@browserbasehq/orca 3.1.0-patch.4 → 3.2.0-middleware.1

package/dist/esm/lib/v3/agent/tools/type.js

@@ -77,38 +77,30 @@ export const typeTool = (v3, provider, variables) => {
             }
         },
         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 };
+            if (result.success === false || result.error !== undefined) {
+                return {
+                    type: "content",
+                    value: [{ type: "text", text: JSON.stringify(result) }],
+                };
             }
-            return {
-                type: "content",
-                value: [
-                    {
-                        type: "text",
-                        text: JSON.stringify({
-                            success: result.success,
-                            error: result.error,
-                        }),
-                    },
-                ],
-            };
+            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 };
         },
     });
 };

package/dist/esm/lib/v3/agent/tools/type.js.map

@@ -1 +1 @@
-{"version":3,"file":"type.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/tools/type.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAC1B,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAQxB,OAAO,EAAE,kBAAkB,EAAE,MAAM,qCAAqC,CAAC;AACzE,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAChD,OAAO,EAAE,wBAAwB,EAAE,MAAM,+BAA+B,CAAC;AACzE,OAAO,EAAE,mBAAmB,EAAE,MAAM,uBAAuB,CAAC;AAE5D,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,EAAM,EAAE,QAAiB,EAAE,SAAqB,EAAE,EAAE;IAC3E,MAAM,YAAY,GAAG,SAAS,IAAI,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;IACpE,MAAM,eAAe,GAAG,YAAY;QAClC,CAAC,CAAC,oGAAoG,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE;QACzI,CAAC,CAAC,mCAAmC,CAAC;IAExC,OAAO,IAAI,CAAC;QACV,WAAW,EACT,2QAA2Q;QAC7Q,WAAW,EAAE,CAAC,CAAC,MAAM,CAAC;YACpB,QAAQ,EAAE,CAAC;iBACR,MAAM,EAAE;iBACR,QAAQ,CACP,4HAA4H,CAC7H;YACH,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,eAAe,CAAC;YAC1C,WAAW,EAAE,CAAC;iBACX,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC;iBACjB,QAAQ,CAAC,iDAAiD,CAAC;SAC/D,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,QAAQ,EACR,WAAW,EACX,IAAI,GACL,EAA2B,EAAE;YAC5B,IAAI,CAAC;gBACH,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC,OAAO,CAAC,eAAe,EAAE,CAAC;gBAChD,MAAM,SAAS,GAAG,kBAAkB,CAClC,WAAW,CAAC,CAAC,CAAC,EACd,WAAW,CAAC,CAAC,CAAC,EACd,QAAQ,EACR,EAAE,CACH,CAAC;gBAEF,mDAAmD;gBACnD,MAAM,UAAU,GAAG,mBAAmB,CAAC,IAAI,EAAE,SAAS,CAAC,CAAC;gBAExD,EAAE,CAAC,MAAM,CAAC;oBACR,QAAQ,EAAE,OAAO;oBACjB,OAAO,EAAE,0BAA0B;oBACnC,KAAK,EAAE,CAAC;oBACR,SAAS,EAAE;wBACT,SAAS,EAAE;4BACT,KAAK,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC;4BACzC,IAAI,EAAE,QAAQ;yBACf;qBACF;iBACF,CAAC,CAAC;gBAEH,8EAA8E;gBAC9E,MAAM,kBAAkB,GAAG,EAAE,CAAC,mBAAmB,EAAE,CAAC;gBACpD,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,EAAE;oBACvD,WAAW,EAAE,kBAAkB;iBAChC,CAAC,CAAC;gBAEH,MAAM,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;gBAE5B,MAAM,gBAAgB,GAAG,MAAM,wBAAwB,CAAC,IAAI,CAAC,CAAC;gBAE9D,0FAA0F;gBAC1F,IAAI,kBAAkB,EAAE,CAAC;oBACvB,MAAM,eAAe,GAAG,WAAW,CAAC,KAAK,CAAC,CAAC;oBAC3C,IAAI,eAAe,EAAE,CAAC;wBACpB,MAAM,MAAM,GAAW;4BACrB,QAAQ,EAAE,eAAe;4BACzB,WAAW,EAAE,QAAQ;4BACrB,MAAM,EAAE,MAAM;4BACd,SAAS,EAAE,CAAC,IAAI,CAAC;yBAClB,CAAC;wBACF,EAAE,CAAC,qBAAqB,CAAC;4BACvB,IAAI,EAAE,KAAK;4BACX,WAAW,EAAE,QAAQ;4BACrB,OAAO,EAAE,CAAC,MAAM,CAAC;4BACjB,iBAAiB,EAAE,QAAQ;yBAC5B,CAAC,CAAC;oBACL,CAAC;gBACH,CAAC;gBAED,OAAO;oBACL,OAAO,EAAE,IAAI;oBACb,QAAQ;oBACR,IAAI,EAAE,8FAA8F;oBACpG,gBAAgB;iBACjB,CAAC;YACJ,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACf,OAAO;oBACL,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,iBAAkB,KAAe,CAAC,OAAO,EAAE;iBACnD,CAAC;YACJ,CAAC;QACH,CAAC;QACD,aAAa,EAAE,CAAC,MAAM,EAAE,EAAE;YACxB,IAAI,MAAM,CAAC,OAAO,EAAE,CAAC;gBACnB,MAAM,OAAO,GAA6B;oBACxC;wBACE,IAAI,EAAE,MAAM;wBACZ,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;4BACnB,OAAO,EAAE,MAAM,CAAC,OAAO;4BACvB,QAAQ,EAAE,MAAM,CAAC,QAAQ;4BACzB,IAAI,EAAE,MAAM,CAAC,IAAI;yBAClB,CAAC;qBACH;iBACF,CAAC;gBACF,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;oBAC5B,OAAO,CAAC,IAAI,CAAC;wBACX,IAAI,EAAE,OAAO;wBACb,SAAS,EAAE,WAAW;wBACtB,IAAI,EAAE,MAAM,CAAC,gBAAgB;qBAC9B,CAAC,CAAC;gBACL,CAAC;gBACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;YAC7C,CAAC;YACD,OAAO;gBACL,IAAI,EAAE,SAAS;gBACf,KAAK,EAAE;oBACL;wBACE,IAAI,EAAE,MAAM;wBACZ,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;4BACnB,OAAO,EAAE,MAAM,CAAC,OAAO;4BACvB,KAAK,EAAE,MAAM,CAAC,KAAK;yBACpB,CAAC;qBACH;iBACF;aACF,CAAC;QACJ,CAAC;KACF,CAAC,CAAC;AACL,CAAC,CAAC","sourcesContent":["import { tool } from \"ai\";\nimport { z } from \"zod\";\nimport type { V3 } from \"../../v3.js\";\nimport type { Action } from \"../../types/public/methods.js\";\nimport type {\n  TypeToolResult,\n  ModelOutputContentItem,\n  Variables,\n} from \"../../types/public/agent.js\";\nimport { processCoordinates } from \"../utils/coordinateNormalization.js\";\nimport { ensureXPath } from \"../utils/xpath.js\";\nimport { waitAndCaptureScreenshot } from \"../utils/screenshotHandler.js\";\nimport { substituteVariables } from \"../utils/variables.js\";\n\nexport const typeTool = (v3: V3, provider?: string, variables?: Variables) => {\n  const hasVariables = variables && Object.keys(variables).length > 0;\n  const textDescription = hasVariables\n    ? `The text to type into the element. Use %variableName% to substitute a variable value. Available: ${Object.keys(variables).join(\", \")}`\n    : \"The text to type into the element\";\n\n  return tool({\n    description:\n      \"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)\",\n    inputSchema: z.object({\n      describe: z\n        .string()\n        .describe(\n          \"Describe the element to type into in a short, specific phrase that mentions the element type and a good visual description\",\n        ),\n      text: z.string().describe(textDescription),\n      coordinates: z\n        .array(z.number())\n        .describe(\"The (x, y) coordinates to type into the element\"),\n    }),\n    execute: async ({\n      describe,\n      coordinates,\n      text,\n    }): Promise<TypeToolResult> => {\n      try {\n        const page = await v3.context.awaitActivePage();\n        const processed = processCoordinates(\n          coordinates[0],\n          coordinates[1],\n          provider,\n          v3,\n        );\n\n        // Substitute any %variableName% tokens in the text\n        const actualText = substituteVariables(text, variables);\n\n        v3.logger({\n          category: \"agent\",\n          message: `Agent calling tool: type`,\n          level: 1,\n          auxiliary: {\n            arguments: {\n              value: JSON.stringify({ describe, text }),\n              type: \"object\",\n            },\n          },\n        });\n\n        // Only request XPath when caching is enabled to avoid unnecessary computation\n        const shouldCollectXpath = v3.isAgentReplayActive();\n        const xpath = await page.click(processed.x, processed.y, {\n          returnXpath: shouldCollectXpath,\n        });\n\n        await page.type(actualText);\n\n        const screenshotBase64 = await waitAndCaptureScreenshot(page);\n\n        // Record as an \"act\" step with proper Action for deterministic replay (only when caching)\n        if (shouldCollectXpath) {\n          const normalizedXpath = ensureXPath(xpath);\n          if (normalizedXpath) {\n            const action: Action = {\n              selector: normalizedXpath,\n              description: describe,\n              method: \"type\",\n              arguments: [text],\n            };\n            v3.recordAgentReplayStep({\n              type: \"act\",\n              instruction: describe,\n              actions: [action],\n              actionDescription: describe,\n            });\n          }\n        }\n\n        return {\n          success: true,\n          describe,\n          text, // Return original text (with %variableName% tokens) to avoid exposing sensitive values to LLM\n          screenshotBase64,\n        };\n      } catch (error) {\n        return {\n          success: false,\n          error: `Error typing: ${(error as Error).message}`,\n        };\n      }\n    },\n    toModelOutput: (result) => {\n      if (result.success) {\n        const content: ModelOutputContentItem[] = [\n          {\n            type: \"text\",\n            text: JSON.stringify({\n              success: result.success,\n              describe: result.describe,\n              text: result.text,\n            }),\n          },\n        ];\n        if (result.screenshotBase64) {\n          content.push({\n            type: \"media\",\n            mediaType: \"image/png\",\n            data: result.screenshotBase64,\n          });\n        }\n        return { type: \"content\", value: content };\n      }\n      return {\n        type: \"content\",\n        value: [\n          {\n            type: \"text\",\n            text: JSON.stringify({\n              success: result.success,\n              error: result.error,\n            }),\n          },\n        ],\n      };\n    },\n  });\n};\n"]}
