (_f = entry.url) != null ? _f : entry.document_uri) != null ? _g : entry.file_name;
         if (uri == null || uri.length === 0) continue;
         if (uri.startsWith("http://") || uri.startsWith("https://")) {
           sources.push({
@@ -3320,7 +3335,7 @@ function buildGoogleInteractionsStreamTransform({
       controller.enqueue({ type: "stream-start", warnings });
     },
     transform(chunk, controller) {
-      var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p;
+      var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r;
       if (includeRawChunks) {
         controller.enqueue({ type: "raw", rawValue: chunk.rawValue });
       }
@@ -3332,7 +3347,7 @@ function buildGoogleInteractionsStreamTransform({
       const value = chunk.value;
       const eventType = value.event_type;
       switch (eventType) {
-        case "interaction.start": {
+        case "interaction.created": {
           const event = value;
           const interaction = event.interaction;
           interactionId = (interaction == null ? void 0 : interaction.id) != null && interaction.id.length > 0 ? interaction.id : void 0;
@@ -3352,91 +3367,106 @@ function buildGoogleInteractionsStreamTransform({
           });
           break;
         }
-        case "content.start": {
+        case "step.start": {
           const event = value;
-          const block = event.content;
+          const step = event.step;
           const index = event.index;
           const blockId = `${interactionId != null ? interactionId : "interaction"}:${index}`;
-          if ((block == null ? void 0 : block.type) === "text") {
-            openBlocks.set(index, {
-              kind: "text",
-              id: blockId,
-              emittedSourceKeys: /* @__PURE__ */ new Set()
-            });
-            controller.enqueue({ type: "text-start", id: blockId });
-            const initialSources = annotationsToSources({
-              annotations: block.annotations,
-              generateId: generateId3
-            });
-            for (const source of initialSources) {
-              const key = sourceKey(source);
-              if (emittedSourceKeys.has(key)) continue;
-              emittedSourceKeys.add(key);
-              controller.enqueue(source);
+          const stepType = step == null ? void 0 : step.type;
+          if (stepType === "model_output") {
+            const initial = (_a = step == null ? void 0 : step.content) == null ? void 0 : _a[0];
+            if ((initial == null ? void 0 : initial.type) === "text") {
+              openBlocks.set(index, {
+                kind: "text",
+                id: blockId,
+                emittedSourceKeys: /* @__PURE__ */ new Set()
+              });
+              controller.enqueue({ type: "text-start", id: blockId });
+              const initialSources = annotationsToSources({
+                annotations: initial.annotations,
+                generateId: generateId3
+              });
+              for (const source of initialSources) {
+                const key = sourceKey(source);
+                if (emittedSourceKeys.has(key)) continue;
+                emittedSourceKeys.add(key);
+                controller.enqueue(source);
+              }
+            } else if ((initial == null ? void 0 : initial.type) === "image") {
+              openBlocks.set(index, {
+                kind: "image",
+                id: blockId,
+                ...initial.data != null ? { data: initial.data } : {},
+                ...initial.mime_type != null ? { mimeType: initial.mime_type } : {},
+                ...initial.uri != null ? { uri: initial.uri } : {}
+              });
+            } else {
+              openBlocks.set(index, {
+                kind: "pending_model_output",
+                id: blockId
+              });
             }
-          } else if ((block == null ? void 0 : block.type) === "image") {
-            const img = block;
-            openBlocks.set(index, {
-              kind: "image",
-              id: blockId,
-              ...img.data != null ? { data: img.data } : {},
-              ...img.mime_type != null ? { mimeType: img.mime_type } : {},
-              ...img.uri != null ? { uri: img.uri } : {}
-            });
-          } else if ((block == null ? void 0 : block.type) === "thought") {
-            const signature = block.signature;
+          } else if (stepType === "thought") {
+            const signature = step == null ? void 0 : step.signature;
             openBlocks.set(index, {
               kind: "reasoning",
               id: blockId,
               ...signature != null ? { signature } : {}
             });
             controller.enqueue({ type: "reasoning-start", id: blockId });
-          } else if ((block == null ? void 0 : block.type) === "function_call") {
-            const fc = block;
-            const toolCallId = (_a = fc.id) != null ? _a : blockId;
+            if (Array.isArray(step == null ? void 0 : step.summary)) {
+              for (const item of step.summary) {
+                if ((item == null ? void 0 : item.type) === "text" && typeof item.text === "string") {
+                  controller.enqueue({
+                    type: "reasoning-delta",
+                    id: blockId,
+                    delta: item.text
+                  });
+                }
+              }
+            }
+          } else if (stepType === "function_call") {
+            const toolCallId = (_b = step == null ? void 0 : step.id) != null ? _b : blockId;
+            const toolName = (_c = step == null ? void 0 : step.name) != null ? _c : "unknown";
             hasFunctionCall = true;
             const state = {
               kind: "function_call",
               id: blockId,
               toolCallId,
-              toolName: fc.name,
-              arguments: (_b = fc.arguments) != null ? _b : {},
-              ...fc.signature != null ? { signature: fc.signature } : {},
-              startEmitted: false
+              toolName,
+              argumentsAccum: "",
+              ...(step == null ? void 0 : step.signature) != null ? { signature: step.signature } : {}
             };
             openBlocks.set(index, state);
-            if (state.toolName != null) {
-              controller.enqueue({
-                type: "tool-input-start",
-                id: toolCallId,
-                toolName: state.toolName
-              });
-              state.startEmitted = true;
-            }
-          } else if ((block == null ? void 0 : block.type) != null && BUILTIN_TOOL_CALL_TYPES.has(block.type)) {
-            const toolName = block.type === "mcp_server_tool_call" ? (_c = block.name) != null ? _c : "mcp_server_tool" : builtinToolNameFromCallType(block.type);
-            const toolCallId = (_d = block.id) != null ? _d : blockId;
+            controller.enqueue({
+              type: "tool-input-start",
+              id: toolCallId,
+              toolName
+            });
+          } else if (stepType != null && BUILTIN_TOOL_CALL_TYPES.has(stepType)) {
+            const toolName = stepType === "mcp_server_tool_call" ? (_d = step == null ? void 0 : step.name) != null ? _d : "mcp_server_tool" : builtinToolNameFromCallType(stepType);
+            const toolCallId = (_e = step == null ? void 0 : step.id) != null ? _e : blockId;
             const state = {
               kind: "builtin_tool_call",
               id: blockId,
-              blockType: block.type,
