@browserbasehq/orca 3.0.7-alpha-1 → 3.0.8-google-cua-1

package/dist/index.js

@@ -179,7 +179,7 @@ var __forAwait = (obj, it, method) => (it = obj[__knownSymbol("asyncIterator")])
 var STAGEHAND_VERSION;
 var init_version = __esm({
   "lib/version.ts"() {
-    STAGEHAND_VERSION = "3.0.7-alpha-1";
+    STAGEHAND_VERSION = "3.0.8-google-cua-1";
   }
 });
 
@@ -201,7 +201,7 @@ var init_sdkErrors = __esm({
             `
 Hey! We're sorry you ran into an error. 
 Stagehand version: ${STAGEHAND_VERSION} 
-If you need help, please open a Github issue or reach out to us on Slack: https://stagehand.dev/slack
+If you need help, please open a Github issue or reach out to us on Discord: https://stagehand.dev/discord
 
 Full error:
 ${error.message}`
@@ -26278,6 +26278,16 @@ var init_page = __esm({
           yield this.waitForMainLoadState(state, timeoutMs != null ? timeoutMs : 15e3);
         });
       }
+      /**
+       * Wait for a specified amount of time.
+       *
+       * @param ms The number of milliseconds to wait.
+       */
+      waitForTimeout(ms) {
+        return __async(this, null, function* () {
+          return new Promise((resolve3) => setTimeout(resolve3, ms));
+        });
+      }
       evaluate(pageFunctionOrExpression, arg) {
         return __async(this, null, function* () {
           var _a4;
@@ -26803,6 +26813,7 @@ var init_page = __esm({
           // Modifier keys
           case "cmd":
           case "command":
+          case "controlormeta":
             return this.isMacOS() ? "Meta" : "Control";
           case "win":
           case "windows":
@@ -27145,6 +27156,7 @@ __export(v3_exports, {
   providerEnvVarMap: () => providerEnvVarMap,
   toGeminiSchema: () => toGeminiSchema,
   toJsonSchema: () => toJsonSchema,
+  tool: () => import_ai24.tool,
   transformSchema: () => transformSchema,
   trimTrailingTextNode: () => trimTrailingTextNode,
   validateZodSchema: () => validateZodSchema
@@ -27718,17 +27730,20 @@ var providerEnvVarMap = {
   xai: "XAI_API_KEY",
   google_legacy: "GOOGLE_API_KEY"
 };
+var providersWithoutApiKey = /* @__PURE__ */ new Set(["ollama"]);
 function loadApiKeyFromEnv(provider, logger) {
   if (!provider) {
     return void 0;
   }
   const envVarName = providerEnvVarMap[provider];
   if (!envVarName) {
-    logger({
-      category: "init",
-      message: `No known environment variable for provider '${provider}'`,
-      level: 0
-    });
+    if (!providersWithoutApiKey.has(provider)) {
+      logger({
+        category: "init",
+        message: `No known environment variable for provider '${provider}'`,
+        level: 0
+      });
+    }
     return void 0;
   }
   const apiKeyFromEnv = Array.isArray(envVarName) ? envVarName.map((name) => process.env[name]).find((key) => key && key.length > 0) : process.env[envVarName];
@@ -28117,7 +28132,7 @@ var ActCache = class {
       };
     });
   }
-  tryReplay(context, page, timeout) {
+  tryReplay(context, page, timeout, llmClientOverride) {
     return __async(this, null, function* () {
       var _a4;
       if (!this.enabled) return null;
@@ -28170,7 +28185,13 @@ var ActCache = class {
           }
         }
       });
-      return yield this.replayCachedActions(context, entry, page, timeout);
+      return yield this.replayCachedActions(
+        context,
+        entry,
+        page,
+        timeout,
+        llmClientOverride
+      );
     });
   }
   store(context, result) {
@@ -28220,12 +28241,13 @@ var ActCache = class {
     });
     return (0, import_crypto.createHash)("sha256").update(payload).digest("hex");
   }
-  replayCachedActions(context, entry, page, timeout) {
+  replayCachedActions(context, entry, page, timeout, llmClientOverride) {
     return __async(this, null, function* () {
       const handler = this.getActHandler();
       if (!handler) {
         throw new StagehandNotInitializedError("act()");
       }
+      const effectiveClient = llmClientOverride != null ? llmClientOverride : this.getDefaultLlmClient();
       const execute = () => __async(this, null, function* () {
         var _a4, _b, _c;
         const actionResults = [];
@@ -28234,7 +28256,7 @@ var ActCache = class {
             action,
             page,
             this.domSettleTimeoutMs,
-            this.getDefaultLlmClient(),
+            effectiveClient,
             void 0,
             context.variables
           );
@@ -28472,7 +28494,7 @@ var AgentCache = class {
       };
     });
   }
-  tryReplay(context) {
+  tryReplay(context, llmClientOverride) {
     return __async(this, null, function* () {
       if (!this.enabled) return null;
       const {
@@ -28505,7 +28527,7 @@ var AgentCache = class {
           url: { value: context.startUrl, type: "string" }
         }
       });
-      return yield this.replayAgentCacheEntry(context, entry);
+      return yield this.replayAgentCacheEntry(context, entry, llmClientOverride);
     });
   }
   /**
@@ -28524,9 +28546,9 @@ var AgentCache = class {
    * and await `result` even when the response comes from cache, maintaining
    * API consistency regardless of whether the result was cached or live.
    */
-  tryReplayAsStream(context) {
+  tryReplayAsStream(context, llmClientOverride) {
     return __async(this, null, function* () {
-      const result = yield this.tryReplay(context);
+      const result = yield this.tryReplay(context, llmClientOverride);
       if (!result) return null;
       return this.createCachedStreamResult(result);
     });
@@ -28752,17 +28774,23 @@ var AgentCache = class {
     }
     return value;
   }
-  replayAgentCacheEntry(context, entry) {
+  replayAgentCacheEntry(context, entry, llmClientOverride) {
     return __async(this, null, function* () {
       var _a4, _b, _c;
       const ctx = this.getContext();
       const handler = this.getActHandler();
       if (!ctx || !handler) return null;
+      const effectiveClient = llmClientOverride != null ? llmClientOverride : this.getDefaultLlmClient();
       try {
         const updatedSteps = [];
         let stepsChanged = false;
         for (const step of (_a4 = entry.steps) != null ? _a4 : []) {
-          const replayedStep = (_b = yield this.executeAgentReplayStep(step, ctx, handler)) != null ? _b : step;
+          const replayedStep = (_b = yield this.executeAgentReplayStep(
+            step,
+            ctx,
+            handler,
+            effectiveClient
+          )) != null ? _b : step;
           stepsChanged || (stepsChanged = replayedStep !== step);
           updatedSteps.push(replayedStep);
         }
@@ -28795,20 +28823,22 @@ var AgentCache = class {
       }
     });
   }
-  executeAgentReplayStep(step, ctx, handler) {
+  executeAgentReplayStep(step, ctx, handler, llmClient) {
     return __async(this, null, function* () {
       switch (step.type) {
         case "act":
           return yield this.replayAgentActStep(
             step,
             ctx,
-            handler
+            handler,
+            llmClient
           );
         case "fillForm":
           return yield this.replayAgentFillFormStep(
             step,
             ctx,
-            handler
+            handler,
+            llmClient
           );
         case "goto":
           yield this.replayAgentGotoStep(step, ctx);
@@ -28840,7 +28870,7 @@ var AgentCache = class {
       }
     });
   }
-  replayAgentActStep(step, ctx, handler) {
+  replayAgentActStep(step, ctx, handler, llmClient) {
     return __async(this, null, function* () {
       const actions = Array.isArray(step.actions) ? step.actions : [];
       if (actions.length > 0) {
@@ -28851,7 +28881,7 @@ var AgentCache = class {
             action,
             page,
             this.domSettleTimeoutMs,
-            this.getDefaultLlmClient()
+            llmClient
           );
           if (result.success && Array.isArray(result.actions)) {
             updatedActions.push(...cloneForCache(result.actions));
@@ -28868,7 +28898,7 @@ var AgentCache = class {
       return step;
     });
   }
-  replayAgentFillFormStep(step, ctx, handler) {
+  replayAgentFillFormStep(step, ctx, handler, llmClient) {
     return __async(this, null, function* () {
       var _a4;
       const actions = Array.isArray(step.actions) && step.actions.length > 0 ? step.actions : (_a4 = step.observeResults) != null ? _a4 : [];
@@ -28882,7 +28912,7 @@ var AgentCache = class {
           action,
           page,
           this.domSettleTimeoutMs,
-          this.getDefaultLlmClient()
+          llmClient
         );
         if (result.success && Array.isArray(result.actions)) {
           updatedActions.push(...cloneForCache(result.actions));
@@ -31115,7 +31145,25 @@ var screenshotTool = (v3) => (0, import_ai3.tool)({
 // lib/v3/agent/tools/wait.ts
 var import_ai4 = require("ai");
 var import_zod8 = require("zod");
-var waitTool = (v3) => (0, import_ai4.tool)({
+
+// lib/v3/agent/utils/screenshotHandler.ts
+var DEFAULT_DELAY_MS = 500;
+function waitAndCaptureScreenshot(_0) {
+  return __async(this, arguments, function* (page, delayMs = DEFAULT_DELAY_MS) {
+    if (delayMs > 0) {
+      yield page.waitForTimeout(delayMs);
+    }
+    try {
+      const buffer = yield page.screenshot({ fullPage: false });
+      return buffer.toString("base64");
+    } catch (e2) {
+      return void 0;
+    }
+  });
+}
+
+// lib/v3/agent/tools/wait.ts
+var waitTool = (v3, mode) => (0, import_ai4.tool)({
   description: "Wait for a specified time",
   inputSchema: import_zod8.z.object({
     timeMs: import_zod8.z.number().describe("Time in milliseconds")
@@ -31136,8 +31184,32 @@ var waitTool = (v3) => (0, import_ai4.tool)({
     if (timeMs > 0) {
       v3.recordAgentReplayStep({ type: "wait", timeMs });
     }
+    if (mode === "hybrid") {
+      const page = yield v3.context.awaitActivePage();
+      const screenshotBase64 = yield waitAndCaptureScreenshot(page, 0);
+      return { success: true, waited: timeMs, screenshotBase64 };
+    }
     return { success: true, waited: timeMs };
-  })
+  }),
+  toModelOutput: (result) => {
+    const content = [
+      {
+        type: "text",
+        text: JSON.stringify({
+          success: result.success,
+          waited: result.waited
+        })
+      }
+    ];
+    if (result.screenshotBase64) {
+      content.push({
+        type: "media",
+        mediaType: "image/png",
+        data: result.screenshotBase64
+      });
+    }
+    return { type: "content", value: content };
+  }
 });
 
 // lib/v3/agent/tools/navback.ts
@@ -31164,26 +31236,12 @@ var navBackTool = (v3) => (0, import_ai5.tool)({
   })
 });
 
-// lib/v3/agent/tools/close.ts
+// lib/v3/agent/tools/ariaTree.ts
 var import_ai6 = require("ai");
 var import_zod10 = require("zod");
-var closeTool = () => (0, import_ai6.tool)({
-  description: "Complete the task and close",
-  inputSchema: import_zod10.z.object({
-    reasoning: import_zod10.z.string().describe("Summary of what was accomplished"),
-    taskComplete: import_zod10.z.boolean().describe("Whether the task was completed successfully")
-  }),
-  execute: (_0) => __async(null, [_0], function* ({ reasoning, taskComplete }) {
-    return { success: true, reasoning, taskComplete };
-  })
-});
-
-// lib/v3/agent/tools/ariaTree.ts
-var import_ai7 = require("ai");
-var import_zod11 = require("zod");
-var ariaTreeTool = (v3) => (0, import_ai7.tool)({
+var ariaTreeTool = (v3) => (0, import_ai6.tool)({
   description: "gets the accessibility (ARIA) hybrid tree text for the current page. use this to understand structure and content.",
-  inputSchema: import_zod11.z.object({}),
+  inputSchema: import_zod10.z.object({}),
   execute: () => __async(null, null, function* () {
     v3.logger({
       category: "agent",
@@ -31210,18 +31268,18 @@ ${result.content}` }]
 });
 
 // lib/v3/agent/tools/fillform.ts
-var import_ai8 = require("ai");
-var import_zod12 = require("zod");
-var fillFormTool = (v3, executionModel) => (0, import_ai8.tool)({
+var import_ai7 = require("ai");
+var import_zod11 = require("zod");
+var fillFormTool = (v3, executionModel) => (0, import_ai7.tool)({
   description: `\u{1F4DD} FORM FILL - MULTI-FIELD INPUT TOOL
 For any form with 2+ inputs/textareas. Faster than individual typing.`,
-  inputSchema: import_zod12.z.object({
-    fields: import_zod12.z.array(
-      import_zod12.z.object({
-        action: import_zod12.z.string().describe(
+  inputSchema: import_zod11.z.object({
+    fields: import_zod11.z.array(
+      import_zod11.z.object({
+        action: import_zod11.z.string().describe(
           'Description of typing action, e.g. "type foo into the email field"'
         ),
-        value: import_zod12.z.string().describe("Text to type into the target")
+        value: import_zod11.z.string().describe("Text to type into the target")
       })
     ).min(1, "Provide at least one field to fill")
   }),
@@ -31264,8 +31322,8 @@ For any form with 2+ inputs/textareas. Faster than individual typing.`,
 });
 
 // lib/v3/agent/tools/scroll.ts
-var import_ai9 = require("ai");
-var import_zod13 = require("zod");
+var import_ai8 = require("ai");
+var import_zod12 = require("zod");
 
 // lib/v3/agent/utils/coordinateNormalization.ts
 var DEFAULT_VIEWPORT = { width: 1288, height: 711 };
@@ -31289,11 +31347,11 @@ function processCoordinates(x2, y, provider) {
 }
 