+{"version":3,"file":"type.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/tools/type.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAC1B,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAQxB,OAAO,EAAE,kBAAkB,EAAE,MAAM,qCAAqC,CAAC;AACzE,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAChD,OAAO,EAAE,wBAAwB,EAAE,MAAM,+BAA+B,CAAC;AACzE,OAAO,EAAE,mBAAmB,EAAE,MAAM,uBAAuB,CAAC;AAE5D,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,EAAM,EAAE,QAAiB,EAAE,SAAqB,EAAE,EAAE;IAC3E,MAAM,YAAY,GAAG,SAAS,IAAI,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC;IACpE,MAAM,eAAe,GAAG,YAAY;QAClC,CAAC,CAAC,oGAAoG,MAAM,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE;QACzI,CAAC,CAAC,mCAAmC,CAAC;IAExC,OAAO,IAAI,CAAC;QACV,WAAW,EACT,2QAA2Q;QAC7Q,WAAW,EAAE,CAAC,CAAC,MAAM,CAAC;YACpB,QAAQ,EAAE,CAAC;iBACR,MAAM,EAAE;iBACR,QAAQ,CACP,4HAA4H,CAC7H;YACH,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,eAAe,CAAC;YAC1C,WAAW,EAAE,CAAC;iBACX,KAAK,CAAC,CAAC,CAAC,MAAM,EAAE,CAAC;iBACjB,QAAQ,CAAC,iDAAiD,CAAC;SAC/D,CAAC;QACF,OAAO,EAAE,KAAK,EAAE,EACd,QAAQ,EACR,WAAW,EACX,IAAI,GACL,EAA2B,EAAE;YAC5B,IAAI,CAAC;gBACH,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC,OAAO,CAAC,eAAe,EAAE,CAAC;gBAChD,MAAM,SAAS,GAAG,kBAAkB,CAClC,WAAW,CAAC,CAAC,CAAC,EACd,WAAW,CAAC,CAAC,CAAC,EACd,QAAQ,EACR,EAAE,CACH,CAAC;gBAEF,mDAAmD;gBACnD,MAAM,UAAU,GAAG,mBAAmB,CAAC,IAAI,EAAE,SAAS,CAAC,CAAC;gBAExD,EAAE,CAAC,MAAM,CAAC;oBACR,QAAQ,EAAE,OAAO;oBACjB,OAAO,EAAE,0BAA0B;oBACnC,KAAK,EAAE,CAAC;oBACR,SAAS,EAAE;wBACT,SAAS,EAAE;4BACT,KAAK,EAAE,IAAI,CAAC,SAAS,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC;4BACzC,IAAI,EAAE,QAAQ;yBACf;qBACF;iBACF,CAAC,CAAC;gBAEH,8EAA8E;gBAC9E,MAAM,kBAAkB,GAAG,EAAE,CAAC,mBAAmB,EAAE,CAAC;gBACpD,MAAM,KAAK,GAAG,MAAM,IAAI,CAAC,KAAK,CAAC,SAAS,CAAC,CAAC,EAAE,SAAS,CAAC,CAAC,EAAE;oBACvD,WAAW,EAAE,kBAAkB;iBAChC,CAAC,CAAC;gBAEH,MAAM,IAAI,CAAC,IAAI,CAAC,UAAU,CAAC,CAAC;gBAE5B,MAAM,gBAAgB,GAAG,MAAM,wBAAwB,CAAC,IAAI,CAAC,CAAC;gBAE9D,0FAA0F;gBAC1F,IAAI,kBAAkB,EAAE,CAAC;oBACvB,MAAM,eAAe,GAAG,WAAW,CAAC,KAAK,CAAC,CAAC;oBAC3C,IAAI,eAAe,EAAE,CAAC;wBACpB,MAAM,MAAM,GAAW;4BACrB,QAAQ,EAAE,eAAe;4BACzB,WAAW,EAAE,QAAQ;4BACrB,MAAM,EAAE,MAAM;4BACd,SAAS,EAAE,CAAC,IAAI,CAAC;yBAClB,CAAC;wBACF,EAAE,CAAC,qBAAqB,CAAC;4BACvB,IAAI,EAAE,KAAK;4BACX,WAAW,EAAE,QAAQ;4BACrB,OAAO,EAAE,CAAC,MAAM,CAAC;4BACjB,iBAAiB,EAAE,QAAQ;yBAC5B,CAAC,CAAC;oBACL,CAAC;gBACH,CAAC;gBAED,OAAO;oBACL,OAAO,EAAE,IAAI;oBACb,QAAQ;oBACR,IAAI,EAAE,8FAA8F;oBACpG,gBAAgB;iBACjB,CAAC;YACJ,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACf,OAAO;oBACL,OAAO,EAAE,KAAK;oBACd,KAAK,EAAE,iBAAiB,KAAK,CAAC,OAAO,EAAE;iBACxC,CAAC;YACJ,CAAC;QACH,CAAC;QACD,aAAa,EAAE,CAAC,MAAM,EAAE,EAAE;YACxB,IAAI,MAAM,CAAC,OAAO,KAAK,KAAK,IAAI,MAAM,CAAC,KAAK,KAAK,SAAS,EAAE,CAAC;gBAC3D,OAAO;oBACL,IAAI,EAAE,SAAS;oBACf,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC;iBACxD,CAAC;YACJ,CAAC;YAED,MAAM,OAAO,GAA6B;gBACxC;oBACE,IAAI,EAAE,MAAM;oBACZ,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;wBACnB,OAAO,EAAE,MAAM,CAAC,OAAO;wBACvB,QAAQ,EAAE,MAAM,CAAC,QAAQ;wBACzB,IAAI,EAAE,MAAM,CAAC,IAAI;qBAClB,CAAC;iBACH;aACF,CAAC;YACF,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;gBAC5B,OAAO,CAAC,IAAI,CAAC;oBACX,IAAI,EAAE,OAAO;oBACb,SAAS,EAAE,WAAW;oBACtB,IAAI,EAAE,MAAM,CAAC,gBAAgB;iBAC9B,CAAC,CAAC;YACL,CAAC;YACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;QAC7C,CAAC;KACF,CAAC,CAAC;AACL,CAAC,CAAC","sourcesContent":["import { tool } from \"ai\";\nimport { z } from \"zod\";\nimport type { V3 } from \"../../v3.js\";\nimport type { Action } from \"../../types/public/methods.js\";\nimport type {\n  TypeToolResult,\n  ModelOutputContentItem,\n  Variables,\n} from \"../../types/public/agent.js\";\nimport { processCoordinates } from \"../utils/coordinateNormalization.js\";\nimport { ensureXPath } from \"../utils/xpath.js\";\nimport { waitAndCaptureScreenshot } from \"../utils/screenshotHandler.js\";\nimport { substituteVariables } from \"../utils/variables.js\";\n\nexport const typeTool = (v3: V3, provider?: string, variables?: Variables) => {\n  const hasVariables = variables && Object.keys(variables).length > 0;\n  const textDescription = hasVariables\n    ? `The text to type into the element. Use %variableName% to substitute a variable value. Available: ${Object.keys(variables).join(\", \")}`\n    : \"The text to type into the element\";\n\n  return tool({\n    description:\n      \"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)\",\n    inputSchema: z.object({\n      describe: z\n        .string()\n        .describe(\n          \"Describe the element to type into in a short, specific phrase that mentions the element type and a good visual description\",\n        ),\n      text: z.string().describe(textDescription),\n      coordinates: z\n        .array(z.number())\n        .describe(\"The (x, y) coordinates to type into the element\"),\n    }),\n    execute: async ({\n      describe,\n      coordinates,\n      text,\n    }): Promise<TypeToolResult> => {\n      try {\n        const page = await v3.context.awaitActivePage();\n        const processed = processCoordinates(\n          coordinates[0],\n          coordinates[1],\n          provider,\n          v3,\n        );\n\n        // Substitute any %variableName% tokens in the text\n        const actualText = substituteVariables(text, variables);\n\n        v3.logger({\n          category: \"agent\",\n          message: `Agent calling tool: type`,\n          level: 1,\n          auxiliary: {\n            arguments: {\n              value: JSON.stringify({ describe, text }),\n              type: \"object\",\n            },\n          },\n        });\n\n        // Only request XPath when caching is enabled to avoid unnecessary computation\n        const shouldCollectXpath = v3.isAgentReplayActive();\n        const xpath = await page.click(processed.x, processed.y, {\n          returnXpath: shouldCollectXpath,\n        });\n\n        await page.type(actualText);\n\n        const screenshotBase64 = await waitAndCaptureScreenshot(page);\n\n        // Record as an \"act\" step with proper Action for deterministic replay (only when caching)\n        if (shouldCollectXpath) {\n          const normalizedXpath = ensureXPath(xpath);\n          if (normalizedXpath) {\n            const action: Action = {\n              selector: normalizedXpath,\n              description: describe,\n              method: \"type\",\n              arguments: [text],\n            };\n            v3.recordAgentReplayStep({\n              type: \"act\",\n              instruction: describe,\n              actions: [action],\n              actionDescription: describe,\n            });\n          }\n        }\n\n        return {\n          success: true,\n          describe,\n          text, // Return original text (with %variableName% tokens) to avoid exposing sensitive values to LLM\n          screenshotBase64,\n        };\n      } catch (error) {\n        return {\n          success: false,\n          error: `Error typing: ${error.message}`,\n        };\n      }\n    },\n    toModelOutput: (result) => {\n      if (result.success === false || result.error !== undefined) {\n        return {\n          type: \"content\",\n          value: [{ type: \"text\", text: JSON.stringify(result) }],\n        };\n      }\n\n      const content: ModelOutputContentItem[] = [\n        {\n          type: \"text\",\n          text: JSON.stringify({\n            success: result.success,\n            describe: result.describe,\n            text: result.text,\n          }),\n        },\n      ];\n      if (result.screenshotBase64) {\n        content.push({\n          type: \"media\",\n          mediaType: \"image/png\",\n          data: result.screenshotBase64,\n        });\n      }\n      return { type: \"content\", value: content };\n    },\n  });\n};\n"]}

