illuma-agents 1.0.30 → 1.0.31

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -186,18 +186,43 @@ function createBrowserTools(options) {
186
186
  }
187
187
  };
188
188
  };
189
- // browser_click
189
+ // browser_click - PRIMARY action for buttons and links
190
190
  tools$1.push(tools.tool(createToolFunction('click'), {
191
191
  name: EBrowserTools.CLICK,
192
- description: `Click an element by its index number from the element list (e.g., [0], [1]).
193
- Returns state verification showing if UI changed (dialog closed, form submitted, etc.).`,
192
+ description: `Click an element by its [index] from the element list.
193
+
194
+ **WHEN TO USE (prioritize this over keypress):**
195
+ - Buttons labeled "Send", "Submit", "Save", "OK", "Confirm" → CLICK them
196
+ - Links (a tags) to navigate
197
+ - Checkboxes, radio buttons to select
198
+ - Any clickable element that performs an action
199
+
200
+ **SEMANTIC ELEMENT IDENTIFICATION:**
201
+ - Look for aria-label="Send", aria-label="Submit" for action buttons
202
+ - Look for type="submit" for form submission
203
+ - Look for button text like "Send", "Submit", "Save"
204
+ - For email compose: Find the Send button, DON'T use keyboard shortcuts
205
+
206
+ Returns verification: dialog closed, form submitted, URL changed, or NO CHANGE (action may have failed).`,
194
207
  schema: BrowserClickSchema,
195
208
  }));
196
209
  // browser_type
197
210
  tools$1.push(tools.tool(createToolFunction('type'), {
198
211
  name: EBrowserTools.TYPE,
199
- description: `Type text into an input element by index. Set pressEnter: true to submit.
200
- Returns state verification showing if form submitted or errors appeared.`,
212
+ description: `Type text into an input/textarea element by [index].
213
+
214
+ **WHEN TO USE:**
215
+ - Input fields (fieldType="email", "search", "phone", etc.)
216
+ - Textareas for longer content
217
+ - Contenteditable divs (like Gmail compose body)
218
+
219
+ **OPTIONS:**
220
+ - pressEnter: true → Press Enter after typing (for search forms)
221
+ - pressEnter: false → Just type without submitting
222
+
223
+ **DO NOT use pressEnter to send emails** - find and CLICK the Send button instead.
224
+
225
+ Returns verification showing if text was entered or form was submitted.`,
201
226
  schema: BrowserTypeSchema,
202
227
  }));
203
228
  // browser_navigate
@@ -268,14 +293,26 @@ The system will download the file and attach it to the input.
268
293
  After upload, you receive an updated element list.`,
269
294
  schema: BrowserUploadFileSchema,
270
295
  }));
271
- // browser_keypress - For keyboard shortcuts and special keys
296
+ // browser_keypress - LAST RESORT for keyboard shortcuts
272
297
  tools$1.push(tools.tool(createToolFunction('keypress'), {
273
298
  name: EBrowserTools.KEYPRESS,
274
- description: `Press keyboard key(s) on the page.
275
- Single keys: "Enter", "Escape", "Tab", "ArrowDown", "ArrowUp", "Backspace", "Delete"
276
- Key combos: "Control+A" (select all), "Control+C" (copy), "Shift+Enter" (newline)
277
- Use this for form submission, closing modals, navigating dropdowns.
278
- After keypress, you receive an updated element list.`,
299
+ description: `Press keyboard key(s) - USE ONLY AS LAST RESORT after clicking fails.
300
+
301
+ **PREFER browser_click OVER keypress:**
302
+ - To send email Find and CLICK the Send button
303
+ - To submit form CLICK the Submit button
304
+ - To close modal → CLICK the X or Cancel button
305
+
306
+ **WHEN TO USE keypress:**
307
+ - Navigation: "Tab", "ArrowDown", "ArrowUp" in dropdowns
308
+ - Escape to close popups if no X button exists
309
+ - Control+A, Control+C for select/copy operations
310
+ - ONLY use Enter to submit if there's NO visible submit button
311
+
312
+ Single keys: "Enter", "Escape", "Tab", "ArrowDown"
313
+ Key combos: "Control+A", "Shift+Enter"
314
+
315
+ Returns verification showing if page state changed.`,
279
316
  schema: BrowserKeypressSchema,
280
317
  }));
281
318
  return tools$1;
@@ -1 +1 @@
1
- {"version":3,"file":"BrowserTools.cjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\nimport type * as _t from '@/types';\n\n/**\n * Browser tool names - keep in sync with ranger-browser extension\n * These tools execute locally in the browser extension, NOT on the server\n */\nexport const EBrowserTools = {\n CLICK: 'browser_click',\n TYPE: 'browser_type',\n NAVIGATE: 'browser_navigate',\n SCROLL: 'browser_scroll',\n EXTRACT: 'browser_extract',\n HOVER: 'browser_hover',\n WAIT: 'browser_wait',\n BACK: 'browser_back',\n SCREENSHOT: 'browser_screenshot',\n GET_PAGE_STATE: 'browser_get_page_state',\n // Skyvern-inspired additions for robust form handling\n SELECT_OPTION: 'browser_select_option',\n UPLOAD_FILE: 'browser_upload_file',\n KEYPRESS: 'browser_keypress',\n} as const;\n\nexport type BrowserToolName =\n (typeof EBrowserTools)[keyof typeof EBrowserTools];\n\n/**\n * Callback function type for waiting on browser action results\n * This allows the server (Ranger) to provide a callback that waits for the extension\n * to POST results back to the server before returning to the LLM.\n *\n * @param action - The browser action (click, type, navigate, etc.)\n * @param args - Arguments for the action\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\n * @returns Promise that resolves with the actual browser result (page state, etc.)\n */\nexport type BrowserToolCallback = (\n action: string,\n args: Record<string, unknown>,\n toolCallId: string\n) => Promise<BrowserActionResult>;\n\n/**\n * Action status - matches Skyvern's ActionStatus\n */\nexport type ActionStatus = 'pending' | 'completed' | 'failed' | 'skipped';\n\n/**\n * Verification status - matches Skyvern's VerificationStatus\n * - complete: Goal achieved\n * - terminate: Goal cannot be achieved, stop\n * - continue: Goal not yet achieved, keep going\n */\nexport type VerificationStatus = 'complete' | 'terminate' | 'continue';\n\n/**\n * Result returned from browser action execution\n */\nexport interface BrowserActionResult {\n success: boolean;\n status?: ActionStatus;\n url?: string;\n title?: string;\n elementList?: string;\n error?: string;\n errorType?: string; // Typed error classification\n screenshot?: string;\n elementDescription?: string;\n // State verification (before/after comparison)\n verification?: {\n urlChanged: boolean;\n titleChanged: boolean;\n elementCountDelta: number;\n significantChange: boolean;\n dialogClosed: boolean;\n formGone: boolean;\n changeDescription: string;\n };\n}\n\n/**\n * Check if browser capability is available based on request headers or context\n * The browser extension sets these headers when connected:\n * - X-Ranger-Browser-Extension: true\n * - X-Ranger-Browser-Capable: true\n */\nexport function hasBrowserCapability(req?: {\n headers?: Record<string, string | string[] | undefined>;\n}): boolean {\n if (!req?.headers) {\n return false;\n }\n\n const browserExtension = req.headers['x-ranger-browser-extension'];\n const browserCapable = req.headers['x-ranger-browser-capable'];\n\n return browserExtension === 'true' || browserCapable === 'true';\n}\n\n// Tool schemas\nconst BrowserClickSchema = z.object({\n index: z\n .number()\n .describe(\n 'The index number [0], [1], etc. of the element to click from the page state element list'\n ),\n});\n\nconst BrowserTypeSchema = z.object({\n index: z\n .number()\n .describe('The index number of the input element to type into'),\n text: z.string().describe('The text to type into the element'),\n pressEnter: z\n .boolean()\n .optional()\n .describe('Whether to press Enter after typing (useful for search forms)'),\n});\n\nconst BrowserNavigateSchema = z.object({\n url: z\n .string()\n .describe('The full URL to navigate to (must include https://)'),\n});\n\nconst BrowserScrollSchema = z.object({\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .describe('Direction to scroll'),\n amount: z\n .number()\n .optional()\n .describe('Pixels to scroll (default: one viewport height)'),\n});\n\nconst BrowserExtractSchema = z.object({\n query: z\n .string()\n .optional()\n .describe('Optional: specific content to extract from the page'),\n});\n\nconst BrowserHoverSchema = z.object({\n index: z.number().describe('The index number of the element to hover over'),\n});\n\nconst BrowserWaitSchema = z.object({\n duration: z\n .number()\n .optional()\n .describe('Milliseconds to wait (default: 1000)'),\n});\n\nconst BrowserBackSchema = z.object({});\n\nconst BrowserScreenshotSchema = z.object({});\n\nconst BrowserGetPageStateSchema = z.object({});\n\n// Skyvern-inspired schemas for robust form handling\nconst BrowserSelectOptionSchema = z.object({\n index: z\n .number()\n .describe('The index number of the select/dropdown element'),\n value: z\n .string()\n .optional()\n .describe('The value or label of the option to select. For native <select>, use the option text. For custom dropdowns, this is the option label to click.'),\n});\n\nconst BrowserUploadFileSchema = z.object({\n index: z\n .number()\n .describe('The index number of the file input element'),\n fileUrl: z\n .string()\n .describe('URL of the file to upload (the system will download and upload it)'),\n});\n\nconst BrowserKeypressSchema = z.object({\n keys: z\n .string()\n .describe('Key(s) to press. Single key: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\". Combo: \"Control+A\", \"Shift+Enter\"'),\n});\n\n/**\n * Browser tool response interface\n * This is what the extension returns after executing the action\n */\nexport interface BrowserToolResponse {\n requiresBrowserExecution: true;\n action: string;\n args: Record<string, unknown>;\n toolCallId?: string; // Added to help extension correlate with callback\n}\n\n/**\n * Options for creating browser tools\n */\nexport interface CreateBrowserToolsOptions {\n /**\n * Optional callback that waits for browser action results.\n * When provided, tools will await this callback to get actual results from the extension.\n * When not provided, tools return markers immediately (for non-server contexts).\n */\n waitForResult?: BrowserToolCallback;\n}\n\n/**\n * Format browser action result for LLM consumption\n */\nfunction formatResultForLLM(\n result: BrowserActionResult,\n action: string,\n actionArgs?: Record<string, unknown>\n): string {\n if (!result.success && result.error) {\n const errorType = result.errorType ? `[${result.errorType}] ` : '';\n return `${errorType}Action \"${action}\" failed: ${result.error}`;\n }\n\n const parts: string[] = [];\n\n // Verification result (Skyvern pattern)\n if (result.verification) {\n const v = result.verification;\n if (v.dialogClosed || v.formGone || v.urlChanged) {\n parts.push(`✓ ${v.changeDescription}`);\n } else if (!v.significantChange) {\n parts.push(`⚠ No change detected`);\n }\n }\n\n // Page state\n if (result.url) parts.push(`URL: ${result.url}`);\n if (result.title) parts.push(`Title: ${result.title}`);\n if (result.elementList) parts.push(`\\nElements:\\n${result.elementList}`);\n\n return parts.join('\\n');\n}\n\n/**\n * Create browser tools with optional callback for waiting on results\n *\n * When waitForResult callback is provided:\n * 1. Tool returns marker that triggers extension\n * 2. Tool then awaits callback to get actual results\n * 3. Returns real page state to LLM\n *\n * When no callback:\n * 1. Tool returns marker only (for non-server contexts)\n *\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\n */\nexport function createBrowserTools(\n options?: CreateBrowserToolsOptions\n): DynamicStructuredTool[] {\n const { waitForResult } = options || {};\n const tools: DynamicStructuredTool[] = [];\n\n /**\n * Helper to create tool function that optionally waits for results\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\n */\n const createToolFunction = (action: string) => {\n return async (\n args: Record<string, unknown>,\n config?: { toolCall?: { id?: string } }\n ): Promise<string> => {\n const toolCallId =\n config?.toolCall?.id ??