 // lib/v3/agent/tools/scroll.ts
-var scrollTool = (v3) => (0, import_ai9.tool)({
+var scrollTool = (v3) => (0, import_ai8.tool)({
   description: "Scroll the page up or down by a percentage of the viewport height. Default is 80%, and what should be typically used for general page scrolling",
-  inputSchema: import_zod13.z.object({
-    direction: import_zod13.z.enum(["up", "down"]),
-    percentage: import_zod13.z.number().min(1).max(200).optional()
+  inputSchema: import_zod12.z.object({
+    direction: import_zod12.z.enum(["up", "down"]),
+    percentage: import_zod12.z.number().min(1).max(200).optional()
   }),
   execute: (_0) => __async(null, [_0], function* ({ direction, percentage = 80 }) {
     v3.logger({
@@ -31327,16 +31385,20 @@ var scrollTool = (v3) => (0, import_ai9.tool)({
     };
   })
 });
-var scrollVisionTool = (v3, provider) => (0, import_ai9.tool)({
+var scrollVisionTool = (v3, provider) => (0, import_ai8.tool)({
   description: `Scroll the page up or down. For general page scrolling, no coordinates needed. Only provide coordinates when scrolling inside a nested scrollable element (e.g., a dropdown menu, modal with overflow, or scrollable sidebar). Default is 80%, and what should be typically used for general page scrolling`,
-  inputSchema: import_zod13.z.object({
-    direction: import_zod13.z.enum(["up", "down"]),
-    coordinates: import_zod13.z.array(import_zod13.z.number()).optional().describe(
+  inputSchema: import_zod12.z.object({
+    direction: import_zod12.z.enum(["up", "down"]),
+    coordinates: import_zod12.z.array(import_zod12.z.number()).optional().describe(
       "Only use coordinates for scrolling inside a nested scrollable element - provide (x, y) within that element"
     ),
-    percentage: import_zod13.z.number().min(1).max(200).optional()
+    percentage: import_zod12.z.number().min(1).max(200).optional()
   }),
-  execute: (_0) => __async(null, [_0], function* ({ direction, coordinates, percentage = 80 }) {
+  execute: (_0) => __async(null, [_0], function* ({
+    direction,
+    coordinates,
+    percentage = 80
+  }) {
     const page = yield v3.context.awaitActivePage();
     const { w, h: h2 } = yield page.mainFrame().evaluate("({ w: window.innerWidth, h: window.innerHeight })");
     let cx;
@@ -31372,6 +31434,7 @@ var scrollVisionTool = (v3, provider) => (0, import_ai9.tool)({
     const scrollDistance = Math.round(h2 * percentage / 100);
     const deltaY = direction === "up" ? -scrollDistance : scrollDistance;
     yield page.scroll(cx, cy, 0, deltaY);
+    const screenshotBase64 = yield waitAndCaptureScreenshot(page, 100);
     v3.recordAgentReplayStep({
       type: "scroll",
       deltaX: 0,
@@ -31381,29 +31444,50 @@ var scrollVisionTool = (v3, provider) => (0, import_ai9.tool)({
     return {
       success: true,
       message: coordinates ? `Scrolled ${percentage}% ${direction} at (${cx}, ${cy})` : `Scrolled ${percentage}% ${direction}`,
-      scrolledPixels: scrollDistance
+      scrolledPixels: scrollDistance,
+      screenshotBase64
     };
-  })
+  }),
+  toModelOutput: (result) => {
+    const content = [
+      {
+        type: "text",
+        text: JSON.stringify({
+          success: result.success,
+          message: result.message,
+          scrolledPixels: result.scrolledPixels
+        })
+      }
+    ];
+    if (result.screenshotBase64) {
+      content.push({
+        type: "media",
+        mediaType: "image/png",
+        data: result.screenshotBase64
+      });
+    }
+    return { type: "content", value: content };
+  }
 });
 
 // lib/v3/agent/tools/extract.ts
-var import_ai10 = require("ai");
-var import_zod14 = require("zod");
+var import_ai9 = require("ai");
+var import_zod13 = require("zod");
 function evaluateZodSchema(schemaStr, logger) {
   var _a4;
   try {
     const fn = new Function("z", `return ${schemaStr}`);
-    return fn(import_zod14.z);
+    return fn(import_zod13.z);
   } catch (e2) {
     logger == null ? void 0 : logger({
       category: "agent",
       message: `Failed to evaluate schema: ${(_a4 = e2 == null ? void 0 : e2.message) != null ? _a4 : String(e2)}`,
       level: 0
     });
-    return import_zod14.z.any();
+    return import_zod13.z.any();
   }
 }
-var extractTool = (v3, executionModel, logger) => (0, import_ai10.tool)({
+var extractTool = (v3, executionModel, logger) => (0, import_ai9.tool)({
   description: `Extract structured data from the current page based on a provided schema.
     
     USAGE GUIDELINES:
@@ -31422,9 +31506,9 @@ var extractTool = (v3, executionModel, logger) => (0, import_ai10.tool)({
     3. Extract arrays:
        instruction: "extract all product names and prices"
        schema: "z.object({ products: z.array(z.object({ name: z.string(), price: z.number() })) })"`,
-  inputSchema: import_zod14.z.object({
-    instruction: import_zod14.z.string(),
-    schema: import_zod14.z.string().optional().describe("Zod schema as code, e.g. z.object({ title: z.string() })")
+  inputSchema: import_zod13.z.object({
+    instruction: import_zod13.z.string(),
+    schema: import_zod13.z.string().optional().describe("Zod schema as code, e.g. z.object({ title: z.string() })")
   }),
   execute: (_0) => __async(null, [_0], function* ({ instruction, schema }) {
     var _a4;
@@ -31439,8 +31523,8 @@ var extractTool = (v3, executionModel, logger) => (0, import_ai10.tool)({
 });
 
 // lib/v3/agent/tools/click.ts
-var import_ai11 = require("ai");
-var import_zod15 = require("zod");
+var import_ai10 = require("ai");
+var import_zod14 = require("zod");
 
 // lib/v3/agent/utils/xpath.ts
 function ensureXPath(value) {
@@ -31451,16 +31535,13 @@ function ensureXPath(value) {
 }
 
 // lib/v3/agent/tools/click.ts
-function waitForTimeout(ms) {
-  return new Promise((resolve3) => setTimeout(resolve3, ms));
-}
-var clickTool = (v3, provider) => (0, import_ai11.tool)({
+var clickTool = (v3, provider) => (0, import_ai10.tool)({
   description: "Click on an element using its coordinates (this is the most reliable way to click on an element, always use this over act, unless the element is not visible in the screenshot, but shown in ariaTree)",
-  inputSchema: import_zod15.z.object({
-    describe: import_zod15.z.string().describe(
+  inputSchema: import_zod14.z.object({
+    describe: import_zod14.z.string().describe(
       "Describe the element to click on in a short, specific phrase that mentions the element type and a good visual description"
     ),
-    coordinates: import_zod15.z.array(import_zod15.z.number()).describe("The (x, y) coordinates to click on")
+    coordinates: import_zod14.z.array(import_zod14.z.number()).describe("The (x, y) coordinates to click on")
   }),
   execute: (_0) => __async(null, [_0], function* ({ describe, coordinates }) {
     try {
@@ -31476,36 +31557,38 @@ var clickTool = (v3, provider) => (0, import_ai11.tool)({
         level: 1,
         auxiliary: {
           arguments: {
-            value: JSON.stringify({ describe, coordinates, processed }),
-            type: "string"
+            value: JSON.stringify({ describe }),
+            type: "object"
           }
         }
       });
+      const shouldCollectXpath = v3.isAgentReplayActive();
       const xpath = yield page.click(processed.x, processed.y, {
-        returnXpath: true
+        returnXpath: shouldCollectXpath
       });
-      if (isGoogleProvider(provider)) {
-        yield waitForTimeout(1e3);
-      }
-      const normalizedXpath = ensureXPath(xpath);
-      if (normalizedXpath) {
-        const action = {
-          selector: normalizedXpath,
-          description: describe,
-          method: "click",
-          arguments: []
-        };
-        v3.recordAgentReplayStep({
-          type: "act",
-          instruction: describe,
-          actions: [action],
-          actionDescription: describe
-        });
+      const screenshotBase64 = yield waitAndCaptureScreenshot(page);
+      if (shouldCollectXpath) {
+        const normalizedXpath = ensureXPath(xpath);
+        if (normalizedXpath) {
+          const action = {
+            selector: normalizedXpath,
+            description: describe,
+            method: "click",
+            arguments: []
+          };
+          v3.recordAgentReplayStep({
+            type: "act",
+            instruction: describe,
+            actions: [action],
+            actionDescription: describe
+          });
+        }
       }
       return {
         success: true,
         describe,
-        coordinates: [processed.x, processed.y]
+        coordinates: [processed.x, processed.y],
+        screenshotBase64
       };
     } catch (error) {
       return {
@@ -31513,25 +31596,60 @@ var clickTool = (v3, provider) => (0, import_ai11.tool)({
         error: `Error clicking: ${error.message}`
       };
     }
-  })
+  }),
+  toModelOutput: (result) => {
+    if (result.success) {
+      const content = [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            describe: result.describe,
+            coordinates: result.coordinates
+          })
+        }
+      ];
+      if (result.screenshotBase64) {
+        content.push({
+          type: "media",
+          mediaType: "image/png",
+          data: result.screenshotBase64
+        });
+      }
+      return { type: "content", value: content };
+    }
+    return {
+      type: "content",
+      value: [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            error: result.error
+          })
+        }
+      ]
+    };
+  }
 });
 
 // lib/v3/agent/tools/type.ts
-var import_ai12 = require("ai");
-var import_zod16 = require("zod");
-function waitForTimeout2(ms) {
-  return new Promise((resolve3) => setTimeout(resolve3, ms));
-}
-var typeTool = (v3, provider) => (0, import_ai12.tool)({
+var import_ai11 = require("ai");
+var import_zod15 = require("zod");
+var typeTool = (v3, provider) => (0, import_ai11.tool)({
   description: "Type text into an element using its coordinates. This will click the element and then type the text into it (this is the most reliable way to type into an element, always use this over act, unless the element is not visible in the screenshot, but shown in ariaTree)",
-  inputSchema: import_zod16.z.object({
-    describe: import_zod16.z.string().describe(
+  inputSchema: import_zod15.z.object({
+    describe: import_zod15.z.string().describe(
       "Describe the element to type into in a short, specific phrase that mentions the element type and a good visual description"
     ),
-    text: import_zod16.z.string().describe("The text to type into the element"),
-    coordinates: import_zod16.z.array(import_zod16.z.number()).describe("The (x, y) coordinates to type into the element")
+    text: import_zod15.z.string().describe("The text to type into the element"),
+    coordinates: import_zod15.z.array(import_zod15.z.number()).describe("The (x, y) coordinates to type into the element")
   }),
-  execute: (_0) => __async(null, [_0], function* ({ describe, coordinates, text }) {
+  execute: (_0) => __async(null, [_0], function* ({
+    describe,
+    coordinates,
+    text
+  }) {
     try {
       const page = yield v3.context.awaitActivePage();
       const processed = processCoordinates(
@@ -31545,54 +31663,98 @@ var typeTool = (v3, provider) => (0, import_ai12.tool)({
         level: 1,
         auxiliary: {
           arguments: {
-            value: JSON.stringify({ describe, coordinates, processed, text }),
-            type: "string"
+            value: JSON.stringify({ describe, text }),
+            type: "object"
           }
         }
       });
+      const shouldCollectXpath = v3.isAgentReplayActive();
       const xpath = yield page.click(processed.x, processed.y, {
-        returnXpath: true
+        returnXpath: shouldCollectXpath
       });
-      if (isGoogleProvider(provider)) {
-        yield waitForTimeout2(1e3);
-      }
       yield page.type(text);
-      const normalizedXpath = ensureXPath(xpath);
-      if (normalizedXpath) {
-        const action = {
-          selector: normalizedXpath,
-          description: describe,
-          method: "type",
-          arguments: [text]
-        };
-        v3.recordAgentReplayStep({
-          type: "act",
-          instruction: describe,
-          actions: [action],
-          actionDescription: describe
-        });
+      const screenshotBase64 = yield waitAndCaptureScreenshot(page);
+      if (shouldCollectXpath) {
+        const normalizedXpath = ensureXPath(xpath);
+        if (normalizedXpath) {
+          const action = {
+            selector: normalizedXpath,
+            description: describe,
+            method: "type",
+            arguments: [text]
+          };
+          v3.recordAgentReplayStep({
+            type: "act",
+            instruction: describe,
+            actions: [action],
+            actionDescription: describe
+          });
+        }
       }
-      return { success: true, describe, text };
+      return {
+        success: true,
+        describe,
+        text,
+        screenshotBase64
+      };
     } catch (error) {
       return {
         success: false,
         error: `Error typing: ${error.message}`
       };
     }
-  })
+  }),
+  toModelOutput: (result) => {
+    if (result.success) {
+      const content = [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            describe: result.describe,
+            text: result.text
+          })
+        }
+      ];
+      if (result.screenshotBase64) {
+        content.push({
+          type: "media",
+          mediaType: "image/png",
+          data: result.screenshotBase64
+        });
+      }
+      return { type: "content", value: content };
+    }
+    return {
+      type: "content",
+      value: [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            error: result.error
+          })
+        }
+      ]
+    };
+  }
 });
 
