Fix multiple click dispatch on CDP and Anthropic CUA handling

.changeset/blue-sides-roll.md

@@ -0,0 +1,5 @@
+---
+"@browserbasehq/stagehand": patch
+---
+
+Fix multiple click event dispatches on CDP and Anthropic CUA handling (double clicks)

packages/core/lib/v3/agent/AnthropicCUAClient.ts

@@ -779,9 +779,13 @@ export class AnthropicCUAClient extends AgentClient {
           };
         } else if (action === "double_click" || action === "doubleClick") {
           return {
-            type: action,
-            x: input.x as number,
-            y: input.y as number,
+            type: "doubleClick",
+            x:
+              (input.x as number) ||
+              (input.coordinate ? (input.coordinate as number[])[0] : 0),
+            y:
+              (input.y as number) ||
+              (input.coordinate ? (input.coordinate as number[])[1] : 0),
             ...input,
           };
         } else if (action === "scroll") {

packages/core/lib/v3/tests/click-count.spec.ts

@@ -0,0 +1,178 @@
+import { test, expect } from "@playwright/test";
+import { V3 } from "../v3";
+import { v3TestConfig } from "./v3.config";
+
+test.describe("Locator and Page click methods", () => {
+  let v3: V3;
+
+  test.beforeEach(async () => {
+    v3 = new V3(v3TestConfig);
+    await v3.init();
+  });
+
+  test.afterEach(async () => {
+    await v3?.close?.().catch(() => {});
+  });
+
+  test("locator.click() performs single click by default", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    // Get initial count
+    const countDisplay = page.locator("#count");
+    const initialCount = await countDisplay.inputValue();
+    expect(initialCount).toBe("0");
+
+    // Perform single click on the textarea (the clickable area)
+    const clickArea = page.locator("#textarea");
+    await clickArea.click();
+
+    // Verify count incremented by 1
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("1");
+  });
+
+  test("locator.click() with clickCount: 2 performs double-click", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    // Get initial counts
+    const countDisplay = page.locator("#count");
+    const dcCountDisplay = page.locator("#dcCount");
+
+    const initialCount = await countDisplay.inputValue();
+    const initialDcCount = await dcCountDisplay.inputValue();
+    expect(initialCount).toBe("0");
+    expect(initialDcCount).toBe("0");
+
+    // Perform double-click on the textarea
+    const clickArea = page.locator("#textarea");
+    await clickArea.click({ clickCount: 2 });
+
+    // Verify both counters incremented
+    // Regular count should be 2 (one for each click in the double-click)
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("2");
+
+    // Double-click count should be 1 (one double-click event detected)
+    const newDcCount = await dcCountDisplay.inputValue();
+    expect(newDcCount).toBe("1");
+  });
+
+  test("locator.click() with clickCount: 3 performs triple-click", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    const countDisplay = page.locator("#count");
+    const initialCount = await countDisplay.inputValue();
+    expect(initialCount).toBe("0");
+
+    // Perform triple-click on the textarea
+    const clickArea = page.locator("#textarea");
+    await clickArea.click({ clickCount: 3 });
+
+    // Verify count incremented by 3
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("3");
+  });
+
+  test("page.click() performs single click with coordinates", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    // Get initial count
+    const countDisplay = page.locator("#count");
+    const initialCount = await countDisplay.inputValue();
+    expect(initialCount).toBe("0");
+
+    // Get the centroid of the textarea to click
+    const clickArea = page.locator("#textarea");
+    const { x, y } = await clickArea.centroid();
+
+    // Perform single click using page.click() with coordinates
+    await page.click(x, y);
+
+    // Verify count incremented by 1
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("1");
+  });
+
+  test("page.click() with clickCount: 2 performs double-click", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    // Get initial counts
+    const countDisplay = page.locator("#count");
+    const dcCountDisplay = page.locator("#dcCount");
+
+    const initialCount = await countDisplay.inputValue();
+    const initialDcCount = await dcCountDisplay.inputValue();
+    expect(initialCount).toBe("0");
+    expect(initialDcCount).toBe("0");
+
+    // Get the centroid of the textarea to click
+    const clickArea = page.locator("#textarea");
+    const { x, y } = await clickArea.centroid();
+
+    // Perform double-click using page.click() with coordinates
+    await page.click(x, y, { clickCount: 2 });
+
+    // Verify both counters incremented
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("2");
+
+    // Double-click count should be 1
+    const newDcCount = await dcCountDisplay.inputValue();
+    expect(newDcCount).toBe("1");
+  });
+
+  test("page.click() with clickCount: 3 performs triple-click", async () => {
+    const page = v3.context.pages()[0];
+    await page.goto(
+      "https://browserbase.github.io/stagehand-eval-sites/sites/click-test/",
+    );
+
+    // Wait for page to be fully loaded
+    await page.waitForLoadState("domcontentloaded");
+
+    const countDisplay = page.locator("#count");
+    const initialCount = await countDisplay.inputValue();
+    expect(initialCount).toBe("0");
+
+    // Get the centroid of the textarea to click
+    const clickArea = page.locator("#textarea");
+    const { x, y } = await clickArea.centroid();
+
+    // Perform triple-click using page.click() with coordinates
+    await page.click(x, y, { clickCount: 3 });
+
+    // Verify count incremented by 3
+    const newCount = await countDisplay.inputValue();
+    expect(newCount).toBe("3");
+  });
+});