\n `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\n\n // Create marker for extension\n const marker: BrowserToolResponse = {\n requiresBrowserExecution: true,\n action,\n args,\n toolCallId,\n };\n\n // If no callback, return marker immediately (extension handles via SSE interception)\n if (!waitForResult) {\n return JSON.stringify(marker);\n }\n\n // With callback: wait for actual results from extension\n // The marker is still returned initially via SSE, but we wait for the callback\n try {\n const result = await waitForResult(action, args, toolCallId);\n return formatResultForLLM(result, action, args);\n } catch (error) {\n const errorMessage =\n error instanceof Error ? error.message : String(error);\n return `❌ **ACTION FAILED**: Browser action \"${action}\" failed: ${errorMessage}\n\n**REQUIRED**: Analyze why the action failed and try an alternative approach.`;\n }\n };\n };\n\n // browser_click\n tools.push(\n tool(createToolFunction('click'), {\n name: EBrowserTools.CLICK,\n description: `Click an element by its index number from the element list (e.g., [0], [1]).\nReturns state verification showing if UI changed (dialog closed, form submitted, etc.).`,\n schema: BrowserClickSchema,\n })\n );\n\n // browser_type\n tools.push(\n tool(createToolFunction('type'), {\n name: EBrowserTools.TYPE,\n description: `Type text into an input element by index. Set pressEnter: true to submit.\nReturns state verification showing if form submitted or errors appeared.`,\n schema: BrowserTypeSchema,\n })\n );\n\n // browser_navigate\n tools.push(\n tool(createToolFunction('navigate'), {\n name: EBrowserTools.NAVIGATE,\n description: `Navigate to a URL (include https://). Returns new page's element list.`,\n schema: BrowserNavigateSchema,\n })\n );\n\n // browser_scroll\n tools.push(\n tool(createToolFunction('scroll'), {\n name: EBrowserTools.SCROLL,\n description: `Scroll the page. Returns updated element list with newly visible elements.`,\n schema: BrowserScrollSchema,\n })\n );\n\n // browser_extract\n tools.push(\n tool(createToolFunction('extract'), {\n name: EBrowserTools.EXTRACT,\n description: `Extract content from the current page.\nReturns page URL, title, and element list.`,\n schema: BrowserExtractSchema,\n })\n );\n\n // browser_hover\n tools.push(\n tool(createToolFunction('hover'), {\n name: EBrowserTools.HOVER,\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\n schema: BrowserHoverSchema,\n })\n );\n\n // browser_wait\n tools.push(\n tool(createToolFunction('wait'), {\n name: EBrowserTools.WAIT,\n description: `Wait for a specified duration for page content to load.\nUse this after actions that trigger async content loading.\nAfter waiting, you receive an updated element list.`,\n schema: BrowserWaitSchema,\n })\n );\n\n // browser_back\n tools.push(\n tool(createToolFunction('back'), {\n name: EBrowserTools.BACK,\n description: `Go back to the previous page in browser history.\nAfter going back, you receive the previous page's element list.`,\n schema: BrowserBackSchema,\n })\n );\n\n // browser_screenshot\n tools.push(\n tool(createToolFunction('screenshot'), {\n name: EBrowserTools.SCREENSHOT,\n description: `Capture a screenshot. Use browser_get_page_state for element list.`,\n schema: BrowserScreenshotSchema,\n })\n );\n\n // browser_get_page_state\n tools.push(\n tool(createToolFunction('get_page_state'), {\n name: EBrowserTools.GET_PAGE_STATE,\n description: `Get current page URL, title, and interactive elements list with index numbers.`,\n schema: BrowserGetPageStateSchema,\n })\n );\n\n // browser_select_option\n tools.push(\n tool(createToolFunction('select_option'), {\n name: EBrowserTools.SELECT_OPTION,\n description: `Select an option from a dropdown. Handles both native <select> and custom dropdowns.`,\n schema: BrowserSelectOptionSchema,\n })\n );\n\n // browser_upload_file - Skyvern-inspired for file input handling\n tools.push(\n tool(createToolFunction('upload_file'), {\n name: EBrowserTools.UPLOAD_FILE,\n description: `Upload a file to a file input element.\nProvide the index of the file input and the URL of the file to upload.\nThe system will download the file and attach it to the input.\nAfter upload, you receive an updated element list.`,\n schema: BrowserUploadFileSchema,\n })\n );\n\n // browser_keypress - For keyboard shortcuts and special keys\n tools.push(\n tool(createToolFunction('keypress'), {\n name: EBrowserTools.KEYPRESS,\n description: `Press keyboard key(s) on the page.\nSingle keys: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\", \"ArrowUp\", \"Backspace\", \"Delete\"\nKey combos: \"Control+A\" (select all), \"Control+C\" (copy), \"Shift+Enter\" (newline)\nUse this for form submission, closing modals, navigating dropdowns.\nAfter keypress, you receive an updated element list.`,\n schema: BrowserKeypressSchema,\n })\n );\n\n return tools;\n}\n"],"names":["z","tools","tool"],"mappings":";;;;;AAIA;;;AAGG;AACU,MAAA,aAAa,GAAG;AAC3B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,QAAQ,EAAE,kBAAkB;AAC5B,IAAA,MAAM,EAAE,gBAAgB;AACxB,IAAA,OAAO,EAAE,iBAAiB;AAC1B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,UAAU,EAAE,oBAAoB;AAChC,IAAA,cAAc,EAAE,wBAAwB;;AAExC,IAAA,aAAa,EAAE,uBAAuB;AACtC,IAAA,WAAW,EAAE,qBAAqB;AAClC,IAAA,QAAQ,EAAE,kBAAkB;;AA4D9B;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAEpC,EAAA;AACC,IAAA,IAAI,CAAC,GAAG,EAAE,OAAO,EAAE;AACjB,QAAA,OAAO,KAAK;;IAGd,MAAM,gBAAgB,GAAG,GAAG,CAAC,OAAO,CAAC,4BAA4B,CAAC;IAClE,MAAM,cAAc,GAAG,GAAG,CAAC,OAAO,CAAC,0BAA0B,CAAC;AAE9D,IAAA,OAAO,gBAAgB,KAAK,MAAM,IAAI,cAAc,KAAK,MAAM;AACjE;AAEA;AACA,MAAM,kBAAkB,GAAGA,KAAC,CAAC,MAAM,CAAC;AAClC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CACP,0FAA0F,CAC3F;AACJ,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACjC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,oDAAoD,CAAC;IACjE,IAAI,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAEA;AACT,SAAA,OAAO;AACP,SAAA,QAAQ;SACR,QAAQ,CAAC,+DAA+D,CAAC;AAC7E,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACrC,IAAA,GAAG,EAAEA;AACF,SAAA,MAAM;SACN,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAEA;SACR,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC;SACpC,QAAQ,CAAC,qBAAqB,CAAC;AAClC,IAAA,MAAM,EAAEA;AACL,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,iDAAiD,CAAC;AAC/D,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,kBAAkB,GAAGA,KAAC,CAAC,MAAM,CAAC;IAClC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,+CAA+C,CAAC;AAC5E,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACjC,IAAA,QAAQ,EAAEA;AACP,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,sCAAsC,CAAC;AACpD,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAEtC,MAAM,uBAAuB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE5C,MAAM,yBAAyB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE9C;AACA,MAAM,yBAAyB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACzC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,iDAAiD,CAAC;AAC9D,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,gJAAgJ,CAAC;AAC9J,CAAA,CAAC;AAEF,MAAM,uBAAuB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACvC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,4CAA4C,CAAC;AACzD,IAAA,OAAO,EAAEA;AACN,SAAA,MAAM;SACN,QAAQ,CAAC,oEAAoE,CAAC;AAClF,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACrC,IAAA,IAAI,EAAEA;AACH,SAAA,MAAM;SACN,QAAQ,CAAC,uGAAuG,CAAC;AACrH,CAAA,CAAC;AAyBF;;AAEG;AACH,SAAS,kBAAkB,CACzB,MAA2B,EAC3B,MAAc,EACd,UAAoC,EAAA;IAEpC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,MAAM,SAAS,GAAG,MAAM,CAAC,SAAS,GAAG,CAAI,CAAA,EAAA,MAAM,CAAC,SAAS,CAAA,EAAA,CAAI,GAAG,EAAE;QAClE,OAAO,CAAA,EAAG,SAAS,CAAW,QAAA,EAAA,MAAM,aAAa,MAAM,CAAC,KAAK,CAAA,CAAE;;IAGjE,MAAM,KAAK,GAAa,EAAE;;AAG1B,IAAA,IAAI,MAAM,CAAC,YAAY,EAAE;AACvB,QAAA,MAAM,CAAC,GAAG,MAAM,CAAC,YAAY;AAC7B,QAAA,IAAI,CAAC,CAAC,YAAY,IAAI,CAAC,CAAC,QAAQ,IAAI,CAAC,CAAC,UAAU,EAAE;YAChD,KAAK,CAAC,IAAI,CAAC,CAAA,EAAA,EAAK,CAAC,CAAC,iBAAiB,CAAE,CAAA,CAAC;;AACjC,aAAA,IAAI,CAAC,CAAC,CAAC,iBAAiB,EAAE;AAC/B,YAAA,KAAK,CAAC,IAAI,CAAC,CAAA,oBAAA,CAAsB,CAAC;;;;IAKtC,IAAI,MAAM,CAAC,GAAG;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,KAAA,EAAQ,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;IAChD,IAAI,MAAM,CAAC,KAAK;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,OAAA,EAAU,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;IACtD,IAAI,MAAM,CAAC,WAAW;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,aAAA,EAAgB,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;AAExE,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAChC,OAAmC,EAAA;AAEnC,IAAA,MAAM,EAAE,aAAa,EAAE,GAAG,OAAO,IAAI,EAAE;IACvC,MAAMC,OAAK,GAA4B,EAAE;AAEzC;;;AAGG;AACH,IAAA,MAAM,kBAAkB,GAAG,CAAC,MAAc,KAAI;AAC5C,QAAA,OAAO,OACL,IAA6B,EAC7B,MAAuC,KACpB;AACnB,YAAA,MAAM,UAAU,GACd,MAAM,EAAE,QAAQ,EAAE,EAAE;gBACpB,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAG7D,YAAA,MAAM,MAAM,GAAwB;AAClC,gBAAA,wBAAwB,EAAE,IAAI;gBAC9B,MAAM;gBACN,IAAI;gBACJ,UAAU;aACX;;YAGD,IAAI,CAAC,aAAa,EAAE;AAClB,gBAAA,OAAO,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC;;;;AAK/B,YAAA,IAAI;gBACF,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,MAAM,EAAE,IAAI,EAAE,UAAU,CAAC;gBAC5D,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,CAAC;;YAC/C,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAChB,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;gBACxD,OAAO,CAAA,qCAAA,EAAwC,MAAM,CAAA,UAAA,EAAa,YAAY;;6EAET;;AAEzE,SAAC;AACH,KAAC;;IAGDA,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACqE,uFAAA,CAAA;AAClF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AACsD,wEAAA,CAAA;AACnE,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAwE,sEAAA,CAAA;AACrF,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,QAAQ,CAAC,EAAE;QACjC,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAA4E,0EAAA,CAAA;AACzF,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,SAAS,CAAC,EAAE;QAClC,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACwB,0CAAA,CAAA;AACrC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACmE,qFAAA,CAAA;AAChF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAEiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC6C,+DAAA,CAAA;AAC1D,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,YAAY,CAAC,EAAE;QACrC,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAoE,kEAAA,CAAA;AACjF,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,gBAAgB,CAAC,EAAE;QACzC,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAgF,8EAAA,CAAA;AAC7F,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,eAAe,CAAC,EAAE;QACxC,IAAI,EAAE,aAAa,CAAC,aAAa;AACjC,QAAA,WAAW,EAAE,CAAsF,oFAAA,CAAA;AACnG,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,aAAa,CAAC,EAAE;QACtC,IAAI,EAAE,aAAa,CAAC,WAAW;AAC/B,QAAA,WAAW,EAAE,CAAA;;;AAGgC,kDAAA,CAAA;AAC7C,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;;;;AAIkC,oDAAA,CAAA;AAC/C,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;AAED,IAAA,OAAOD,OAAK;AACd;;;;;;"}