 // lib/v3/agent/tools/dragAndDrop.ts
-var import_ai13 = require("ai");
-var import_zod17 = require("zod");
-var dragAndDropTool = (v3, provider) => (0, import_ai13.tool)({
+var import_ai12 = require("ai");
+var import_zod16 = require("zod");
+var dragAndDropTool = (v3, provider) => (0, import_ai12.tool)({
   description: "Drag and drop an element using its coordinates (this is the most reliable way to drag and drop an element, always use this over act, unless the element is not visible in the screenshot, but shown in ariaTree)",
-  inputSchema: import_zod17.z.object({
-    describe: import_zod17.z.string().describe("Describe the element to drag and drop"),
-    startCoordinates: import_zod17.z.array(import_zod17.z.number()).describe("The (x, y) coordinates to start the drag and drop from"),
-    endCoordinates: import_zod17.z.array(import_zod17.z.number()).describe("The (x, y) coordinates to end the drag and drop at")
+  inputSchema: import_zod16.z.object({
+    describe: import_zod16.z.string().describe("Describe the element to drag and drop"),
+    startCoordinates: import_zod16.z.array(import_zod16.z.number()).describe("The (x, y) coordinates to start the drag and drop from"),
+    endCoordinates: import_zod16.z.array(import_zod16.z.number()).describe("The (x, y) coordinates to end the drag and drop at")
   }),
-  execute: (_0) => __async(null, [_0], function* ({ describe, startCoordinates, endCoordinates }) {
+  execute: (_0) => __async(null, [_0], function* ({
+    describe,
+    startCoordinates,
+    endCoordinates
+  }) {
     try {
       const page = yield v3.context.awaitActivePage();
       const processedStart = processCoordinates(
@@ -31612,60 +31774,97 @@ var dragAndDropTool = (v3, provider) => (0, import_ai13.tool)({
         auxiliary: {
           arguments: {
             value: JSON.stringify({
-              describe,
-              startCoordinates,
-              endCoordinates,
-              processedStart,
-              processedEnd
+              describe
             }),
-            type: "string"
+            type: "object"
           }
         }
       });
+      const shouldCollectXpath = v3.isAgentReplayActive();
       const [fromXpath, toXpath] = yield page.dragAndDrop(
         processedStart.x,
         processedStart.y,
         processedEnd.x,
         processedEnd.y,
-        { returnXpath: true }
+        { returnXpath: shouldCollectXpath }
       );
-      const normalizedFrom = ensureXPath(fromXpath);
-      const normalizedTo = ensureXPath(toXpath);
-      if (normalizedFrom && normalizedTo) {
-        const action = {
-          selector: normalizedFrom,
-          description: describe,
-          method: "dragAndDrop",
-          arguments: [normalizedTo]
-        };
-        v3.recordAgentReplayStep({
-          type: "act",
-          instruction: describe,
-          actions: [action],
-          actionDescription: describe
-        });
+      const screenshotBase64 = yield waitAndCaptureScreenshot(page);
+      if (shouldCollectXpath) {
+        const normalizedFrom = ensureXPath(fromXpath);
+        const normalizedTo = ensureXPath(toXpath);
+        if (normalizedFrom && normalizedTo) {
+          const action = {
+            selector: normalizedFrom,
+            description: describe,
+            method: "dragAndDrop",
+            arguments: [normalizedTo]
+          };
+          v3.recordAgentReplayStep({
+            type: "act",
+            instruction: describe,
+            actions: [action],
+            actionDescription: describe
+          });
+        }
       }
-      return { success: true, describe };
+      return {
+        success: true,
+        describe,
+        screenshotBase64
+      };
     } catch (error) {
       return {
         success: false,
         error: `Error dragging: ${error.message}`
       };
     }
-  })
+  }),
+  toModelOutput: (result) => {
+    if (result.success) {
+      const content = [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            describe: result.describe
+          })
+        }
+      ];
+      if (result.screenshotBase64) {
+        content.push({
+          type: "media",
+          mediaType: "image/png",
+          data: result.screenshotBase64
+        });
+      }
+      return { type: "content", value: content };
+    }
+    return {
+      type: "content",
+      value: [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            error: result.error
+          })
+        }
+      ]
+    };
+  }
 });
 
 // lib/v3/agent/tools/clickAndHold.ts
-var import_ai14 = require("ai");
-var import_zod18 = require("zod");
-var clickAndHoldTool = (v3, provider) => (0, import_ai14.tool)({
+var import_ai13 = require("ai");
+var import_zod17 = require("zod");
+var clickAndHoldTool = (v3, provider) => (0, import_ai13.tool)({
   description: "Click and hold on an element using its coordinates",
-  inputSchema: import_zod18.z.object({
-    describe: import_zod18.z.string().describe(
+  inputSchema: import_zod17.z.object({
+    describe: import_zod17.z.string().describe(
       "Describe the element to click on in a short, specific phrase that mentions the element type and a good visual description"
     ),
-    duration: import_zod18.z.number().describe("The duration to hold the element in milliseconds"),
-    coordinates: import_zod18.z.array(import_zod18.z.number()).describe("The (x, y) coordinates to click on")
+    duration: import_zod17.z.number().describe("The duration to hold the element in milliseconds"),
+    coordinates: import_zod17.z.array(import_zod17.z.number()).describe("The (x, y) coordinates to click on")
   }),
   execute: (_0) => __async(null, [_0], function* ({ describe, coordinates, duration }) {
     try {
@@ -31683,35 +31882,36 @@ var clickAndHoldTool = (v3, provider) => (0, import_ai14.tool)({
           arguments: {
             value: JSON.stringify({
               describe,
-              coordinates,
-              processed,
               duration
             }),
-            type: "string"
+            type: "object"
           }
         }
       });
+      const shouldCollectXpath = v3.isAgentReplayActive();
       const [xpath] = yield page.dragAndDrop(
         processed.x,
         processed.y,
         processed.x,
         processed.y,
-        { delay: duration, returnXpath: true }
+        { delay: duration, returnXpath: shouldCollectXpath }
       );
-      const normalizedXpath = ensureXPath(xpath);
-      if (normalizedXpath) {
-        const action = {
-          selector: normalizedXpath,
-          description: describe,
-          method: "clickAndHold",
-          arguments: [String(duration)]
-        };
-        v3.recordAgentReplayStep({
-          type: "act",
-          instruction: describe,
-          actions: [action],
-          actionDescription: describe
-        });
+      if (shouldCollectXpath) {
+        const normalizedXpath = ensureXPath(xpath);
+        if (normalizedXpath) {
+          const action = {
+            selector: normalizedXpath,
+            description: describe,
+            method: "clickAndHold",
+            arguments: [String(duration)]
+          };
+          v3.recordAgentReplayStep({
+            type: "act",
+            instruction: describe,
+            actions: [action],
+            actionDescription: describe
+          });
+        }
       }
       return { success: true, describe };
     } catch (error) {
@@ -31724,20 +31924,20 @@ var clickAndHoldTool = (v3, provider) => (0, import_ai14.tool)({
 });
 
 // lib/v3/agent/tools/keys.ts
-var import_ai15 = require("ai");
-var import_zod19 = require("zod");
-var keysTool = (v3) => (0, import_ai15.tool)({
+var import_ai14 = require("ai");
+var import_zod18 = require("zod");
+var keysTool = (v3) => (0, import_ai14.tool)({
   description: `Send keyboard input to the page without targeting a specific element. Unlike the type tool which clicks then types into coordinates, this sends keystrokes directly to wherever focus currently is.
 
 Use method="type" to enter text into the currently focused element. Preferred when: input is already focused, text needs to flow across multiple fields (e.g., verification codes)
 
 Use method="press" for navigation keys (Enter, Tab, Escape, Backspace, arrows) and keyboard shortcuts (Cmd+A, Ctrl+C, Shift+Tab).`,
-  inputSchema: import_zod19.z.object({
-    method: import_zod19.z.enum(["press", "type"]),
-    value: import_zod19.z.string().describe(
+  inputSchema: import_zod18.z.object({
+    method: import_zod18.z.enum(["press", "type"]),
+    value: import_zod18.z.string().describe(
       "The text to type, or the key/combo to press (Enter, Tab, Cmd+A)"
     ),
-    repeat: import_zod19.z.number().optional()
+    repeat: import_zod18.z.number().optional()
   }),
   execute: (_0) => __async(null, [_0], function* ({ method, value, repeat }) {
     try {
@@ -31749,7 +31949,7 @@ Use method="press" for navigation keys (Enter, Tab, Escape, Backspace, arrows) a
         auxiliary: {
           arguments: {
             value: JSON.stringify({ method, value, repeat }),
-            type: "string"
+            type: "object"
           }
         }
       });
@@ -31784,9 +31984,9 @@ Use method="press" for navigation keys (Enter, Tab, Escape, Backspace, arrows) a
 });
 
 // lib/v3/agent/tools/fillFormVision.ts
-var import_ai16 = require("ai");
-var import_zod20 = require("zod");
-var fillFormVisionTool = (v3, provider) => (0, import_ai16.tool)({
+var import_ai15 = require("ai");
+var import_zod19 = require("zod");
+var fillFormVisionTool = (v3, provider) => (0, import_ai15.tool)({
   description: `FORM FILL - SPECIALIZED MULTI-FIELD INPUT TOOL
 
 CRITICAL: Use this for ANY form with 2+ input fields (text inputs, textareas, etc.)
@@ -31804,16 +32004,16 @@ MANDATORY USE CASES (always use fillFormVision for these):
 - Checkout forms: address, payment info fields
 - Profile updates: multiple user data fields
 - Search filters: multiple criteria inputs`,
-  inputSchema: import_zod20.z.object({
-    fields: import_zod20.z.array(
-      import_zod20.z.object({
-        action: import_zod20.z.string().describe(
+  inputSchema: import_zod19.z.object({
+    fields: import_zod19.z.array(
+      import_zod19.z.object({
+        action: import_zod19.z.string().describe(
           "Description of the typing action, e.g. 'type foo into the bar field'"
         ),
-        value: import_zod20.z.string().describe("Text to type into the target field"),
-        coordinates: import_zod20.z.object({
-          x: import_zod20.z.number(),
-          y: import_zod20.z.number()
+        value: import_zod19.z.string().describe("Text to type into the target field"),
+        coordinates: import_zod19.z.object({
+          x: import_zod19.z.number(),
+          y: import_zod19.z.number()
         }).describe("Coordinates of the target field")
       })
     ).min(2, "Provide at least two fields to fill")
@@ -31838,32 +32038,36 @@ MANDATORY USE CASES (always use fillFormVision for these):
         auxiliary: {
           arguments: {
             value: JSON.stringify({ fields, processedFields }),
-            type: "string"
+            type: "object"
           }
         }
       });
+      const shouldCollectXpath = v3.isAgentReplayActive();
       const actions = [];
       for (const field of processedFields) {
         const xpath = yield page.click(
           field.coordinates.x,
           field.coordinates.y,
           {
-            returnXpath: true
+            returnXpath: shouldCollectXpath
           }
         );
         yield page.type(field.value);
-        const normalizedXpath = ensureXPath(xpath);
-        if (normalizedXpath) {
-          actions.push({
-            selector: normalizedXpath,
-            description: field.action,
-            method: "type",
-            arguments: [field.value]
-          });
+        if (shouldCollectXpath) {
+          const normalizedXpath = ensureXPath(xpath);
+          if (normalizedXpath) {
+            actions.push({
+              selector: normalizedXpath,
+              description: field.action,
+              method: "type",
+              arguments: [field.value]
+            });
+          }
         }
         yield new Promise((resolve3) => setTimeout(resolve3, 100));
       }
-      if (actions.length > 0) {
+      const screenshotBase64 = yield waitAndCaptureScreenshot(page, 100);
+      if (shouldCollectXpath && actions.length > 0) {
         v3.recordAgentReplayStep({
           type: "act",
           instruction: `Fill ${fields.length} form fields`,
@@ -31873,7 +32077,8 @@ MANDATORY USE CASES (always use fillFormVision for these):
       }
       return {
         success: true,
-        playwrightArguments: processedFields
+        playwrightArguments: processedFields,
+        screenshotBase64
       };
     } catch (error) {
       return {
@@ -31881,13 +32086,47 @@ MANDATORY USE CASES (always use fillFormVision for these):
         error: `Error filling form: ${error.message}`
       };
     }
-  })
+  }),
+  toModelOutput: (result) => {
+    var _a4, _b;
+    if (result.success) {
+      const content = [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            fieldsCount: (_b = (_a4 = result.playwrightArguments) == null ? void 0 : _a4.length) != null ? _b : 0
+          })
+        }
+      ];
+      if (result.screenshotBase64) {
+        content.push({
+          type: "media",
+          mediaType: "image/png",
+          data: result.screenshotBase64
+        });
+      }
+      return { type: "content", value: content };
+    }
+    return {
+      type: "content",
+      value: [
+        {
+          type: "text",
+          text: JSON.stringify({
+            success: result.success,
+            error: result.error
+          })
+        }
+      ]
+    };
+  }
 });
 
 // lib/v3/agent/tools/think.ts
-var import_ai17 = require("ai");
-var import_zod21 = require("zod");
-var thinkTool = () => (0, import_ai17.tool)({
+var import_ai16 = require("ai");
+var import_zod20 = require("zod");
+var thinkTool = () => (0, import_ai16.tool)({
   description: `Use this tool to think through complex problems or plan a sequence of steps. This is for internal reasoning only and doesn't perform any actions. Use this to:
 
 1. Plan a multi-step approach before taking action
@@ -31896,8 +32135,8 @@ var thinkTool = () => (0, import_ai17.tool)({
 4. Evaluate options when you're unsure what to do next
 
 The output is only visible to you; use it to track your own reasoning process.`,
-  inputSchema: import_zod21.z.object({
-    reasoning: import_zod21.z.string().describe(
+  inputSchema: import_zod20.z.object({
+    reasoning: import_zod20.z.string().describe(
       "Your step-by-step reasoning or planning process. Be as detailed as needed."
     )
   }),
@@ -31910,8 +32149,8 @@ The output is only visible to you; use it to track your own reasoning process.`,
 });
 
 // lib/v3/agent/tools/search.ts
-var import_ai18 = require("ai");
-var import_zod22 = require("zod");
+var import_ai17 = require("ai");
+var import_zod21 = require("zod");
 function performBraveSearch(query) {
   return __async(this, null, function* () {
     var _a4;
@@ -31957,10 +32196,10 @@ function performBraveSearch(query) {
     }
   });
 }
-var searchTool = (v3) => (0, import_ai18.tool)({
+var searchTool = (v3) => (0, import_ai17.tool)({
   description: "Perform a web search and returns results. Use this tool when you need information from the web or when you are unsure of the exact URL you want to navigate to. This can be used to find the ideal entry point, resulting in a task that is easier to complete due to starting further in the process.",
-  inputSchema: import_zod22.z.object({
-    query: import_zod22.z.string().describe("The search query to look for on the web")
+  inputSchema: import_zod21.z.object({
+    query: import_zod21.z.string().describe("The search query to look for on the web")
   }),
   execute: (_0) => __async(null, [_0], function* ({ query }) {
     var _a4, _b, _c;
@@ -31971,7 +32210,7 @@ var searchTool = (v3) => (0, import_ai18.tool)({
       auxiliary: {
         arguments: {
           value: JSON.stringify({ query }),
-          type: "string"
+          type: "object"
         }
       }
     });
@@ -32018,7 +32257,7 @@ function createAgentTools(v3, options) {
     ariaTree: ariaTreeTool(v3),
     click: clickTool(v3, provider),
     clickAndHold: clickAndHoldTool(v3, provider),
-    close: closeTool(),
+    //close: closeTool(),
     dragAndDrop: dragAndDropTool(v3, provider),
     extract: extractTool(v3, executionModel, options == null ? void 0 : options.logger),
     fillForm: fillFormTool(v3, executionModel),
@@ -32030,7 +32269,7 @@ function createAgentTools(v3, options) {
     scroll: mode === "hybrid" ? scrollVisionTool(v3, provider) : scrollTool(v3),
     think: thinkTool(),
     type: typeTool(v3, provider),
-    wait: waitTool(v3)
+    wait: waitTool(v3, mode)
   };
   if (process.env.BRAVE_API_KEY) {
     allTools.search = searchTool(v3);
@@ -32074,8 +32313,7 @@ function buildToolsSection(isHybridMode, hasSearch, excludeTools) {
     { name: "goto", description: "Navigate to a URL" },
     { name: "wait", description: "Wait for a specified time" },
     { name: "navback", description: "Navigate back in browser history" },
-    { name: "scroll", description: "Scroll the page x pixels up or down" },
-    { name: "close", description: "Mark the task as complete or failed" }
+    { name: "scroll", description: "Scroll the page x pixels up or down" }
   ];
   const domTools = [
     {
@@ -32097,8 +32335,7 @@ function buildToolsSection(isHybridMode, hasSearch, excludeTools) {
     { name: "goto", description: "Navigate to a URL" },
     { name: "wait", description: "Wait for a specified time" },
     { name: "navback", description: "Navigate back in browser history" },
-    { name: "scroll", description: "Scroll the page x pixels up or down" },
-    { name: "close", description: "Mark the task as complete or failed" }
+    { name: "scroll", description: "Scroll the page x pixels up or down" }
   ];
   const baseTools = isHybridMode ? hybridTools : domTools;
   if (hasSearch) {
@@ -32107,8 +32344,8 @@ function buildToolsSection(isHybridMode, hasSearch, excludeTools) {
       description: "Perform a web search and return results. Prefer this over navigating to Google and searching within the page for reliability and efficiency."
     });
   }
-  const filteredTools = baseTools.filter((tool21) => !excludeSet.has(tool21.name));
-  const toolLines = filteredTools.map((tool21) => `    <tool name="${tool21.name}">${tool21.description}</tool>`).join("\n");
+  const filteredTools = baseTools.filter((tool22) => !excludeSet.has(tool22.name));
+  const toolLines = filteredTools.map((tool22) => `    <tool name="${tool22.name}">${tool22.description}</tool>`).join("\n");
   return `<tools>
 ${toolLines}
   </tools>`;
@@ -32200,8 +32437,6 @@ function buildAgentSystemPrompt(options) {
     <item>Always start by understanding the current page state</item>
     <item>Use the screenshot tool to verify page state when needed</item>
     <item>Use appropriate tools for each action</item>
-    <item>When the task is complete, use the "close" tool with taskComplete: true</item>
-    <item>If the task cannot be completed, use "close" with taskComplete: false</item>
   </guidelines>
   ${pageUnderstandingProtocol}
   <navigation>
@@ -32225,132 +32460,143 @@ function buildAgentSystemPrompt(options) {
 }
 