+              blockType: stepType,
               toolCallId,
               toolName,
-              arguments: (_e = block.arguments) != null ? _e : {},
+              arguments: (_f = step == null ? void 0 : step.arguments) != null ? _f : {},
               callEmitted: false
             };
             openBlocks.set(index, state);
-          } else if ((block == null ? void 0 : block.type) != null && BUILTIN_TOOL_RESULT_TYPES.has(block.type)) {
-            const toolName = block.type === "mcp_server_tool_result" ? (_f = block.name) != null ? _f : "mcp_server_tool" : builtinToolNameFromResultType(block.type);
-            const callId = (_g = block.call_id) != null ? _g : blockId;
+          } else if (stepType != null && BUILTIN_TOOL_RESULT_TYPES.has(stepType)) {
+            const toolName = stepType === "mcp_server_tool_result" ? (_g = step == null ? void 0 : step.name) != null ? _g : "mcp_server_tool" : builtinToolNameFromResultType(stepType);
+            const callId = (_h = step == null ? void 0 : step.call_id) != null ? _h : blockId;
             const state = {
               kind: "builtin_tool_result",
               id: blockId,
-              blockType: block.type,
+              blockType: stepType,
               callId,
               toolName,
-              result: (_h = block.result) != null ? _h : null,
-              ...block.is_error != null ? { isError: block.is_error } : {},
+              result: (_i = step == null ? void 0 : step.result) != null ? _i : null,
+              ...(step == null ? void 0 : step.is_error) != null ? { isError: step.is_error } : {},
               resultEmitted: false
             };
             openBlocks.set(index, state);
@@ -3445,13 +3475,58 @@ function buildGoogleInteractionsStreamTransform({
           }
           break;
         }
-        case "content.delta": {
+        case "step.delta": {
           const event = value;
-          const open = openBlocks.get(event.index);
+          let open = openBlocks.get(event.index);
           if (open == null) break;
+          const dtype = (_j = event.delta) == null ? void 0 : _j.type;
+          if (open.kind === "pending_model_output") {
+            if (dtype === "text" || dtype === "text_annotation" || dtype === "text_annotation_delta") {
+              const promoted = {
+                kind: "text",
+                id: open.id,
+                emittedSourceKeys: /* @__PURE__ */ new Set()
+              };
+              openBlocks.set(event.index, promoted);
+              open = promoted;
+              controller.enqueue({ type: "text-start", id: promoted.id });
+            }
+          }
+          if (dtype === "image" && (open.kind === "pending_model_output" || open.kind === "text" || open.kind === "image")) {
+            const img = event.delta;
+            const google2 = {};
+            if (interactionId != null) google2.interactionId = interactionId;
+            const providerMetadata = Object.keys(google2).length > 0 ? { google: google2 } : void 0;
+            if ((img == null ? void 0 : img.data) != null && img.data.length > 0) {
+              controller.enqueue({
+                type: "file",
+                mediaType: (_k = img.mime_type) != null ? _k : "image/png",
+                data: img.data,
+                ...providerMetadata ? { providerMetadata } : {}
+              });
+            } else if ((img == null ? void 0 : img.uri) != null && img.uri.length > 0) {
+              const uriProviderMetadata = {
+                google: {
+                  ...interactionId != null ? { interactionId } : {},
+                  imageUri: img.uri
+                }
+              };
+              controller.enqueue({
+                type: "file",
+                mediaType: (_l = img.mime_type) != null ? _l : "image/png",
+                data: "",
+                providerMetadata: uriProviderMetadata
+              });
+            }
+            if (open.kind === "image") {
+              open.data = void 0;
+              open.uri = void 0;
+            }
+            break;
+          }
           const delta = event.delta;
           if (open.kind === "text" && (delta == null ? void 0 : delta.type) === "text") {
-            const text = (_i = delta.text) != null ? _i : "";
+            const text = (_m = delta.text) != null ? _m : "";
             if (text.length > 0) {
               controller.enqueue({
                 type: "text-delta",
@@ -3459,7 +3534,7 @@ function buildGoogleInteractionsStreamTransform({
                 delta: text
               });
             }
-          } else if (open.kind === "text" && (delta == null ? void 0 : delta.type) === "text_annotation") {
+          } else if (open.kind === "text" && ((delta == null ? void 0 : delta.type) === "text_annotation" || (delta == null ? void 0 : delta.type) === "text_annotation_delta")) {
             const sources = annotationsToSources({
               annotations: delta.annotations,
               generateId: generateId3
@@ -3491,31 +3566,28 @@ function buildGoogleInteractionsStreamTransform({
                 open.signature = signature;
               }
             }
-          } else if (open.kind === "function_call" && (delta == null ? void 0 : delta.type) === "function_call") {
+          } else if (open.kind === "function_call" && (delta == null ? void 0 : delta.type) === "arguments_delta") {
+            const slice = typeof delta.arguments === "string" ? delta.arguments : "";
+            if (slice.length > 0) {
+              open.argumentsAccum += slice;
+              controller.enqueue({
+                type: "tool-input-delta",
+                id: open.toolCallId,
+                delta: slice
+              });
+            }
             if (delta.id != null) {
               open.toolCallId = delta.id;
             }
-            if (delta.name != null) {
-              open.toolName = delta.name;
-            }
-            if (delta.arguments != null) {
-              open.arguments = delta.arguments;
-            }
             if (delta.signature != null) {
               open.signature = delta.signature;
             }
-            if (!open.startEmitted && open.toolName != null) {
-              controller.enqueue({
-                type: "tool-input-start",
-                id: open.toolCallId,
-                toolName: open.toolName
-              });
-              open.startEmitted = true;
-            }
             hasFunctionCall = true;
           } else if (open.kind === "builtin_tool_call" && (delta == null ? void 0 : delta.type) === open.blockType) {
             if (delta.id != null) open.toolCallId = delta.id;
-            if (delta.arguments != null) open.arguments = delta.arguments;
+            if (delta.arguments != null && typeof delta.arguments === "object") {
+              open.arguments = delta.arguments;
+            }
             if (delta.name != null && open.blockType === "mcp_server_tool_call") {
               open.toolName = delta.name;
             }
@@ -3529,7 +3601,7 @@ function buildGoogleInteractionsStreamTransform({
           }
           break;
         }
-        case "content.stop": {
+        case "step.stop": {
           const event = value;
           const open = openBlocks.get(event.index);
           if (open == null) break;
@@ -3557,7 +3629,7 @@ function buildGoogleInteractionsStreamTransform({
             if (open.data != null && open.data.length > 0) {
               controller.enqueue({