1
+ {"version":3,"file":"BrowserTools.cjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\nimport type * as _t from '@/types';\n\n/**\n * Browser tool names - keep in sync with ranger-browser extension\n * These tools execute locally in the browser extension, NOT on the server\n */\nexport const EBrowserTools = {\n CLICK: 'browser_click',\n TYPE: 'browser_type',\n NAVIGATE: 'browser_navigate',\n SCROLL: 'browser_scroll',\n EXTRACT: 'browser_extract',\n HOVER: 'browser_hover',\n WAIT: 'browser_wait',\n BACK: 'browser_back',\n SCREENSHOT: 'browser_screenshot',\n GET_PAGE_STATE: 'browser_get_page_state',\n // Skyvern-inspired additions for robust form handling\n SELECT_OPTION: 'browser_select_option',\n UPLOAD_FILE: 'browser_upload_file',\n KEYPRESS: 'browser_keypress',\n} as const;\n\nexport type BrowserToolName =\n (typeof EBrowserTools)[keyof typeof EBrowserTools];\n\n/**\n * Callback function type for waiting on browser action results\n * This allows the server (Ranger) to provide a callback that waits for the extension\n * to POST results back to the server before returning to the LLM.\n *\n * @param action - The browser action (click, type, navigate, etc.)\n * @param args - Arguments for the action\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\n * @returns Promise that resolves with the actual browser result (page state, etc.)\n */\nexport type BrowserToolCallback = (\n action: string,\n args: Record<string, unknown>,\n toolCallId: string\n) => Promise<BrowserActionResult>;\n\n/**\n * Action status - matches Skyvern's ActionStatus\n */\nexport type ActionStatus = 'pending' | 'completed' | 'failed' | 'skipped';\n\n/**\n * Verification status - matches Skyvern's VerificationStatus\n * - complete: Goal achieved\n * - terminate: Goal cannot be achieved, stop\n * - continue: Goal not yet achieved, keep going\n */\nexport type VerificationStatus = 'complete' | 'terminate' | 'continue';\n\n/**\n * Result returned from browser action execution\n */\nexport interface BrowserActionResult {\n success: boolean;\n status?: ActionStatus;\n url?: string;\n title?: string;\n elementList?: string;\n error?: string;\n errorType?: string; // Typed error classification\n screenshot?: string;\n elementDescription?: string;\n // State verification (before/after comparison)\n verification?: {\n urlChanged: boolean;\n titleChanged: boolean;\n elementCountDelta: number;\n significantChange: boolean;\n dialogClosed: boolean;\n formGone: boolean;\n changeDescription: string;\n };\n}\n\n/**\n * Check if browser capability is available based on request headers or context\n * The browser extension sets these headers when connected:\n * - X-Ranger-Browser-Extension: true\n * - X-Ranger-Browser-Capable: true\n */\nexport function hasBrowserCapability(req?: {\n headers?: Record<string, string | string[] | undefined>;\n}): boolean {\n if (!req?.headers) {\n return false;\n }\n\n const browserExtension = req.headers['x-ranger-browser-extension'];\n const browserCapable = req.headers['x-ranger-browser-capable'];\n\n return browserExtension === 'true' || browserCapable === 'true';\n}\n\n// Tool schemas\nconst BrowserClickSchema = z.object({\n index: z\n .number()\n .describe(\n 'The index number [0], [1], etc. of the element to click from the page state element list'\n ),\n});\n\nconst BrowserTypeSchema = z.object({\n index: z\n .number()\n .describe('The index number of the input element to type into'),\n text: z.string().describe('The text to type into the element'),\n pressEnter: z\n .boolean()\n .optional()\n .describe('Whether to press Enter after typing (useful for search forms)'),\n});\n\nconst BrowserNavigateSchema = z.object({\n url: z\n .string()\n .describe('The full URL to navigate to (must include https://)'),\n});\n\nconst BrowserScrollSchema = z.object({\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .describe('Direction to scroll'),\n amount: z\n .number()\n .optional()\n .describe('Pixels to scroll (default: one viewport height)'),\n});\n\nconst BrowserExtractSchema = z.object({\n query: z\n .string()\n .optional()\n .describe('Optional: specific content to extract from the page'),\n});\n\nconst BrowserHoverSchema = z.object({\n index: z.number().describe('The index number of the element to hover over'),\n});\n\nconst BrowserWaitSchema = z.object({\n duration: z\n .number()\n .optional()\n .describe('Milliseconds to wait (default: 1000)'),\n});\n\nconst BrowserBackSchema = z.object({});\n\nconst BrowserScreenshotSchema = z.object({});\n\nconst BrowserGetPageStateSchema = z.object({});\n\n// Skyvern-inspired schemas for robust form handling\nconst BrowserSelectOptionSchema = z.object({\n index: z\n .number()\n .describe('The index number of the select/dropdown element'),\n value: z\n .string()\n .optional()\n .describe('The value or label of the option to select. For native <select>, use the option text. For custom dropdowns, this is the option label to click.'),\n});\n\nconst BrowserUploadFileSchema = z.object({\n index: z\n .number()\n .describe('The index number of the file input element'),\n fileUrl: z\n .string()\n .describe('URL of the file to upload (the system will download and upload it)'),\n});\n\nconst BrowserKeypressSchema = z.object({\n keys: z\n .string()\n .describe('Key(s) to press. Single key: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\". Combo: \"Control+A\", \"Shift+Enter\"'),\n});\n\n/**\n * Browser tool response interface\n * This is what the extension returns after executing the action\n */\nexport interface BrowserToolResponse {\n requiresBrowserExecution: true;\n action: string;\n args: Record<string, unknown>;\n toolCallId?: string; // Added to help extension correlate with callback\n}\n\n/**\n * Options for creating browser tools\n */\nexport interface CreateBrowserToolsOptions {\n /**\n * Optional callback that waits for browser action results.\n * When provided, tools will await this callback to get actual results from the extension.\n * When not provided, tools return markers immediately (for non-server contexts).\n */\n waitForResult?: BrowserToolCallback;\n}\n\n/**\n * Format browser action result for LLM consumption\n */\nfunction formatResultForLLM(\n result: BrowserActionResult,\n action: string,\n actionArgs?: Record<string, unknown>\n): string {\n if (!result.success && result.error) {\n const errorType = result.errorType ? `[${result.errorType}] ` : '';\n return `${errorType}Action \"${action}\" failed: ${result.error}`;\n }\n\n const parts: string[] = [];\n\n // Verification result (Skyvern pattern)\n if (result.verification) {\n const v = result.verification;\n if (v.dialogClosed || v.formGone || v.urlChanged) {\n parts.push(`✓ ${v.changeDescription}`);\n } else if (!v.significantChange) {\n parts.push(`⚠ No change detected`);\n }\n }\n\n // Page state\n if (result.url) parts.push(`URL: ${result.url}`);\n if (result.title) parts.push(`Title: ${result.title}`);\n if (result.elementList) parts.push(`\\nElements:\\n${result.elementList}`);\n\n return parts.join('\\n');\n}\n\n/**\n * Create browser tools with optional callback for waiting on results\n *\n * When waitForResult callback is provided:\n * 1. Tool returns marker that triggers extension\n * 2. Tool then awaits callback to get actual results\n * 3. Returns real page state to LLM\n *\n * When no callback:\n * 1. Tool returns marker only (for non-server contexts)\n *\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\n */\nexport function createBrowserTools(\n options?: CreateBrowserToolsOptions\n): DynamicStructuredTool[] {\n const { waitForResult } = options || {};\n const tools: DynamicStructuredTool[] = [];\n\n /**\n * Helper to create tool function that optionally waits for results\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\n */\n const createToolFunction = (action: string) => {\n return async (\n args: Record<string, unknown>,\n config?: { toolCall?: { id?: string } }\n ): Promise<string> => {\n const toolCallId =\n config?.toolCall?.id ??\n `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\n\n // Create marker for extension\n const marker: BrowserToolResponse = {\n requiresBrowserExecution: true,\n action,\n args,\n toolCallId,\n };\n\n // If no callback, return marker immediately (extension handles via SSE interception)\n if (!waitForResult) {\n return JSON.stringify(marker);\n }\n\n // With callback: wait for actual results from extension\n // The marker is still returned initially via SSE, but we wait for the callback\n try {\n const result = await waitForResult(action, args, toolCallId);\n return formatResultForLLM(result, action, args);\n } catch (error) {\n const errorMessage =\n error instanceof Error ? error.message : String(error);\n return `❌ **ACTION FAILED**: Browser action \"${action}\" failed: ${errorMessage}\n\n**REQUIRED**: Analyze why the action failed and try an alternative approach.`;\n }\n };\n };\n\n // browser_click - PRIMARY action for buttons and links\n tools.push(\n tool(createToolFunction('click'), {\n name: EBrowserTools.CLICK,\n description: `Click an element by its [index] from the element list.\n\n**WHEN TO USE (prioritize this over keypress):**\n- Buttons labeled \"Send\", \"Submit\", \"Save\", \"OK\", \"Confirm\" → CLICK them\n- Links (a tags) to navigate\n- Checkboxes, radio buttons to select\n- Any clickable element that performs an action\n\n**SEMANTIC ELEMENT IDENTIFICATION:**\n- Look for aria-label=\"Send\", aria-label=\"Submit\" for action buttons\n- Look for type=\"submit\" for form submission\n- Look for button text like \"Send\", \"Submit\", \"Save\"\n- For email compose: Find the Send button, DON'T use keyboard shortcuts\n\nReturns verification: dialog closed, form submitted, URL changed, or NO CHANGE (action may have failed).`,\n schema: BrowserClickSchema,\n })\n );\n\n // browser_type\n tools.push(\n tool(createToolFunction('type'), {\n name: EBrowserTools.TYPE,\n description: `Type text into an input/textarea element by [index].\n\n**WHEN TO USE:**\n- Input fields (fieldType=\"email\", \"search\", \"phone\", etc.)\n- Textareas for longer content\n- Contenteditable divs (like Gmail compose body)\n\n**OPTIONS:**\n- pressEnter: true → Press Enter after typing (for search forms)\n- pressEnter: false → Just type without submitting\n\n**DO NOT use pressEnter to send emails** - find and CLICK the Send button instead.\n\nReturns verification showing if text was entered or form was submitted.`,\n schema: BrowserTypeSchema,\n })\n );\n\n // browser_navigate\n tools.push(\n tool(createToolFunction('navigate'), {\n name: EBrowserTools.NAVIGATE,\n description: `Navigate to a URL (include https://). Returns new page's element list.`,\n schema: BrowserNavigateSchema,\n })\n );\n\n // browser_scroll\n tools.push(\n tool(createToolFunction('scroll'), {\n name: EBrowserTools.SCROLL,\n description: `Scroll the page. Returns updated element list with newly visible elements.`,\n schema: BrowserScrollSchema,\n })\n );\n\n // browser_extract\n tools.push(\n tool(createToolFunction('extract'), {\n name: EBrowserTools.EXTRACT,\n description: `Extract content from the current page.\nReturns page URL, title, and element list.