 // lib/v3/handlers/v3AgentHandler.ts
-var import_ai19 = require("ai");
+var import_ai20 = require("ai");
 
 // lib/v3/agent/utils/messageProcessing.ts
+var VISION_ACTION_TOOLS = [
+  "click",
+  "type",
+  "dragAndDrop",
+  "wait",
+  "fillFormVision",
+  "scroll"
+];
 function isToolMessage(message) {
   return !!message && typeof message === "object" && message.role === "tool" && Array.isArray(message.content);
 }
 function isScreenshotPart(part) {
   return !!part && typeof part === "object" && part.toolName === "screenshot";
 }
+function isVisionActionPart(part) {
+  if (!part || typeof part !== "object") return false;
+  const toolName = part.toolName;
+  return typeof toolName === "string" && VISION_ACTION_TOOLS.includes(toolName);
+}
+function isVisionPart(part) {
+  return isScreenshotPart(part) || isVisionActionPart(part);
+}
 function isAriaTreePart(part) {
   return !!part && typeof part === "object" && part.toolName === "ariaTree";
 }
-function processMessages(params) {
-  const originalContentSize = JSON.stringify(params.prompt).length;
-  const screenshotIndices = findToolIndices(params.prompt, "screenshot");
-  const ariaTreeIndices = findToolIndices(params.prompt, "ariaTree");
-  const processedPrompt = params.prompt.map(
-    (message, index) => {
+function processMessages(messages) {
+  let compressedCount = 0;
+  const visionIndices = [];
+  const ariaTreeIndices = [];
+  for (let i2 = 0; i2 < messages.length; i2++) {
+    const message = messages[i2];
+    if (isToolMessage(message)) {
+      const content = message.content;
+      if (content.some(isVisionPart)) {
+        visionIndices.push(i2);
+      }
+      if (content.some(isAriaTreePart)) {
+        ariaTreeIndices.push(i2);
+      }
+    }
+  }
+  if (visionIndices.length > 2) {
+    const toCompress = visionIndices.slice(0, visionIndices.length - 2);
+    for (const index of toCompress) {
+      const message = messages[index];
       if (isToolMessage(message)) {
-        if (message.content.some((part) => isScreenshotPart(part))) {
-          const shouldCompress = shouldCompressScreenshot(
-            index,
-            screenshotIndices
-          );
-          if (shouldCompress) {
-            return compressScreenshotMessage(message);
-          }
-        }
-        if (message.content.some((part) => isAriaTreePart(part))) {
-          const shouldCompress = shouldCompressAriaTree(index, ariaTreeIndices);
-          if (shouldCompress) {
-            return compressAriaTreeMessage(message);
-          }
-        }
+        compressScreenshotMessage(message);
+        compressVisionActionMessage(message);
+        compressedCount++;
       }
-      return message;
     }
-  );
-  const compressedContentSize = JSON.stringify(processedPrompt).length;
-  const stats = calculateCompressionStats(
-    originalContentSize,
-    compressedContentSize,
-    screenshotIndices.length,
-    ariaTreeIndices.length
-  );
-  return {
-    processedPrompt,
-    stats
-  };
-}
-function findToolIndices(prompt, toolName) {
-  const screenshotIndices = [];
-  prompt.forEach((message, index) => {
-    if (isToolMessage(message)) {
-      const hasMatch = message.content.some(
-        (part) => toolName === "screenshot" ? isScreenshotPart(part) : isAriaTreePart(part)
-      );
-      if (hasMatch) {
-        screenshotIndices.push(index);
+  }
+  if (ariaTreeIndices.length > 1) {
+    const toCompress = ariaTreeIndices.slice(0, ariaTreeIndices.length - 1);
+    for (const idx of toCompress) {
+      const message = messages[idx];
+      if (isToolMessage(message)) {
+        compressAriaTreeMessage(message);
+        compressedCount++;
       }
     }
-  });
-  return screenshotIndices;
-}
-function shouldCompressScreenshot(index, screenshotIndices) {
-  const isNewestScreenshot = index === Math.max(...screenshotIndices);
-  const isSecondNewestScreenshot = screenshotIndices.length > 1 && index === screenshotIndices.sort((a, b) => b - a)[1];
-  return !isNewestScreenshot && !isSecondNewestScreenshot;
-}
-function shouldCompressAriaTree(index, ariaTreeIndices) {
-  const isNewestAriaTree = index === Math.max(...ariaTreeIndices);
-  return !isNewestAriaTree;
+  }
+  return compressedCount;
 }
 function compressScreenshotMessage(message) {
-  const updatedContent = message.content.map((part) => {
+  var _a4;
+  for (const part of message.content) {
     if (isScreenshotPart(part)) {
-      return __spreadProps(__spreadValues({}, part), {
-        result: [
-          {
-            type: "text",
-            text: "screenshot taken"
-          }
-        ]
-      });
+      const typedPart = part;
+      const placeholder = [{ type: "text", text: "screenshot taken" }];
+      if ((_a4 = typedPart.output) == null ? void 0 : _a4.value) {
+        typedPart.output.value = placeholder;
+      }
+      if (typedPart.result) {
+        typedPart.result = placeholder;
+      }
     }
-    return part;
-  });
-  return __spreadProps(__spreadValues({}, message), {
-    content: updatedContent
-  });
+  }
+}
+function compressVisionActionMessage(message) {
+  var _a4;
+  for (const part of message.content) {
+    if (isVisionActionPart(part)) {
+      const typedPart = part;
+      if (((_a4 = typedPart.output) == null ? void 0 : _a4.value) && Array.isArray(typedPart.output.value)) {
+        typedPart.output.value = typedPart.output.value.filter(
+          (item) => item && typeof item === "object" && item.type !== "media"
+        );
+      }
+      if (typedPart.result && Array.isArray(typedPart.result)) {
+        typedPart.result = typedPart.result.filter(
+          (item) => item && typeof item === "object" && item.type !== "media"
+        );
+      }
+    }
+  }
 }
 function compressAriaTreeMessage(message) {
-  const updatedContent = message.content.map((part) => {
+  var _a4;
+  for (const part of message.content) {
     if (isAriaTreePart(part)) {
-      return __spreadProps(__spreadValues({}, part), {
-        result: [
-          {
-            type: "text",
-            text: "ARIA tree extracted for context of page elements"
-          }
-        ]
-      });
+      const typedPart = part;
+      const placeholder = [
+        {
+          type: "text",
+          text: "ARIA tree extracted for context of page elements"
+        }
+      ];
+      if ((_a4 = typedPart.output) == null ? void 0 : _a4.value) {
+        typedPart.output.value = placeholder;
+      }
+      if (typedPart.result) {
+        typedPart.result = placeholder;
+      }
     }
-    return part;
-  });
-  return __spreadProps(__spreadValues({}, message), {
-    content: updatedContent
-  });
-}
-function calculateCompressionStats(originalSize, compressedSize, screenshotCount, ariaTreeCount) {
-  const savedChars = originalSize - compressedSize;
-  const compressionRatio = originalSize > 0 ? (originalSize - compressedSize) / originalSize * 100 : 0;
-  return {
-    originalSize,
-    compressedSize,
-    savedChars,
-    compressionRatio,
-    screenshotCount,
-    ariaTreeCount
-  };
+  }
 }
 
 // lib/v3/handlers/v3AgentHandler.ts
 init_flowLogger();
 
 // lib/v3/agent/utils/actionMapping.ts
+var EXCLUDED_OUTPUT_KEYS = ["screenshotBase64"];
+function stripExcludedKeys(output) {
+  const result = {};
+  for (const [key, value] of Object.entries(output)) {
+    if (!EXCLUDED_OUTPUT_KEYS.includes(
+      key
+    )) {
+      result[key] = value;
+    }
+  }
+  return result;
+}
 function mapToolResultToActions({
   toolCallName,
   toolResult,
@@ -32416,14 +32662,112 @@ function createStandardAction(toolCallName, toolResult, args, reasoning) {
     return action;
   }
   if (toolCallName !== "ariaTree" && toolResult) {
-    const { output } = toolResult;
-    Object.assign(action, output);
+    const result = toolResult;
+    const output = result.output;
+    if (output && typeof output === "object" && !Array.isArray(output)) {
+      const cleanedOutput = stripExcludedKeys(
+        output
+      );
+      Object.assign(action, cleanedOutput);
+    }
   }
   return action;
 }
 