                 type: "file",
-                mediaType: (_j = open.mimeType) != null ? _j : "image/png",
+                mediaType: (_n = open.mimeType) != null ? _n : "image/png",
                 data: open.data,
                 ...providerMetadata ? { providerMetadata } : {}
               });
@@ -3570,26 +3642,13 @@ function buildGoogleInteractionsStreamTransform({
               };
               controller.enqueue({
                 type: "file",
-                mediaType: (_k = open.mimeType) != null ? _k : "image/png",
+                mediaType: (_o = open.mimeType) != null ? _o : "image/png",
                 data: "",
                 providerMetadata: uriProviderMetadata
               });
             }
           } else if (open.kind === "function_call") {
-            const toolName = (_l = open.toolName) != null ? _l : "unknown";
-            const argsJson = JSON.stringify((_m = open.arguments) != null ? _m : {});
-            if (!open.startEmitted) {
-              controller.enqueue({
-                type: "tool-input-start",
-                id: open.toolCallId,
-                toolName
-              });
-            }
-            controller.enqueue({
-              type: "tool-input-delta",
-              id: open.toolCallId,
-              delta: argsJson
-            });
+            const accumulated = open.argumentsAccum.length > 0 ? open.argumentsAccum : "{}";
             controller.enqueue({
               type: "tool-input-end",
               id: open.toolCallId
@@ -3601,8 +3660,8 @@ function buildGoogleInteractionsStreamTransform({
             controller.enqueue({
               type: "tool-call",
               toolCallId: open.toolCallId,
-              toolName,
-              input: argsJson,
+              toolName: open.toolName,
+              input: accumulated,
               ...providerMetadata ? { providerMetadata } : {}
             });
           } else if (open.kind === "builtin_tool_call" && !open.callEmitted) {
@@ -3610,7 +3669,7 @@ function buildGoogleInteractionsStreamTransform({
               type: "tool-call",
               toolCallId: open.toolCallId,
               toolName: open.toolName,
-              input: JSON.stringify((_n = open.arguments) != null ? _n : {}),
+              input: JSON.stringify((_p = open.arguments) != null ? _p : {}),
               providerExecuted: true
             });
             open.callEmitted = true;
@@ -3619,7 +3678,7 @@ function buildGoogleInteractionsStreamTransform({
               type: "tool-result",
               toolCallId: open.callId,
               toolName: open.toolName,
-              result: (_o = open.result) != null ? _o : null
+              result: (_q = open.result) != null ? _q : null
             });
             open.resultEmitted = true;
             const sources = builtinToolResultToSources({
@@ -3640,12 +3699,20 @@ function buildGoogleInteractionsStreamTransform({
           openBlocks.delete(event.index);
           break;
         }
-        case "interaction.status_update": {
+        case "interaction.status_update":
+        case "interaction.in_progress":
+        case "interaction.requires_action": {
           const event = value;
-          finishStatus = event.status;
+          if (event.status != null) {
+            finishStatus = event.status;
+          } else if (eventType === "interaction.requires_action") {
+            finishStatus = "requires_action";
+          } else {
+            finishStatus = "in_progress";
+          }
           break;
         }
-        case "interaction.complete": {
+        case "interaction.completed": {
           const event = value;
           const interaction = event.interaction;
           if ((interaction == null ? void 0 : interaction.id) != null && interaction.id.length > 0) {
@@ -3665,7 +3732,7 @@ function buildGoogleInteractionsStreamTransform({
         case "error": {
           const event = value;
           finishStatus = "failed";
-          const errorPayload = (_p = event.error) != null ? _p : {
+          const errorPayload = (_r = event.error) != null ? _r : {
             message: "Unknown interaction error"
           };
           controller.enqueue({ type: "error", error: errorPayload });
@@ -3734,7 +3801,7 @@ function convertToGoogleInteractionsInput({
     previousInteractionId
   }) : prompt;
   const systemTexts = [];
-  const turns = [];
+  const steps = [];
   for (const message of compactedPrompt) {
     switch (message.role) {
       case "system": {
@@ -3745,11 +3812,7 @@ function convertToGoogleInteractionsInput({
         const content = [];
         for (const part of message.content) {
           if (part.type === "text") {
-            const block = {
-              type: "text",
-              text: part.text
-            };
-            content.push(block);
+            content.push({ type: "text", text: part.text });
           } else if (part.type === "file") {
             const fileBlock = convertFilePartToContent({
               part,
@@ -3763,18 +3826,25 @@ function convertToGoogleInteractionsInput({
         }
         const merged = mergeAdjacentTextContent(content);
         if (merged.length > 0) {
-          turns.push({ role: "user", content: merged });
+          steps.push({ type: "user_input", content: merged });
         }
         break;
       }
       case "assistant": {
-        const content = [];
+        let pendingModelOutput = [];
+        const flushModelOutput = () => {
+          if (pendingModelOutput.length > 0) {
+            steps.push({ type: "model_output", content: pendingModelOutput });
+            pendingModelOutput = [];
+          }
+        };
         for (const part of message.content) {
           if (part.type === "text") {
-            content.push({ type: "text", text: part.text });
+            pendingModelOutput.push({ type: "text", text: part.text });
           } else if (part.type === "reasoning") {
+            flushModelOutput();
             const signature = (_b = (_a = part.providerOptions) == null ? void 0 : _a.google) == null ? void 0 : _b.signature;
-            content.push({
+            steps.push({
               type: "thought",
               ...signature != null ? { signature } : {},
               summary: part.text.length > 0 ? [{ type: "text", text: part.text }] : void 0
@@ -3786,12 +3856,13 @@ function convertToGoogleInteractionsInput({
               mediaResolution
             });
             if (fileBlock != null) {
-              content.push(fileBlock);
+              pendingModelOutput.push(fileBlock);
             }
           } else if (part.type === "tool-call") {
+            flushModelOutput();
             const signature = (_d = (_c = part.providerOptions) == null ? void 0 : _c.google) == null ? void 0 : _d.signature;
             const args = typeof part.input === "string" ? safeParseToolArgs(part.input) : (_e = part.input) != null ? _e : {};
-            content.push({
+            steps.push({
               type: "function_call",
               id: part.toolCallId,
               name: part.toolName,
@@ -3805,9 +3876,7 @@ function convertToGoogleInteractionsInput({
             });
           }
         }
-        if (content.length > 0) {
-          turns.push({ role: "model", content });
-        }
+        flushModelOutput();
         break;
       }
       case "tool": {
@@ -3830,22 +3899,14 @@ function convertToGoogleInteractionsInput({