`,\n schema: BrowserExtractSchema,\n })\n );\n\n // browser_hover\n tools.push(\n tool(createToolFunction('hover'), {\n name: EBrowserTools.HOVER,\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\n schema: BrowserHoverSchema,\n })\n );\n\n // browser_wait\n tools.push(\n tool(createToolFunction('wait'), {\n name: EBrowserTools.WAIT,\n description: `Wait for a specified duration for page content to load.\nUse this after actions that trigger async content loading.\nAfter waiting, you receive an updated element list.`,\n schema: BrowserWaitSchema,\n })\n );\n\n // browser_back\n tools.push(\n tool(createToolFunction('back'), {\n name: EBrowserTools.BACK,\n description: `Go back to the previous page in browser history.\nAfter going back, you receive the previous page's element list.`,\n schema: BrowserBackSchema,\n })\n );\n\n // browser_screenshot\n tools.push(\n tool(createToolFunction('screenshot'), {\n name: EBrowserTools.SCREENSHOT,\n description: `Capture a screenshot. Use browser_get_page_state for element list.`,\n schema: BrowserScreenshotSchema,\n })\n );\n\n // browser_get_page_state\n tools.push(\n tool(createToolFunction('get_page_state'), {\n name: EBrowserTools.GET_PAGE_STATE,\n description: `Get current page URL, title, and interactive elements list with index numbers.`,\n schema: BrowserGetPageStateSchema,\n })\n );\n\n // browser_select_option\n tools.push(\n tool(createToolFunction('select_option'), {\n name: EBrowserTools.SELECT_OPTION,\n description: `Select an option from a dropdown. Handles both native <select> and custom dropdowns.`,\n schema: BrowserSelectOptionSchema,\n })\n );\n\n // browser_upload_file - Skyvern-inspired for file input handling\n tools.push(\n tool(createToolFunction('upload_file'), {\n name: EBrowserTools.UPLOAD_FILE,\n description: `Upload a file to a file input element.\nProvide the index of the file input and the URL of the file to upload.\nThe system will download the file and attach it to the input.\nAfter upload, you receive an updated element list.`,\n schema: BrowserUploadFileSchema,\n })\n );\n\n // browser_keypress - LAST RESORT for keyboard shortcuts\n tools.push(\n tool(createToolFunction('keypress'), {\n name: EBrowserTools.KEYPRESS,\n description: `Press keyboard key(s) - USE ONLY AS LAST RESORT after clicking fails.\n\n**PREFER browser_click OVER keypress:**\n- To send email → Find and CLICK the Send button\n- To submit form → CLICK the Submit button\n- To close modal → CLICK the X or Cancel button\n\n**WHEN TO USE keypress:**\n- Navigation: \"Tab\", \"ArrowDown\", \"ArrowUp\" in dropdowns\n- Escape to close popups if no X button exists\n- Control+A, Control+C for select/copy operations\n- ONLY use Enter to submit if there's NO visible submit button\n\nSingle keys: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\"\nKey combos: \"Control+A\", \"Shift+Enter\"\n\nReturns verification showing if page state changed.`,\n schema: BrowserKeypressSchema,\n })\n );\n\n return tools;\n}\n"],"names":["z","tools","tool"],"mappings":";;;;;AAIA;;;AAGG;AACU,MAAA,aAAa,GAAG;AAC3B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,QAAQ,EAAE,kBAAkB;AAC5B,IAAA,MAAM,EAAE,gBAAgB;AACxB,IAAA,OAAO,EAAE,iBAAiB;AAC1B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,UAAU,EAAE,oBAAoB;AAChC,IAAA,cAAc,EAAE,wBAAwB;;AAExC,IAAA,aAAa,EAAE,uBAAuB;AACtC,IAAA,WAAW,EAAE,qBAAqB;AAClC,IAAA,QAAQ,EAAE,kBAAkB;;AA4D9B;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAEpC,EAAA;AACC,IAAA,IAAI,CAAC,GAAG,EAAE,OAAO,EAAE;AACjB,QAAA,OAAO,KAAK;;IAGd,MAAM,gBAAgB,GAAG,GAAG,CAAC,OAAO,CAAC,4BAA4B,CAAC;IAClE,MAAM,cAAc,GAAG,GAAG,CAAC,OAAO,CAAC,0BAA0B,CAAC;AAE9D,IAAA,OAAO,gBAAgB,KAAK,MAAM,IAAI,cAAc,KAAK,MAAM;AACjE;AAEA;AACA,MAAM,kBAAkB,GAAGA,KAAC,CAAC,MAAM,CAAC;AAClC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CACP,0FAA0F,CAC3F;AACJ,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACjC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,oDAAoD,CAAC;IACjE,IAAI,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAEA;AACT,SAAA,OAAO;AACP,SAAA,QAAQ;SACR,QAAQ,CAAC,+DAA+D,CAAC;AAC7E,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACrC,IAAA,GAAG,EAAEA;AACF,SAAA,MAAM;SACN,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAEA;SACR,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC;SACpC,QAAQ,CAAC,qBAAqB,CAAC;AAClC,IAAA,MAAM,EAAEA;AACL,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,iDAAiD,CAAC;AAC/D,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,kBAAkB,GAAGA,KAAC,CAAC,MAAM,CAAC;IAClC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,+CAA+C,CAAC;AAC5E,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACjC,IAAA,QAAQ,EAAEA;AACP,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,sCAAsC,CAAC;AACpD,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAEtC,MAAM,uBAAuB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE5C,MAAM,yBAAyB,GAAGA,KAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE9C;AACA,MAAM,yBAAyB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACzC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,iDAAiD,CAAC;AAC9D,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,gJAAgJ,CAAC;AAC9J,CAAA,CAAC;AAEF,MAAM,uBAAuB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACvC,IAAA,KAAK,EAAEA;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,4CAA4C,CAAC;AACzD,IAAA,OAAO,EAAEA;AACN,SAAA,MAAM;SACN,QAAQ,CAAC,oEAAoE,CAAC;AAClF,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACrC,IAAA,IAAI,EAAEA;AACH,SAAA,MAAM;SACN,QAAQ,CAAC,uGAAuG,CAAC;AACrH,CAAA,CAAC;AAyBF;;AAEG;AACH,SAAS,kBAAkB,CACzB,MAA2B,EAC3B,MAAc,EACd,UAAoC,EAAA;IAEpC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,MAAM,SAAS,GAAG,MAAM,CAAC,SAAS,GAAG,CAAI,CAAA,EAAA,MAAM,CAAC,SAAS,CAAA,EAAA,CAAI,GAAG,EAAE;QAClE,OAAO,CAAA,EAAG,SAAS,CAAW,QAAA,EAAA,MAAM,aAAa,MAAM,CAAC,KAAK,CAAA,CAAE;;IAGjE,MAAM,KAAK,GAAa,EAAE;;AAG1B,IAAA,IAAI,MAAM,CAAC,YAAY,EAAE;AACvB,QAAA,MAAM,CAAC,GAAG,MAAM,CAAC,YAAY;AAC7B,QAAA,IAAI,CAAC,CAAC,YAAY,IAAI,CAAC,CAAC,QAAQ,IAAI,CAAC,CAAC,UAAU,EAAE;YAChD,KAAK,CAAC,IAAI,CAAC,CAAA,EAAA,EAAK,CAAC,CAAC,iBAAiB,CAAE,CAAA,CAAC;;AACjC,aAAA,IAAI,CAAC,CAAC,CAAC,iBAAiB,EAAE;AAC/B,YAAA,KAAK,CAAC,IAAI,CAAC,CAAA,oBAAA,CAAsB,CAAC;;;;IAKtC,IAAI,MAAM,CAAC,GAAG;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,KAAA,EAAQ,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;IAChD,IAAI,MAAM,CAAC,KAAK;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,OAAA,EAAU,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;IACtD,IAAI,MAAM,CAAC,WAAW;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,aAAA,EAAgB,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;AAExE,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAChC,OAAmC,EAAA;AAEnC,IAAA,MAAM,EAAE,aAAa,EAAE,GAAG,OAAO,IAAI,EAAE;IACvC,MAAMC,OAAK,GAA4B,EAAE;AAEzC;;;AAGG;AACH,IAAA,MAAM,kBAAkB,GAAG,CAAC,MAAc,KAAI;AAC5C,QAAA,OAAO,OACL,IAA6B,EAC7B,MAAuC,KACpB;AACnB,YAAA,MAAM,UAAU,GACd,MAAM,EAAE,QAAQ,EAAE,EAAE;gBACpB,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAG7D,YAAA,MAAM,MAAM,GAAwB;AAClC,gBAAA,wBAAwB,EAAE,IAAI;gBAC9B,MAAM;gBACN,IAAI;gBACJ,UAAU;aACX;;YAGD,IAAI,CAAC,aAAa,EAAE;AAClB,gBAAA,OAAO,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC;;;;AAK/B,YAAA,IAAI;gBACF,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,MAAM,EAAE,IAAI,EAAE,UAAU,CAAC;gBAC5D,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,CAAC;;YAC/C,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAChB,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;gBACxD,OAAO,CAAA,qCAAA,EAAwC,MAAM,CAAA,UAAA,EAAa,YAAY;;6EAET;;AAEzE,SAAC;AACH,KAAC;;IAGDA,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;;AAcsF,wGAAA,CAAA;AACnG,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;AAaqD,uEAAA,CAAA;AAClE,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAwE,sEAAA,CAAA;AACrF,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,QAAQ,CAAC,EAAE;QACjC,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAA4E,0EAAA,CAAA;AACzF,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,SAAS,CAAC,EAAE;QAClC,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACwB,0CAAA,CAAA;AACrC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACmE,qFAAA,CAAA;AAChF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAEiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC6C,+DAAA,CAAA;AAC1D,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,YAAY,CAAC,EAAE;QACrC,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAoE,kEAAA,CAAA;AACjF,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,gBAAgB,CAAC,EAAE;QACzC,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAgF,8EAAA,CAAA;AAC7F,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,eAAe,CAAC,EAAE;QACxC,IAAI,EAAE,aAAa,CAAC,aAAa;AACjC,QAAA,WAAW,EAAE,CAAsF,oFAAA,CAAA;AACnG,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,aAAa,CAAC,EAAE;QACtC,IAAI,EAAE,aAAa,CAAC,WAAW;AAC/B,QAAA,WAAW,EAAE,CAAA;;;AAGgC,kDAAA,CAAA;AAC7C,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;;;;AAgBiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;AAED,IAAA,OAAOD,OAAK;AACd;;;;;;"}
@@ -184,18 +184,43 @@ function createBrowserTools(options) {
184
184
  }
185
185
  };
186
186
  };
187
- // browser_click
187
+ // browser_click - PRIMARY action for buttons and links
188
188
  tools.push(tool(createToolFunction('click'), {
189
189
  name: EBrowserTools.CLICK,
190
- description: `Click an element by its index number from the element list (e.g., [0], [1]).
191
- Returns state verification showing if UI changed (dialog closed, form submitted, etc.).`,
190
+ description: `Click an element by its [index] from the element list.
191
+
192
+ **WHEN TO USE (prioritize this over keypress):**
193
+ - Buttons labeled "Send", "Submit", "Save", "OK", "Confirm" → CLICK them
194
+ - Links (a tags) to navigate
195
+ - Checkboxes, radio buttons to select
196
+ - Any clickable element that performs an action
197
+
198
+ **SEMANTIC ELEMENT IDENTIFICATION:**
199
+ - Look for aria-label="Send", aria-label="Submit" for action buttons
200
+ - Look for type="submit" for form submission
201
+ - Look for button text like "Send", "Submit", "Save"
202
+ - For email compose: Find the Send button, DON'T use keyboard shortcuts
203
+
204
+ Returns verification: dialog closed, form submitted, URL changed, or NO CHANGE (action may have failed).`,
192
205
  schema: BrowserClickSchema,
193
206
  }));
194
207
  // browser_type
195
208
  tools.push(tool(createToolFunction('type'), {
196
209
  name: EBrowserTools.TYPE,
197
- description: `Type text into an input element by index. Set pressEnter: true to submit.
198
- Returns state verification showing if form submitted or errors appeared.`,
210
+ description: `Type text into an input/textarea element by [index].
211
+
212
+ **WHEN TO USE:**
213
+ - Input fields (fieldType="email", "search", "phone", etc.)
214
+ - Textareas for longer content
215
+ - Contenteditable divs (like Gmail compose body)
216
+
217
+ **OPTIONS:**
218
+ - pressEnter: true → Press Enter after typing (for search forms)
219
+ - pressEnter: false → Just type without submitting
220
+
221
+ **DO NOT use pressEnter to send emails** - find and CLICK the Send button instead.
222
+
223
+ Returns verification showing if text was entered or form was submitted.`,
199
224
  schema: BrowserTypeSchema,
200
225
  }));
201
226
  // browser_navigate
@@ -266,14 +291,26 @@ The system will download the file and attach it to the input.