 // lib/v3/handlers/v3AgentHandler.ts
 init_sdkErrors();
+
+// lib/v3/agent/utils/handleCloseToolCall.ts
+var import_ai18 = require("ai");
+var import_zod22 = require("zod");
+var import_ai19 = require("ai");
+var baseCloseSchema = import_zod22.z.object({
+  reasoning: import_zod22.z.string().describe("Brief summary of what actions were taken and the outcome"),
+  taskComplete: import_zod22.z.boolean().describe("true if the task was fully completed, false otherwise")
+});
+function handleCloseToolCall(options) {
+  return __async(this, null, function* () {
+    var _a4;
+    const { model, inputMessages, instruction, outputSchema, logger } = options;
+    logger({
+      category: "agent",
+      message: "Agent calling tool: close",
+      level: 1
+    });
+    const closeToolSchema = outputSchema ? baseCloseSchema.extend({
+      output: outputSchema.describe(
+        "The specific data the user requested from this task"
+      )
+    }) : baseCloseSchema;
+    const outputInstructions = outputSchema ? `
+
+The user also requested the following information from this task. Provide it in the "output" field:
+${JSON.stringify(
+      Object.fromEntries(
+        Object.entries(outputSchema.shape).map(([key, value]) => [
+          key,
+          value.description || "no description"
+        ])
+      ),
+      null,
+      2
+    )}` : "";
+    const systemPrompt = `You are a web automation assistant that was tasked with completing a task.
+
+The task was:
+"${instruction}"
+
+Review what was accomplished and provide your final assessment in whether the task was completed successfully. you have been provided with the history of the actions taken so far, use this to determine if the task was completed successfully.${outputInstructions}
+
+Call the "close" tool with:
+1. A brief summary of what was done
+2. Whether the task was completed successfully${outputSchema ? "\n3. The requested output data based on what you found" : ""}`;
+    const closeTool = (0, import_ai19.tool)({
+      description: outputSchema ? "Complete the task with your assessment and the requested output data." : "Complete the task with your final assessment.",
+      inputSchema: closeToolSchema,
+      execute: (params) => __async(null, null, function* () {
+        return __spreadValues({ success: true }, params);
+      })
+    });
+    const userPrompt = {
+      role: "user",
+      content: outputSchema ? "Provide your final assessment and the requested output data." : "Provide your final assessment."
+    };
+    const result = yield (0, import_ai18.generateText)({
+      model,
+      system: systemPrompt,
+      messages: [...inputMessages, userPrompt],
+      tools: { close: closeTool },
+      toolChoice: { type: "tool", toolName: "close" }
+    });
+    const closeToolCall = result.toolCalls.find((tc) => tc.toolName === "close");
+    const outputMessages = [
+      userPrompt,
+      ...((_a4 = result.response) == null ? void 0 : _a4.messages) || []
+    ];
+    if (!closeToolCall) {
+      return {
+        reasoning: result.text || "Task execution completed",
+        taskComplete: false,
+        messages: outputMessages
+      };
+    }
+    const input = closeToolCall.input;
+    logger({
+      category: "agent",
+      message: `Task completed`,
+      level: 1
+    });
+    return {
+      reasoning: input.reasoning,
+      taskComplete: input.taskComplete,
+      messages: outputMessages,
+      output: input.output
+    };
+  });
+}
+
+// lib/v3/handlers/v3AgentHandler.ts
 function getErrorMessage(error) {
   return error instanceof Error ? error.message : String(error);
 }
@@ -32459,14 +32803,9 @@ var V3AgentHandler = class {
           throw new MissingLLMConfigurationError();
         }
         const baseModel = this.llmClient.getLanguageModel();
-        const wrappedModel = (0, import_ai19.wrapLanguageModel)({
+        const wrappedModel = (0, import_ai20.wrapLanguageModel)({
           model: baseModel,
-          middleware: __spreadValues({
-            transformParams: (_0) => __async(this, [_0], function* ({ params }) {
-              const { processedPrompt } = processMessages(params);
-              return __spreadProps(__spreadValues({}, params), { prompt: processedPrompt });
-            })
-          }, SessionFileLogger.createLlmLoggingMiddleware(baseModel.modelId))
+          middleware: __spreadValues({}, SessionFileLogger.createLlmLoggingMiddleware(baseModel.modelId))
         });
         return {
           options,
@@ -32487,6 +32826,15 @@ var V3AgentHandler = class {
       }
     });
   }
+  createPrepareStep(userCallback) {
+    return (options) => __async(null, null, function* () {
+      processMessages(options.messages);
+      if (userCallback) {
+        return userCallback(options);
+      }
+      return options;
+    });
+  }
   createStepHandler(state, userCallback) {
     return (event) => __async(this, null, function* () {
       var _a4;
@@ -32548,7 +32896,7 @@ var V3AgentHandler = class {
   }
   execute(instructionOrOptions) {
     return __async(this, null, function* () {
-      var _a4;
+      var _a4, _b;
       const startTime = Date.now();
       const options = typeof instructionOrOptions === "object" ? instructionOrOptions : null;
       const signal = options == null ? void 0 : options.signal;
@@ -32601,7 +32949,7 @@ var V3AgentHandler = class {
           stopWhen: (result2) => this.handleStop(result2, maxSteps),
           temperature: 1,
           toolChoice: "auto",
-          prepareStep: callbacks == null ? void 0 : callbacks.prepareStep,
+          prepareStep: this.createPrepareStep(callbacks == null ? void 0 : callbacks.prepareStep),
           onStepFinish: this.createStepHandler(state, callbacks == null ? void 0 : callbacks.onStepFinish),
           abortSignal: preparedOptions.signal,
           providerOptions: wrappedModel.modelId.includes("gemini-3") ? {
@@ -32610,11 +32958,22 @@ var V3AgentHandler = class {
             }
           } : void 0
         });
+        const allMessages = [...messages, ...((_b = result.response) == null ? void 0 : _b.messages) || []];
+        const closeResult = yield this.ensureClosed(
+          state,
+          wrappedModel,
+          allMessages,
+          preparedOptions.instruction,
+          preparedOptions.output,
+          this.logger
+        );
         return this.consolidateMetricsAndResult(
           startTime,
           state,
-          messages,
-          result
+          closeResult.messages,
+          result,
+          maxSteps,
+          closeResult.output
         );
       } catch (error) {
         if (error instanceof StreamingCallbacksInNonStreamingModeError) {
@@ -32691,7 +33050,7 @@ var V3AgentHandler = class {
         stopWhen: (result) => this.handleStop(result, maxSteps),
         temperature: 1,
         toolChoice: "auto",
-        prepareStep: callbacks == null ? void 0 : callbacks.prepareStep,
+        prepareStep: this.createPrepareStep(callbacks == null ? void 0 : callbacks.prepareStep),
         onStepFinish: this.createStepHandler(state, callbacks == null ? void 0 : callbacks.onStepFinish),
         onError: (event) => {
           if (callbacks == null ? void 0 : callbacks.onError) {
@@ -32701,16 +33060,29 @@ var V3AgentHandler = class {
         },
         onChunk: callbacks == null ? void 0 : callbacks.onChunk,
         onFinish: (event) => {
+          var _a5;
           if (callbacks == null ? void 0 : callbacks.onFinish) {
             callbacks.onFinish(event);
           }
-          const result = this.consolidateMetricsAndResult(
-            startTime,
+          const allMessages = [...messages, ...((_a5 = event.response) == null ? void 0 : _a5.messages) || []];
+          this.ensureClosed(
             state,
-            messages,
-            event
-          );
-          resolveResult(result);
+            wrappedModel,
+            allMessages,
+            options.instruction,
+            options.output,
+            this.logger
+          ).then((closeResult) => {
+            const result = this.consolidateMetricsAndResult(
+              startTime,
+              state,
+              closeResult.messages,
+              event,
+              maxSteps,
+              closeResult.output
+            );
+            resolveResult(result);
+          });
         },
         onAbort: (event) => {
           var _a5;
@@ -32732,11 +33104,20 @@ var V3AgentHandler = class {
       return agentStreamResult;
     });
   }
-  consolidateMetricsAndResult(startTime, state, inputMessages, result) {
+  consolidateMetricsAndResult(startTime, state, inputMessages, result, maxSteps, output) {
     var _a4;
     if (!state.finalMessage) {
       const allReasoning = state.collectedReasoning.join(" ").trim();
-      state.finalMessage = allReasoning || result.text || "";
+      if (!state.completed && maxSteps && ((_a4 = result.steps) == null ? void 0 : _a4.length) >= maxSteps) {
+        this.logger({
+          category: "agent",
+          message: `Agent stopped: reached maximum steps (${maxSteps})`,
+          level: 1
+        });
+        state.finalMessage = `Agent stopped: reached maximum steps (${maxSteps})`;
+      } else {
+        state.finalMessage = allReasoning || result.text || "";
+      }
     }
     const endTime = Date.now();
     const inferenceTimeMs = endTime - startTime;
@@ -32750,16 +33131,12 @@ var V3AgentHandler = class {
         inferenceTimeMs
       );
     }
-    const responseMessages = ((_a4 = result.response) == null ? void 0 : _a4.messages) || [];
-    const fullMessages = [
-      ...inputMessages,
-      ...responseMessages
-    ];
     return {
       success: state.completed,
       message: state.finalMessage || "Task execution completed",
       actions: state.actions,
       completed: state.completed,
+      output,
       usage: result.usage ? {
         input_tokens: result.usage.inputTokens || 0,
         output_tokens: result.usage.outputTokens || 0,
@@ -32767,7 +33144,7 @@ var V3AgentHandler = class {
         cached_input_tokens: result.usage.cachedInputTokens || 0,
         inference_time_ms: inferenceTimeMs
       } : void 0,
-      messages: fullMessages
+      messages: inputMessages
     };
   }
   createTools(excludeTools) {
@@ -32787,7 +33164,47 @@ var V3AgentHandler = class {
     if ((_a4 = lastStep == null ? void 0 : lastStep.toolCalls) == null ? void 0 : _a4.some((tc) => tc.toolName === "close")) {
       return true;
     }
-    return (0, import_ai19.stepCountIs)(maxSteps)(result);
+    return (0, import_ai20.stepCountIs)(maxSteps)(result);
+  }
+  /**
+   * Ensures the close tool is called at the end of agent execution.
+   * Returns the messages and any extracted output from the close call.
+   */
+  ensureClosed(state, model, messages, instruction, outputSchema, logger) {
+    return __async(this, null, function* () {
+      if (state.completed) return { messages };
+      const closeResult = yield handleCloseToolCall({
+        model,
+        inputMessages: messages,
+        instruction,
+        outputSchema,
+        logger
+      });
+      state.completed = closeResult.taskComplete;
+      state.finalMessage = closeResult.reasoning;
+      const closeAction = mapToolResultToActions({
+        toolCallName: "close",
+        toolResult: {
+          success: true,
+          reasoning: closeResult.reasoning,
+          taskComplete: closeResult.taskComplete
+        },
+        args: {
+          reasoning: closeResult.reasoning,
+          taskComplete: closeResult.taskComplete
+        },
+        reasoning: closeResult.reasoning
+      });
+      for (const action of closeAction) {
+        action.pageUrl = state.currentPageUrl;
+        action.timestamp = Date.now();
+        state.actions.push(action);
+      }
+      return {
+        messages: [...messages, ...closeResult.messages],
+        output: closeResult.output
+      };
+    });
   }
   /**
    * Capture a screenshot and emit it via the event bus
@@ -33269,8 +33686,8 @@ var AnthropicCUAClient = class extends AgentClient {
           betas: ["computer-use-2025-01-24"]
         };
         if (this.tools && Object.keys(this.tools).length > 0) {
-          const customTools = Object.entries(this.tools).map(([name, tool21]) => {
-            const schema = tool21.inputSchema;
+          const customTools = Object.entries(this.tools).map(([name, tool22]) => {
+            const schema = tool22.inputSchema;
             const jsonSchema3 = toJsonSchema(schema);
             const inputSchema = {
               type: "object",
@@ -33279,7 +33696,7 @@ var AnthropicCUAClient = class extends AgentClient {
             };
             return {
               name,
-              description: tool21.description,
+              description: tool22.description,
               input_schema: inputSchema
             };
           });
@@ -33397,13 +33814,13 @@ var AnthropicCUAClient = class extends AgentClient {
             let toolResult = "Tool executed successfully";
             if (this.tools && item.name in this.tools) {
               try {
-                const tool21 = this.tools[item.name];
+                const tool22 = this.tools[item.name];
                 logger({
                   category: "agent",
                   message: `Executing tool call: ${item.name} with args: ${JSON.stringify(item.input)}`,
                   level: 1
                 });
-                const result = yield tool21.execute(item.input, {
+                const result = yield tool22.execute(item.input, {
                   toolCallId: item.id,
                   messages: []
                 });
@@ -33674,7 +34091,6 @@ var OpenAICUAClient = class extends AgentClient {
   constructor(type, modelName, userProvidedInstructions, clientOptions, tools) {
     super(type, modelName, userProvidedInstructions);
     this.currentViewport = { width: 1288, height: 711 };
-    this.actualScreenshotSize = { width: 1288, height: 711 };
     this.reasoningItems = /* @__PURE__ */ new Map();
     this.environment = "browser";
     this.apiKey = (clientOptions == null ? void 0 : clientOptions.apiKey) || process.env.OPENAI_API_KEY || "";
@@ -33695,9 +34111,6 @@ var OpenAICUAClient = class extends AgentClient {
   setViewport(width, height) {
     this.currentViewport = { width, height };
   }
-  setScreenshotSize(width, height) {
-    this.actualScreenshotSize = { width, height };
-  }
   setCurrentUrl(url) {
     this.currentUrl = url;
   }
@@ -33965,13 +34378,13 @@ var OpenAICUAClient = class extends AgentClient {
           truncation: "auto"
         };
         if (this.tools && Object.keys(this.tools).length > 0) {
-          const customTools = Object.entries(this.tools).map(([name, tool21]) => ({
+          const customTools = Object.entries(this.tools).map(([name, tool22]) => ({
             type: "function",
             name,
             function: {
               name,
-              description: tool21.description,
-              parameters: tool21.inputSchema
+              description: tool22.description,
+              parameters: tool22.inputSchema
             }
           }));
           requestParams.tools = [
@@ -34123,14 +34536,14 @@ var OpenAICUAClient = class extends AgentClient {
             let toolResult = "Tool executed successfully";
             if (this.tools && item.name in this.tools) {
               try {
-                const tool21 = this.tools[item.name];
+                const tool22 = this.tools[item.name];
                 const args = JSON.parse(item.arguments);
                 logger({
                   category: "agent",
                   message: `Executing tool call: ${item.name} with args: ${item.arguments}`,
                   level: 1
                 });
-                const result = yield tool21.execute(args, {
+                const result = yield tool22.execute(args, {
                   toolCallId: item.call_id,
                   messages: []
                 });
@@ -34180,16 +34593,9 @@ var OpenAICUAClient = class extends AgentClient {
   }
   convertComputerCallToAction(call) {
     const { action } = call;
-    const scaledAction = __spreadValues({}, action);
-    if (action.x !== void 0 && action.y !== void 0) {
-      const scaleX = this.currentViewport.width / this.actualScreenshotSize.width;
-      const scaleY = this.currentViewport.height / this.actualScreenshotSize.height;
-      scaledAction.x = Math.floor(action.x * scaleX);
-      scaledAction.y = Math.floor(action.y * scaleY);
-    }
     return __spreadValues({
       type: action.type
-    }, scaledAction);
+    }, action);
   }
   convertFunctionCallToAction(call) {
     try {
@@ -34293,8 +34699,8 @@ function executeGoogleCustomTool(toolName, toolArgs, tools, functionCall, logger
         message: `Executing custom tool: ${toolName} with args: ${JSON.stringify(toolArgs)}`,
         level: 1
       });
-      const tool21 = tools[toolName];
-      const toolResult = yield tool21.execute(toolArgs, {
+      const tool22 = tools[toolName];
+      const toolResult = yield tool22.execute(toolArgs, {
         toolCallId: `tool_${Date.now()}`,
         messages: []
       });
@@ -34342,22 +34748,22 @@ function isCustomTool(functionCall, tools) {
 }
 function convertToolSetToFunctionDeclarations(tools) {
   const functionDeclarations = [];
-  for (const [name, tool21] of Object.entries(tools)) {
-    const functionDeclaration = convertToolToFunctionDeclaration(name, tool21);
+  for (const [name, tool22] of Object.entries(tools)) {
+    const functionDeclaration = convertToolToFunctionDeclaration(name, tool22);
     if (functionDeclaration) {
       functionDeclarations.push(functionDeclaration);
     }
   }
   return functionDeclarations;
 }
-function convertToolToFunctionDeclaration(name, tool21) {
+function convertToolToFunctionDeclaration(name, tool22) {
   try {
-    const schema = tool21.inputSchema;
+    const schema = tool22.inputSchema;
     const jsonSchema3 = toJsonSchema(schema);
     const parameters = convertJsonSchemaToGoogleParameters(jsonSchema3);
     return {
       name,
-      description: tool21.description || `Execute ${name}`,
+      description: tool22.description || `Execute ${name}`,
       parameters
     };
   } catch (error) {
@@ -34408,7 +34814,6 @@ var GoogleCUAClient = class extends AgentClient {
   constructor(type, modelName, userProvidedInstructions, clientOptions, tools) {
     super(type, modelName, userProvidedInstructions);
     this.currentViewport = { width: 1288, height: 711 };
-    this.actualScreenshotSize = { width: 1288, height: 711 };
     this.history = [];
     this.environment = "ENVIRONMENT_BROWSER";
     this.tools = tools;
@@ -34447,9 +34852,6 @@ var GoogleCUAClient = class extends AgentClient {
   setViewport(width, height) {
     this.currentViewport = { width, height };
   }
-  setScreenshotSize(width, height) {
-    this.actualScreenshotSize = { width, height };
-  }
   setCurrentUrl(url) {
     this.currentUrl = url;
   }
@@ -35011,9 +35413,26 @@ var GoogleCUAClient = class extends AgentClient {
       }
       case "scroll_document": {
         const direction = args.direction.toLowerCase();
+        const magnitude = typeof args.magnitude === "number" ? args.magnitude : 800;
+        let scroll_x = 0;
+        let scroll_y = 0;
+        if (direction === "up") {
+          scroll_y = -magnitude;
+        } else if (direction === "down") {
+          scroll_y = magnitude;
+        } else if (direction === "left") {
+          scroll_x = -magnitude;
+        } else if (direction === "right") {
+          scroll_x = magnitude;
+        } else {
+          scroll_y = magnitude;
+        }
         return {
-          type: "keypress",
-          keys: [direction === "up" ? "PageUp" : "PageDown"]
+          type: "scroll",
+          x: 0,
+          y: 0,
+          scroll_x,
+          scroll_y
         };
       }
       case "scroll_at": {
@@ -35111,18 +35530,17 @@ var GoogleCUAClient = class extends AgentClient {
     }
   }
   /**
-   * Normalize coordinates from Google's 0-1000 range to actual viewport dimensions
+   * Normalize coordinates from Google's 0-1000 range to fixed viewport dimensions.
+   * Google CUA outputs coordinates in 0-1000 range based on the screenshot it sees,
+   * which is always 1288x711 regardless of browser-reported viewport size.
    */
   normalizeCoordinates(x2, y) {
+    const GOOGLE_CUA_VIEWPORT = { width: 1288, height: 711 };
     x2 = Math.min(999, Math.max(0, x2));
     y = Math.min(999, Math.max(0, y));
-    const screenshotX = x2 / 1e3 * this.actualScreenshotSize.width;
-    const screenshotY = y / 1e3 * this.actualScreenshotSize.height;
-    const scaleX = this.currentViewport.width / this.actualScreenshotSize.width;
-    const scaleY = this.currentViewport.height / this.actualScreenshotSize.height;
     return {
-      x: Math.floor(screenshotX * scaleX),
-      y: Math.floor(screenshotY * scaleY)
+      x: Math.floor(x2 / 1e3 * GOOGLE_CUA_VIEWPORT.width),
+      y: Math.floor(y / 1e3 * GOOGLE_CUA_VIEWPORT.height)
     };
   }
   captureScreenshot(options) {
@@ -35173,8 +35591,6 @@ var MicrosoftCUAClient = class extends AgentClient {
     };
     // Resized dimensions for model input
     this.resizedViewport = { width: 1288, height: 711 };
-    // Actual screenshot dimensions (tracked separately from viewport)
-    this.actualScreenshotSize = { width: 1288, height: 711 };
     this.apiKey = (clientOptions == null ? void 0 : clientOptions.apiKey) || process.env.AZURE_API_KEY || process.env.FIREWORKS_API_KEY || "";
     this.baseURL = (clientOptions == null ? void 0 : clientOptions.baseURL) || process.env.AZURE_ENDPOINT || process.env.FIREWORKS_ENDPOINT || "";
     this.clientOptions = {
@@ -35201,9 +35617,6 @@ var MicrosoftCUAClient = class extends AgentClient {
     this.currentViewport = { width, height };
     this.resizedViewport = this.smartResize(width, height);
   }
-  setScreenshotSize(width, height) {
-    this.actualScreenshotSize = { width, height };
-  }
   setCurrentUrl(url) {
     this.currentUrl = url;
   }
@@ -35241,7 +35654,7 @@ var MicrosoftCUAClient = class extends AgentClient {
    * Simplified to match Python's minimal approach
    */
   generateSystemPrompt() {
-    const { width, height } = this.actualScreenshotSize;
+    const { width, height } = this.resizedViewport;
     let basePrompt = "You are a helpful assistant.";
     if (this.userProvidedInstructions) {
       basePrompt = `${basePrompt}
@@ -35402,8 +35815,8 @@ ${functionCallTemplate}`;
     const transformCoordinate = (coord) => {
       if (!coord || coord.length !== 2) return coord;
       const [x2, y] = coord;
-      const scaleX = this.currentViewport.width / this.actualScreenshotSize.width;
-      const scaleY = this.currentViewport.height / this.actualScreenshotSize.height;
+      const scaleX = this.currentViewport.width / this.resizedViewport.width;
+      const scaleY = this.currentViewport.height / this.resizedViewport.height;
       return [Math.round(x2 * scaleX), Math.round(y * scaleY)];
     };
     const baseAction = {
@@ -35831,6 +36244,7 @@ var modelToAgentProviderMap = {
   "claude-opus-4-5-20251101": "anthropic",
   "claude-haiku-4-5-20251001": "anthropic",
   "gemini-2.5-computer-use-preview-10-2025": "google",
+  "gemini-3-flash-computer-use": "google",
   "fara-7b": "microsoft"
 };
 var AgentProvider = class _AgentProvider {
@@ -35912,14 +36326,6 @@ var AgentProvider = class _AgentProvider {
 // lib/v3/handlers/v3CuaAgentHandler.ts
 init_flowLogger();
 init_sdkErrors();
-function getPNGDimensions(buffer) {
-  if (buffer.length < 24 || buffer[0] !== 137 || buffer[1] !== 80 || buffer[2] !== 78 || buffer[3] !== 71) {
-    throw new Error("Invalid PNG file");
-  }
-  const width = buffer.readUInt32BE(16);
-  const height = buffer.readUInt32BE(20);
-  return { width, height };
-}
 var V3CuaAgentHandler = class {
   constructor(v3, logger, options, tools) {
     this.v3 = v3;
@@ -35950,21 +36356,6 @@ var V3CuaAgentHandler = class {
       this.ensureNotClosed();
       const page = yield this.v3.context.awaitActivePage();
       const screenshotBuffer = yield page.screenshot({ fullPage: false });
-      if (this.agentClient instanceof GoogleCUAClient || this.agentClient instanceof OpenAICUAClient || this.agentClient instanceof MicrosoftCUAClient) {
-        try {
-          const dimensions = getPNGDimensions(screenshotBuffer);
-          this.agentClient.setScreenshotSize(
-            dimensions.width,
-            dimensions.height
-          );
-        } catch (e2) {
-          this.logger({
-            category: "agent",
-            message: `Could not read screenshot dimensions: ${e2}`,
-            level: 1
-          });
-        }
-      }
       return screenshotBuffer.toString("base64");
     }));
     this.agentClient.setActionHandler((action) => __async(this, null, function* () {
@@ -36418,21 +36809,6 @@ var V3CuaAgentHandler = class {
       try {
         const page = yield this.v3.context.awaitActivePage();
         const screenshotBuffer = yield page.screenshot({ fullPage: false });
-        if (this.agentClient instanceof GoogleCUAClient || this.agentClient instanceof OpenAICUAClient || this.agentClient instanceof MicrosoftCUAClient) {
-          try {
-            const dimensions = getPNGDimensions(screenshotBuffer);
-            this.agentClient.setScreenshotSize(
-              dimensions.width,
-              dimensions.height
-            );
-          } catch (e2) {
-            this.logger({
-              category: "agent",
-              message: `Could not read screenshot dimensions: ${e2}`,
-              level: 1
-            });
-          }
-        }
         this.v3.bus.emit("agent_screenshot_taken_event", screenshotBuffer);
         const currentUrl = page.url();
         return yield this.agentClient.captureScreenshot({
@@ -37609,23 +37985,23 @@ function waitForWebSocketDebuggerUrl(port, timeoutMs) {
 init_sdkErrors();
 
