illuma-agents 1.0.26 → 1.0.27

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.
@@ -62,25 +62,79 @@ const BrowserBackSchema = zod.z.object({});
62
62
  const BrowserScreenshotSchema = zod.z.object({});
63
63
  const BrowserGetPageStateSchema = zod.z.object({});
64
64
  /**
65
- * Create placeholder browser tools that signal to the extension to execute locally
66
- * The server creates these tools to give the LLM the schema, but actual execution
67
- * happens in the browser extension which intercepts tool calls
65
+ * Format browser action result for LLM consumption
66
+ */
67
+ function formatResultForLLM(result, action) {
68
+ if (!result.success && result.error) {
69
+ return `Browser action "${action}" failed: ${result.error}`;
70
+ }
71
+ const parts = [];
72
+ if (result.url) {
73
+ parts.push(`**Current URL:** ${result.url}`);
74
+ }
75
+ if (result.title) {
76
+ parts.push(`**Page Title:** ${result.title}`);
77
+ }
78
+ if (result.elementList) {
79
+ parts.push(`\n**Interactive Elements:**\n${result.elementList}`);
80
+ }
81
+ if (result.screenshot) {
82
+ parts.push(`\n[Screenshot captured and displayed to user]`);
83
+ }
84
+ if (parts.length === 0) {
85
+ return `Browser action "${action}" completed successfully.`;
86
+ }
87
+ return parts.join('\n');
88
+ }
89
+ /**
90
+ * Create browser tools with optional callback for waiting on results
91
+ *
92
+ * When waitForResult callback is provided:
93
+ * 1. Tool returns marker that triggers extension
94
+ * 2. Tool then awaits callback to get actual results
95
+ * 3. Returns real page state to LLM
96
+ *
97
+ * When no callback:
98
+ * 1. Tool returns marker only (for non-server contexts)
68
99
  *
69
100
  * NOTE: These tools use TEXT-BASED element lists, NOT screenshots
70
101
  * Screenshots would be 100K+ tokens each - element lists are ~100 tokens
71
102
  */
72
- function createBrowserTools() {
103
+ function createBrowserTools(options) {
104
+ const { waitForResult } = options || {};
73
105
  const tools$1 = [];
74
- // browser_click
75
- tools$1.push(tools.tool(async (args) => {
76
- // This returns a marker that tells the extension to execute locally
77
- const response = {
78
- requiresBrowserExecution: true,
79
- action: 'click',
80
- args,
106
+ /**
107
+ * Helper to create tool function that optionally waits for results
108
+ * The toolCallId is extracted from the RunnableConfig passed by LangChain
109
+ */
110
+ const createToolFunction = (action) => {
111
+ return async (args, config) => {
112
+ const toolCallId = config?.toolCall?.id || `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;
113
+ // Create marker for extension
114
+ const marker = {
115
+ requiresBrowserExecution: true,
116
+ action,
117
+ args,
118
+ toolCallId,
119
+ };
120
+ // If no callback, return marker immediately (extension handles via SSE interception)
121
+ if (!waitForResult) {
122
+ return JSON.stringify(marker);
123
+ }
124
+ // With callback: wait for actual results from extension
125
+ // The marker is still returned initially via SSE, but we wait for the callback
126
+ try {
127
+ const result = await waitForResult(action, args, toolCallId);
128
+ return formatResultForLLM(result, action);
129
+ }
130
+ catch (error) {
131
+ const errorMessage = error instanceof Error ? error.message : String(error);
132
+ return `Browser action "${action}" failed: ${errorMessage}`;
133
+ }
81
134
  };
82
- return JSON.stringify(response);
83
- }, {
135
+ };
136
+ // browser_click
137
+ tools$1.push(tools.tool(createToolFunction('click'), {
84
138
  name: EBrowserTools.CLICK,
85
139
  description: `Click an element on the current web page by its index number.
86
140
  The element list shows clickable items like: [0]<button>Submit</button> [1]<a href="/home">Home</a>
@@ -89,14 +143,7 @@ After clicking, you receive an updated element list showing the new page state.`
89
143
  schema: BrowserClickSchema,
90
144
  }));
91
145
  // browser_type
92
- tools$1.push(tools.tool(async (args) => {
93
- const response = {
94
- requiresBrowserExecution: true,
95
- action: 'type',
96
- args,
97
- };
98
- return JSON.stringify(response);
99
- }, {
146
+ tools$1.push(tools.tool(createToolFunction('type'), {
100
147
  name: EBrowserTools.TYPE,
101
148
  description: `Type text into an input element on the page.