266
291
  After upload, you receive an updated element list.`,
267
292
  schema: BrowserUploadFileSchema,
268
293
  }));
269
- // browser_keypress - For keyboard shortcuts and special keys
294
+ // browser_keypress - LAST RESORT for keyboard shortcuts
270
295
  tools.push(tool(createToolFunction('keypress'), {
271
296
  name: EBrowserTools.KEYPRESS,
272
- description: `Press keyboard key(s) on the page.
273
- Single keys: "Enter", "Escape", "Tab", "ArrowDown", "ArrowUp", "Backspace", "Delete"
274
- Key combos: "Control+A" (select all), "Control+C" (copy), "Shift+Enter" (newline)
275
- Use this for form submission, closing modals, navigating dropdowns.
276
- After keypress, you receive an updated element list.`,
297
+ description: `Press keyboard key(s) - USE ONLY AS LAST RESORT after clicking fails.
298
+
299
+ **PREFER browser_click OVER keypress:**
300
+ - To send email Find and CLICK the Send button
301
+ - To submit form CLICK the Submit button
302
+ - To close modal → CLICK the X or Cancel button
303
+
304
+ **WHEN TO USE keypress:**
305
+ - Navigation: "Tab", "ArrowDown", "ArrowUp" in dropdowns
306
+ - Escape to close popups if no X button exists
307
+ - Control+A, Control+C for select/copy operations
308
+ - ONLY use Enter to submit if there's NO visible submit button
309
+
310
+ Single keys: "Enter", "Escape", "Tab", "ArrowDown"
311
+ Key combos: "Control+A", "Shift+Enter"
312
+
313
+ Returns verification showing if page state changed.`,
277
314
  schema: BrowserKeypressSchema,
278
315
  }));
279
316
  return tools;
@@ -1 +1 @@
1
- {"version":3,"file":"BrowserTools.mjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\nimport type * as _t from '@/types';\n\n/**\n * Browser tool names - keep in sync with ranger-browser extension\n * These tools execute locally in the browser extension, NOT on the server\n */\nexport const EBrowserTools = {\n CLICK: 'browser_click',\n TYPE: 'browser_type',\n NAVIGATE: 'browser_navigate',\n SCROLL: 'browser_scroll',\n EXTRACT: 'browser_extract',\n HOVER: 'browser_hover',\n WAIT: 'browser_wait',\n BACK: 'browser_back',\n SCREENSHOT: 'browser_screenshot',\n GET_PAGE_STATE: 'browser_get_page_state',\n // Skyvern-inspired additions for robust form handling\n SELECT_OPTION: 'browser_select_option',\n UPLOAD_FILE: 'browser_upload_file',\n KEYPRESS: 'browser_keypress',\n} as const;\n\nexport type BrowserToolName =\n (typeof EBrowserTools)[keyof typeof EBrowserTools];\n\n/**\n * Callback function type for waiting on browser action results\n * This allows the server (Ranger) to provide a callback that waits for the extension\n * to POST results back to the server before returning to the LLM.\n *\n * @param action - The browser action (click, type, navigate, etc.)\n * @param args - Arguments for the action\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\n * @returns Promise that resolves with the actual browser result (page state, etc.)\n */\nexport type BrowserToolCallback = (\n action: string,\n args: Record<string, unknown>,\n toolCallId: string\n) => Promise<BrowserActionResult>;\n\n/**\n * Action status - matches Skyvern's ActionStatus\n */\nexport type ActionStatus = 'pending' | 'completed' | 'failed' | 'skipped';\n\n/**\n * Verification status - matches Skyvern's VerificationStatus\n * - complete: Goal achieved\n * - terminate: Goal cannot be achieved, stop\n * - continue: Goal not yet achieved, keep going\n */\nexport type VerificationStatus = 'complete' | 'terminate' | 'continue';\n\n/**\n * Result returned from browser action execution\n */\nexport interface BrowserActionResult {\n success: boolean;\n status?: ActionStatus;\n url?: string;\n title?: string;\n elementList?: string;\n error?: string;\n errorType?: string; // Typed error classification\n screenshot?: string;\n elementDescription?: string;\n // State verification (before/after comparison)\n verification?: {\n urlChanged: boolean;\n titleChanged: boolean;\n elementCountDelta: number;\n significantChange: boolean;\n dialogClosed: boolean;\n formGone: boolean;\n changeDescription: string;\n };\n}\n\n/**\n * Check if browser capability is available based on request headers or context\n * The browser extension sets these headers when connected:\n * - X-Ranger-Browser-Extension: true\n * - X-Ranger-Browser-Capable: true\n */\nexport function hasBrowserCapability(req?: {\n headers?: Record<string, string | string[] | undefined>;\n}): boolean {\n if (!req?.headers) {\n return false;\n }\n\n const browserExtension = req.headers['x-ranger-browser-extension'];\n const browserCapable = req.headers['x-ranger-browser-capable'];\n\n return browserExtension === 'true' || browserCapable === 'true';\n}\n\n// Tool schemas\nconst BrowserClickSchema = z.object({\n index: z\n .number()\n .describe(\n 'The index number [0], [1], etc. of the element to click from the page state element list'\n ),\n});\n\nconst BrowserTypeSchema = z.object({\n index: z\n .number()\n .describe('The index number of the input element to type into'),\n text: z.string().describe('The text to type into the element'),\n pressEnter: z\n .boolean()\n .optional()\n .describe('Whether to press Enter after typing (useful for search forms)'),\n});\n\nconst BrowserNavigateSchema = z.object({\n url: z\n .string()\n .describe('The full URL to navigate to (must include https://)'),\n});\n\nconst BrowserScrollSchema = z.object({\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .describe('Direction to scroll'),\n amount: z\n .number()\n .optional()\n .describe('Pixels to scroll (default: one viewport height)'),\n});\n\nconst BrowserExtractSchema = z.object({\n query: z\n .string()\n .optional()\n .describe('Optional: specific content to extract from the page'),\n});\n\nconst BrowserHoverSchema = z.object({\n index: z.number().describe('The index number of the element to hover over'),\n});\n\nconst BrowserWaitSchema = z.object({\n duration: z\n .number()\n .optional()\n .describe('Milliseconds to wait (default: 1000)'),\n});\n\nconst BrowserBackSchema = z.object({});\n\nconst BrowserScreenshotSchema = z.object({});\n\nconst BrowserGetPageStateSchema = z.object({});\n\n// Skyvern-inspired schemas for robust form handling\nconst BrowserSelectOptionSchema = z.object({\n index: z\n .number()\n .describe('The index number of the select/dropdown element'),\n value: z\n .string()\n .optional()\n .describe('The value or label of the option to select. For native <select>, use the option text. For custom dropdowns, this is the option label to click.'),\n});\n\nconst BrowserUploadFileSchema = z.object({\n index: z\n .number()\n .describe('The index number of the file input element'),\n fileUrl: z\n .string()\n .describe('URL of the file to upload (the system will download and upload it)'),\n});\n\nconst BrowserKeypressSchema = z.object({\n keys: z\n .string()\n .describe('Key(s) to press. Single key: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\". Combo: \"Control+A\", \"Shift+Enter\"'),\n});\n\n/**\n * Browser tool response interface\n * This is what the extension returns after executing the action\n */\nexport interface BrowserToolResponse {\n requiresBrowserExecution: true;\n action: string;\n args: Record<string, unknown>;\n toolCallId?: string; // Added to help extension correlate with callback\n}\n\n/**\n * Options for creating browser tools\n */\nexport interface CreateBrowserToolsOptions {\n /**\n * Optional callback that waits for browser action results.\n * When provided, tools will await this callback to get actual results from the extension.\n * When not provided, tools return markers immediately (for non-server contexts).\n */\n waitForResult?: BrowserToolCallback;\n}\n\n/**\n * Format browser action result for LLM consumption\n */\nfunction formatResultForLLM(\n result: BrowserActionResult,\n action: string,\n actionArgs?: Record<string, unknown>\n): string {\n if (!result.success && result.error) {\n const errorType = result.errorType ? `[${result.errorType}] ` : '';\n return `${errorType}Action \"${action}\" failed: ${result.error}`;\n }\n\n const parts: string[] = [];\n\n // Verification result (Skyvern pattern)\n if (result.verification) {\n const v = result.verification;\n if (v.dialogClosed || v.formGone || v.urlChanged) {\n parts.push(`✓ ${v.changeDescription}`);\n } else if (!v.significantChange) {\n parts.push(`⚠ No change detected`);\n }\n }\n\n // Page state\n if (result.url) parts.push(`URL: ${result.url}`);\n if (result.title) parts.push(`Title: ${result.title}`);\n if (result.elementList) parts.push(`\\nElements:\\n${result.elementList}`);\n\n return parts.join('\\n');\n}\n\n/**\n * Create browser tools with optional callback for waiting on results\n *\n * When waitForResult callback is provided:\n * 1. Tool returns marker that triggers extension\n * 2. Tool then awaits callback to get actual results\n * 3. Returns real page state to LLM\n *\n * When no callback:\n * 1. Tool returns marker only (for non-server contexts)\n *\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\n */\nexport function createBrowserTools(\n options?: CreateBrowserToolsOptions\n): DynamicStructuredTool[] {\n const { waitForResult } = options || {};\n const tools: DynamicStructuredTool[] = [];\n\n /**\n * Helper to create tool function that optionally waits for results\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\n */\n const createToolFunction = (action: string) => {\n return async (\n args: Record<string, unknown>,\n config?: { toolCall?: { id?: string } }\n ): Promise<string> => {\n const toolCallId =\n config?.toolCall?.id ??\n `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\n\n // Create marker for extension\n const marker: BrowserToolResponse = {\n requiresBrowserExecution: true,\n action,\n args,\n toolCallId,\n };\n\n // If no callback, return marker immediately (extension handles via SSE interception)\n if (!waitForResult) {\n return JSON.stringify(marker);\n }\n\n // With callback: wait for actual results from extension\n // The marker is still returned initially via SSE, but we wait for the callback\n try {\n const result = await waitForResult(action, args, toolCallId);\n return formatResultForLLM(result, action, args);\n } catch (error) {\n const errorMessage =\n error instanceof Error ? error.message : String(error);\n return `❌ **ACTION FAILED**: Browser action \"${action}\" failed: ${errorMessage}\n\n**REQUIRED**: Analyze why the action failed and try an alternative approach.`;\n }\n };\n };\n\n // browser_click\n tools.push(\n tool(createToolFunction('click'), {\n name: EBrowserTools.CLICK,\n description: `Click an element by its index number from the element list (e.g., [0], [1]).\nReturns state verification showing if UI changed (dialog closed, form submitted, etc.).`,\n schema: BrowserClickSchema,\n })\n );\n\n // browser_type\n tools.push(\n tool(createToolFunction('type'), {\n name: EBrowserTools.TYPE,\n description: `Type text into an input element by index. Set pressEnter: true to submit.\nReturns state verification showing if form submitted or errors appeared.`,\n schema: BrowserTypeSchema,\n })\n );\n\n // browser_navigate\n tools.push(\n tool(createToolFunction('navigate'), {\n name: EBrowserTools.NAVIGATE,\n description: `Navigate to a URL (include https://). Returns new page's element list.