package/dist/esm/lib/v3/agent/tools/wait.js

@@ -31,6 +31,12 @@ export const waitTool = (v3, mode) => tool({
         return { success: true, waited: timeMs };
     },
     toModelOutput: (result) => {
+        if (result.success === false || result.error !== undefined) {
+            return {
+                type: "content",
+                value: [{ type: "text", text: JSON.stringify(result) }],
+            };
+        }
         const content = [
             {
                 type: "text",

package/dist/esm/lib/v3/agent/tools/wait.js.map

@@ -1 +1 @@
-{"version":3,"file":"wait.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/tools/wait.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAC1B,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAOxB,OAAO,EAAE,wBAAwB,EAAE,MAAM,+BAA+B,CAAC;AAEzE,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,EAAM,EAAE,IAAoB,EAAE,EAAE,CACvD,IAAI,CAAC;IACH,WAAW,EAAE,2BAA2B;IACxC,WAAW,EAAE,CAAC,CAAC,MAAM,CAAC;QACpB,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,sBAAsB,CAAC;KACpD,CAAC;IACF,OAAO,EAAE,KAAK,EAAE,EAAE,MAAM,EAAE,EAA2B,EAAE;QACrD,EAAE,CAAC,MAAM,CAAC;YACR,QAAQ,EAAE,OAAO;YACjB,OAAO,EAAE,0BAA0B;YACnC,KAAK,EAAE,CAAC;YACR,SAAS,EAAE;gBACT,SAAS,EAAE;oBACT,KAAK,EAAE,eAAe,MAAM,eAAe;oBAC3C,IAAI,EAAE,QAAQ;iBACf;aACF;SACF,CAAC,CAAC;QACH,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,UAAU,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;QAC5D,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;YACf,EAAE,CAAC,qBAAqB,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;QACrD,CAAC;QAED,gEAAgE;QAChE,IAAI,IAAI,KAAK,QAAQ,EAAE,CAAC;YACtB,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC,OAAO,CAAC,eAAe,EAAE,CAAC;YAChD,MAAM,gBAAgB,GAAG,MAAM,wBAAwB,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;YACjE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,gBAAgB,EAAE,CAAC;QAC7D,CAAC;QAED,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC;IAC3C,CAAC;IACD,aAAa,EAAE,CAAC,MAAM,EAAE,EAAE;QACxB,MAAM,OAAO,GAA6B;YACxC;gBACE,IAAI,EAAE,MAAM;gBACZ,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACnB,OAAO,EAAE,MAAM,CAAC,OAAO;oBACvB,MAAM,EAAE,MAAM,CAAC,MAAM;iBACtB,CAAC;aACH;SACF,CAAC;QACF,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;YAC5B,OAAO,CAAC,IAAI,CAAC;gBACX,IAAI,EAAE,OAAO;gBACb,SAAS,EAAE,WAAW;gBACtB,IAAI,EAAE,MAAM,CAAC,gBAAgB;aAC9B,CAAC,CAAC;QACL,CAAC;QACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;IAC7C,CAAC;CACF,CAAC,CAAC","sourcesContent":["import { tool } from \"ai\";\nimport { z } from \"zod\";\nimport type { V3 } from \"../../v3.js\";\nimport type {\n  AgentToolMode,\n  WaitToolResult,\n  ModelOutputContentItem,\n} from \"../../types/public/agent.js\";\nimport { waitAndCaptureScreenshot } from \"../utils/screenshotHandler.js\";\n\nexport const waitTool = (v3: V3, mode?: AgentToolMode) =>\n  tool({\n    description: \"Wait for a specified time\",\n    inputSchema: z.object({\n      timeMs: z.number().describe(\"Time in milliseconds\"),\n    }),\n    execute: async ({ timeMs }): Promise<WaitToolResult> => {\n      v3.logger({\n        category: \"agent\",\n        message: `Agent calling tool: wait`,\n        level: 1,\n        auxiliary: {\n          arguments: {\n            value: `Waiting for ${timeMs} milliseconds`,\n            type: \"string\",\n          },\n        },\n      });\n      await new Promise((resolve) => setTimeout(resolve, timeMs));\n      if (timeMs > 0) {\n        v3.recordAgentReplayStep({ type: \"wait\", timeMs });\n      }\n\n      // Take screenshot after wait in hybrid mode for visual feedback\n      if (mode === \"hybrid\") {\n        const page = await v3.context.awaitActivePage();\n        const screenshotBase64 = await waitAndCaptureScreenshot(page, 0);\n        return { success: true, waited: timeMs, screenshotBase64 };\n      }\n\n      return { success: true, waited: timeMs };\n    },\n    toModelOutput: (result) => {\n      const content: ModelOutputContentItem[] = [\n        {\n          type: \"text\",\n          text: JSON.stringify({\n            success: result.success,\n            waited: result.waited,\n          }),\n        },\n      ];\n      if (result.screenshotBase64) {\n        content.push({\n          type: \"media\",\n          mediaType: \"image/png\",\n          data: result.screenshotBase64,\n        });\n      }\n      return { type: \"content\", value: content };\n    },\n  });\n"]}
+{"version":3,"file":"wait.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/tools/wait.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAC1B,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAOxB,OAAO,EAAE,wBAAwB,EAAE,MAAM,+BAA+B,CAAC;AAEzE,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,EAAM,EAAE,IAAoB,EAAE,EAAE,CACvD,IAAI,CAAC;IACH,WAAW,EAAE,2BAA2B;IACxC,WAAW,EAAE,CAAC,CAAC,MAAM,CAAC;QACpB,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,sBAAsB,CAAC;KACpD,CAAC;IACF,OAAO,EAAE,KAAK,EAAE,EAAE,MAAM,EAAE,EAA2B,EAAE;QACrD,EAAE,CAAC,MAAM,CAAC;YACR,QAAQ,EAAE,OAAO;YACjB,OAAO,EAAE,0BAA0B;YACnC,KAAK,EAAE,CAAC;YACR,SAAS,EAAE;gBACT,SAAS,EAAE;oBACT,KAAK,EAAE,eAAe,MAAM,eAAe;oBAC3C,IAAI,EAAE,QAAQ;iBACf;aACF;SACF,CAAC,CAAC;QACH,MAAM,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,UAAU,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC,CAAC;QAC5D,IAAI,MAAM,GAAG,CAAC,EAAE,CAAC;YACf,EAAE,CAAC,qBAAqB,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC,CAAC;QACrD,CAAC;QAED,gEAAgE;QAChE,IAAI,IAAI,KAAK,QAAQ,EAAE,CAAC;YACtB,MAAM,IAAI,GAAG,MAAM,EAAE,CAAC,OAAO,CAAC,eAAe,EAAE,CAAC;YAChD,MAAM,gBAAgB,GAAG,MAAM,wBAAwB,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC;YACjE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,gBAAgB,EAAE,CAAC;QAC7D,CAAC;QAED,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,CAAC;IAC3C,CAAC;IACD,aAAa,EAAE,CAAC,MAAM,EAAE,EAAE;QACxB,IAAI,MAAM,CAAC,OAAO,KAAK,KAAK,IAAI,MAAM,CAAC,KAAK,KAAK,SAAS,EAAE,CAAC;YAC3D,OAAO;gBACL,IAAI,EAAE,SAAS;gBACf,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,EAAE,CAAC;aACxD,CAAC;QACJ,CAAC;QAED,MAAM,OAAO,GAA6B;YACxC;gBACE,IAAI,EAAE,MAAM;gBACZ,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACnB,OAAO,EAAE,MAAM,CAAC,OAAO;oBACvB,MAAM,EAAE,MAAM,CAAC,MAAM;iBACtB,CAAC;aACH;SACF,CAAC;QACF,IAAI,MAAM,CAAC,gBAAgB,EAAE,CAAC;YAC5B,OAAO,CAAC,IAAI,CAAC;gBACX,IAAI,EAAE,OAAO;gBACb,SAAS,EAAE,WAAW;gBACtB,IAAI,EAAE,MAAM,CAAC,gBAAgB;aAC9B,CAAC,CAAC;QACL,CAAC;QACD,OAAO,EAAE,IAAI,EAAE,SAAS,EAAE,KAAK,EAAE,OAAO,EAAE,CAAC;IAC7C,CAAC;CACF,CAAC,CAAC","sourcesContent":["import { tool } from \"ai\";\nimport { z } from \"zod\";\nimport type { V3 } from \"../../v3.js\";\nimport type {\n  AgentToolMode,\n  WaitToolResult,\n  ModelOutputContentItem,\n} from \"../../types/public/agent.js\";\nimport { waitAndCaptureScreenshot } from \"../utils/screenshotHandler.js\";\n\nexport const waitTool = (v3: V3, mode?: AgentToolMode) =>\n  tool({\n    description: \"Wait for a specified time\",\n    inputSchema: z.object({\n      timeMs: z.number().describe(\"Time in milliseconds\"),\n    }),\n    execute: async ({ timeMs }): Promise<WaitToolResult> => {\n      v3.logger({\n        category: \"agent\",\n        message: `Agent calling tool: wait`,\n        level: 1,\n        auxiliary: {\n          arguments: {\n            value: `Waiting for ${timeMs} milliseconds`,\n            type: \"string\",\n          },\n        },\n      });\n      await new Promise((resolve) => setTimeout(resolve, timeMs));\n      if (timeMs > 0) {\n        v3.recordAgentReplayStep({ type: \"wait\", timeMs });\n      }\n\n      // Take screenshot after wait in hybrid mode for visual feedback\n      if (mode === \"hybrid\") {\n        const page = await v3.context.awaitActivePage();\n        const screenshotBase64 = await waitAndCaptureScreenshot(page, 0);\n        return { success: true, waited: timeMs, screenshotBase64 };\n      }\n\n      return { success: true, waited: timeMs };\n    },\n    toModelOutput: (result) => {\n      if (result.success === false || result.error !== undefined) {\n        return {\n          type: \"content\",\n          value: [{ type: \"text\", text: JSON.stringify(result) }],\n        };\n      }\n\n      const content: ModelOutputContentItem[] = [\n        {\n          type: \"text\",\n          text: JSON.stringify({\n            success: result.success,\n            waited: result.waited,\n          }),\n        },\n      ];\n      if (result.screenshotBase64) {\n        content.push({\n          type: \"media\",\n          mediaType: \"image/png\",\n          data: result.screenshotBase64,\n        });\n      }\n      return { type: \"content\", value: content };\n    },\n  });\n"]}