102
149
  Find the input element in the list by its index (e.g., [5]<input placeholder="Search">).
@@ -105,28 +152,14 @@ After typing, you receive an updated element list.`,
105
152
  schema: BrowserTypeSchema,
106
153
  }));
107
154
  // browser_navigate
108
- tools$1.push(tools.tool(async (args) => {
109
- const response = {
110
- requiresBrowserExecution: true,
111
- action: 'navigate',
112
- args,
113
- };
114
- return JSON.stringify(response);
115
- }, {
155
+ tools$1.push(tools.tool(createToolFunction('navigate'), {
116
156
  name: EBrowserTools.NAVIGATE,
117
157
  description: `Navigate to a URL. Always include the full URL with https://.
118
158
  After navigation, you receive the new page's element list.`,
119
159
  schema: BrowserNavigateSchema,
120
160
  }));
121
161
  // browser_scroll
122
- tools$1.push(tools.tool(async (args) => {
123
- const response = {
124
- requiresBrowserExecution: true,
125
- action: 'scroll',
126
- args,
127
- };
128
- return JSON.stringify(response);
129
- }, {
162
+ tools$1.push(tools.tool(createToolFunction('scroll'), {
130
163
  name: EBrowserTools.SCROLL,
131
164
  description: `Scroll the page to reveal more content.
132
165
  Use 'down' to scroll down, 'up' to scroll up.
@@ -134,42 +167,21 @@ After scrolling, you receive an updated element list with newly visible elements
134
167
  schema: BrowserScrollSchema,
135
168
  }));
136
169
  // browser_extract
137
- tools$1.push(tools.tool(async (args) => {
138
- const response = {
139
- requiresBrowserExecution: true,
140
- action: 'extract',
141
- args,
142
- };
143
- return JSON.stringify(response);
144
- }, {
170
+ tools$1.push(tools.tool(createToolFunction('extract'), {
145
171
  name: EBrowserTools.EXTRACT,
146
172
  description: `Extract content from the current page.
147
173
  Returns page URL, title, and element list.`,
148
174
  schema: BrowserExtractSchema,
149
175
  }));
150
176
  // browser_hover
151
- tools$1.push(tools.tool(async (args) => {
152
- const response = {
153
- requiresBrowserExecution: true,
154
- action: 'hover',
155
- args,
156
- };
157
- return JSON.stringify(response);
158
- }, {
177
+ tools$1.push(tools.tool(createToolFunction('hover'), {
159
178
  name: EBrowserTools.HOVER,
160
179
  description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.
161
180
  After hovering, you receive an updated element list with any newly revealed elements.`,
162
181
  schema: BrowserHoverSchema,
163
182
  }));
164
183
  // browser_wait
165
- tools$1.push(tools.tool(async (args) => {
166
- const response = {
167
- requiresBrowserExecution: true,
168
- action: 'wait',
169
- args,
170
- };
171
- return JSON.stringify(response);
172
- }, {
184
+ tools$1.push(tools.tool(createToolFunction('wait'), {
173
185
  name: EBrowserTools.WAIT,
174
186
  description: `Wait for a specified duration for page content to load.
175
187
  Use this after actions that trigger async content loading.
@@ -177,47 +189,26 @@ After waiting, you receive an updated element list.`,
177
189
  schema: BrowserWaitSchema,
178
190
  }));
179
191
  // browser_back
180
- tools$1.push(tools.tool(async (args) => {
181
- const response = {
182
- requiresBrowserExecution: true,
183
- action: 'back',
184
- args,
185
- };
186
- return JSON.stringify(response);
187
- }, {
192
+ tools$1.push(tools.tool(createToolFunction('back'), {
188
193
  name: EBrowserTools.BACK,
189
194
  description: `Go back to the previous page in browser history.