`,\n schema: BrowserNavigateSchema,\n })\n );\n\n // browser_scroll\n tools.push(\n tool(createToolFunction('scroll'), {\n name: EBrowserTools.SCROLL,\n description: `Scroll the page. Returns updated element list with newly visible elements.`,\n schema: BrowserScrollSchema,\n })\n );\n\n // browser_extract\n tools.push(\n tool(createToolFunction('extract'), {\n name: EBrowserTools.EXTRACT,\n description: `Extract content from the current page.\nReturns page URL, title, and element list.`,\n schema: BrowserExtractSchema,\n })\n );\n\n // browser_hover\n tools.push(\n tool(createToolFunction('hover'), {\n name: EBrowserTools.HOVER,\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\n schema: BrowserHoverSchema,\n })\n );\n\n // browser_wait\n tools.push(\n tool(createToolFunction('wait'), {\n name: EBrowserTools.WAIT,\n description: `Wait for a specified duration for page content to load.\nUse this after actions that trigger async content loading.\nAfter waiting, you receive an updated element list.`,\n schema: BrowserWaitSchema,\n })\n );\n\n // browser_back\n tools.push(\n tool(createToolFunction('back'), {\n name: EBrowserTools.BACK,\n description: `Go back to the previous page in browser history.\nAfter going back, you receive the previous page's element list.`,\n schema: BrowserBackSchema,\n })\n );\n\n // browser_screenshot\n tools.push(\n tool(createToolFunction('screenshot'), {\n name: EBrowserTools.SCREENSHOT,\n description: `Capture a screenshot. Use browser_get_page_state for element list.`,\n schema: BrowserScreenshotSchema,\n })\n );\n\n // browser_get_page_state\n tools.push(\n tool(createToolFunction('get_page_state'), {\n name: EBrowserTools.GET_PAGE_STATE,\n description: `Get current page URL, title, and interactive elements list with index numbers.`,\n schema: BrowserGetPageStateSchema,\n })\n );\n\n // browser_select_option\n tools.push(\n tool(createToolFunction('select_option'), {\n name: EBrowserTools.SELECT_OPTION,\n description: `Select an option from a dropdown. Handles both native <select> and custom dropdowns.`,\n schema: BrowserSelectOptionSchema,\n })\n );\n\n // browser_upload_file - Skyvern-inspired for file input handling\n tools.push(\n tool(createToolFunction('upload_file'), {\n name: EBrowserTools.UPLOAD_FILE,\n description: `Upload a file to a file input element.\nProvide the index of the file input and the URL of the file to upload.\nThe system will download the file and attach it to the input.\nAfter upload, you receive an updated element list.`,\n schema: BrowserUploadFileSchema,\n })\n );\n\n // browser_keypress - For keyboard shortcuts and special keys\n tools.push(\n tool(createToolFunction('keypress'), {\n name: EBrowserTools.KEYPRESS,\n description: `Press keyboard key(s) on the page.\nSingle keys: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\", \"ArrowUp\", \"Backspace\", \"Delete\"\nKey combos: \"Control+A\" (select all), \"Control+C\" (copy), \"Shift+Enter\" (newline)\nUse this for form submission, closing modals, navigating dropdowns.\nAfter keypress, you receive an updated element list.`,\n schema: BrowserKeypressSchema,\n })\n );\n\n return tools;\n}\n"],"names":[],"mappings":";;;AAIA;;;AAGG;AACU,MAAA,aAAa,GAAG;AAC3B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,QAAQ,EAAE,kBAAkB;AAC5B,IAAA,MAAM,EAAE,gBAAgB;AACxB,IAAA,OAAO,EAAE,iBAAiB;AAC1B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,UAAU,EAAE,oBAAoB;AAChC,IAAA,cAAc,EAAE,wBAAwB;;AAExC,IAAA,aAAa,EAAE,uBAAuB;AACtC,IAAA,WAAW,EAAE,qBAAqB;AAClC,IAAA,QAAQ,EAAE,kBAAkB;;AA4D9B;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAEpC,EAAA;AACC,IAAA,IAAI,CAAC,GAAG,EAAE,OAAO,EAAE;AACjB,QAAA,OAAO,KAAK;;IAGd,MAAM,gBAAgB,GAAG,GAAG,CAAC,OAAO,CAAC,4BAA4B,CAAC;IAClE,MAAM,cAAc,GAAG,GAAG,CAAC,OAAO,CAAC,0BAA0B,CAAC;AAE9D,IAAA,OAAO,gBAAgB,KAAK,MAAM,IAAI,cAAc,KAAK,MAAM;AACjE;AAEA;AACA,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;AAClC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CACP,0FAA0F,CAC3F;AACJ,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;AACjC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,oDAAoD,CAAC;IACjE,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAE;AACT,SAAA,OAAO;AACP,SAAA,QAAQ;SACR,QAAQ,CAAC,+DAA+D,CAAC;AAC7E,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;AACrC,IAAA,GAAG,EAAE;AACF,SAAA,MAAM;SACN,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAE;SACR,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC;SACpC,QAAQ,CAAC,qBAAqB,CAAC;AAClC,IAAA,MAAM,EAAE;AACL,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,iDAAiD,CAAC;AAC/D,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAClC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,+CAA+C,CAAC;AAC5E,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;AACjC,IAAA,QAAQ,EAAE;AACP,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,sCAAsC,CAAC;AACpD,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAEtC,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE5C,MAAM,yBAAyB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE9C;AACA,MAAM,yBAAyB,GAAG,CAAC,CAAC,MAAM,CAAC;AACzC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,iDAAiD,CAAC;AAC9D,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,gJAAgJ,CAAC;AAC9J,CAAA,CAAC;AAEF,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC;AACvC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,4CAA4C,CAAC;AACzD,IAAA,OAAO,EAAE;AACN,SAAA,MAAM;SACN,QAAQ,CAAC,oEAAoE,CAAC;AAClF,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;AACrC,IAAA,IAAI,EAAE;AACH,SAAA,MAAM;SACN,QAAQ,CAAC,uGAAuG,CAAC;AACrH,CAAA,CAAC;AAyBF;;AAEG;AACH,SAAS,kBAAkB,CACzB,MAA2B,EAC3B,MAAc,EACd,UAAoC,EAAA;IAEpC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,MAAM,SAAS,GAAG,MAAM,CAAC,SAAS,GAAG,CAAI,CAAA,EAAA,MAAM,CAAC,SAAS,CAAA,EAAA,CAAI,GAAG,EAAE;QAClE,OAAO,CAAA,EAAG,SAAS,CAAW,QAAA,EAAA,MAAM,aAAa,MAAM,CAAC,KAAK,CAAA,CAAE;;IAGjE,MAAM,KAAK,GAAa,EAAE;;AAG1B,IAAA,IAAI,MAAM,CAAC,YAAY,EAAE;AACvB,QAAA,MAAM,CAAC,GAAG,MAAM,CAAC,YAAY;AAC7B,QAAA,IAAI,CAAC,CAAC,YAAY,IAAI,CAAC,CAAC,QAAQ,IAAI,CAAC,CAAC,UAAU,EAAE;YAChD,KAAK,CAAC,IAAI,CAAC,CAAA,EAAA,EAAK,CAAC,CAAC,iBAAiB,CAAE,CAAA,CAAC;;AACjC,aAAA,IAAI,CAAC,CAAC,CAAC,iBAAiB,EAAE;AAC/B,YAAA,KAAK,CAAC,IAAI,CAAC,CAAA,oBAAA,CAAsB,CAAC;;;;IAKtC,IAAI,MAAM,CAAC,GAAG;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,KAAA,EAAQ,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;IAChD,IAAI,MAAM,CAAC,KAAK;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,OAAA,EAAU,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;IACtD,IAAI,MAAM,CAAC,WAAW;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,aAAA,EAAgB,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;AAExE,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAChC,OAAmC,EAAA;AAEnC,IAAA,MAAM,EAAE,aAAa,EAAE,GAAG,OAAO,IAAI,EAAE;IACvC,MAAM,KAAK,GAA4B,EAAE;AAEzC;;;AAGG;AACH,IAAA,MAAM,kBAAkB,GAAG,CAAC,MAAc,KAAI;AAC5C,QAAA,OAAO,OACL,IAA6B,EAC7B,MAAuC,KACpB;AACnB,YAAA,MAAM,UAAU,GACd,MAAM,EAAE,QAAQ,EAAE,EAAE;gBACpB,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAG7D,YAAA,MAAM,MAAM,GAAwB;AAClC,gBAAA,wBAAwB,EAAE,IAAI;gBAC9B,MAAM;gBACN,IAAI;gBACJ,UAAU;aACX;;YAGD,IAAI,CAAC,aAAa,EAAE;AAClB,gBAAA,OAAO,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC;;;;AAK/B,YAAA,IAAI;gBACF,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,MAAM,EAAE,IAAI,EAAE,UAAU,CAAC;gBAC5D,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,CAAC;;YAC/C,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAChB,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;gBACxD,OAAO,CAAA,qCAAA,EAAwC,MAAM,CAAA,UAAA,EAAa,YAAY;;6EAET;;AAEzE,SAAC;AACH,KAAC;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACqE,uFAAA,CAAA;AAClF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AACsD,wEAAA,CAAA;AACnE,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAwE,sEAAA,CAAA;AACrF,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,QAAQ,CAAC,EAAE;QACjC,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAA4E,0EAAA,CAAA;AACzF,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,SAAS,CAAC,EAAE;QAClC,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACwB,0CAAA,CAAA;AACrC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACmE,qFAAA,CAAA;AAChF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAEiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC6C,+DAAA,CAAA;AAC1D,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,YAAY,CAAC,EAAE;QACrC,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAoE,kEAAA,CAAA;AACjF,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,gBAAgB,CAAC,EAAE;QACzC,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAgF,8EAAA,CAAA;AAC7F,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,eAAe,CAAC,EAAE;QACxC,IAAI,EAAE,aAAa,CAAC,aAAa;AACjC,QAAA,WAAW,EAAE,CAAsF,oFAAA,CAAA;AACnG,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,aAAa,CAAC,EAAE;QACtC,IAAI,EAAE,aAAa,CAAC,WAAW;AAC/B,QAAA,WAAW,EAAE,CAAA;;;AAGgC,kDAAA,CAAA;AAC7C,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;;;;AAIkC,oDAAA,CAAA;AAC/C,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;AAED,IAAA,OAAO,KAAK;AACd;;;;"}
1
+ {"version":3,"file":"BrowserTools.mjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\nimport type * as _t from '@/types';\n\n/**\n * Browser tool names - keep in sync with ranger-browser extension\n * These tools execute locally in the browser extension, NOT on the server\n */\nexport const EBrowserTools = {\n CLICK: 'browser_click',\n TYPE: 'browser_type',\n NAVIGATE: 'browser_navigate',\n SCROLL: 'browser_scroll',\n EXTRACT: 'browser_extract',\n HOVER: 'browser_hover',\n WAIT: 'browser_wait',\n BACK: 'browser_back',\n SCREENSHOT: 'browser_screenshot',\n GET_PAGE_STATE: 'browser_get_page_state',\n // Skyvern-inspired additions for robust form handling\n SELECT_OPTION: 'browser_select_option',\n UPLOAD_FILE: 'browser_upload_file',\n KEYPRESS: 'browser_keypress',\n} as const;\n\nexport type BrowserToolName =\n (typeof EBrowserTools)[keyof typeof EBrowserTools];\n\n/**\n * Callback function type for waiting on browser action results\n * This allows the server (Ranger) to provide a callback that waits for the extension\n * to POST results back to the server before returning to the LLM.\n *\n * @param action - The browser action (click, type, navigate, etc.)\n * @param args - Arguments for the action\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\n * @returns Promise that resolves with the actual browser result (page state, etc.)