package/dist/esm/lib/v3/agent/utils/captchaSolver.d.ts

@@ -0,0 +1,76 @@
+import type { Page } from "../../understudy/page.js";
+/** Injected into the agent message stream after a successful captcha solve. */
+export declare const CAPTCHA_SOLVED_MSG = "A captcha was automatically detected and solved \u2014 no further interaction with the captcha is needed, even if it does not visually appear solved. Do not click the captcha checkbox, widget, or challenge again. Continue with your task.";
+/** Injected into the agent message stream when the captcha solver fails. */
+export declare const CAPTCHA_ERRORED_MSG = "A captcha was detected but the automatic captcha solver failed to solve it. You may need to try a different approach or navigate around the captcha.";
+/** Appended to the system prompt (DOM/hybrid agents) when captchas auto-solve. */
+export declare const CAPTCHA_SYSTEM_PROMPT_NOTE = "Captchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself \u2014 they will be handled for you. Do not click the captcha checkbox, widget, or challenge again after it has been solved, even if it still looks unresolved. Continue with your task as if the captcha does not exist.";
+/** Appended to the CUA system prompt when captchas auto-solve. */
+export declare const CAPTCHA_CUA_SYSTEM_PROMPT_NOTE = "\n\nCaptchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself \u2014 they will be handled for you. Continue with your task as if the captcha does not exist.";
+/**
+ * Tracks Browserbase captcha solver state via console messages and provides
+ * a blocking `waitIfSolving()` that agents call before each step/action.
+ *
+ * Accepts a page-provider callback so the listener is automatically
+ * re-attached when the active page changes (e.g. popup / new tab).
+ *
+ * All concurrent callers of `waitIfSolving()` share the same underlying
+ * promise, so multiple waiters are safely resolved together.
+ */
+export declare class CaptchaSolver {
+    private solving;
+    private _solvedSinceLastConsume;
+    private _erroredSinceLastConsume;
+    private listener;
+    private attachedPage;
+    private pageProvider;
+    /** Shared promise that all concurrent waitIfSolving() callers await. */
+    private waitPromise;
+    /** Resolves the shared waitPromise. */
+    private resolveWait;
+    /** Timeout handle for the 90s deadline. */
+    private waitTimer;
+    /**
+     * Initialise with a callback that returns the current active page.
+     * The listener is lazily (re-)attached whenever the active page changes.
+     */
+    init(pageProvider: () => Promise<Page>): void;
+    /** Whether a captcha solve is currently in progress. */
+    isSolving(): boolean;
+    /**
+     * Ensure the console listener is attached to the current active page.
+     * If the active page has changed since the last call, the old listener
+     * is removed and a new one is installed.
+     */
+    ensureAttached(): Promise<void>;
+    /**
+     * Returns a promise that resolves immediately if no captcha is being
+     * solved, or blocks until the solver finishes, errors, or the 90s
+     * timeout is reached.
+     *
+     * Also re-attaches the listener to the current active page if it has
+     * changed since the last call.
+     *
+     * All concurrent callers share the same promise, so no waiter is
+     * orphaned.
+     */
+    waitIfSolving(): Promise<void>;
+    /**
+     * Returns and resets the solve event flags.
+     * Call after `waitIfSolving()` to check whether a captcha was solved
+     * (or errored) since the last consume.  This captures events even if
+     * the solve completed between two `waitIfSolving()` calls.
+     */
+    consumeSolveResult(): {
+        solved: boolean;
+        errored: boolean;
+    };
+    /**
+     * Remove the console listener and reset all state.
+     */
+    dispose(): void;
+    /** Remove the console listener from the currently attached page. */
+    private detachListener;
+    /** Resolve the shared wait promise and clear the timeout. */
+    private settle;
+}

package/dist/esm/lib/v3/agent/utils/captchaSolver.js

@@ -0,0 +1,171 @@
+const SOLVING_STARTED = "browserbase-solving-started";
+const SOLVING_FINISHED = "browserbase-solving-finished";
+const SOLVING_ERRORED = "browserbase-solving-errored";
+/** Maximum time (ms) to wait for the captcha solver before giving up. */
+const SOLVE_TIMEOUT_MS = 90_000;
+// ---------------------------------------------------------------------------
+// Shared captcha notification strings
+// ---------------------------------------------------------------------------
+/** Injected into the agent message stream after a successful captcha solve. */
+export const CAPTCHA_SOLVED_MSG = "A captcha was automatically detected and solved — no further interaction with the captcha is needed, even if it does not visually appear solved. Do not click the captcha checkbox, widget, or challenge again. Continue with your task.";
+/** Injected into the agent message stream when the captcha solver fails. */
+export const CAPTCHA_ERRORED_MSG = "A captcha was detected but the automatic captcha solver failed to solve it. You may need to try a different approach or navigate around the captcha.";
+/** Appended to the system prompt (DOM/hybrid agents) when captchas auto-solve. */
+export const CAPTCHA_SYSTEM_PROMPT_NOTE = "Captchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself — they will be handled for you. Do not click the captcha checkbox, widget, or challenge again after it has been solved, even if it still looks unresolved. Continue with your task as if the captcha does not exist.";
+/** Appended to the CUA system prompt when captchas auto-solve. */
+export const CAPTCHA_CUA_SYSTEM_PROMPT_NOTE = "\n\nCaptchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself — they will be handled for you. Continue with your task as if the captcha does not exist.";
+/**
+ * Tracks Browserbase captcha solver state via console messages and provides
+ * a blocking `waitIfSolving()` that agents call before each step/action.
+ *
+ * Accepts a page-provider callback so the listener is automatically
+ * re-attached when the active page changes (e.g. popup / new tab).
+ *
+ * All concurrent callers of `waitIfSolving()` share the same underlying
+ * promise, so multiple waiters are safely resolved together.
+ */
+export class CaptchaSolver {
+    solving = false;
+    _solvedSinceLastConsume = false;
+    _erroredSinceLastConsume = false;
+    listener = null;
+    attachedPage = null;
+    pageProvider = null;
+    /** Shared promise that all concurrent waitIfSolving() callers await. */
+    waitPromise = null;
+    /** Resolves the shared waitPromise. */
+    resolveWait = null;
+    /** Timeout handle for the 90s deadline. */
+    waitTimer = null;
+    /**
+     * Initialise with a callback that returns the current active page.
+     * The listener is lazily (re-)attached whenever the active page changes.
+     */
+    init(pageProvider) {
+        this.pageProvider = pageProvider;
+    }
+    /** Whether a captcha solve is currently in progress. */
+    isSolving() {
+        return this.solving;
+    }
+    /**
+     * Ensure the console listener is attached to the current active page.
+     * If the active page has changed since the last call, the old listener
+     * is removed and a new one is installed.
+     */
+    async ensureAttached() {
+        if (!this.pageProvider)
+            return;
+        const page = await this.pageProvider();
+        if (page === this.attachedPage)
+            return;
+        // Detach from the old page
+        this.detachListener();
+        this.attachedPage = page;
+        this.listener = (msg) => {
+            const text = msg.text();
+            if (text === SOLVING_STARTED) {
+                this.solving = true;
+            }
+            else if (text === SOLVING_FINISHED) {
+                this.solving = false;
+                this._solvedSinceLastConsume = true;
+                this.settle();
+            }
+            else if (text === SOLVING_ERRORED) {
+                this.solving = false;
+                this._erroredSinceLastConsume = true;
+                this.settle();
+            }
+        };
+        page.on("console", this.listener);
+    }
+    /**
+     * Returns a promise that resolves immediately if no captcha is being
+     * solved, or blocks until the solver finishes, errors, or the 90s
+     * timeout is reached.
+     *
+     * Also re-attaches the listener to the current active page if it has
+     * changed since the last call.
+     *
+     * All concurrent callers share the same promise, so no waiter is
+     * orphaned.
+     */
+    async waitIfSolving() {
+        await this.ensureAttached();
+        if (!this.solving)
+            return;
+        // Return the existing shared promise if one is already pending
+        if (this.waitPromise)
+            return this.waitPromise;
+        this.waitPromise = new Promise((resolve) => {
+            this.resolveWait = resolve;
+            this.waitTimer = setTimeout(() => {
+                this.solving = false;
+                this._erroredSinceLastConsume = true;
+                this.settle();
+            }, SOLVE_TIMEOUT_MS);
+        });
+        return this.waitPromise;
+    }
+    /**
+     * Returns and resets the solve event flags.
+     * Call after `waitIfSolving()` to check whether a captcha was solved
+     * (or errored) since the last consume.  This captures events even if
+     * the solve completed between two `waitIfSolving()` calls.
+     */
+    consumeSolveResult() {
+        const result = {
+            solved: this._solvedSinceLastConsume,
+            errored: this._erroredSinceLastConsume,
+        };
+        this._solvedSinceLastConsume = false;
+        this._erroredSinceLastConsume = false;
+        return result;
+    }
+    /**
+     * Remove the console listener and reset all state.
+     */
+    dispose() {
+        this.detachListener();
+        this.attachedPage = null;
+        this.pageProvider = null;
+        this.solving = false;
+        this._solvedSinceLastConsume = false;
+        this._erroredSinceLastConsume = false;
+        this.settle();
+    }
+    // ------------------------------------------------------------------
+    // Internal helpers
+    // ------------------------------------------------------------------
+    /** Remove the console listener from the currently attached page. */
+    detachListener() {
+        if (this.attachedPage && this.listener) {
+            this.attachedPage.off("console", this.listener);
+        }
+        this.listener = null;
+        // If a solve was in progress, mark it as errored so consumers
+        // know it was interrupted (consistent with the timeout path).
+        if (this.solving) {
+            this._erroredSinceLastConsume = true;
+        }
+        // Reset solving state so waiters aren't stuck waiting for events
+        // that can never arrive from the detached page.
+        this.solving = false;
+        this.settle();
+    }
+    /** Resolve the shared wait promise and clear the timeout. */
+    settle() {
+        if (this.waitTimer) {
+            clearTimeout(this.waitTimer);
+            this.waitTimer = null;
+        }
+        if (this.resolveWait) {
+            const resolve = this.resolveWait;
+            this.resolveWait = null;
+            this.waitPromise = null;
+            resolve();
+        }
+    }
+}
+//# sourceMappingURL=captchaSolver.js.map