190
195
  After going back, you receive the previous page's element list.`,
191
196
  schema: BrowserBackSchema,
192
197
  }));
193
198
  // browser_screenshot
194
- tools$1.push(tools.tool(async (args) => {
195
- const response = {
196
- requiresBrowserExecution: true,
197
- action: 'screenshot',
198
- args,
199
- };
200
- return JSON.stringify(response);
201
- }, {
199
+ tools$1.push(tools.tool(createToolFunction('screenshot'), {
202
200
  name: EBrowserTools.SCREENSHOT,
203
201
  description: `Capture a screenshot of the current page.
204
- NOTE: Screenshot is displayed to the USER only, not returned to you.
205
- Use browser_get_page_state instead to get the element list.`,
202
+ Returns the page state with a note that screenshot was displayed to the user.
203
+ Use browser_get_page_state to get the element list for automation.`,
206
204
  schema: BrowserScreenshotSchema,
207
205
  }));
208
206
  // browser_get_page_state
209
- tools$1.push(tools.tool(async (args) => {
210
- const response = {
211
- requiresBrowserExecution: true,
212
- action: 'get_page_state',
213
- args,
214
- };
215
- return JSON.stringify(response);
216
- }, {
207
+ tools$1.push(tools.tool(createToolFunction('get_page_state'), {
217
208
  name: EBrowserTools.GET_PAGE_STATE,
218
209
  description: `Get the current page state including URL, title, and all interactive elements.
219
210
  Use this at the start of a task to see what elements are available.
220
- Returns a text list of elements with their index numbers.`,
211
+ Returns a text list of elements with their index numbers for interaction.`,
221
212
  schema: BrowserGetPageStateSchema,
222
213
  }));
223
214
  return tools$1;
@@ -1 +1 @@
1
- {"version":3,"file":"BrowserTools.cjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\r\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\r\nimport type * as t from '@/types';\r\n\r\n/**\r\n * Browser tool names - keep in sync with ranger-browser extension\r\n * These tools execute locally in the browser extension, NOT on the server\r\n */\r\nexport const EBrowserTools = {\r\n CLICK: 'browser_click',\r\n TYPE: 'browser_type',\r\n NAVIGATE: 'browser_navigate',\r\n SCROLL: 'browser_scroll',\r\n EXTRACT: 'browser_extract',\r\n HOVER: 'browser_hover',\r\n WAIT: 'browser_wait',\r\n BACK: 'browser_back',\r\n SCREENSHOT: 'browser_screenshot',\r\n GET_PAGE_STATE: 'browser_get_page_state',\r\n} as const;\r\n\r\nexport type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];\r\n\r\n/**\r\n * Check if browser capability is available based on request headers or context\r\n * The browser extension sets these headers when connected:\r\n * - X-Ranger-Browser-Extension: true\r\n * - X-Ranger-Browser-Capable: true\r\n */\r\nexport function hasBrowserCapability(req?: { headers?: Record<string, string | string[] | undefined> }): boolean {\r\n if (!req?.headers) {\r\n return false;\r\n }\r\n \r\n const browserExtension = req.headers['x-ranger-browser-extension'];\r\n const browserCapable = req.headers['x-ranger-browser-capable'];\r\n \r\n return browserExtension === 'true' || browserCapable === 'true';\r\n}\r\n\r\n// Tool schemas\r\nconst BrowserClickSchema = z.object({\r\n index: z.number().describe('The index number [0], [1], etc. of the element to click from the page state element list'),\r\n});\r\n\r\nconst BrowserTypeSchema = z.object({\r\n index: z.number().describe('The index number of the input element to type into'),\r\n text: z.string().describe('The text to type into the element'),\r\n pressEnter: z.boolean().optional().describe('Whether to press Enter after typing (useful for search forms)'),\r\n});\r\n\r\nconst BrowserNavigateSchema = z.object({\r\n url: z.string().describe('The full URL to navigate to (must include https://)'),\r\n});\r\n\r\nconst BrowserScrollSchema = z.object({\r\n direction: z.enum(['up', 'down', 'left', 'right']).describe('Direction to scroll'),\r\n amount: z.number().optional().describe('Pixels to scroll (default: one viewport height)'),\r\n});\r\n\r\nconst BrowserExtractSchema = z.object({\r\n query: z.string().optional().describe('Optional: specific content to extract from the page'),\r\n});\r\n\r\nconst BrowserHoverSchema = z.object({\r\n index: z.number().describe('The index number of the element to hover over'),\r\n});\r\n\r\nconst BrowserWaitSchema = z.object({\r\n duration: z.number().optional().describe('Milliseconds to wait (default: 1000)'),\r\n});\r\n\r\nconst BrowserBackSchema = z.object({});\r\n\r\nconst BrowserScreenshotSchema = z.object({});\r\n\r\nconst BrowserGetPageStateSchema = z.object({});\r\n\r\n/**\r\n * Browser tool response interface\r\n * This is what the extension returns after executing the action\r\n */\r\nexport interface BrowserToolResponse {\r\n requiresBrowserExecution: true;\r\n action: string;\r\n args: Record<string, unknown>;\r\n}\r\n\r\n/**\r\n * Create placeholder browser tools that signal to the extension to execute locally\r\n * The server creates these tools to give the LLM the schema, but actual execution\r\n * happens in the browser extension which intercepts tool calls\r\n * \r\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\r\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\r\n */\r\nexport function createBrowserTools(): DynamicStructuredTool[] {\r\n const tools: DynamicStructuredTool[] = [];\r\n\r\n // browser_click\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n // This returns a marker that tells the extension to execute locally\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'click',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.CLICK,\r\n description: `Click an element on the current web page by its index number.\r\nThe element list shows clickable items like: [0]<button>Submit</button> [1]<a href=\"/home\">Home</a>\r\nUse the index number in brackets to click that element.\r\nAfter clicking, you receive an updated element list showing the new page state.`,\r\n schema: BrowserClickSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_type\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'type',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.TYPE,\r\n description: `Type text into an input element on the page.\r\nFind the input element in the list by its index (e.g., [5]<input placeholder=\"Search\">).\r\nSet pressEnter: true to submit forms after typing.\r\nAfter typing, you receive an updated element list.`,\r\n schema: BrowserTypeSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_navigate\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'navigate',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.NAVIGATE,\r\n description: `Navigate to a URL. Always include the full URL with https://.\r\nAfter navigation, you receive the new page's element list.`,\r\n schema: BrowserNavigateSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_scroll\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'scroll',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.SCROLL,\r\n description: `Scroll the page to reveal more content.\r\nUse 'down' to scroll down, 'up' to scroll up.\r\nAfter scrolling, you receive an updated element list with newly visible elements.`,\r\n schema: BrowserScrollSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_extract\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'extract',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.EXTRACT,\r\n description: `Extract content from the current page.\r\nReturns page URL, title, and element list.`,\r\n schema: BrowserExtractSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_hover\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'hover',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.HOVER,\r\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\r\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\r\n schema: BrowserHoverSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_wait\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'wait',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.WAIT,\r\n description: `Wait for a specified duration for page content to load.\r\nUse this after actions that trigger async content loading.\r\nAfter waiting, you receive an updated element list.`,\r\n schema: BrowserWaitSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_back\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'back',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.BACK,\r\n description: `Go back to the previous page in browser history.\r\nAfter going back, you receive the previous page's element list.`,\r\n schema: BrowserBackSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_screenshot\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'screenshot',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.SCREENSHOT,\r\n description: `Capture a screenshot of the current page.\r\nNOTE: Screenshot is displayed to the USER only, not returned to you.\r\nUse browser_get_page_state instead to get the element list.`,\r\n schema: BrowserScreenshotSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_get_page_state\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'get_page_state',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.GET_PAGE_STATE,\r\n description: `Get the current page state including URL, title, and all interactive elements.\r\nUse this at the start of a task to see what elements are available.\r\nReturns a text list of elements with their index numbers.