 // lib/v3/llm/aisdk.ts
-var import_ai21 = require("ai");
+var import_ai22 = require("ai");
 var import_uuid5 = require("uuid");
 
 // lib/v3/llm/LLMClient.ts
-var import_ai20 = require("ai");
+var import_ai21 = require("ai");
 var AnnotatedScreenshotText = "This is a screenshot of the current page state with the elements annotated on it. Each element id is annotated with a number to the top left of it. Duplicate annotations at the same location are under each other vertically.";
 var LLMClient = class {
   constructor(modelName, userProvidedInstructions) {
-    this.generateObject = import_ai20.generateObject;
-    this.generateText = import_ai20.generateText;
-    this.streamText = import_ai20.streamText;
-    this.streamObject = import_ai20.streamObject;
-    this.generateImage = import_ai20.experimental_generateImage;
-    this.embed = import_ai20.embed;
-    this.embedMany = import_ai20.embedMany;
-    this.transcribe = import_ai20.experimental_transcribe;
-    this.generateSpeech = import_ai20.experimental_generateSpeech;
+    this.generateObject = import_ai21.generateObject;
+    this.generateText = import_ai21.generateText;
+    this.streamText = import_ai21.streamText;
+    this.streamObject = import_ai21.streamObject;
+    this.generateImage = import_ai21.experimental_generateImage;
+    this.embed = import_ai21.embed;
+    this.embedMany = import_ai21.embedMany;
+    this.transcribe = import_ai21.experimental_transcribe;
+    this.generateSpeech = import_ai21.experimental_generateSpeech;
     this.modelName = modelName;
     this.userProvidedInstructions = userProvidedInstructions;
   }
@@ -37750,7 +38126,7 @@ You must respond in JSON format. respond WITH JSON. Do not include any other tex
           });
         }
         try {
-          objectResponse = yield (0, import_ai21.generateObject)({
+          objectResponse = yield (0, import_ai22.generateObject)({
             model: this.model,
             messages: formattedMessages,
             schema: options.response_model.schema,
@@ -37770,7 +38146,7 @@ You must respond in JSON format. respond WITH JSON. Do not include any other tex
             operation: "generateObject",
             output: `[error: ${err instanceof Error ? err.message : "unknown"}]`
           });
-          if (import_ai21.NoObjectGeneratedError.isInstance(err)) {
+          if (import_ai22.NoObjectGeneratedError.isInstance(err)) {
             (_g = this.logger) == null ? void 0 : _g.call(this, {
               category: "AISDK error",
               message: err.message,
@@ -37848,10 +38224,10 @@ You must respond in JSON format. respond WITH JSON. Do not include any other tex
       }
       const tools = {};
       if (options.tools && options.tools.length > 0) {
-        for (const tool21 of options.tools) {
-          tools[tool21.name] = {
-            description: tool21.description,
-            inputSchema: tool21.parameters
+        for (const tool22 of options.tools) {
+          tools[tool22.name] = {
+            description: tool22.description,
+            inputSchema: tool22.parameters
           };
         }
       }
@@ -37868,7 +38244,7 @@ You must respond in JSON format. respond WITH JSON. Do not include any other tex
       });
       let textResponse;
       try {
-        textResponse = yield (0, import_ai21.generateText)({
+        textResponse = yield (0, import_ai22.generateText)({
           model: this.model,
           messages: formattedMessages,
           tools: Object.keys(tools).length > 0 ? tools : void 0,
@@ -38050,14 +38426,14 @@ var AnthropicClient = class extends LLMClient {
         }
         formattedMessages.push(screenshotMessage);
       }
-      let anthropicTools2 = (_a4 = options.tools) == null ? void 0 : _a4.map((tool21) => {
+      let anthropicTools2 = (_a4 = options.tools) == null ? void 0 : _a4.map((tool22) => {
         return {
-          name: tool21.name,
-          description: tool21.description,
+          name: tool22.name,
+          description: tool22.description,
           input_schema: {
             type: "object",
-            properties: tool21.parameters.properties,
-            required: tool21.parameters.required
+            properties: tool22.parameters.properties,
+            required: tool22.parameters.required
           }
         };
       });
@@ -38244,15 +38620,15 @@ var CerebrasClient = class extends LLMClient {
           return __spreadProps(__spreadValues({}, baseMessage), { role: "user" });
         }
       });
-      let tools = (_a4 = options.tools) == null ? void 0 : _a4.map((tool21) => ({
+      let tools = (_a4 = options.tools) == null ? void 0 : _a4.map((tool22) => ({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
+          name: tool22.name,
+          description: tool22.description,
           parameters: {
             type: "object",
-            properties: tool21.parameters.properties,
-            required: tool21.parameters.required
+            properties: tool22.parameters.properties,
+            required: tool22.parameters.required
           }
         }
       }));
@@ -38545,18 +38921,18 @@ ${firstPartText.text}`;
     }
     return [
       {
-        functionDeclarations: tools.map((tool21) => {
+        functionDeclarations: tools.map((tool22) => {
           let parameters = void 0;
-          if (tool21.parameters) {
+          if (tool22.parameters) {
             parameters = {
               type: import_genai4.Type.OBJECT,
-              properties: tool21.parameters.properties,
-              required: tool21.parameters.required
+              properties: tool22.parameters.properties,
+              required: tool22.parameters.required
             };
           }
           return {
-            name: tool21.name,
-            description: tool21.description,
+            name: tool22.name,
+            description: tool22.description,
             parameters
           };
         })
@@ -38832,15 +39208,15 @@ var GroqClient = class extends LLMClient {
           return __spreadProps(__spreadValues({}, baseMessage), { role: "user" });
         }
       });
-      let tools = (_a4 = options.tools) == null ? void 0 : _a4.map((tool21) => ({
+      let tools = (_a4 = options.tools) == null ? void 0 : _a4.map((tool22) => ({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
+          name: tool22.name,
+          description: tool22.description,
           parameters: {
             type: "object",
-            properties: tool21.parameters.properties,
-            required: tool21.parameters.required
+            properties: tool22.parameters.properties,
+            required: tool22.parameters.required
           }
         }
       }));
@@ -39217,11 +39593,11 @@ ${parsedSchema}
         messages: formattedMessages,
         response_format: responseFormat,
         stream: false,
-        tools: (_e = options.tools) == null ? void 0 : _e.map((tool21) => ({
+        tools: (_e = options.tools) == null ? void 0 : _e.map((tool22) => ({
           function: {
-            name: tool21.name,
-            description: tool21.description,
-            parameters: tool21.parameters
+            name: tool22.name,
+            description: tool22.description,
+            parameters: tool22.parameters
           },
           type: "function"
         }))
@@ -41887,21 +42263,21 @@ function prepareChatTools({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const openaiTools2 = [];
-  for (const tool21 of tools) {
-    switch (tool21.type) {
+  for (const tool22 of tools) {
+    switch (tool22.type) {
       case "function":
         openaiTools2.push({
           type: "function",
           function: {
-            name: tool21.name,
-            description: tool21.description,
-            parameters: tool21.inputSchema,
+            name: tool22.name,
+            description: tool22.description,
+            parameters: tool22.inputSchema,
             strict: structuredOutputs ? strictJsonSchema : void 0
           }
         });
         break;
       default:
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
         break;
     }
   }
@@ -44217,22 +44593,22 @@ function prepareResponsesTools(_0) {
       return { tools: void 0, toolChoice: void 0, toolWarnings };
     }
     const openaiTools2 = [];
-    for (const tool21 of tools) {
-      switch (tool21.type) {
+    for (const tool22 of tools) {
+      switch (tool22.type) {
         case "function":
           openaiTools2.push({
             type: "function",
-            name: tool21.name,
-            description: tool21.description,
-            parameters: tool21.inputSchema,
+            name: tool22.name,
+            description: tool22.description,
+            parameters: tool22.inputSchema,
             strict: strictJsonSchema
           });
           break;
         case "provider-defined": {
-          switch (tool21.id) {
+          switch (tool22.id) {
             case "openai.file_search": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: fileSearchArgsSchema
               });
               openaiTools2.push({
@@ -44255,7 +44631,7 @@ function prepareResponsesTools(_0) {
             }
             case "openai.web_search_preview": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webSearchPreviewArgsSchema
               });
               openaiTools2.push({
@@ -44267,7 +44643,7 @@ function prepareResponsesTools(_0) {
             }
             case "openai.web_search": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webSearchArgsSchema
               });
               openaiTools2.push({
@@ -44280,7 +44656,7 @@ function prepareResponsesTools(_0) {
             }
             case "openai.code_interpreter": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: codeInterpreterArgsSchema
               });
               openaiTools2.push({
@@ -44291,7 +44667,7 @@ function prepareResponsesTools(_0) {
             }
             case "openai.image_generation": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: imageGenerationArgsSchema
               });
               openaiTools2.push({
@@ -44315,7 +44691,7 @@ function prepareResponsesTools(_0) {
           break;
         }
         default:
-          toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+          toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
           break;
       }
     }
@@ -44420,7 +44796,7 @@ var OpenAIResponsesLanguageModel = class {
       }
       function hasOpenAITool(id) {
         return (tools == null ? void 0 : tools.find(
-          (tool21) => tool21.type === "provider-defined" && tool21.id === id
+          (tool22) => tool22.type === "provider-defined" && tool22.id === id
         )) != null;
       }
       const topLogprobs = typeof (openaiOptions == null ? void 0 : openaiOptions.logprobs) === "number" ? openaiOptions == null ? void 0 : openaiOptions.logprobs : (openaiOptions == null ? void 0 : openaiOptions.logprobs) === true ? TOP_LOGPROBS_MAX : void 0;
@@ -44428,7 +44804,7 @@ var OpenAIResponsesLanguageModel = class {
         addInclude("message.output_text.logprobs");
       }
       const webSearchToolName = (_c = tools == null ? void 0 : tools.find(
-        (tool21) => tool21.type === "provider-defined" && (tool21.id === "openai.web_search" || tool21.id === "openai.web_search_preview")
+        (tool22) => tool22.type === "provider-defined" && (tool22.id === "openai.web_search" || tool22.id === "openai.web_search_preview")
       )) == null ? void 0 : _c.name;
       if (webSearchToolName) {
         addInclude("web_search_call.action.sources");
@@ -48022,25 +48398,25 @@ function prepareTools({
   if (tools == null) {
     return { tools: void 0, toolConfig: void 0, toolWarnings };
   }
-  const hasFunctionTools = tools.some((tool21) => tool21.type === "function");
+  const hasFunctionTools = tools.some((tool22) => tool22.type === "function");
   const hasProviderDefinedTools = tools.some(
-    (tool21) => tool21.type === "provider-defined"
+    (tool22) => tool22.type === "provider-defined"
   );
   if (hasFunctionTools && hasProviderDefinedTools) {
-    const functionTools = tools.filter((tool21) => tool21.type === "function");
+    const functionTools = tools.filter((tool22) => tool22.type === "function");
     toolWarnings.push({
       type: "unsupported-tool",
-      tool: tools.find((tool21) => tool21.type === "function"),
+      tool: tools.find((tool22) => tool22.type === "function"),
       details: `Cannot mix function tools with provider-defined tools in the same request. Falling back to provider-defined tools only. The following function tools will be ignored: ${functionTools.map((t2) => t2.name).join(", ")}. Please use either function tools or provider-defined tools, but not both.`
     });
   }
   if (hasProviderDefinedTools) {
     const googleTools22 = [];
     const providerDefinedTools = tools.filter(
-      (tool21) => tool21.type === "provider-defined"
+      (tool22) => tool22.type === "provider-defined"
     );
-    providerDefinedTools.forEach((tool21) => {
-      switch (tool21.id) {
+    providerDefinedTools.forEach((tool22) => {
+      switch (tool22.id) {
         case "google.google_search":
           if (isGemini2orNewer) {
             googleTools22.push({ googleSearch: {} });
@@ -48048,8 +48424,8 @@ function prepareTools({
             googleTools22.push({
               googleSearchRetrieval: {
                 dynamicRetrievalConfig: {
-                  mode: tool21.args.mode,
-                  dynamicThreshold: tool21.args.dynamicThreshold
+                  mode: tool22.args.mode,
+                  dynamicThreshold: tool22.args.dynamicThreshold
                 }
               }
             });
@@ -48063,7 +48439,7 @@ function prepareTools({
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The URL context tool is not supported with other Gemini models than Gemini 2."
             });
           }
@@ -48074,18 +48450,18 @@ function prepareTools({
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The code execution tools is not supported with other Gemini models than Gemini 2."
             });
           }
           break;
         case "google.file_search":
           if (supportsFileSearch) {
-            googleTools22.push({ fileSearch: __spreadValues({}, tool21.args) });
+            googleTools22.push({ fileSearch: __spreadValues({}, tool22.args) });
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The file search tool is only supported with Gemini 2.5 models."
             });
           }
@@ -48096,22 +48472,22 @@ function prepareTools({
               retrieval: {
                 vertex_rag_store: {
                   rag_resources: {
-                    rag_corpus: tool21.args.ragCorpus
+                    rag_corpus: tool22.args.ragCorpus
                   },
-                  similarity_top_k: tool21.args.topK
+                  similarity_top_k: tool22.args.topK
                 }
               }
             });
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The RAG store tool is not supported with other Gemini models than Gemini 2."
             });
           }
           break;
         default:
-          toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+          toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
           break;
       }
     });
@@ -48122,17 +48498,17 @@ function prepareTools({
     };
   }
   const functionDeclarations = [];
-  for (const tool21 of tools) {
-    switch (tool21.type) {
+  for (const tool22 of tools) {
+    switch (tool22.type) {
       case "function":
         functionDeclarations.push({
-          name: tool21.name,
-          description: (_a4 = tool21.description) != null ? _a4 : "",
-          parameters: convertJSONSchemaToOpenAPISchema(tool21.inputSchema)
+          name: tool22.name,
+          description: (_a4 = tool22.description) != null ? _a4 : "",
+          parameters: convertJSONSchemaToOpenAPISchema(tool22.inputSchema)
         });
         break;
       default:
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
         break;
     }
   }
@@ -48246,7 +48622,7 @@ var GoogleGenerativeAILanguageModel = class {
         schema: googleGenerativeAIProviderOptions
       });
       if ((tools == null ? void 0 : tools.some(
-        (tool21) => tool21.type === "provider-defined" && tool21.id === "google.vertex_rag_store"
+        (tool22) => tool22.type === "provider-defined" && tool22.id === "google.vertex_rag_store"
       )) && !this.config.provider.startsWith("google.vertex.")) {
         warnings.push({
           type: "other",
@@ -50108,23 +50484,23 @@ function prepareTools2(_0) {
       return { tools: void 0, toolChoice: void 0, toolWarnings, betas };
     }
     const anthropicTools2 = [];
-    for (const tool21 of tools) {
-      switch (tool21.type) {
+    for (const tool22 of tools) {
+      switch (tool22.type) {
         case "function": {
-          const cacheControl = validator3.getCacheControl(tool21.providerOptions, {
+          const cacheControl = validator3.getCacheControl(tool22.providerOptions, {
             type: "tool definition",
             canCache: true
           });
           anthropicTools2.push({
-            name: tool21.name,
-            description: tool21.description,
-            input_schema: tool21.inputSchema,
+            name: tool22.name,
+            description: tool22.description,
+            input_schema: tool22.inputSchema,
             cache_control: cacheControl
           });
           break;
         }
         case "provider-defined": {
-          switch (tool21.id) {
+          switch (tool22.id) {
             case "anthropic.code_execution_20250522": {