           content.push(block);
         }
         if (content.length > 0) {
-          turns.push({ role: "user", content });
+          steps.push({ type: "user_input", content });
         }
         break;
       }
     }
   }
   const systemInstruction = systemTexts.length > 0 ? systemTexts.join("\n\n") : void 0;
-  let input;
-  if (turns.length === 0) {
-    input = "";
-  } else if (turns.length === 1 && turns[0].role === "user" && Array.isArray(turns[0].content)) {
-    input = turns[0].content;
-  } else {
-    input = turns;
-  }
-  return { input, systemInstruction, warnings };
+  return { input: steps, systemInstruction, warnings };
 }
 function convertFilePartToContent({
   part,
@@ -4140,7 +4201,7 @@ var annotationSchema = () => {
     type: import_v415.z.literal("file_citation"),
     file_name: import_v415.z.string().nullish(),
     document_uri: import_v415.z.string().nullish(),
-    source: import_v415.z.string().nullish(),
+    url: import_v415.z.string().nullish(),
     page_number: import_v415.z.number().nullish(),
     media_id: import_v415.z.string().nullish(),
     start_index: import_v415.z.number().nullish(),
@@ -4174,34 +4235,58 @@ var contentBlockSchema = () => {
     text: import_v415.z.string(),
     annotations: import_v415.z.array(annotationSchema()).nullish()
   }).loose();
-  const thoughtContent = import_v415.z.object({
-    type: import_v415.z.literal("thought"),
-    signature: import_v415.z.string().nullish(),
-    summary: import_v415.z.array(thoughtSummaryItemSchema()).nullish()
+  const imageContent = import_v415.z.object({
+    type: import_v415.z.literal("image"),
+    data: import_v415.z.string().nullish(),
+    mime_type: import_v415.z.string().nullish(),
+    resolution: import_v415.z.enum(["low", "medium", "high", "ultra_high"]).nullish(),
+    uri: import_v415.z.string().nullish()
+  }).loose();
+  return import_v415.z.union([
+    textContent,
+    imageContent,
+    import_v415.z.object({ type: import_v415.z.string() }).loose()
+  ]);
+};
+var BUILTIN_TOOL_CALL_STEP_TYPES = [
+  "google_search_call",
+  "code_execution_call",
+  "url_context_call",
+  "file_search_call",
+  "google_maps_call",
+  "mcp_server_tool_call"
+];
+var BUILTIN_TOOL_RESULT_STEP_TYPES = [
+  "google_search_result",
+  "code_execution_result",
+  "url_context_result",
+  "file_search_result",
+  "google_maps_result",
+  "mcp_server_tool_result"
+];
+var stepSchema = () => {
+  const userInputStep = import_v415.z.object({
+    type: import_v415.z.literal("user_input"),
+    content: import_v415.z.array(contentBlockSchema()).nullish()
+  }).loose();
+  const modelOutputStep = import_v415.z.object({
+    type: import_v415.z.literal("model_output"),
+    content: import_v415.z.array(contentBlockSchema()).nullish()
   }).loose();
-  const functionCallContent = import_v415.z.object({
+  const functionCallStep = import_v415.z.object({
     type: import_v415.z.literal("function_call"),
     id: import_v415.z.string(),
     name: import_v415.z.string(),
     arguments: import_v415.z.record(import_v415.z.string(), import_v415.z.unknown()).nullish(),
     signature: import_v415.z.string().nullish()
   }).loose();
-  const imageContent = import_v415.z.object({
-    type: import_v415.z.literal("image"),
-    data: import_v415.z.string().nullish(),
-    mime_type: import_v415.z.string().nullish(),
-    resolution: import_v415.z.enum(["low", "medium", "high", "ultra_high"]).nullish(),
-    uri: import_v415.z.string().nullish()
+  const thoughtStep = import_v415.z.object({
+    type: import_v415.z.literal("thought"),
+    signature: import_v415.z.string().nullish(),
+    summary: import_v415.z.array(thoughtSummaryItemSchema()).nullish()
   }).loose();
-  const builtinToolCall = import_v415.z.object({
-    type: import_v415.z.enum([
-      "google_search_call",
-      "code_execution_call",
-      "url_context_call",
-      "file_search_call",
-      "google_maps_call",
-      "mcp_server_tool_call"
-    ]),
+  const builtinToolCallStep = import_v415.z.object({
+    type: import_v415.z.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
     id: import_v415.z.string(),
     arguments: import_v415.z.record(import_v415.z.string(), import_v415.z.unknown()).nullish(),
     name: import_v415.z.string().nullish(),
@@ -4209,15 +4294,8 @@ var contentBlockSchema = () => {
     search_type: import_v415.z.string().nullish(),
     signature: import_v415.z.string().nullish()
   }).loose();
-  const builtinToolResult = import_v415.z.object({
-    type: import_v415.z.enum([
-      "google_search_result",
-      "code_execution_result",
-      "url_context_result",
-      "file_search_result",
-      "google_maps_result",
-      "mcp_server_tool_result"
-    ]),
+  const builtinToolResultStep = import_v415.z.object({
+    type: import_v415.z.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
     call_id: import_v415.z.string(),
     result: import_v415.z.unknown().nullish(),
     is_error: import_v415.z.boolean().nullish(),
@@ -4226,12 +4304,12 @@ var contentBlockSchema = () => {
     signature: import_v415.z.string().nullish()
   }).loose();
   return import_v415.z.union([
-    textContent,
-    imageContent,
-    thoughtContent,
-    functionCallContent,
-    builtinToolCall,
-    builtinToolResult,
+    userInputStep,
+    modelOutputStep,
+    functionCallStep,
+    thoughtStep,
+    builtinToolCallStep,
+    builtinToolResultStep,
     import_v415.z.object({ type: import_v415.z.string() }).loose()
   ]);
 };
@@ -4249,7 +4327,7 @@ var googleInteractionsResponseSchema = (0, import_provider_utils17.lazySchema)(
       status: interactionStatusSchema(),
       model: import_v415.z.string().nullish(),
       agent: import_v415.z.string().nullish(),
-      outputs: import_v415.z.array(contentBlockSchema()).nullish(),
+      steps: import_v415.z.array(stepSchema()).nullish(),
       usage: usageSchema2().nullish(),
       service_tier: import_v415.z.string().nullish(),
       previous_interaction_id: import_v415.z.string().nullish(),
@@ -4263,8 +4341,8 @@ var googleInteractionsEventSchema = (0, import_provider_utils17.lazySchema)(
       const status = interactionStatusSchema();
       const annotation = annotationSchema();
       const thoughtSummaryItem = thoughtSummaryItemSchema();
-      const interactionStartEvent = import_v415.z.object({
-        event_type: import_v415.z.literal("interaction.start"),
+      const interactionCreatedEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("interaction.created"),
         event_id: import_v415.z.string().nullish(),
         interaction: import_v415.z.object({
           /*
@@ -4278,94 +4356,79 @@ var googleInteractionsEventSchema = (0, import_provider_utils17.lazySchema)(
           status: status.nullish()
         }).loose()
       }).loose();
-      const contentStartEvent = import_v415.z.object({
-        event_type: import_v415.z.literal("content.start"),
+      const stepStartEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("step.start"),
         event_id: import_v415.z.string().nullish(),
         index: import_v415.z.number(),
-        content: contentBlockSchema()
+        step: stepSchema()
       }).loose();
-      const contentDeltaText = import_v415.z.object({
+      const stepDeltaText = import_v415.z.object({
         type: import_v415.z.literal("text"),
         text: import_v415.z.string()
       }).loose();
-      const contentDeltaThoughtSummary = import_v415.z.object({
+      const stepDeltaThoughtSummary = import_v415.z.object({