\n */\nexport type BrowserToolCallback = (\n action: string,\n args: Record<string, unknown>,\n toolCallId: string\n) => Promise<BrowserActionResult>;\n\n/**\n * Action status - matches Skyvern's ActionStatus\n */\nexport type ActionStatus = 'pending' | 'completed' | 'failed' | 'skipped';\n\n/**\n * Verification status - matches Skyvern's VerificationStatus\n * - complete: Goal achieved\n * - terminate: Goal cannot be achieved, stop\n * - continue: Goal not yet achieved, keep going\n */\nexport type VerificationStatus = 'complete' | 'terminate' | 'continue';\n\n/**\n * Result returned from browser action execution\n */\nexport interface BrowserActionResult {\n success: boolean;\n status?: ActionStatus;\n url?: string;\n title?: string;\n elementList?: string;\n error?: string;\n errorType?: string; // Typed error classification\n screenshot?: string;\n elementDescription?: string;\n // State verification (before/after comparison)\n verification?: {\n urlChanged: boolean;\n titleChanged: boolean;\n elementCountDelta: number;\n significantChange: boolean;\n dialogClosed: boolean;\n formGone: boolean;\n changeDescription: string;\n };\n}\n\n/**\n * Check if browser capability is available based on request headers or context\n * The browser extension sets these headers when connected:\n * - X-Ranger-Browser-Extension: true\n * - X-Ranger-Browser-Capable: true\n */\nexport function hasBrowserCapability(req?: {\n headers?: Record<string, string | string[] | undefined>;\n}): boolean {\n if (!req?.headers) {\n return false;\n }\n\n const browserExtension = req.headers['x-ranger-browser-extension'];\n const browserCapable = req.headers['x-ranger-browser-capable'];\n\n return browserExtension === 'true' || browserCapable === 'true';\n}\n\n// Tool schemas\nconst BrowserClickSchema = z.object({\n index: z\n .number()\n .describe(\n 'The index number [0], [1], etc. of the element to click from the page state element list'\n ),\n});\n\nconst BrowserTypeSchema = z.object({\n index: z\n .number()\n .describe('The index number of the input element to type into'),\n text: z.string().describe('The text to type into the element'),\n pressEnter: z\n .boolean()\n .optional()\n .describe('Whether to press Enter after typing (useful for search forms)'),\n});\n\nconst BrowserNavigateSchema = z.object({\n url: z\n .string()\n .describe('The full URL to navigate to (must include https://)'),\n});\n\nconst BrowserScrollSchema = z.object({\n direction: z\n .enum(['up', 'down', 'left', 'right'])\n .describe('Direction to scroll'),\n amount: z\n .number()\n .optional()\n .describe('Pixels to scroll (default: one viewport height)'),\n});\n\nconst BrowserExtractSchema = z.object({\n query: z\n .string()\n .optional()\n .describe('Optional: specific content to extract from the page'),\n});\n\nconst BrowserHoverSchema = z.object({\n index: z.number().describe('The index number of the element to hover over'),\n});\n\nconst BrowserWaitSchema = z.object({\n duration: z\n .number()\n .optional()\n .describe('Milliseconds to wait (default: 1000)'),\n});\n\nconst BrowserBackSchema = z.object({});\n\nconst BrowserScreenshotSchema = z.object({});\n\nconst BrowserGetPageStateSchema = z.object({});\n\n// Skyvern-inspired schemas for robust form handling\nconst BrowserSelectOptionSchema = z.object({\n index: z\n .number()\n .describe('The index number of the select/dropdown element'),\n value: z\n .string()\n .optional()\n .describe('The value or label of the option to select. For native <select>, use the option text. For custom dropdowns, this is the option label to click.'),\n});\n\nconst BrowserUploadFileSchema = z.object({\n index: z\n .number()\n .describe('The index number of the file input element'),\n fileUrl: z\n .string()\n .describe('URL of the file to upload (the system will download and upload it)'),\n});\n\nconst BrowserKeypressSchema = z.object({\n keys: z\n .string()\n .describe('Key(s) to press. Single key: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\". Combo: \"Control+A\", \"Shift+Enter\"'),\n});\n\n/**\n * Browser tool response interface\n * This is what the extension returns after executing the action\n */\nexport interface BrowserToolResponse {\n requiresBrowserExecution: true;\n action: string;\n args: Record<string, unknown>;\n toolCallId?: string; // Added to help extension correlate with callback\n}\n\n/**\n * Options for creating browser tools\n */\nexport interface CreateBrowserToolsOptions {\n /**\n * Optional callback that waits for browser action results.\n * When provided, tools will await this callback to get actual results from the extension.\n * When not provided, tools return markers immediately (for non-server contexts).\n */\n waitForResult?: BrowserToolCallback;\n}\n\n/**\n * Format browser action result for LLM consumption\n */\nfunction formatResultForLLM(\n result: BrowserActionResult,\n action: string,\n actionArgs?: Record<string, unknown>\n): string {\n if (!result.success && result.error) {\n const errorType = result.errorType ? `[${result.errorType}] ` : '';\n return `${errorType}Action \"${action}\" failed: ${result.error}`;\n }\n\n const parts: string[] = [];\n\n // Verification result (Skyvern pattern)\n if (result.verification) {\n const v = result.verification;\n if (v.dialogClosed || v.formGone || v.urlChanged) {\n parts.push(`✓ ${v.changeDescription}`);\n } else if (!v.significantChange) {\n parts.push(`⚠ No change detected`);\n }\n }\n\n // Page state\n if (result.url) parts.push(`URL: ${result.url}`);\n if (result.title) parts.push(`Title: ${result.title}`);\n if (result.elementList) parts.push(`\\nElements:\\n${result.elementList}`);\n\n return parts.join('\\n');\n}\n\n/**\n * Create browser tools with optional callback for waiting on results\n *\n * When waitForResult callback is provided:\n * 1. Tool returns marker that triggers extension\n * 2. Tool then awaits callback to get actual results\n * 3. Returns real page state to LLM\n *\n * When no callback:\n * 1. Tool returns marker only (for non-server contexts)\n *\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\n */\nexport function createBrowserTools(\n options?: CreateBrowserToolsOptions\n): DynamicStructuredTool[] {\n const { waitForResult } = options || {};\n const tools: DynamicStructuredTool[] = [];\n\n /**\n * Helper to create tool function that optionally waits for results\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\n */\n const createToolFunction = (action: string) => {\n return async (\n args: Record<string, unknown>,\n config?: { toolCall?: { id?: string } }\n ): Promise<string> => {\n const toolCallId =\n config?.toolCall?.id ??\n `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\n\n // Create marker for extension\n const marker: BrowserToolResponse = {\n requiresBrowserExecution: true,\n action,\n args,\n toolCallId,\n };\n\n // If no callback, return marker immediately (extension handles via SSE interception)\n if (!waitForResult) {\n return JSON.stringify(marker);\n }\n\n // With callback: wait for actual results from extension\n // The marker is still returned initially via SSE, but we wait for the callback\n try {\n const result = await waitForResult(action, args, toolCallId);\n return formatResultForLLM(result, action, args);\n } catch (error) {\n const errorMessage =\n error instanceof Error ? error.message : String(error);\n return `❌ **ACTION FAILED**: Browser action \"${action}\" failed: ${errorMessage}\n\n**REQUIRED**: Analyze why the action failed and try an alternative approach.`;\n }\n };\n };\n\n // browser_click - PRIMARY action for buttons and links\n tools.push(\n tool(createToolFunction('click'), {\n name: EBrowserTools.CLICK,\n description: `Click an element by its [index] from the element list.\n\n**WHEN TO USE (prioritize this over keypress):**\n- Buttons labeled \"Send\", \"Submit\", \"Save\", \"OK\", \"Confirm\" → CLICK them\n- Links (a tags) to navigate\n- Checkboxes, radio buttons to select\n- Any clickable element that performs an action\n\n**SEMANTIC ELEMENT IDENTIFICATION:**\n- Look for aria-label=\"Send\", aria-label=\"Submit\" for action buttons\n- Look for type=\"submit\" for form submission\n- Look for button text like \"Send\", \"Submit\", \"Save\"\n- For email compose: Find the Send button, DON'T use keyboard shortcuts\n\nReturns verification: dialog closed, form submitted, URL changed, or NO CHANGE (action may have failed).`,\n schema: BrowserClickSchema,\n })\n );\n\n // browser_type\n tools.push(\n tool(createToolFunction('type'), {\n name: EBrowserTools.TYPE,\n description: `Type text into an input/textarea element by [index].\n\n**WHEN TO USE:**\n- Input fields (fieldType=\"email\", \"search\", \"phone\", etc.)\n- Textareas for longer content\n- Contenteditable divs (like Gmail compose body)\n\n**OPTIONS:**\n- pressEnter: true → Press Enter after typing (for search forms)\n- pressEnter: false → Just type without submitting\n\n**DO NOT use pressEnter to send emails** - find and CLICK the Send button instead.\n\nReturns verification showing if text was entered or form was submitted.`,\n schema: BrowserTypeSchema,\n })\n );\n\n // browser_navigate\n tools.push(\n tool(createToolFunction('navigate'), {\n name: EBrowserTools.NAVIGATE,\n description: `Navigate to a URL (include https://). Returns new page's element list.`,\n schema: BrowserNavigateSchema,\n })\n );\n\n // browser_scroll\n tools.push(\n tool(createToolFunction('scroll'), {\n name: EBrowserTools.SCROLL,\n description: `Scroll the page. Returns updated element list with newly visible elements.`,\n schema: BrowserScrollSchema,\n })\n );\n\n // browser_extract\n tools.push(\n tool(createToolFunction('extract'), {\n name: EBrowserTools.EXTRACT,\n description: `Extract content from the current page.\nReturns page URL, title, and element list.`,\n schema: BrowserExtractSchema,\n })\n );\n\n // browser_hover\n tools.push(\n tool(createToolFunction('hover'), {\n name: EBrowserTools.HOVER,\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\n schema: BrowserHoverSchema,\n })\n );\n\n // browser_wait\n tools.push(\n tool(createToolFunction('wait'), {\n name: EBrowserTools.WAIT,\n description: `Wait for a specified duration for page content to load.\nUse this after actions that trigger async content loading.\nAfter waiting, you receive an updated element list.`,\n schema: BrowserWaitSchema,\n })\n );\n\n // browser_back\n tools.push(\n tool(createToolFunction('back'), {\n name: EBrowserTools.BACK,\n description: `Go back to the previous page in browser history.\nAfter going back, you receive the previous page's element list.`,\n schema: BrowserBackSchema,\n })\n );\n\n // browser_screenshot\n tools.push(\n tool(createToolFunction('screenshot'), {\n name: EBrowserTools.SCREENSHOT,\n description: `Capture a screenshot. Use browser_get_page_state for element list.`,\n schema: BrowserScreenshotSchema,\n })\n );\n\n // browser_get_page_state\n tools.push(\n tool(createToolFunction('get_page_state'), {\n name: EBrowserTools.GET_PAGE_STATE,\n description: `Get current page URL, title, and interactive elements list with index numbers.`,\n schema: BrowserGetPageStateSchema,\n })\n );\n\n // browser_select_option\n tools.push(\n tool(createToolFunction('select_option'), {\n name: EBrowserTools.SELECT_OPTION,\n description: `Select an option from a dropdown. Handles both native <select> and custom dropdowns.`,\n schema: BrowserSelectOptionSchema,\n })\n );\n\n // browser_upload_file - Skyvern-inspired for file input handling\n tools.push(\n tool(createToolFunction('upload_file'), {\n name: EBrowserTools.UPLOAD_FILE,\n description: `Upload a file to a file input element.\nProvide the index of the file input and the URL of the file to upload.\nThe system will download the file and attach it to the input.\nAfter upload, you receive an updated element list.`,\n schema: BrowserUploadFileSchema,\n })\n );\n\n // browser_keypress - LAST RESORT for keyboard shortcuts\n tools.push(\n tool(createToolFunction('keypress'), {\n name: EBrowserTools.KEYPRESS,\n description: `Press keyboard key(s) - USE ONLY AS LAST RESORT after clicking fails.