               betas.add("code-execution-2025-05-22");
               anthropicTools2.push({
@@ -50147,9 +50523,9 @@ function prepareTools2(_0) {
               anthropicTools2.push({
                 name: "computer",
                 type: "computer_20250124",
-                display_width_px: tool21.args.displayWidthPx,
-                display_height_px: tool21.args.displayHeightPx,
-                display_number: tool21.args.displayNumber,
+                display_width_px: tool22.args.displayWidthPx,
+                display_height_px: tool22.args.displayHeightPx,
+                display_number: tool22.args.displayNumber,
                 cache_control: void 0
               });
               break;
@@ -50159,9 +50535,9 @@ function prepareTools2(_0) {
               anthropicTools2.push({
                 name: "computer",
                 type: "computer_20241022",
-                display_width_px: tool21.args.displayWidthPx,
-                display_height_px: tool21.args.displayHeightPx,
-                display_number: tool21.args.displayNumber,
+                display_width_px: tool22.args.displayWidthPx,
+                display_height_px: tool22.args.displayHeightPx,
+                display_number: tool22.args.displayNumber,
                 cache_control: void 0
               });
               break;
@@ -50195,7 +50571,7 @@ function prepareTools2(_0) {
             }
             case "anthropic.text_editor_20250728": {
               const args = yield validateTypes2({
-                value: tool21.args,
+                value: tool22.args,
                 schema: textEditor_20250728ArgsSchema
               });
               anthropicTools2.push({
@@ -50235,7 +50611,7 @@ function prepareTools2(_0) {
             case "anthropic.web_fetch_20250910": {
               betas.add("web-fetch-2025-09-10");
               const args = yield validateTypes2({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webFetch_20250910ArgsSchema
               });
               anthropicTools2.push({
@@ -50252,7 +50628,7 @@ function prepareTools2(_0) {
             }
             case "anthropic.web_search_20250305": {
               const args = yield validateTypes2({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webSearch_20250305ArgsSchema
               });
               anthropicTools2.push({
@@ -50267,14 +50643,14 @@ function prepareTools2(_0) {
               break;
             }
             default: {
-              toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+              toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
               break;
             }
           }
           break;
         }
         default: {
-          toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+          toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
           break;
         }
       }
@@ -51233,7 +51609,7 @@ var AnthropicMessagesLanguageModel = class {
         betas.add("skills-2025-10-02");
         betas.add("files-api-2025-04-14");
         if (!(tools == null ? void 0 : tools.some(
-          (tool21) => tool21.type === "provider-defined" && tool21.id === "anthropic.code_execution_20250825"
+          (tool22) => tool22.type === "provider-defined" && tool22.id === "anthropic.code_execution_20250825"
         ))) {
           warnings.push({
             type: "other",
@@ -53071,24 +53447,24 @@ function prepareTools3({
   if (tools == null) {
     return { tools: void 0, toolConfig: void 0, toolWarnings };
   }
-  const hasFunctionTools = tools.some((tool21) => tool21.type === "function");
+  const hasFunctionTools = tools.some((tool22) => tool22.type === "function");
   const hasProviderDefinedTools = tools.some(
-    (tool21) => tool21.type === "provider-defined"
+    (tool22) => tool22.type === "provider-defined"
   );
   if (hasFunctionTools && hasProviderDefinedTools) {
     toolWarnings.push({
       type: "unsupported-tool",
-      tool: tools.find((tool21) => tool21.type === "function"),
+      tool: tools.find((tool22) => tool22.type === "function"),
       details: "Cannot mix function tools with provider-defined tools in the same request. Please use either function tools or provider-defined tools, but not both."
     });
   }
   if (hasProviderDefinedTools) {
     const googleTools22 = [];
     const providerDefinedTools = tools.filter(
-      (tool21) => tool21.type === "provider-defined"
+      (tool22) => tool22.type === "provider-defined"
     );
-    providerDefinedTools.forEach((tool21) => {
-      switch (tool21.id) {
+    providerDefinedTools.forEach((tool22) => {
+      switch (tool22.id) {
         case "google.google_search":
           if (isGemini2) {
             googleTools22.push({ googleSearch: {} });
@@ -53096,8 +53472,8 @@ function prepareTools3({
             googleTools22.push({
               googleSearchRetrieval: {
                 dynamicRetrievalConfig: {
-                  mode: tool21.args.mode,
-                  dynamicThreshold: tool21.args.dynamicThreshold
+                  mode: tool22.args.mode,
+                  dynamicThreshold: tool22.args.dynamicThreshold
                 }
               }
             });
@@ -53111,7 +53487,7 @@ function prepareTools3({
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The URL context tool is not supported with other Gemini models than Gemini 2."
             });
           }
@@ -53122,13 +53498,13 @@ function prepareTools3({
           } else {
             toolWarnings.push({
               type: "unsupported-tool",
-              tool: tool21,
+              tool: tool22,
               details: "The code execution tools is not supported with other Gemini models than Gemini 2."
             });
           }
           break;
         default:
-          toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+          toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
           break;
       }
     });
@@ -53139,17 +53515,17 @@ function prepareTools3({
     };
   }
   const functionDeclarations = [];
-  for (const tool21 of tools) {
-    switch (tool21.type) {
+  for (const tool22 of tools) {
+    switch (tool22.type) {
       case "function":
         functionDeclarations.push({
-          name: tool21.name,
-          description: (_a4 = tool21.description) != null ? _a4 : "",
-          parameters: convertJSONSchemaToOpenAPISchema2(tool21.inputSchema)
+          name: tool22.name,
+          description: (_a4 = tool22.description) != null ? _a4 : "",
+          parameters: convertJSONSchemaToOpenAPISchema2(tool22.inputSchema)
         });
         break;
       default:
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
         break;
     }
   }
@@ -54272,16 +54648,16 @@ function prepareTools4({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const openaiCompatTools = [];
-  for (const tool21 of tools) {
-    if (tool21.type === "provider-defined") {
-      toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+  for (const tool22 of tools) {
+    if (tool22.type === "provider-defined") {
+      toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
     } else {
       openaiCompatTools.push({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
-          parameters: tool21.inputSchema
+          name: tool22.name,
+          description: tool22.description,
+          parameters: tool22.inputSchema
         }
       });
     }
@@ -55626,16 +56002,16 @@ function prepareTools5({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const xaiTools = [];
-  for (const tool21 of tools) {
-    if (tool21.type === "provider-defined") {
-      toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+  for (const tool22 of tools) {
+    if (tool22.type === "provider-defined") {
+      toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
     } else {
       xaiTools.push({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
-          parameters: tool21.inputSchema
+          name: tool22.name,
+          description: tool22.description,
+          parameters: tool22.inputSchema
         }
       });
     }
@@ -56635,21 +57011,21 @@ function prepareChatTools2({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const openaiTools2 = [];
-  for (const tool21 of tools) {
-    switch (tool21.type) {
+  for (const tool22 of tools) {
+    switch (tool22.type) {
       case "function":
         openaiTools2.push({
           type: "function",
           function: {
-            name: tool21.name,
-            description: tool21.description,
-            parameters: tool21.inputSchema,
+            name: tool22.name,
+            description: tool22.description,
+            parameters: tool22.inputSchema,
             strict: structuredOutputs ? strictJsonSchema : void 0
           }
         });
         break;
       default:
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
         break;
     }
   }
@@ -59231,22 +59607,22 @@ function prepareResponsesTools2(_0) {
       return { tools: void 0, toolChoice: void 0, toolWarnings };
     }
     const openaiTools2 = [];
-    for (const tool21 of tools) {
-      switch (tool21.type) {
+    for (const tool22 of tools) {
+      switch (tool22.type) {
         case "function":
           openaiTools2.push({
             type: "function",
-            name: tool21.name,
-            description: tool21.description,
-            parameters: tool21.inputSchema,
+            name: tool22.name,
+            description: tool22.description,
+            parameters: tool22.inputSchema,
             strict: strictJsonSchema
           });
           break;
         case "provider-defined": {
-          switch (tool21.id) {
+          switch (tool22.id) {
             case "openai.file_search": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: fileSearchArgsSchema3
               });
               openaiTools2.push({
@@ -59269,7 +59645,7 @@ function prepareResponsesTools2(_0) {
             }
             case "openai.web_search_preview": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webSearchPreviewArgsSchema2
               });
               openaiTools2.push({
@@ -59281,7 +59657,7 @@ function prepareResponsesTools2(_0) {
             }
             case "openai.web_search": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: webSearchArgsSchema2
               });
               openaiTools2.push({
@@ -59294,7 +59670,7 @@ function prepareResponsesTools2(_0) {
             }
             case "openai.code_interpreter": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: codeInterpreterArgsSchema2
               });
               openaiTools2.push({
@@ -59305,7 +59681,7 @@ function prepareResponsesTools2(_0) {
             }
             case "openai.image_generation": {
               const args = yield validateTypes({
-                value: tool21.args,
+                value: tool22.args,
                 schema: imageGenerationArgsSchema2
               });
               openaiTools2.push({
@@ -59329,7 +59705,7 @@ function prepareResponsesTools2(_0) {
           break;
         }
         default:
-          toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+          toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
           break;
       }
     }
@@ -59434,7 +59810,7 @@ var OpenAIResponsesLanguageModel2 = class {
       }
       function hasOpenAITool(id) {
         return (tools == null ? void 0 : tools.find(
-          (tool21) => tool21.type === "provider-defined" && tool21.id === id
+          (tool22) => tool22.type === "provider-defined" && tool22.id === id
         )) != null;
       }
       const topLogprobs = typeof (openaiOptions == null ? void 0 : openaiOptions.logprobs) === "number" ? openaiOptions == null ? void 0 : openaiOptions.logprobs : (openaiOptions == null ? void 0 : openaiOptions.logprobs) === true ? TOP_LOGPROBS_MAX2 : void 0;
@@ -59442,7 +59818,7 @@ var OpenAIResponsesLanguageModel2 = class {
         addInclude("message.output_text.logprobs");
       }
       const webSearchToolName = (_c = tools == null ? void 0 : tools.find(
-        (tool21) => tool21.type === "provider-defined" && (tool21.id === "openai.web_search" || tool21.id === "openai.web_search_preview")
+        (tool22) => tool22.type === "provider-defined" && (tool22.id === "openai.web_search" || tool22.id === "openai.web_search_preview")
       )) == null ? void 0 : _c.name;
       if (webSearchToolName) {
         addInclude("web_search_call.action.sources");
@@ -60665,13 +61041,13 @@ function prepareTools6({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const groqTools2 = [];
-  for (const tool21 of tools) {
-    if (tool21.type === "provider-defined") {
-      if (tool21.id === "groq.browser_search") {
+  for (const tool22 of tools) {
+    if (tool22.type === "provider-defined") {
+      if (tool22.id === "groq.browser_search") {
         if (!isBrowserSearchSupportedModel(modelId)) {
           toolWarnings.push({
             type: "unsupported-tool",
-            tool: tool21,
+            tool: tool22,
             details: `Browser search is only supported on the following models: ${getSupportedModelsString()}. Current model: ${modelId}`
           });
         } else {
@@ -60680,15 +61056,15 @@ function prepareTools6({
           });
         }
       } else {
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
       }
     } else {
       groqTools2.push({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
-          parameters: tool21.inputSchema
+          name: tool22.name,
+          description: tool22.description,
+          parameters: tool22.inputSchema
         }
       });
     }
@@ -61740,16 +62116,16 @@ function prepareTools7({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const mistralTools = [];
-  for (const tool21 of tools) {
-    if (tool21.type === "provider-defined") {
-      toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+  for (const tool22 of tools) {
+    if (tool22.type === "provider-defined") {
+      toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
     } else {
       mistralTools.push({
         type: "function",
         function: {
-          name: tool21.name,
-          description: tool21.description,
-          parameters: tool21.inputSchema
+          name: tool22.name,
+          description: tool22.description,
+          parameters: tool22.inputSchema
         }
       });
     }
@@ -61769,7 +62145,7 @@ function prepareTools7({
     case "tool":
       return {
         tools: mistralTools.filter(
-          (tool21) => tool21.function.name === toolChoice.toolName
+          (tool22) => tool22.function.name === toolChoice.toolName
         ),
         toolChoice: "any",
         toolWarnings
@@ -63552,10 +63928,10 @@ function prepareResponsesTools3({
     return { tools: void 0, toolChoice: void 0, toolWarnings };
   }
   const ollamaTools = [];
-  for (const tool21 of tools) {
-    switch (tool21.type) {
+  for (const tool22 of tools) {
+    switch (tool22.type) {
       case "function": {
-        let parameters = tool21.inputSchema;
+        let parameters = tool22.inputSchema;
         if (!parameters) {
           parameters = {
             type: "object",
@@ -63571,15 +63947,15 @@ function prepareResponsesTools3({
         ollamaTools.push({
           type: "function",
           function: {
-            name: tool21.name,
-            description: tool21.description,
+            name: tool22.name,
+            description: tool22.description,
             parameters
           }
         });
         break;
       }
       default:
-        toolWarnings.push({ type: "unsupported-tool", tool: tool21 });
+        toolWarnings.push({ type: "unsupported-tool", tool: tool22 });
         break;
     }
   }
@@ -64183,7 +64559,8 @@ var AISDKProvidersWithAPIKey = {
   togetherai: createTogetherAI,
   mistral: createMistral,
   deepseek: createDeepSeek,
-  perplexity: createPerplexity
+  perplexity: createPerplexity,
+  ollama: createOllama
 };
 var modelToProviderMap = {
   "gpt-4.1": "openai",
@@ -64220,7 +64597,8 @@ var modelToProviderMap = {
   "gemini-2.5-pro-preview-03-25": "google"
 };
 function getAISDKLanguageModel(subProvider, subModelName, clientOptions) {
-  if (clientOptions && Object.keys(clientOptions).length > 0) {
+  const hasValidOptions = clientOptions && Object.values(clientOptions).some((v) => v !== void 0 && v !== null);
+  if (hasValidOptions) {
     const creator = AISDKProvidersWithAPIKey[subProvider];
     if (!creator) {
       throw new UnsupportedAISDKModelProviderError(
@@ -64377,13 +64755,13 @@ var resolveTools = (clients, userTools) => __async(null, null, function* () {
       const clientTools = yield clientInstance.listTools({
         cursor: nextCursor
       });
-      for (const tool21 of clientTools.tools) {
-        tools[tool21.name] = {
-          description: tool21.description,
-          inputSchema: jsonSchemaToZod(tool21.inputSchema),
+      for (const tool22 of clientTools.tools) {
+        tools[tool22.name] = {
+          description: tool22.description,
+          inputSchema: jsonSchemaToZod(tool22.inputSchema),
           execute: (input) => __async(null, null, function* () {
             const result = yield clientInstance.callTool({
-              name: tool21.name,
+              name: tool22.name,
               arguments: input
             });
             return result;
@@ -64406,6 +64784,7 @@ var AVAILABLE_CUA_MODELS = [
   "anthropic/claude-sonnet-4-20250514",
   "anthropic/claude-sonnet-4-5-20250929",
   "google/gemini-2.5-computer-use-preview-10-2025",
+  "google/gemini-3-flash-computer-use",
   "microsoft/fara-7b"
 ];
 