package/dist/esm/lib/v3/agent/utils/captchaSolver.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"captchaSolver.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/utils/captchaSolver.ts"],"names":[],"mappings":"AAGA,MAAM,eAAe,GAAG,6BAA6B,CAAC;AACtD,MAAM,gBAAgB,GAAG,8BAA8B,CAAC;AACxD,MAAM,eAAe,GAAG,6BAA6B,CAAC;AAEtD,yEAAyE;AACzE,MAAM,gBAAgB,GAAG,MAAM,CAAC;AAEhC,8EAA8E;AAC9E,sCAAsC;AACtC,8EAA8E;AAE9E,+EAA+E;AAC/E,MAAM,CAAC,MAAM,kBAAkB,GAC7B,0OAA0O,CAAC;AAE7O,4EAA4E;AAC5E,MAAM,CAAC,MAAM,mBAAmB,GAC9B,sJAAsJ,CAAC;AAEzJ,kFAAkF;AAClF,MAAM,CAAC,MAAM,0BAA0B,GACrC,6WAA6W,CAAC;AAEhX,kEAAkE;AAClE,MAAM,CAAC,MAAM,8BAA8B,GACzC,sPAAsP,CAAC;AAEzP;;;;;;;;;GASG;AACH,MAAM,OAAO,aAAa;IAChB,OAAO,GAAG,KAAK,CAAC;IAChB,uBAAuB,GAAG,KAAK,CAAC;IAChC,wBAAwB,GAAG,KAAK,CAAC;IACjC,QAAQ,GAA2C,IAAI,CAAC;IACxD,YAAY,GAAgB,IAAI,CAAC;IACjC,YAAY,GAAiC,IAAI,CAAC;IAE1D,wEAAwE;IAChE,WAAW,GAAyB,IAAI,CAAC;IACjD,uCAAuC;IAC/B,WAAW,GAAwB,IAAI,CAAC;IAChD,2CAA2C;IACnC,SAAS,GAAyC,IAAI,CAAC;IAE/D;;;OAGG;IACH,IAAI,CAAC,YAAiC;QACpC,IAAI,CAAC,YAAY,GAAG,YAAY,CAAC;IACnC,CAAC;IAED,wDAAwD;IACxD,SAAS;QACP,OAAO,IAAI,CAAC,OAAO,CAAC;IACtB,CAAC;IAED;;;;OAIG;IACH,KAAK,CAAC,cAAc;QAClB,IAAI,CAAC,IAAI,CAAC,YAAY;YAAE,OAAO;QAC/B,MAAM,IAAI,GAAG,MAAM,IAAI,CAAC,YAAY,EAAE,CAAC;QACvC,IAAI,IAAI,KAAK,IAAI,CAAC,YAAY;YAAE,OAAO;QAEvC,2BAA2B;QAC3B,IAAI,CAAC,cAAc,EAAE,CAAC;QAEtB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;QACzB,IAAI,CAAC,QAAQ,GAAG,CAAC,GAAmB,EAAE,EAAE;YACtC,MAAM,IAAI,GAAG,GAAG,CAAC,IAAI,EAAE,CAAC;YACxB,IAAI,IAAI,KAAK,eAAe,EAAE,CAAC;gBAC7B,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;YACtB,CAAC;iBAAM,IAAI,IAAI,KAAK,gBAAgB,EAAE,CAAC;gBACrC,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;gBACrB,IAAI,CAAC,uBAAuB,GAAG,IAAI,CAAC;gBACpC,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,CAAC;iBAAM,IAAI,IAAI,KAAK,eAAe,EAAE,CAAC;gBACpC,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;gBACrB,IAAI,CAAC,wBAAwB,GAAG,IAAI,CAAC;gBACrC,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,CAAC;QACH,CAAC,CAAC;QACF,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC;IACpC,CAAC;IAED;;;;;;;;;;OAUG;IACH,KAAK,CAAC,aAAa;QACjB,MAAM,IAAI,CAAC,cAAc,EAAE,CAAC;QAE5B,IAAI,CAAC,IAAI,CAAC,OAAO;YAAE,OAAO;QAE1B,+DAA+D;QAC/D,IAAI,IAAI,CAAC,WAAW;YAAE,OAAO,IAAI,CAAC,WAAW,CAAC;QAE9C,IAAI,CAAC,WAAW,GAAG,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,EAAE;YAC/C,IAAI,CAAC,WAAW,GAAG,OAAO,CAAC;YAC3B,IAAI,CAAC,SAAS,GAAG,UAAU,CAAC,GAAG,EAAE;gBAC/B,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;gBACrB,IAAI,CAAC,wBAAwB,GAAG,IAAI,CAAC;gBACrC,IAAI,CAAC,MAAM,EAAE,CAAC;YAChB,CAAC,EAAE,gBAAgB,CAAC,CAAC;QACvB,CAAC,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,WAAW,CAAC;IAC1B,CAAC;IAED;;;;;OAKG;IACH,kBAAkB;QAChB,MAAM,MAAM,GAAG;YACb,MAAM,EAAE,IAAI,CAAC,uBAAuB;YACpC,OAAO,EAAE,IAAI,CAAC,wBAAwB;SACvC,CAAC;QACF,IAAI,CAAC,uBAAuB,GAAG,KAAK,CAAC;QACrC,IAAI,CAAC,wBAAwB,GAAG,KAAK,CAAC;QACtC,OAAO,MAAM,CAAC;IAChB,CAAC;IAED;;OAEG;IACH,OAAO;QACL,IAAI,CAAC,cAAc,EAAE,CAAC;QACtB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;QACzB,IAAI,CAAC,YAAY,GAAG,IAAI,CAAC;QACzB,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;QACrB,IAAI,CAAC,uBAAuB,GAAG,KAAK,CAAC;QACrC,IAAI,CAAC,wBAAwB,GAAG,KAAK,CAAC;QACtC,IAAI,CAAC,MAAM,EAAE,CAAC;IAChB,CAAC;IAED,qEAAqE;IACrE,mBAAmB;IACnB,qEAAqE;IAErE,oEAAoE;IAC5D,cAAc;QACpB,IAAI,IAAI,CAAC,YAAY,IAAI,IAAI,CAAC,QAAQ,EAAE,CAAC;YACvC,IAAI,CAAC,YAAY,CAAC,GAAG,CAAC,SAAS,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC;QAClD,CAAC;QACD,IAAI,CAAC,QAAQ,GAAG,IAAI,CAAC;QACrB,8DAA8D;QAC9D,8DAA8D;QAC9D,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;YACjB,IAAI,CAAC,wBAAwB,GAAG,IAAI,CAAC;QACvC,CAAC;QACD,iEAAiE;QACjE,gDAAgD;QAChD,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;QACrB,IAAI,CAAC,MAAM,EAAE,CAAC;IAChB,CAAC;IAED,6DAA6D;IACrD,MAAM;QACZ,IAAI,IAAI,CAAC,SAAS,EAAE,CAAC;YACnB,YAAY,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;YAC7B,IAAI,CAAC,SAAS,GAAG,IAAI,CAAC;QACxB,CAAC;QACD,IAAI,IAAI,CAAC,WAAW,EAAE,CAAC;YACrB,MAAM,OAAO,GAAG,IAAI,CAAC,WAAW,CAAC;YACjC,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;YACxB,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;YACxB,OAAO,EAAE,CAAC;QACZ,CAAC;IACH,CAAC;CACF","sourcesContent":["import type { Page } from \"../../understudy/page.js\";\nimport type { ConsoleMessage } from \"../../understudy/consoleMessage.js\";\n\nconst SOLVING_STARTED = \"browserbase-solving-started\";\nconst SOLVING_FINISHED = \"browserbase-solving-finished\";\nconst SOLVING_ERRORED = \"browserbase-solving-errored\";\n\n/** Maximum time (ms) to wait for the captcha solver before giving up. */\nconst SOLVE_TIMEOUT_MS = 90_000;\n\n// ---------------------------------------------------------------------------\n// Shared captcha notification strings\n// ---------------------------------------------------------------------------\n\n/** Injected into the agent message stream after a successful captcha solve. */\nexport const CAPTCHA_SOLVED_MSG =\n  \"A captcha was automatically detected and solved — no further interaction with the captcha is needed, even if it does not visually appear solved. Do not click the captcha checkbox, widget, or challenge again. Continue with your task.\";\n\n/** Injected into the agent message stream when the captcha solver fails. */\nexport const CAPTCHA_ERRORED_MSG =\n  \"A captcha was detected but the automatic captcha solver failed to solve it. You may need to try a different approach or navigate around the captcha.\";\n\n/** Appended to the system prompt (DOM/hybrid agents) when captchas auto-solve. */\nexport const CAPTCHA_SYSTEM_PROMPT_NOTE =\n  \"Captchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself — they will be handled for you. Do not click the captcha checkbox, widget, or challenge again after it has been solved, even if it still looks unresolved. Continue with your task as if the captcha does not exist.