packages/core/lib/v3/understudy/locator.ts

@@ -407,21 +407,24 @@ export class Locator {
         button: "none",
       } as Protocol.Input.DispatchMouseEventRequest);
 
-      await session.send<never>("Input.dispatchMouseEvent", {
-        type: "mousePressed",
-        x: cx,
-        y: cy,
-        button,
-        clickCount,
-      } as Protocol.Input.DispatchMouseEventRequest);
-
-      await session.send<never>("Input.dispatchMouseEvent", {
-        type: "mouseReleased",
-        x: cx,
-        y: cy,
-        button,
-        clickCount,
-      } as Protocol.Input.DispatchMouseEventRequest);
+      // Dispatch mouse pressed and released events for the given click count
+      for (let i = 1; i <= clickCount; i++) {
+        await session.send<never>("Input.dispatchMouseEvent", {
+          type: "mousePressed",
+          x: cx,
+          y: cy,
+          button,
+          clickCount: i,
+        } as Protocol.Input.DispatchMouseEventRequest);
+
+        await session.send<never>("Input.dispatchMouseEvent", {
+          type: "mouseReleased",
+          x: cx,
+          y: cy,
+          button,
+          clickCount: i,
+        } as Protocol.Input.DispatchMouseEventRequest);
+      }
     } finally {
       // release the element handle
       try {

packages/core/lib/v3/understudy/page.ts

@@ -1350,22 +1350,24 @@ export class Page {
       button: "none",
     } as Protocol.Input.DispatchMouseEventRequest);
 
-    await this.mainSession.send<never>("Input.dispatchMouseEvent", {
-      type: "mousePressed",
-      x,
-      y,
-      button,
-      clickCount,
-    } as Protocol.Input.DispatchMouseEventRequest);
-
-    await this.mainSession.send<never>("Input.dispatchMouseEvent", {
-      type: "mouseReleased",
-      x,
-      y,
-      button,
-      clickCount,
-    } as Protocol.Input.DispatchMouseEventRequest);
+    // Dispatch mouse pressed and released events for the given click count
+    for (let i = 1; i <= clickCount; i++) {
+      await this.mainSession.send<never>("Input.dispatchMouseEvent", {
+        type: "mousePressed",
+        x,
+        y,
+        button,
+        clickCount: i,
+      } as Protocol.Input.DispatchMouseEventRequest);
 
+      await this.mainSession.send<never>("Input.dispatchMouseEvent", {
+        type: "mouseReleased",
+        x,
+        y,
+        button,
+        clickCount: i,
+      } as Protocol.Input.DispatchMouseEventRequest);
+    }
     if (options?.returnXpath) return xpathResult ?? "";
   }