         type: import_v415.z.literal("thought_summary"),
         content: thoughtSummaryItem.nullish()
       }).loose();
-      const contentDeltaThoughtSignature = import_v415.z.object({
+      const stepDeltaThoughtSignature = import_v415.z.object({
         type: import_v415.z.literal("thought_signature"),
         signature: import_v415.z.string().nullish()
       }).loose();
-      const contentDeltaFunctionCall = import_v415.z.object({
-        type: import_v415.z.literal("function_call"),
-        id: import_v415.z.string(),
-        name: import_v415.z.string(),
-        arguments: import_v415.z.record(import_v415.z.string(), import_v415.z.unknown()).nullish(),
+      const stepDeltaArgumentsDelta = import_v415.z.object({
+        type: import_v415.z.literal("arguments_delta"),
+        arguments: import_v415.z.string().nullish(),
+        id: import_v415.z.string().nullish(),
         signature: import_v415.z.string().nullish()
       }).loose();
-      const contentDeltaTextAnnotation = import_v415.z.object({
-        type: import_v415.z.literal("text_annotation"),
+      const stepDeltaTextAnnotation = import_v415.z.object({
+        type: import_v415.z.enum(["text_annotation_delta", "text_annotation"]),
         annotations: import_v415.z.array(annotation).nullish()
       }).loose();
-      const contentDeltaImage = import_v415.z.object({
+      const stepDeltaImage = import_v415.z.object({
         type: import_v415.z.literal("image"),
         data: import_v415.z.string().nullish(),
         mime_type: import_v415.z.string().nullish(),
         resolution: import_v415.z.enum(["low", "medium", "high", "ultra_high"]).nullish(),
         uri: import_v415.z.string().nullish()
       }).loose();
-      const contentDeltaBuiltinToolCall = import_v415.z.object({
-        type: import_v415.z.enum([
-          "google_search_call",
-          "code_execution_call",
-          "url_context_call",
-          "file_search_call",
-          "google_maps_call",
-          "mcp_server_tool_call"
-        ]),
-        id: import_v415.z.string(),
+      const stepDeltaBuiltinToolCall = import_v415.z.object({
+        type: import_v415.z.enum(BUILTIN_TOOL_CALL_STEP_TYPES),
+        id: import_v415.z.string().nullish(),
         arguments: import_v415.z.record(import_v415.z.string(), import_v415.z.unknown()).nullish(),
         name: import_v415.z.string().nullish(),
         server_name: import_v415.z.string().nullish(),
         search_type: import_v415.z.string().nullish(),
         signature: import_v415.z.string().nullish()
       }).loose();
-      const contentDeltaBuiltinToolResult = import_v415.z.object({
-        type: import_v415.z.enum([
-          "google_search_result",
-          "code_execution_result",
-          "url_context_result",
-          "file_search_result",
-          "google_maps_result",
-          "mcp_server_tool_result"
-        ]),
-        call_id: import_v415.z.string(),
+      const stepDeltaBuiltinToolResult = import_v415.z.object({
+        type: import_v415.z.enum(BUILTIN_TOOL_RESULT_STEP_TYPES),
+        call_id: import_v415.z.string().nullish(),
         result: import_v415.z.unknown().nullish(),
         is_error: import_v415.z.boolean().nullish(),
         name: import_v415.z.string().nullish(),
         server_name: import_v415.z.string().nullish(),
         signature: import_v415.z.string().nullish()
       }).loose();
-      const contentDeltaUnknown = import_v415.z.object({ type: import_v415.z.string() }).loose();
-      const contentDeltaUnion = import_v415.z.union([
-        contentDeltaText,
-        contentDeltaImage,
-        contentDeltaThoughtSummary,
-        contentDeltaThoughtSignature,
-        contentDeltaFunctionCall,
-        contentDeltaTextAnnotation,
-        contentDeltaBuiltinToolCall,
-        contentDeltaBuiltinToolResult,
-        contentDeltaUnknown
+      const stepDeltaUnknown = import_v415.z.object({ type: import_v415.z.string() }).loose();
+      const stepDeltaUnion = import_v415.z.union([
+        stepDeltaText,
+        stepDeltaImage,
+        stepDeltaThoughtSummary,
+        stepDeltaThoughtSignature,
+        stepDeltaArgumentsDelta,
+        stepDeltaTextAnnotation,
+        stepDeltaBuiltinToolCall,
+        stepDeltaBuiltinToolResult,
+        stepDeltaUnknown
       ]);
-      const contentDeltaEvent = import_v415.z.object({
-        event_type: import_v415.z.literal("content.delta"),
+      const stepDeltaEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("step.delta"),
         event_id: import_v415.z.string().nullish(),
         index: import_v415.z.number(),
-        delta: contentDeltaUnion
+        delta: stepDeltaUnion
       }).loose();
-      const contentStopEvent = import_v415.z.object({
-        event_type: import_v415.z.literal("content.stop"),
+      const stepStopEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("step.stop"),
         event_id: import_v415.z.string().nullish(),
         index: import_v415.z.number()
       }).loose();
@@ -4373,10 +4436,22 @@ var googleInteractionsEventSchema = (0, import_provider_utils17.lazySchema)(
         event_type: import_v415.z.literal("interaction.status_update"),
         event_id: import_v415.z.string().nullish(),
         interaction_id: import_v415.z.string().nullish(),
-        status
+        status: status.nullish()
+      }).loose();
+      const interactionInProgressEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("interaction.in_progress"),
+        event_id: import_v415.z.string().nullish(),
+        interaction_id: import_v415.z.string().nullish(),
+        status: status.nullish()
       }).loose();
-      const interactionCompleteEvent = import_v415.z.object({
-        event_type: import_v415.z.literal("interaction.complete"),
+      const interactionRequiresActionEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("interaction.requires_action"),
+        event_id: import_v415.z.string().nullish(),
+        interaction_id: import_v415.z.string().nullish(),
+        status: status.nullish()
+      }).loose();
+      const interactionCompletedEvent = import_v415.z.object({
+        event_type: import_v415.z.literal("interaction.completed"),
         event_id: import_v415.z.string().nullish(),
         interaction: import_v415.z.object({
           id: import_v415.z.string().nullish(),
@@ -4395,12 +4470,14 @@ var googleInteractionsEventSchema = (0, import_provider_utils17.lazySchema)(
       }).loose();
       const unknownEvent = import_v415.z.object({ event_type: import_v415.z.string() }).loose();
       return import_v415.z.union([
-        interactionStartEvent,
-        contentStartEvent,
-        contentDeltaEvent,
-        contentStopEvent,
+        interactionCreatedEvent,
+        stepStartEvent,
+        stepDeltaEvent,
+        stepStopEvent,
         interactionStatusUpdateEvent,
-        interactionCompleteEvent,
+        interactionInProgressEvent,
+        interactionRequiresActionEvent,
+        interactionCompletedEvent,
         errorEvent,
         unknownEvent
       ]);
@@ -4430,6 +4507,55 @@ var googleInteractionsLanguageModelOptions = (0, import_provider_utils18.lazySch
       ]).nullish(),
       thinkingLevel: import_v416.z.enum(["minimal", "low", "medium", "high"]).nullish(),
       thinkingSummaries: import_v416.z.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
+       * with full control over `mime_type`, `aspect_ratio`, and `image_size`.