`,\r\n schema: BrowserGetPageStateSchema,\r\n }\r\n )\r\n );\r\n\r\n return tools;\r\n}\r\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;;AAK1C;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAAiE,EAAA;AACpG,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;IAClC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,0FAA0F,CAAC;AACvH,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;IACjC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,oDAAoD,CAAC;IAChF,IAAI,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAEA,KAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,+DAA+D,CAAC;AAC7G,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;IACrC,GAAG,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAChF,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAEA,KAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,QAAQ,CAAC,qBAAqB,CAAC;AAClF,IAAA,MAAM,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,iDAAiD,CAAC;AAC1F,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAC7F,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,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,sCAAsC,CAAC;AACjF,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;AAY9C;;;;;;;AAOG;SACa,kBAAkB,GAAA;IAChC,MAAMC,OAAK,GAA4B,EAAE;;IAGzCA,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;;AAE9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,OAAO;YACf,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;AAG2D,+EAAA,CAAA;AACxE,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;AAG8B,kDAAA,CAAA;AAC3C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,UAAU;YAClB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;AACsC,0DAAA,CAAA;AACnD,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,QAAQ;YAChB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAAA;;AAE6D,iFAAA,CAAA;AAC1E,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,SAAS;YACjB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACsB,0CAAA,CAAA;AACnC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,OAAO;YACf,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACiE,qFAAA,CAAA;AAC9E,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAE+B,mDAAA,CAAA;AAC5C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC2C,+DAAA,CAAA;AACxD,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,YAAY;YACpB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAA;;AAEuC,2DAAA,CAAA;AACpD,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,gBAAgB;YACxB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAA;;AAEqC,yDAAA,CAAA;AAClD,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CACF,CACF;AAED,IAAA,OAAOD,OAAK;AACd;;;;;;"}
1
+ {"version":3,"file":"BrowserTools.cjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\r\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\r\nimport type * as t from '@/types';\r\n\r\n/**\r\n * Browser tool names - keep in sync with ranger-browser extension\r\n * These tools execute locally in the browser extension, NOT on the server\r\n */\r\nexport const EBrowserTools = {\r\n CLICK: 'browser_click',\r\n TYPE: 'browser_type',\r\n NAVIGATE: 'browser_navigate',\r\n SCROLL: 'browser_scroll',\r\n EXTRACT: 'browser_extract',\r\n HOVER: 'browser_hover',\r\n WAIT: 'browser_wait',\r\n BACK: 'browser_back',\r\n SCREENSHOT: 'browser_screenshot',\r\n GET_PAGE_STATE: 'browser_get_page_state',\r\n} as const;\r\n\r\nexport type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];\r\n\r\n/**\r\n * Callback function type for waiting on browser action results\r\n * This allows the server (Ranger) to provide a callback that waits for the extension\r\n * to POST results back to the server before returning to the LLM.\r\n * \r\n * @param action - The browser action (click, type, navigate, etc.)\r\n * @param args - Arguments for the action\r\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\r\n * @returns Promise that resolves with the actual browser result (page state, etc.)\r\n */\r\nexport type BrowserToolCallback = (\r\n action: string,\r\n args: Record<string, unknown>,\r\n toolCallId: string\r\n) => Promise<BrowserActionResult>;\r\n\r\n/**\r\n * Result returned from browser action execution\r\n */\r\nexport interface BrowserActionResult {\r\n success: boolean;\r\n url?: string;\r\n title?: string;\r\n elementList?: string; // Text-based element list\r\n error?: string;\r\n screenshot?: string; // Base64 screenshot (if requested)\r\n}\r\n\r\n/**\r\n * Check if browser capability is available based on request headers or context\r\n * The browser extension sets these headers when connected:\r\n * - X-Ranger-Browser-Extension: true\r\n * - X-Ranger-Browser-Capable: true\r\n */\r\nexport function hasBrowserCapability(req?: { headers?: Record<string, string | string[] | undefined> }): boolean {\r\n if (!req?.headers) {\r\n return false;\r\n }\r\n \r\n const browserExtension = req.headers['x-ranger-browser-extension'];\r\n const browserCapable = req.headers['x-ranger-browser-capable'];\r\n \r\n return browserExtension === 'true' || browserCapable === 'true';\r\n}\r\n\r\n// Tool schemas\r\nconst BrowserClickSchema = z.object({\r\n index: z.number().describe('The index