\";\n\n/** Appended to the CUA system prompt when captchas auto-solve. */\nexport const CAPTCHA_CUA_SYSTEM_PROMPT_NOTE =\n  \"\\n\\nCaptchas on this page are automatically detected and solved by the browser environment. Do not interact with or attempt to solve any captchas yourself — they will be handled for you. Continue with your task as if the captcha does not exist.\";\n\n/**\n * Tracks Browserbase captcha solver state via console messages and provides\n * a blocking `waitIfSolving()` that agents call before each step/action.\n *\n * Accepts a page-provider callback so the listener is automatically\n * re-attached when the active page changes (e.g. popup / new tab).\n *\n * All concurrent callers of `waitIfSolving()` share the same underlying\n * promise, so multiple waiters are safely resolved together.\n */\nexport class CaptchaSolver {\n  private solving = false;\n  private _solvedSinceLastConsume = false;\n  private _erroredSinceLastConsume = false;\n  private listener: ((msg: ConsoleMessage) => void) | null = null;\n  private attachedPage: Page | null = null;\n  private pageProvider: (() => Promise<Page>) | null = null;\n\n  /** Shared promise that all concurrent waitIfSolving() callers await. */\n  private waitPromise: Promise<void> | null = null;\n  /** Resolves the shared waitPromise. */\n  private resolveWait: (() => void) | null = null;\n  /** Timeout handle for the 90s deadline. */\n  private waitTimer: ReturnType<typeof setTimeout> | null = null;\n\n  /**\n   * Initialise with a callback that returns the current active page.\n   * The listener is lazily (re-)attached whenever the active page changes.\n   */\n  init(pageProvider: () => Promise<Page>): void {\n    this.pageProvider = pageProvider;\n  }\n\n  /** Whether a captcha solve is currently in progress. */\n  isSolving(): boolean {\n    return this.solving;\n  }\n\n  /**\n   * Ensure the console listener is attached to the current active page.\n   * If the active page has changed since the last call, the old listener\n   * is removed and a new one is installed.\n   */\n  async ensureAttached(): Promise<void> {\n    if (!this.pageProvider) return;\n    const page = await this.pageProvider();\n    if (page === this.attachedPage) return;\n\n    // Detach from the old page\n    this.detachListener();\n\n    this.attachedPage = page;\n    this.listener = (msg: ConsoleMessage) => {\n      const text = msg.text();\n      if (text === SOLVING_STARTED) {\n        this.solving = true;\n      } else if (text === SOLVING_FINISHED) {\n        this.solving = false;\n        this._solvedSinceLastConsume = true;\n        this.settle();\n      } else if (text === SOLVING_ERRORED) {\n        this.solving = false;\n        this._erroredSinceLastConsume = true;\n        this.settle();\n      }\n    };\n    page.on(\"console\", this.listener);\n  }\n\n  /**\n   * Returns a promise that resolves immediately if no captcha is being\n   * solved, or blocks until the solver finishes, errors, or the 90s\n   * timeout is reached.\n   *\n   * Also re-attaches the listener to the current active page if it has\n   * changed since the last call.\n   *\n   * All concurrent callers share the same promise, so no waiter is\n   * orphaned.\n   */\n  async waitIfSolving(): Promise<void> {\n    await this.ensureAttached();\n\n    if (!this.solving) return;\n\n    // Return the existing shared promise if one is already pending\n    if (this.waitPromise) return this.waitPromise;\n\n    this.waitPromise = new Promise<void>((resolve) => {\n      this.resolveWait = resolve;\n      this.waitTimer = setTimeout(() => {\n        this.solving = false;\n        this._erroredSinceLastConsume = true;\n        this.settle();\n      }, SOLVE_TIMEOUT_MS);\n    });\n\n    return this.waitPromise;\n  }\n\n  /**\n   * Returns and resets the solve event flags.\n   * Call after `waitIfSolving()` to check whether a captcha was solved\n   * (or errored) since the last consume.  This captures events even if\n   * the solve completed between two `waitIfSolving()` calls.\n   */\n  consumeSolveResult(): { solved: boolean; errored: boolean } {\n    const result = {\n      solved: this._solvedSinceLastConsume,\n      errored: this._erroredSinceLastConsume,\n    };\n    this._solvedSinceLastConsume = false;\n    this._erroredSinceLastConsume = false;\n    return result;\n  }\n\n  /**\n   * Remove the console listener and reset all state.\n   */\n  dispose(): void {\n    this.detachListener();\n    this.attachedPage = null;\n    this.pageProvider = null;\n    this.solving = false;\n    this._solvedSinceLastConsume = false;\n    this._erroredSinceLastConsume = false;\n    this.settle();\n  }\n\n  // ------------------------------------------------------------------\n  // Internal helpers\n  // ------------------------------------------------------------------\n\n  /** Remove the console listener from the currently attached page. */\n  private detachListener(): void {\n    if (this.attachedPage && this.listener) {\n      this.attachedPage.off(\"console\", this.listener);\n    }\n    this.listener = null;\n    // If a solve was in progress, mark it as errored so consumers\n    // know it was interrupted (consistent with the timeout path).\n    if (this.solving) {\n      this._erroredSinceLastConsume = true;\n    }\n    // Reset solving state so waiters aren't stuck waiting for events\n    // that can never arrive from the detached page.\n    this.solving = false;\n    this.settle();\n  }\n\n  /** Resolve the shared wait promise and clear the timeout. */\n  private settle(): void {\n    if (this.waitTimer) {\n      clearTimeout(this.waitTimer);\n      this.waitTimer = null;\n    }\n    if (this.resolveWait) {\n      const resolve = this.resolveWait;\n      this.resolveWait = null;\n      this.waitPromise = null;\n      resolve();\n    }\n  }\n}\n"]}