+       *
+       * Entries are sent in order. The AI SDK call-level `responseFormat: {
+       * type: 'json', schema }` still drives JSON-mode and adds a matching
+       * text entry automatically; entries listed here are appended.
+       */
+      responseFormat: import_v416.z.array(
+        import_v416.z.union([
+          import_v416.z.object({
+            type: import_v416.z.literal("text"),
+            mimeType: import_v416.z.string().nullish(),
+            schema: import_v416.z.unknown().nullish()
+          }).loose(),
+          import_v416.z.object({
+            type: import_v416.z.literal("image"),
+            mimeType: import_v416.z.string().nullish(),
+            aspectRatio: import_v416.z.enum([
+              "1:1",
+              "2:3",
+              "3:2",
+              "3:4",
+              "4:3",
+              "4:5",
+              "5:4",
+              "9:16",
+              "16:9",
+              "21:9",
+              "1:8",
+              "8:1",
+              "1:4",
+              "4:1"
+            ]).nullish(),
+            imageSize: import_v416.z.enum(["1K", "2K", "4K", "512"]).nullish()
+          }).loose(),
+          import_v416.z.object({
+            type: import_v416.z.literal("audio"),
+            mimeType: import_v416.z.string().nullish()
+          }).loose()
+        ])
+      ).nullish(),
+      /**
+       * @deprecated Use `responseFormat` with a `{ type: 'image', ... }`
+       * entry instead. Retained for backwards compatibility; the SDK
+       * translates it into a matching `response_format` image entry and
+       * emits a warning when set.
+       */
       imageConfig: import_v416.z.object({
         aspectRatio: import_v416.z.enum([
           "1:1",
@@ -4517,37 +4643,69 @@ function builtinToolNameFromResultType2(type) {
   return type.replace(/_result$/, "");
 }
 function parseGoogleInteractionsOutputs({
-  outputs,
+  steps,
   generateId: generateId3,
   interactionId
 }) {
-  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j;
+  var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k;
   const content = [];
   let hasFunctionCall = false;
-  if (outputs == null) {
+  if (steps == null) {
     return { content, hasFunctionCall };
   }
-  for (const block of outputs) {
-    if (block == null || typeof block !== "object") continue;
-    const type = block.type;
+  for (const step of steps) {
+    if (step == null || typeof step !== "object") continue;
+    const type = step.type;
     if (typeof type !== "string") continue;
     switch (type) {
-      case "text": {
-        const text = (_a = block.text) != null ? _a : "";
-        const annotations = block.annotations;
-        content.push({
-          type: "text",
-          text,
-          ...googleProviderMetadata({ interactionId })
-        });
-        const sources = annotationsToSources({ annotations, generateId: generateId3 });
-        for (const source of sources) {
-          content.push(source);
+      case "user_input": {
+        break;
+      }
+      case "model_output": {
+        const blocks = (_a = step.content) != null ? _a : [];
+        for (const block of blocks) {
+          if (block == null || typeof block !== "object") continue;
+          const blockType = block.type;
+          if (blockType === "text") {
+            const text = (_b = block.text) != null ? _b : "";
+            const annotations = block.annotations;
+            content.push({
+              type: "text",
+              text,
+              ...googleProviderMetadata({ interactionId })
+            });
+            const sources = annotationsToSources({ annotations, generateId: generateId3 });
+            for (const source of sources) {
+              content.push(source);
+            }
+          } else if (blockType === "image") {
+            const image = block;
+            if (image.data != null && image.data.length > 0) {
+              content.push({
+                type: "file",
+                mediaType: (_c = image.mime_type) != null ? _c : "image/png",
+                data: image.data,
+                ...googleProviderMetadata({ interactionId })
+              });
+            } else if (image.uri != null && image.uri.length > 0) {
+              content.push({
+                type: "file",
+                mediaType: (_d = image.mime_type) != null ? _d : "image/png",
+                data: "",
+                providerMetadata: {
+                  google: {
+                    ...interactionId != null ? { interactionId } : {},
+                    imageUri: image.uri
+                  }
+                }
+              });
+            }
+          }
         }
         break;
       }
       case "thought": {
-        const thought = block;
+        const thought = step;
         const summary = Array.isArray(thought.summary) ? thought.summary : [];
         const text = summary.filter(
           (item) => (item == null ? void 0 : item.type) === "text" && typeof item.text === "string"
@@ -4562,38 +4720,14 @@ function parseGoogleInteractionsOutputs({
         });
         break;
       }
-      case "image": {
-        const image = block;
-        if (image.data != null && image.data.length > 0) {
-          content.push({
-            type: "file",
-            mediaType: (_b = image.mime_type) != null ? _b : "image/png",
-            data: image.data,
-            ...googleProviderMetadata({ interactionId })
-          });
-        } else if (image.uri != null && image.uri.length > 0) {
-          content.push({
-            type: "file",
-            mediaType: (_c = image.mime_type) != null ? _c : "image/png",
-            data: "",
-            providerMetadata: {
-              google: {
-                ...interactionId != null ? { interactionId } : {},
-                imageUri: image.uri
-              }
-            }
-          });
-        }
-        break;
-      }
       case "function_call": {
         hasFunctionCall = true;
-        const call = block;
+        const call = step;
         content.push({
           type: "tool-call",
           toolCallId: call.id,
           toolName: call.name,
-          input: JSON.stringify((_d = call.arguments) != null ? _d : {}),
+          input: JSON.stringify((_e = call.arguments) != null ? _e : {}),
           ...googleProviderMetadata({
             signature: call.signature,
             interactionId
@@ -4603,27 +4737,27 @@ function parseGoogleInteractionsOutputs({
       }
       default: {
         if (BUILTIN_TOOL_CALL_TYPES2.has(type)) {
-          const call = block;
-          const toolName = type === "mcp_server_tool_call" ? (_e = call.name) != null ? _e : "mcp_server_tool" : builtinToolNameFromCallType2(type);
-          const input = JSON.stringify((_f = call.arguments) != null ? _f : {});
+          const call = step;
+          const toolName = type === "mcp_server_tool_call" ? (_f = call.name) != null ? _f : "mcp_server_tool" : builtinToolNameFromCallType2(type);
+          const input = JSON.stringify((_g = call.arguments) != null ? _g : {});
           content.push({
             type: "tool-call",
-            toolCallId: (_g = call.id) != null ? _g : generateId3(),
+            toolCallId: (_h = call.id) != null ? _h : generateId3(),
             toolName,
             input,
             providerExecuted: true
           });
         } else if (BUILTIN_TOOL_RESULT_TYPES2.has(type)) {
-          const result = block;
-          const toolName = type === "mcp_server_tool_result" ? (_h = result.name) != null ? _h : "mcp_server_tool" : builtinToolNameFromResultType2(type);