\n\n**PREFER browser_click OVER keypress:**\n- To send email → Find and CLICK the Send button\n- To submit form → CLICK the Submit button\n- To close modal → CLICK the X or Cancel button\n\n**WHEN TO USE keypress:**\n- Navigation: \"Tab\", \"ArrowDown\", \"ArrowUp\" in dropdowns\n- Escape to close popups if no X button exists\n- Control+A, Control+C for select/copy operations\n- ONLY use Enter to submit if there's NO visible submit button\n\nSingle keys: \"Enter\", \"Escape\", \"Tab\", \"ArrowDown\"\nKey combos: \"Control+A\", \"Shift+Enter\"\n\nReturns verification showing if page state changed.`,\n schema: BrowserKeypressSchema,\n })\n );\n\n return tools;\n}\n"],"names":[],"mappings":";;;AAIA;;;AAGG;AACU,MAAA,aAAa,GAAG;AAC3B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,QAAQ,EAAE,kBAAkB;AAC5B,IAAA,MAAM,EAAE,gBAAgB;AACxB,IAAA,OAAO,EAAE,iBAAiB;AAC1B,IAAA,KAAK,EAAE,eAAe;AACtB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,IAAI,EAAE,cAAc;AACpB,IAAA,UAAU,EAAE,oBAAoB;AAChC,IAAA,cAAc,EAAE,wBAAwB;;AAExC,IAAA,aAAa,EAAE,uBAAuB;AACtC,IAAA,WAAW,EAAE,qBAAqB;AAClC,IAAA,QAAQ,EAAE,kBAAkB;;AA4D9B;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAEpC,EAAA;AACC,IAAA,IAAI,CAAC,GAAG,EAAE,OAAO,EAAE;AACjB,QAAA,OAAO,KAAK;;IAGd,MAAM,gBAAgB,GAAG,GAAG,CAAC,OAAO,CAAC,4BAA4B,CAAC;IAClE,MAAM,cAAc,GAAG,GAAG,CAAC,OAAO,CAAC,0BAA0B,CAAC;AAE9D,IAAA,OAAO,gBAAgB,KAAK,MAAM,IAAI,cAAc,KAAK,MAAM;AACjE;AAEA;AACA,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;AAClC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CACP,0FAA0F,CAC3F;AACJ,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;AACjC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,oDAAoD,CAAC;IACjE,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAE;AACT,SAAA,OAAO;AACP,SAAA,QAAQ;SACR,QAAQ,CAAC,+DAA+D,CAAC;AAC7E,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;AACrC,IAAA,GAAG,EAAE;AACF,SAAA,MAAM;SACN,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAE;SACR,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC;SACpC,QAAQ,CAAC,qBAAqB,CAAC;AAClC,IAAA,MAAM,EAAE;AACL,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,iDAAiD,CAAC;AAC/D,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,qDAAqD,CAAC;AACnE,CAAA,CAAC;AAEF,MAAM,kBAAkB,GAAG,CAAC,CAAC,MAAM,CAAC;IAClC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,+CAA+C,CAAC;AAC5E,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;AACjC,IAAA,QAAQ,EAAE;AACP,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,sCAAsC,CAAC;AACpD,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAEtC,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE5C,MAAM,yBAAyB,GAAG,CAAC,CAAC,MAAM,CAAC,EAAE,CAAC;AAE9C;AACA,MAAM,yBAAyB,GAAG,CAAC,CAAC,MAAM,CAAC;AACzC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,iDAAiD,CAAC;AAC9D,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;AACN,SAAA,QAAQ;SACR,QAAQ,CAAC,gJAAgJ,CAAC;AAC9J,CAAA,CAAC;AAEF,MAAM,uBAAuB,GAAG,CAAC,CAAC,MAAM,CAAC;AACvC,IAAA,KAAK,EAAE;AACJ,SAAA,MAAM;SACN,QAAQ,CAAC,4CAA4C,CAAC;AACzD,IAAA,OAAO,EAAE;AACN,SAAA,MAAM;SACN,QAAQ,CAAC,oEAAoE,CAAC;AAClF,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;AACrC,IAAA,IAAI,EAAE;AACH,SAAA,MAAM;SACN,QAAQ,CAAC,uGAAuG,CAAC;AACrH,CAAA,CAAC;AAyBF;;AAEG;AACH,SAAS,kBAAkB,CACzB,MAA2B,EAC3B,MAAc,EACd,UAAoC,EAAA;IAEpC,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,MAAM,SAAS,GAAG,MAAM,CAAC,SAAS,GAAG,CAAI,CAAA,EAAA,MAAM,CAAC,SAAS,CAAA,EAAA,CAAI,GAAG,EAAE;QAClE,OAAO,CAAA,EAAG,SAAS,CAAW,QAAA,EAAA,MAAM,aAAa,MAAM,CAAC,KAAK,CAAA,CAAE;;IAGjE,MAAM,KAAK,GAAa,EAAE;;AAG1B,IAAA,IAAI,MAAM,CAAC,YAAY,EAAE;AACvB,QAAA,MAAM,CAAC,GAAG,MAAM,CAAC,YAAY;AAC7B,QAAA,IAAI,CAAC,CAAC,YAAY,IAAI,CAAC,CAAC,QAAQ,IAAI,CAAC,CAAC,UAAU,EAAE;YAChD,KAAK,CAAC,IAAI,CAAC,CAAA,EAAA,EAAK,CAAC,CAAC,iBAAiB,CAAE,CAAA,CAAC;;AACjC,aAAA,IAAI,CAAC,CAAC,CAAC,iBAAiB,EAAE;AAC/B,YAAA,KAAK,CAAC,IAAI,CAAC,CAAA,oBAAA,CAAsB,CAAC;;;;IAKtC,IAAI,MAAM,CAAC,GAAG;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,KAAA,EAAQ,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;IAChD,IAAI,MAAM,CAAC,KAAK;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,OAAA,EAAU,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;IACtD,IAAI,MAAM,CAAC,WAAW;QAAE,KAAK,CAAC,IAAI,CAAC,CAAA,aAAA,EAAgB,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;AAExE,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAChC,OAAmC,EAAA;AAEnC,IAAA,MAAM,EAAE,aAAa,EAAE,GAAG,OAAO,IAAI,EAAE;IACvC,MAAM,KAAK,GAA4B,EAAE;AAEzC;;;AAGG;AACH,IAAA,MAAM,kBAAkB,GAAG,CAAC,MAAc,KAAI;AAC5C,QAAA,OAAO,OACL,IAA6B,EAC7B,MAAuC,KACpB;AACnB,YAAA,MAAM,UAAU,GACd,MAAM,EAAE,QAAQ,EAAE,EAAE;gBACpB,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,IAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAG7D,YAAA,MAAM,MAAM,GAAwB;AAClC,gBAAA,wBAAwB,EAAE,IAAI;gBAC9B,MAAM;gBACN,IAAI;gBACJ,UAAU;aACX;;YAGD,IAAI,CAAC,aAAa,EAAE;AAClB,gBAAA,OAAO,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC;;;;AAK/B,YAAA,IAAI;gBACF,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,MAAM,EAAE,IAAI,EAAE,UAAU,CAAC;gBAC5D,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,EAAE,IAAI,CAAC;;YAC/C,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAChB,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;gBACxD,OAAO,CAAA,qCAAA,EAAwC,MAAM,CAAA,UAAA,EAAa,YAAY;;6EAET;;AAEzE,SAAC;AACH,KAAC;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;;AAcsF,wGAAA,CAAA;AACnG,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;AAaqD,uEAAA,CAAA;AAClE,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAwE,sEAAA,CAAA;AACrF,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,QAAQ,CAAC,EAAE;QACjC,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAA4E,0EAAA,CAAA;AACzF,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,SAAS,CAAC,EAAE;QAClC,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACwB,0CAAA,CAAA;AACrC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,OAAO,CAAC,EAAE;QAChC,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACmE,qFAAA,CAAA;AAChF,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAEiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,MAAM,CAAC,EAAE;QAC/B,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC6C,+DAAA,CAAA;AAC1D,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,YAAY,CAAC,EAAE;QACrC,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAoE,kEAAA,CAAA;AACjF,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,gBAAgB,CAAC,EAAE;QACzC,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAgF,8EAAA,CAAA;AAC7F,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,eAAe,CAAC,EAAE;QACxC,IAAI,EAAE,aAAa,CAAC,aAAa;AACjC,QAAA,WAAW,EAAE,CAAsF,oFAAA,CAAA;AACnG,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,aAAa,CAAC,EAAE;QACtC,IAAI,EAAE,aAAa,CAAC,WAAW;AAC/B,QAAA,WAAW,EAAE,CAAA;;;AAGgC,kDAAA,CAAA;AAC7C,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CAAC,CACH;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CAAC,kBAAkB,CAAC,UAAU,CAAC,EAAE;QACnC,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;;;;;;;;;;;;;;;;AAgBiC,mDAAA,CAAA;AAC9C,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CAAC,CACH;AAED,IAAA,OAAO,KAAK;AACd;;;;"}
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "illuma-agents",
3
- "version": "1.0.30",
3
+ "version": "1.0.31",
4
4
  "main": "./dist/cjs/main.cjs",
5
5
  "module": "./dist/esm/main.mjs",
6
6
  "types": "./dist/types/index.d.ts",
@@ -302,12 +302,25 @@ export function createBrowserTools(
302
302
  };
303
303
  };
304
304
 
305
- // browser_click
305
+ // browser_click - PRIMARY action for buttons and links
306
306
  tools.push(
307
307
  tool(createToolFunction('click'), {
308
308
  name: EBrowserTools.CLICK,
309
- description: `Click an element by its index number from the element list (e.g., [0], [1]).
310
- Returns state verification showing if UI changed (dialog closed, form submitted, etc.).`,
309
+ description: `Click an element by its [index] from the element list.
310
+
311
+ **WHEN TO USE (prioritize this over keypress):**
312
+ - Buttons labeled "Send", "Submit", "Save", "OK", "Confirm" → CLICK them
313
+ - Links (a tags) to navigate
314
+ - Checkboxes, radio buttons to select
315
+ - Any clickable element that performs an action
316
+
317
+ **SEMANTIC ELEMENT IDENTIFICATION:**
318
+ - Look for aria-label="Send", aria-label="Submit" for action buttons
319
+ - Look for type="submit" for form submission
320
+ - Look for button text like "Send", "Submit", "Save"
321
+ - For email compose: Find the Send button, DON'T use keyboard shortcuts
322
+
323
+ Returns verification: dialog closed, form submitted, URL changed, or NO CHANGE (action may have failed).`,
311
324
  schema: BrowserClickSchema,
312
325
  })
313
326
  );
@@ -316,8 +329,20 @@ Returns state verification showing if UI changed (dialog closed, form submitted,
316
329
  tools.push(
317
330
  tool(createToolFunction('type'), {
318
331
  name: EBrowserTools.TYPE,
319
- description: `Type text into an input element by index. Set pressEnter: true to submit.
320
- Returns state verification showing if form submitted or errors appeared.`,
332
+ description: `Type text into an input/textarea element by [index].
333
+
334
+ **WHEN TO USE:**
335
+ - Input fields (fieldType="email", "search", "phone", etc.)
336
+ - Textareas for longer content
337
+ - Contenteditable divs (like Gmail compose body)
338
+
339
+ **OPTIONS:**
340
+ - pressEnter: true → Press Enter after typing (for search forms)
341
+ - pressEnter: false → Just type without submitting
342
+
343
+ **DO NOT use pressEnter to send emails** - find and CLICK the Send button instead.
344
+
345
+ Returns verification showing if text was entered or form was submitted.`,
321
346
  schema: BrowserTypeSchema,
322
347
  })
323
348
  );
@@ -420,15 +445,27 @@ After upload, you receive an updated element list.`,
420
445
  })
421
446
  );
422
447
 
423
- // browser_keypress - For keyboard shortcuts and special keys
448
+ // browser_keypress - LAST RESORT for keyboard shortcuts
424
449
  tools.push(
425
450
  tool(createToolFunction('keypress'), {
426
451
  name: EBrowserTools.KEYPRESS,
427
- description: `Press keyboard key(s) on the page.
428
- Single keys: "Enter", "Escape", "Tab", "ArrowDown", "ArrowUp", "Backspace", "Delete"
429
- Key combos: "Control+A" (select all), "Control+C" (copy), "Shift+Enter" (newline)
430
- Use this for form submission, closing modals, navigating dropdowns.
431
- After keypress, you receive an updated element list.`,
452
+ description: `Press keyboard key(s) - USE ONLY AS LAST RESORT after clicking fails.
453
+
454
+ **PREFER browser_click OVER keypress:**
455
+ - To send email Find and CLICK the Send button
456
+ - To submit form CLICK the Submit button
457
+ - To close modal → CLICK the X or Cancel button
458
+
459
+ **WHEN TO USE keypress:**
460
+ - Navigation: "Tab", "ArrowDown", "ArrowUp" in dropdowns
461
+ - Escape to close popups if no X button exists
462
+ - Control+A, Control+C for select/copy operations
463
+ - ONLY use Enter to submit if there's NO visible submit button
464
+
465
+ Single keys: "Enter", "Escape", "Tab", "ArrowDown"
466
+ Key combos: "Control+A", "Shift+Enter"
467
+
468
+ Returns verification showing if page state changed.`,
432
469
  schema: BrowserKeypressSchema,
433
470
  })
434
471
  );