@@ -64459,6 +64838,7 @@ __export(api_exports, {
   ReplayPageSchema: () => ReplayPageSchema,
   ReplayResponseSchema: () => ReplayResponseSchema,
   ReplayResultSchema: () => ReplayResultSchema,
+  SessionEndRequestSchema: () => SessionEndRequestSchema,
   SessionEndResponseSchema: () => SessionEndResponseSchema,
   SessionEndResultSchema: () => SessionEndResultSchema,
   SessionHeadersSchema: () => SessionHeadersSchema,
@@ -64718,6 +65098,11 @@ var SessionStartResponseSchema = wrapResponse(
   SessionStartResultSchema,
   "SessionStartResponse"
 );
+var SessionEndRequestSchema = import_v4107.z.object({
+  // Dummy property to ensure Stainless generates body parameter
+  // The server accepts {} (this field should be omitted)
+  _forceBody: import_v4107.z.undefined().optional()
+}).strict().meta({ id: "SessionEndRequest" });
 var SessionEndResultSchema = import_v4107.z.object({}).strict().meta({ id: "SessionEndResult" });
 var SessionEndResponseSchema = import_v4107.z.object({
   success: import_v4107.z.boolean().meta({
@@ -65182,7 +65567,7 @@ init_response2();
 init_sdkErrors();
 
 // examples/external_clients/aisdk.ts
-var import_ai22 = require("ai");
+var import_ai23 = require("ai");
 var AISdkClient2 = class extends LLMClient {
   constructor({ model }) {
     super(model.modelId);
@@ -65244,7 +65629,7 @@ var AISdkClient2 = class extends LLMClient {
         }
       );
       if (options.response_model) {
-        const response2 = yield (0, import_ai22.generateObject)({
+        const response2 = yield (0, import_ai23.generateObject)({
           model: this.model,
           messages: formattedMessages,
           schema: options.response_model.schema
@@ -65267,7 +65652,7 @@ var AISdkClient2 = class extends LLMClient {
           inputSchema: rawTool.parameters
         };
       }
-      const response = yield (0, import_ai22.generateText)({
+      const response = yield (0, import_ai23.generateText)({
         model: this.model,
         messages: formattedMessages,
         tools
@@ -66629,6 +67014,9 @@ function validateExperimentalFeatures(options) {
     if ((executeOptions == null ? void 0 : executeOptions.excludeTools) && executeOptions.excludeTools.length > 0) {
       unsupportedFeatures.push("excludeTools");
     }
+    if (executeOptions == null ? void 0 : executeOptions.output) {
+      unsupportedFeatures.push("output schema");
+    }
     if (unsupportedFeatures.length > 0) {
       throw new StagehandInvalidArgumentError(
         `${unsupportedFeatures.join(", ")} ${unsupportedFeatures.length === 1 ? "is" : "are"} not supported with CUA (Computer Use Agent) mode.`
@@ -66658,6 +67046,9 @@ function validateExperimentalFeatures(options) {
     if (executeOptions.excludeTools && executeOptions.excludeTools.length > 0) {
       features.push("excludeTools");
     }
+    if (executeOptions.output) {
+      features.push("output schema");
+    }
   }
   if (features.length > 0) {
     throw new ExperimentalNotConfiguredError(`Agent ${features.join(", ")}`);
@@ -67440,6 +67831,7 @@ var _V3 = class _V3 {
           );
         }
         const page = yield this.resolvePage(options == null ? void 0 : options.page);
+        const actCacheLlmClient = (options == null ? void 0 : options.model) ? this.resolveLlmClient(options.model) : void 0;
         let actCacheContext = null;
         const canUseCache = typeof input === "string" && !this.isAgentReplayRecording() && this.actCache.enabled;
         if (canUseCache) {
@@ -67452,7 +67844,8 @@ var _V3 = class _V3 {
             const cachedResult = yield this.actCache.tryReplay(
               actCacheContext,
               page,
-              options == null ? void 0 : options.timeout
+              options == null ? void 0 : options.timeout,
+              actCacheLlmClient
             );
             if (cachedResult) {
               this.addToHistory(
@@ -67855,7 +68248,13 @@ var _V3 = class _V3 {
         configSignature: agentConfigSignature,
         page: yield this.ctx.awaitActivePage()
       }) : null;
-      return { handler, resolvedOptions, instruction, cacheContext };
+      return {
+        handler,
+        resolvedOptions,
+        instruction,
+        cacheContext,
+        llmClient: agentLlmClient
+      };
     });
   }
   agent(options) {
@@ -68005,13 +68404,16 @@ Do not ask follow up questions, the user will trust your judgement.`
             args: [instructionOrOptions]
           });
           if (isStreaming) {
-            const { handler: handler2, resolvedOptions: resolvedOptions2, cacheContext: cacheContext2 } = yield this.prepareAgentExecution(
+            const { handler: handler2, resolvedOptions: resolvedOptions2, cacheContext: cacheContext2, llmClient: llmClient2 } = yield this.prepareAgentExecution(
               options,
               instructionOrOptions,
               agentConfigSignature
             );
             if (cacheContext2) {
-              const replayed = yield this.agentCache.tryReplayAsStream(cacheContext2);
+              const replayed = yield this.agentCache.tryReplayAsStream(
+                cacheContext2,
+                llmClient2
+              );
               if (replayed) {
                 SessionFileLogger.logAgentTaskCompleted({ cacheHit: true });
                 return replayed;
@@ -68034,13 +68436,16 @@ Do not ask follow up questions, the user will trust your judgement.`
             SessionFileLogger.logAgentTaskCompleted();
             return streamResult;
           }
-          const { handler, resolvedOptions, cacheContext } = yield this.prepareAgentExecution(
+          const { handler, resolvedOptions, cacheContext, llmClient } = yield this.prepareAgentExecution(
             options,
             instructionOrOptions,
             agentConfigSignature
           );
           if (cacheContext) {
-            const replayed = yield this.agentCache.tryReplay(cacheContext);
+            const replayed = yield this.agentCache.tryReplay(
+              cacheContext,
+              llmClient
+            );
             if (replayed) {
               SessionFileLogger.logAgentTaskCompleted({ cacheHit: true });
               return replayed;
@@ -68333,6 +68738,9 @@ I'm providing ${screenshots.length} screenshots showing the progression of the t
     });
   }
 };
+
+// lib/v3/index.ts
+var import_ai24 = require("ai");
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   AISdkClient,
@@ -68415,6 +68823,7 @@ I'm providing ${screenshots.length} screenshots showing the progression of the t
   providerEnvVarMap,
   toGeminiSchema,
   toJsonSchema,
+  tool,
   transformSchema,
   trimTrailingTextNode,
   validateZodSchema