+          const result = step;
+          const toolName = type === "mcp_server_tool_result" ? (_i = result.name) != null ? _i : "mcp_server_tool" : builtinToolNameFromResultType2(type);
           content.push({
             type: "tool-result",
-            toolCallId: (_i = result.call_id) != null ? _i : generateId3(),
+            toolCallId: (_j = result.call_id) != null ? _j : generateId3(),
             toolName,
-            result: (_j = result.result) != null ? _j : null
+            result: (_k = result.result) != null ? _k : null
           });
           const sources = builtinToolResultToSources({
-            block,
+            block: step,
             generateId: generateId3
           });
           for (const source of sources) {
@@ -4997,7 +5131,7 @@ function streamGoogleInteractionEvents({
               if (typeof ev.event_id === "string" && ev.event_id.length > 0) {
                 lastEventId = ev.event_id;
               }
-              if (ev.event_type === "interaction.complete" || ev.event_type === "error") {
+              if (ev.event_type === "interaction.completed" || ev.event_type === "error") {
                 complete = true;
               }
             }
@@ -5078,7 +5212,7 @@ function synthesizeGoogleInteractionsAgentStream({
         controller.enqueue({ type: "raw", rawValue: response });
       }
       const { content, hasFunctionCall } = parseGoogleInteractionsOutputs({
-        outputs: (_b = response.outputs) != null ? _b : null,
+        steps: (_b = response.steps) != null ? _b : null,
         generateId: generateId3,
         interactionId
       });
@@ -5220,7 +5354,7 @@ var GoogleInteractionsLanguageModel = class {
     };
   }
   async getArgs(options) {
-    var _a, _b, _c, _d, _e, _f, _g, _h, _i, _j, _k, _l, _m, _n, _o, _p, _q, _r, _s, _t;
+    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;
     const warnings = [];
     const opts = await (0, import_provider_utils22.parseProviderOptions)({
       provider: "google",
@@ -5245,8 +5379,7 @@ var GoogleInteractionsLanguageModel = class {
       toolChoiceForBody = prepared.toolChoice;
       warnings.push(...prepared.toolWarnings);
     }
-    let responseMimeType;
-    let responseFormat;
+    const responseFormatEntries = [];
     if (((_a = options.responseFormat) == null ? void 0 : _a.type) === "json") {
       if (isAgent) {
         warnings.push({
@@ -5254,9 +5387,40 @@ var GoogleInteractionsLanguageModel = class {
           message: "google.interactions: structured output (responseFormat) is not supported when an agent is set; responseFormat will be ignored."
         });
       } else {
-        responseMimeType = "application/json";
-        if (options.responseFormat.schema != null) {
-          responseFormat = options.responseFormat.schema;
+        const entry = {
+          type: "text",
+          mime_type: "application/json",
+          ...options.responseFormat.schema != null ? { schema: options.responseFormat.schema } : {}
+        };
+        responseFormatEntries.push(entry);
+      }
+    }
+    if ((opts == null ? void 0 : opts.responseFormat) != null) {
+      for (const entry of opts.responseFormat) {
+        if (entry.type === "text") {
+          responseFormatEntries.push(
+            pruneUndefined({
+              type: "text",
+              mime_type: (_b = entry.mimeType) != null ? _b : void 0,
+              schema: (_c = entry.schema) != null ? _c : void 0
+            })
+          );
+        } else if (entry.type === "image") {
+          responseFormatEntries.push(
+            pruneUndefined({
+              type: "image",
+              mime_type: (_d = entry.mimeType) != null ? _d : void 0,
+              aspect_ratio: (_e = entry.aspectRatio) != null ? _e : void 0,
+              image_size: (_f = entry.imageSize) != null ? _f : void 0
+            })
+          );
+        } else if (entry.type === "audio") {
+          responseFormatEntries.push(
+            pruneUndefined({
+              type: "audio",
+              mime_type: (_g = entry.mimeType) != null ? _g : void 0
+            })
+          );
         }
       }
     }
@@ -5266,13 +5430,13 @@ var GoogleInteractionsLanguageModel = class {
       warnings: convWarnings
     } = convertToGoogleInteractionsInput({
       prompt: options.prompt,
-      previousInteractionId: (_b = opts == null ? void 0 : opts.previousInteractionId) != null ? _b : void 0,
-      store: (_c = opts == null ? void 0 : opts.store) != null ? _c : void 0,
-      mediaResolution: (_d = opts == null ? void 0 : opts.mediaResolution) != null ? _d : void 0
+      previousInteractionId: (_h = opts == null ? void 0 : opts.previousInteractionId) != null ? _h : void 0,
+      store: (_i = opts == null ? void 0 : opts.store) != null ? _i : void 0,
+      mediaResolution: (_j = opts == null ? void 0 : opts.mediaResolution) != null ? _j : void 0
     });
     warnings.push(...convWarnings);
     let systemInstruction = convertedSystemInstruction;
-    const optionSystemInstruction = (_e = opts == null ? void 0 : opts.systemInstruction) != null ? _e : void 0;
+    const optionSystemInstruction = (_k = opts == null ? void 0 : opts.systemInstruction) != null ? _k : void 0;
     if (systemInstruction != null && optionSystemInstruction != null) {
       warnings.push({
         type: "other",
@@ -5306,19 +5470,32 @@ var GoogleInteractionsLanguageModel = class {
       generationConfig = void 0;
     } else {
       generationConfig = pruneUndefined({
-        temperature: (_f = options.temperature) != null ? _f : void 0,
-        top_p: (_g = options.topP) != null ? _g : void 0,
-        seed: (_h = options.seed) != null ? _h : void 0,
+        temperature: (_l = options.temperature) != null ? _l : void 0,
+        top_p: (_m = options.topP) != null ? _m : void 0,
+        seed: (_n = options.seed) != null ? _n : void 0,
         stop_sequences: options.stopSequences != null && options.stopSequences.length > 0 ? options.stopSequences : void 0,
-        max_output_tokens: (_i = options.maxOutputTokens) != null ? _i : void 0,
-        thinking_level: (_j = opts == null ? void 0 : opts.thinkingLevel) != null ? _j : void 0,
-        thinking_summaries: (_k = opts == null ? void 0 : opts.thinkingSummaries) != null ? _k : void 0,
-        image_config: (opts == null ? void 0 : opts.imageConfig) != null ? pruneUndefined({
-          aspect_ratio: (_l = opts.imageConfig.aspectRatio) != null ? _l : void 0,
-          image_size: (_m = opts.imageConfig.imageSize) != null ? _m : void 0
-        }) : void 0,
+        max_output_tokens: (_o = options.maxOutputTokens) != null ? _o : void 0,
+        thinking_level: (_p = opts == null ? void 0 : opts.thinkingLevel) != null ? _p : void 0,
+        thinking_summaries: (_q = opts == null ? void 0 : opts.thinkingSummaries) != null ? _q : void 0,
         tool_choice: toolChoiceForBody
       });
+      if ((opts == null ? void 0 : opts.imageConfig) != null) {
+        const alreadyHasImageEntry = responseFormatEntries.some(
+          (entry) => entry.type === "image"
+        );
+        warnings.push({
+          type: "other",
+          message: alreadyHasImageEntry ? "google.interactions: providerOptions.google.imageConfig is deprecated and was ignored because providerOptions.google.responseFormat already supplies an image entry. Use responseFormat exclusively." : 'google.interactions: providerOptions.google.imageConfig is deprecated. Use providerOptions.google.responseFormat with a { type: "image", ... } entry instead.'