package/dist/esm/lib/v3/agent/utils/handleDoneToolCall.js

@@ -72,6 +72,10 @@ Call the "done" tool with:
         messages: [...inputMessages, userPrompt],
         tools: { done: doneTool },
         toolChoice: { type: "tool", toolName: "done" },
+        providerOptions: {
+            google: { mediaResolution: "MEDIA_RESOLUTION_HIGH" },
+            openai: { store: false },
+        },
     });
     const doneToolCall = result.toolCalls.find((tc) => tc.toolName === "done");
     const outputMessages = [

package/dist/esm/lib/v3/agent/utils/handleDoneToolCall.js.map

@@ -1 +1 @@
-{"version":3,"file":"handleDoneToolCall.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/utils/handleDoneToolCall.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAwC,MAAM,IAAI,CAAC;AACxE,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAG1B,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAUhD,SAAS,mBAAmB,CAAC,OAAiB;IAC5C,OAAO,OAAO,CAAC,MAAM,CAAC;QACpB,SAAS,EAAE,OAAO;aACf,MAAM,EAAE;aACR,QAAQ,CAAC,0DAA0D,CAAC;QACvE,YAAY,EAAE,OAAO;aAClB,OAAO,EAAE;aACT,QAAQ,CAAC,uDAAuD,CAAC;KACrE,CAAC,CAAC;AACL,CAAC;AAED;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,OAMxC;IACC,MAAM,EAAE,KAAK,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,EAAE,GAAG,OAAO,CAAC;IAE5E,MAAM,CAAC;QACL,QAAQ,EAAE,OAAO;QACjB,OAAO,EAAE,0BAA0B;QACnC,KAAK,EAAE,CAAC;KACT,CAAC,CAAC;IACH,4EAA4E;IAC5E,MAAM,OAAO,GAAG,YAAY;QAC1B,CAAC,CAAC,WAAW,CAAC,YAAkC,CAAC;QACjD,CAAC,CAAC,CAAC,CAAC;IACN,MAAM,cAAc,GAAG,mBAAmB,CAAC,OAAO,CAAC,CAAC;IAEpD,qEAAqE;IACrE,MAAM,cAAc,GAAG,YAAY;QACjC,CAAC,CAAC,cAAc,CAAC,MAAM,CAAC;YACpB,MAAM,EAAE,YAAY,CAAC,QAAQ,CAC3B,qDAAqD,CACtD;SACF,CAAC;QACJ,CAAC,CAAC,cAAc,CAAC;IAEnB,MAAM,kBAAkB,GAAG,YAAY;QACrC,CAAC,CAAC,4GAA4G,IAAI,CAAC,SAAS,CACxH,MAAM,CAAC,WAAW,CAChB,MAAM,CAAC,OAAO,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC,GAAG,CACpC,CAAC,CAAC,GAAG,EAAE,KAAK,CAA+B,EAAE,EAAE,CAAC;YAC9C,GAAG;YACH,KAAK,CAAC,WAAW,IAAI,gBAAgB;SACtC,CACF,CACF,EACD,IAAI,EACJ,CAAC,CACF,EAAE;QACL,CAAC,CAAC,EAAE,CAAC;IAEP,MAAM,YAAY,GAAG;;;GAGpB,WAAW;;mPAEqO,kBAAkB;;;;gDAIrN,YAAY,CAAC,CAAC,CAAC,wDAAwD,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC;IAE7H,MAAM,QAAQ,GAAG,IAAI,CAAC;QACpB,WAAW,EAAE,YAAY;YACvB,CAAC,CAAC,uEAAuE;YACzE,CAAC,CAAC,+CAA+C;QACnD,WAAW,EAAE,cAAc;QAC3B,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,EAAE;YACxB,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,GAAG,MAAM,EAAE,CAAC;QACtC,CAAC;KACF,CAAC,CAAC;IAEH,MAAM,UAAU,GAAiB;QAC/B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,YAAY;YACnB,CAAC,CAAC,8DAA8D;YAChE,CAAC,CAAC,gCAAgC;KACrC,CAAC;IAEF,MAAM,MAAM,GAAG,MAAM,YAAY,CAAC;QAChC,KAAK;QACL,MAAM,EAAE,YAAY;QACpB,QAAQ,EAAE,CAAC,GAAG,aAAa,EAAE,UAAU,CAAC;QACxC,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAa;QACpC,UAAU,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,EAAE;KAC/C,CAAC,CAAC;IAEH,MAAM,YAAY,GAAG,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,QAAQ,KAAK,MAAM,CAAC,CAAC;IAC3E,MAAM,cAAc,GAAmB;QACrC,UAAU;QACV,GAAG,CAAC,MAAM,CAAC,QAAQ,EAAE,QAAQ,IAAI,EAAE,CAAC;KACrC,CAAC;IAEF,IAAI,CAAC,YAAY,EAAE,CAAC;QAClB,OAAO;YACL,SAAS,EAAE,MAAM,CAAC,IAAI,IAAI,0BAA0B;YACpD,YAAY,EAAE,KAAK;YACnB,QAAQ,EAAE,cAAc;SACzB,CAAC;IACJ,CAAC;IAED,MAAM,KAAK,GAAG,YAAY,CAAC,KAI1B,CAAC;IACF,MAAM,CAAC;QACL,QAAQ,EAAE,OAAO;QACjB,OAAO,EAAE,gBAAgB;QACzB,KAAK,EAAE,CAAC;KACT,CAAC,CAAC;IAEH,OAAO;QACL,SAAS,EAAE,KAAK,CAAC,SAAS;QAC1B,YAAY,EAAE,KAAK,CAAC,YAAY;QAChC,QAAQ,EAAE,cAAc;QACxB,MAAM,EAAE,KAAK,CAAC,MAAM;KACrB,CAAC;AACJ,CAAC","sourcesContent":["import { generateText, ModelMessage, LanguageModel, ToolSet } from \"ai\";\nimport { z } from \"zod\";\nimport { tool } from \"ai\";\nimport { LogLine } from \"../../types/public/logs.js\";\nimport { StagehandZodObject } from \"../../zodCompat.js\";\nimport { getZFactory } from \"../../../utils.js\";\nimport type { StagehandZodSchema } from \"../../zodCompat.js\";\n\ninterface DoneResult {\n  reasoning: string;\n  taskComplete: boolean;\n  messages: ModelMessage[];\n  output?: Record<string, unknown>;\n}\n\nfunction buildBaseDoneSchema(factory: typeof z) {\n  return factory.object({\n    reasoning: factory\n      .string()\n      .describe(\"Brief summary of what actions were taken and the outcome\"),\n    taskComplete: factory\n      .boolean()\n      .describe(\"true if the task was fully completed, false otherwise\"),\n  });\n}\n\n/**\n * Force a done tool call at the end of an agent run.\n * This ensures we always get a structured final response,\n * even if the main loop ended without calling done.\n */\nexport async function handleDoneToolCall(options: {\n  model: LanguageModel;\n  inputMessages: ModelMessage[];\n  instruction: string;\n  outputSchema?: StagehandZodObject;\n  logger: (message: LogLine) => void;\n}): Promise<DoneResult> {\n  const { model, inputMessages, instruction, outputSchema, logger } = options;\n\n  logger({\n    category: \"agent\",\n    message: \"Agent calling tool: done\",\n    level: 1,\n  });\n  // Use the same Zod version as the user's outputSchema to avoid v3/v4 mixing\n  const factory = outputSchema\n    ? getZFactory(outputSchema as StagehandZodSchema)\n    : z;\n  const baseDoneSchema = buildBaseDoneSchema(factory);\n\n  // Merge base done schema with user-provided output schema if present\n  const doneToolSchema = outputSchema\n    ? baseDoneSchema.extend({\n        output: outputSchema.describe(\n          \"The specific data the user requested from this task\",\n        ),\n      })\n    : baseDoneSchema;\n\n  const outputInstructions = outputSchema\n    ? `\\n\\nThe user also requested the following information from this task. Provide it in the \"output\" field:\\n${JSON.stringify(\n        Object.fromEntries(\n          Object.entries(outputSchema.shape).map(\n            ([key, value]: [string, StagehandZodSchema]) => [\n              key,\n              value.description || \"no description\",\n            ],\n          ),\n        ),\n        null,\n        2,\n      )}`\n    : \"\";\n\n  const systemPrompt = `You are a web automation assistant that was tasked with completing a task.\n\nThe task was:\n\"${instruction}\"\n\nReview 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.${outputInstructions}\n\nCall the \"done\" tool with:\n1. A brief summary of what was done\n2. Whether the task was completed successfully${outputSchema ? \"\\n3. The requested output data based on what you found\" : \"\"}`;\n\n  const doneTool = tool({\n    description: outputSchema\n      ? \"Complete the task with your assessment and the requested output data.\"\n      : \"Complete the task with your final assessment.\",\n    inputSchema: doneToolSchema,\n    execute: async (params) => {\n      return { success: true, ...params };\n    },\n  });\n\n  const userPrompt: ModelMessage = {\n    role: \"user\",\n    content: outputSchema\n      ? \"Provide your final assessment and the requested output data.\"\n      : \"Provide your final assessment.\",\n  };\n\n  const result = await generateText({\n    model,\n    system: systemPrompt,\n    messages: [...inputMessages, userPrompt],\n    tools: { done: doneTool } as ToolSet,\n    toolChoice: { type: \"tool\", toolName: \"done\" },\n  });\n\n  const doneToolCall = result.toolCalls.find((tc) => tc.toolName === \"done\");\n  const outputMessages: ModelMessage[] = [\n    userPrompt,\n    ...(result.response?.messages || []),\n  ];\n\n  if (!doneToolCall) {\n    return {\n      reasoning: result.text || \"Task execution completed\",\n      taskComplete: false,\n      messages: outputMessages,\n    };\n  }\n\n  const input = doneToolCall.input as {\n    reasoning: string;\n    taskComplete: boolean;\n    output?: Record<string, unknown>;\n  };\n  logger({\n    category: \"agent\",\n    message: `Task completed`,\n    level: 1,\n  });\n\n  return {\n    reasoning: input.reasoning,\n    taskComplete: input.taskComplete,\n    messages: outputMessages,\n    output: input.output,\n  };\n}\n"]}