+        });
+        if (!alreadyHasImageEntry) {
+          responseFormatEntries.push({
+            type: "image",
+            mime_type: "image/png",
+            ...opts.imageConfig.aspectRatio != null ? { aspect_ratio: opts.imageConfig.aspectRatio } : {},
+            ...opts.imageConfig.imageSize != null ? { image_size: opts.imageConfig.imageSize } : {}
+          });
+        }
+      }
     }
     let agentConfig;
     if (isAgent && (opts == null ? void 0 : opts.agentConfig) != null) {
@@ -5326,9 +5503,9 @@ var GoogleInteractionsLanguageModel = class {
       if (ac.type === "deep-research") {
         agentConfig = pruneUndefined({
           type: "deep-research",
-          thinking_summaries: (_n = ac.thinkingSummaries) != null ? _n : void 0,
-          visualization: (_o = ac.visualization) != null ? _o : void 0,
-          collaborative_planning: (_p = ac.collaborativePlanning) != null ? _p : void 0
+          thinking_summaries: (_r = ac.thinkingSummaries) != null ? _r : void 0,
+          visualization: (_s = ac.visualization) != null ? _s : void 0,
+          collaborative_planning: (_t = ac.collaborativePlanning) != null ? _t : void 0
         });
       } else if (ac.type === "dynamic") {
         agentConfig = { type: "dynamic" };
@@ -5339,12 +5516,11 @@ var GoogleInteractionsLanguageModel = class {
       input,
       system_instruction: systemInstruction,
       tools: toolsForBody,
-      response_format: responseFormat,
-      response_mime_type: responseMimeType,
+      response_format: responseFormatEntries.length > 0 ? responseFormatEntries : void 0,
       response_modalities: (opts == null ? void 0 : opts.responseModalities) != null ? opts.responseModalities : void 0,
-      previous_interaction_id: (_q = opts == null ? void 0 : opts.previousInteractionId) != null ? _q : void 0,
-      service_tier: (_r = opts == null ? void 0 : opts.serviceTier) != null ? _r : void 0,
-      store: (_s = opts == null ? void 0 : opts.store) != null ? _s : void 0,
+      previous_interaction_id: (_u = opts == null ? void 0 : opts.previousInteractionId) != null ? _u : void 0,
+      service_tier: (_v = opts == null ? void 0 : opts.serviceTier) != null ? _v : void 0,
+      store: (_w = opts == null ? void 0 : opts.store) != null ? _w : void 0,
       generation_config: generationConfig != null && Object.keys(generationConfig).length > 0 ? generationConfig : void 0,
       agent_config: agentConfig,
       ...isAgent ? { background: true } : {}
@@ -5353,7 +5529,7 @@ var GoogleInteractionsLanguageModel = class {
       args,
       warnings,
       isAgent,
-      pollingTimeoutMs: (_t = opts == null ? void 0 : opts.pollingTimeoutMs) != null ? _t : void 0
+      pollingTimeoutMs: (_x = opts == null ? void 0 : opts.pollingTimeoutMs) != null ? _x : void 0
     };
   }
   async doGenerate(options) {
@@ -5361,6 +5537,7 @@ var GoogleInteractionsLanguageModel = class {
     const { args, warnings, isAgent, pollingTimeoutMs } = await this.getArgs(options);
     const url = `${this.config.baseURL}/interactions`;
     const mergedHeaders = (0, import_provider_utils22.combineHeaders)(
+      INTERACTIONS_API_REVISION_HEADER,
       this.config.headers ? await (0, import_provider_utils22.resolve)(this.config.headers) : void 0,
       options.headers
     );
@@ -5395,7 +5572,7 @@ var GoogleInteractionsLanguageModel = class {
     }
     const interactionId = typeof response.id === "string" && response.id.length > 0 ? response.id : void 0;
     const { content, hasFunctionCall } = parseGoogleInteractionsOutputs({
-      outputs: (_b = response.outputs) != null ? _b : null,
+      steps: (_b = response.steps) != null ? _b : null,
       generateId: (_c = this.config.generateId) != null ? _c : import_provider_utils22.generateId,
       interactionId
     });
@@ -5441,6 +5618,7 @@ var GoogleInteractionsLanguageModel = class {
     const { args, warnings, isAgent, pollingTimeoutMs } = await this.getArgs(options);
     const url = `${this.config.baseURL}/interactions`;
     const mergedHeaders = (0, import_provider_utils22.combineHeaders)(
+      INTERACTIONS_API_REVISION_HEADER,
       this.config.headers ? await (0, import_provider_utils22.resolve)(this.config.headers) : void 0,
       options.headers
     );
@@ -5560,6 +5738,9 @@ var GoogleInteractionsLanguageModel = class {
     };
   }
 };
+var INTERACTIONS_API_REVISION_HEADER = {
+  "Api-Revision": "2026-05-20"
+};
 function pruneUndefined(obj) {
   const result = {};
   for (const [key, value] of Object.entries(obj)) {