+{"version":3,"file":"handleDoneToolCall.js","sourceRoot":"","sources":["../../../../../../lib/v3/agent/utils/handleDoneToolCall.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAwC,MAAM,IAAI,CAAC;AACxE,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AACxB,OAAO,EAAE,IAAI,EAAE,MAAM,IAAI,CAAC;AAG1B,OAAO,EAAE,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAUhD,SAAS,mBAAmB,CAAC,OAAiB;IAC5C,OAAO,OAAO,CAAC,MAAM,CAAC;QACpB,SAAS,EAAE,OAAO;aACf,MAAM,EAAE;aACR,QAAQ,CAAC,0DAA0D,CAAC;QACvE,YAAY,EAAE,OAAO;aAClB,OAAO,EAAE;aACT,QAAQ,CAAC,uDAAuD,CAAC;KACrE,CAAC,CAAC;AACL,CAAC;AAED;;;;GAIG;AACH,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,OAMxC;IACC,MAAM,EAAE,KAAK,EAAE,aAAa,EAAE,WAAW,EAAE,YAAY,EAAE,MAAM,EAAE,GAAG,OAAO,CAAC;IAE5E,MAAM,CAAC;QACL,QAAQ,EAAE,OAAO;QACjB,OAAO,EAAE,0BAA0B;QACnC,KAAK,EAAE,CAAC;KACT,CAAC,CAAC;IACH,4EAA4E;IAC5E,MAAM,OAAO,GAAG,YAAY;QAC1B,CAAC,CAAC,WAAW,CAAC,YAAkC,CAAC;QACjD,CAAC,CAAC,CAAC,CAAC;IACN,MAAM,cAAc,GAAG,mBAAmB,CAAC,OAAO,CAAC,CAAC;IAEpD,qEAAqE;IACrE,MAAM,cAAc,GAAG,YAAY;QACjC,CAAC,CAAC,cAAc,CAAC,MAAM,CAAC;YACpB,MAAM,EAAE,YAAY,CAAC,QAAQ,CAC3B,qDAAqD,CACtD;SACF,CAAC;QACJ,CAAC,CAAC,cAAc,CAAC;IAEnB,MAAM,kBAAkB,GAAG,YAAY;QACrC,CAAC,CAAC,4GAA4G,IAAI,CAAC,SAAS,CACxH,MAAM,CAAC,WAAW,CAChB,MAAM,CAAC,OAAO,CAAC,YAAY,CAAC,KAAK,CAAC,CAAC,GAAG,CACpC,CAAC,CAAC,GAAG,EAAE,KAAK,CAA+B,EAAE,EAAE,CAAC;YAC9C,GAAG;YACH,KAAK,CAAC,WAAW,IAAI,gBAAgB;SACtC,CACF,CACF,EACD,IAAI,EACJ,CAAC,CACF,EAAE;QACL,CAAC,CAAC,EAAE,CAAC;IAEP,MAAM,YAAY,GAAG;;;GAGpB,WAAW;;mPAEqO,kBAAkB;;;;gDAIrN,YAAY,CAAC,CAAC,CAAC,wDAAwD,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC;IAE7H,MAAM,QAAQ,GAAG,IAAI,CAAC;QACpB,WAAW,EAAE,YAAY;YACvB,CAAC,CAAC,uEAAuE;YACzE,CAAC,CAAC,+CAA+C;QACnD,WAAW,EAAE,cAAc;QAC3B,OAAO,EAAE,KAAK,EAAE,MAAM,EAAE,EAAE;YACxB,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,GAAG,MAAM,EAAE,CAAC;QACtC,CAAC;KACF,CAAC,CAAC;IAEH,MAAM,UAAU,GAAiB;QAC/B,IAAI,EAAE,MAAM;QACZ,OAAO,EAAE,YAAY;YACnB,CAAC,CAAC,8DAA8D;YAChE,CAAC,CAAC,gCAAgC;KACrC,CAAC;IAEF,MAAM,MAAM,GAAG,MAAM,YAAY,CAAC;QAChC,KAAK;QACL,MAAM,EAAE,YAAY;QACpB,QAAQ,EAAE,CAAC,GAAG,aAAa,EAAE,UAAU,CAAC;QACxC,KAAK,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAa;QACpC,UAAU,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,MAAM,EAAE;QAC9C,eAAe,EAAE;YACf,MAAM,EAAE,EAAE,eAAe,EAAE,uBAAuB,EAAE;YACpD,MAAM,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE;SACzB;KACF,CAAC,CAAC;IAEH,MAAM,YAAY,GAAG,MAAM,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,QAAQ,KAAK,MAAM,CAAC,CAAC;IAC3E,MAAM,cAAc,GAAmB;QACrC,UAAU;QACV,GAAG,CAAC,MAAM,CAAC,QAAQ,EAAE,QAAQ,IAAI,EAAE,CAAC;KACrC,CAAC;IAEF,IAAI,CAAC,YAAY,EAAE,CAAC;QAClB,OAAO;YACL,SAAS,EAAE,MAAM,CAAC,IAAI,IAAI,0BAA0B;YACpD,YAAY,EAAE,KAAK;YACnB,QAAQ,EAAE,cAAc;SACzB,CAAC;IACJ,CAAC;IAED,MAAM,KAAK,GAAG,YAAY,CAAC,KAI1B,CAAC;IACF,MAAM,CAAC;QACL,QAAQ,EAAE,OAAO;QACjB,OAAO,EAAE,gBAAgB;QACzB,KAAK,EAAE,CAAC;KACT,CAAC,CAAC;IAEH,OAAO;QACL,SAAS,EAAE,KAAK,CAAC,SAAS;QAC1B,YAAY,EAAE,KAAK,CAAC,YAAY;QAChC,QAAQ,EAAE,cAAc;QACxB,MAAM,EAAE,KAAK,CAAC,MAAM;KACrB,CAAC;AACJ,CAAC","sourcesContent":["import { generateText, ModelMessage, LanguageModel, ToolSet } from \"ai\";\nimport { z } from \"zod\";\nimport { tool } from \"ai\";\nimport { LogLine } from \"../../types/public/logs.js\";\nimport { StagehandZodObject } from \"../../zodCompat.js\";\nimport { getZFactory } from \"../../../utils.js\";\nimport type { StagehandZodSchema } from \"../../zodCompat.js\";\n\ninterface DoneResult {\n  reasoning: string;\n  taskComplete: boolean;\n  messages: ModelMessage[];\n  output?: Record<string, unknown>;\n}\n\nfunction buildBaseDoneSchema(factory: typeof z) {\n  return factory.object({\n    reasoning: factory\n      .string()\n      .describe(\"Brief summary of what actions were taken and the outcome\"),\n    taskComplete: factory\n      .boolean()\n      .describe(\"true if the task was fully completed, false otherwise\"),\n  });\n}\n\n/**\n * Force a done tool call at the end of an agent run.\n * This ensures we always get a structured final response,\n * even if the main loop ended without calling done.\n */\nexport async function handleDoneToolCall(options: {\n  model: LanguageModel;\n  inputMessages: ModelMessage[];\n  instruction: string;\n  outputSchema?: StagehandZodObject;\n  logger: (message: LogLine) => void;\n}): Promise<DoneResult> {\n  const { model, inputMessages, instruction, outputSchema, logger } = options;\n\n  logger({\n    category: \"agent\",\n    message: \"Agent calling tool: done\",\n    level: 1,\n  });\n  // Use the same Zod version as the user's outputSchema to avoid v3/v4 mixing\n  const factory = outputSchema\n    ? getZFactory(outputSchema as StagehandZodSchema)\n    : z;\n  const baseDoneSchema = buildBaseDoneSchema(factory);\n\n  // Merge base done schema with user-provided output schema if present\n  const doneToolSchema = outputSchema\n    ? baseDoneSchema.extend({\n        output: outputSchema.describe(\n          \"The specific data the user requested from this task\",\n        ),\n      })\n    : baseDoneSchema;\n\n  const outputInstructions = outputSchema\n    ? `\\n\\nThe user also requested the following information from this task. Provide it in the \"output\" field:\\n${JSON.stringify(\n        Object.fromEntries(\n          Object.entries(outputSchema.shape).map(\n            ([key, value]: [string, StagehandZodSchema]) => [\n              key,\n              value.description || \"no description\",\n            ],\n          ),\n        ),\n        null,\n        2,\n      )}`\n    : \"\";\n\n  const systemPrompt = `You are a web automation assistant that was tasked with completing a task.\n\nThe task was:\n\"${instruction}\"\n\nReview 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.${outputInstructions}\n\nCall the \"done\" tool with:\n1. A brief summary of what was done\n2. Whether the task was completed successfully${outputSchema ? \"\\n3. The requested output data based on what you found\" : \"\"}`;\n\n  const doneTool = tool({\n    description: outputSchema\n      ? \"Complete the task with your assessment and the requested output data.\"\n      : \"Complete the task with your final assessment.\",\n    inputSchema: doneToolSchema,\n    execute: async (params) => {\n      return { success: true, ...params };\n    },\n  });\n\n  const userPrompt: ModelMessage = {\n    role: \"user\",\n    content: outputSchema\n      ? \"Provide your final assessment and the requested output data.\"\n      : \"Provide your final assessment.\",\n  };\n\n  const result = await generateText({\n    model,\n    system: systemPrompt,\n    messages: [...inputMessages, userPrompt],\n    tools: { done: doneTool } as ToolSet,\n    toolChoice: { type: \"tool\", toolName: \"done\" },\n    providerOptions: {\n      google: { mediaResolution: \"MEDIA_RESOLUTION_HIGH\" },\n      openai: { store: false },\n    },\n  });\n\n  const doneToolCall = result.toolCalls.find((tc) => tc.toolName === \"done\");\n  const outputMessages: ModelMessage[] = [\n    userPrompt,\n    ...(result.response?.messages || []),\n  ];\n\n  if (!doneToolCall) {\n    return {\n      reasoning: result.text || \"Task execution completed\",\n      taskComplete: false,\n      messages: outputMessages,\n    };\n  }\n\n  const input = doneToolCall.input as {\n    reasoning: string;\n    taskComplete: boolean;\n    output?: Record<string, unknown>;\n  };\n  logger({\n    category: \"agent\",\n    message: `Task completed`,\n    level: 1,\n  });\n\n  return {\n    reasoning: input.reasoning,\n    taskComplete: input.taskComplete,\n    messages: outputMessages,\n    output: input.output,\n  };\n}\n"]}

package/dist/esm/lib/v3/api.d.ts

@@ -20,7 +20,7 @@ export declare function getApiUrlForRegion(region: BrowserbaseRegion | undefined
  */
 interface StagehandAPIConstructorParams {
     apiKey: string;
-    projectId: string;
+    projectId?: string;
     logger: (message: LogLine) => void;
     /**
      * When true, enables server-side caching by default for all requests.
@@ -73,7 +73,7 @@ interface ClientObserveParameters {
 }
 export declare class StagehandAPIClient {
     private apiKey;
-    private projectId;
+    private projectId?;
     private sessionId?;
     private modelApiKey;
     private modelProvider?;

package/dist/esm/lib/v3/api.js

@@ -542,7 +542,7 @@ export class StagehandAPIClient {
     async request(path, options, serverCache) {
         const defaultHeaders = {
             "x-bb-api-key": this.apiKey,
-            "x-bb-project-id": this.projectId,
+            ...(this.projectId ? { "x-bb-project-id": this.projectId } : {}),
             "x-bb-session-id": this.sessionId,
             // we want real-time logs, so we stream the response
             "x-stream-response": "true",