number [0], [1], etc. of the element to click from the page state element list'),\r\n});\r\n\r\nconst BrowserTypeSchema = z.object({\r\n index: z.number().describe('The index number of the input element to type into'),\r\n text: z.string().describe('The text to type into the element'),\r\n pressEnter: z.boolean().optional().describe('Whether to press Enter after typing (useful for search forms)'),\r\n});\r\n\r\nconst BrowserNavigateSchema = z.object({\r\n url: z.string().describe('The full URL to navigate to (must include https://)'),\r\n});\r\n\r\nconst BrowserScrollSchema = z.object({\r\n direction: z.enum(['up', 'down', 'left', 'right']).describe('Direction to scroll'),\r\n amount: z.number().optional().describe('Pixels to scroll (default: one viewport height)'),\r\n});\r\n\r\nconst BrowserExtractSchema = z.object({\r\n query: z.string().optional().describe('Optional: specific content to extract from the page'),\r\n});\r\n\r\nconst BrowserHoverSchema = z.object({\r\n index: z.number().describe('The index number of the element to hover over'),\r\n});\r\n\r\nconst BrowserWaitSchema = z.object({\r\n duration: z.number().optional().describe('Milliseconds to wait (default: 1000)'),\r\n});\r\n\r\nconst BrowserBackSchema = z.object({});\r\n\r\nconst BrowserScreenshotSchema = z.object({});\r\n\r\nconst BrowserGetPageStateSchema = z.object({});\r\n\r\n/**\r\n * Browser tool response interface\r\n * This is what the extension returns after executing the action\r\n */\r\nexport interface BrowserToolResponse {\r\n requiresBrowserExecution: true;\r\n action: string;\r\n args: Record<string, unknown>;\r\n toolCallId?: string; // Added to help extension correlate with callback\r\n}\r\n\r\n/**\r\n * Options for creating browser tools\r\n */\r\nexport interface CreateBrowserToolsOptions {\r\n /**\r\n * Optional callback that waits for browser action results.\r\n * When provided, tools will await this callback to get actual results from the extension.\r\n * When not provided, tools return markers immediately (for non-server contexts).\r\n */\r\n waitForResult?: BrowserToolCallback;\r\n}\r\n\r\n/**\r\n * Format browser action result for LLM consumption\r\n */\r\nfunction formatResultForLLM(result: BrowserActionResult, action: string): string {\r\n if (!result.success && result.error) {\r\n return `Browser action \"${action}\" failed: ${result.error}`;\r\n }\r\n\r\n const parts: string[] = [];\r\n \r\n if (result.url) {\r\n parts.push(`**Current URL:** ${result.url}`);\r\n }\r\n if (result.title) {\r\n parts.push(`**Page Title:** ${result.title}`);\r\n }\r\n if (result.elementList) {\r\n parts.push(`\\n**Interactive Elements:**\\n${result.elementList}`);\r\n }\r\n if (result.screenshot) {\r\n parts.push(`\\n[Screenshot captured and displayed to user]`);\r\n }\r\n \r\n if (parts.length === 0) {\r\n return `Browser action \"${action}\" completed successfully.`;\r\n }\r\n \r\n return parts.join('\\n');\r\n}\r\n\r\n/**\r\n * Create browser tools with optional callback for waiting on results\r\n * \r\n * When waitForResult callback is provided:\r\n * 1. Tool returns marker that triggers extension\r\n * 2. Tool then awaits callback to get actual results\r\n * 3. Returns real page state to LLM\r\n * \r\n * When no callback:\r\n * 1. Tool returns marker only (for non-server contexts)\r\n * \r\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\r\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\r\n */\r\nexport function createBrowserTools(options?: CreateBrowserToolsOptions): DynamicStructuredTool[] {\r\n const { waitForResult } = options || {};\r\n const tools: DynamicStructuredTool[] = [];\r\n\r\n /**\r\n * Helper to create tool function that optionally waits for results\r\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\r\n */\r\n const createToolFunction = (action: string) => {\r\n return async (args: Record<string, unknown>, config?: { toolCall?: { id?: string } }): Promise<string> => {\r\n const toolCallId = config?.toolCall?.id || `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\r\n \r\n // Create marker for extension\r\n const marker: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action,\r\n args,\r\n toolCallId,\r\n };\r\n \r\n // If no callback, return marker immediately (extension handles via SSE interception)\r\n if (!waitForResult) {\r\n return JSON.stringify(marker);\r\n }\r\n \r\n // With callback: wait for actual results from extension\r\n // The marker is still returned initially via SSE, but we wait for the callback\r\n try {\r\n const result = await waitForResult(action, args, toolCallId);\r\n return formatResultForLLM(result, action);\r\n } catch (error) {\r\n const errorMessage = error instanceof Error ? error.message : String(error);\r\n return `Browser action \"${action}\" failed: ${errorMessage}`;\r\n }\r\n };\r\n };\r\n\r\n // browser_click\r\n tools.push(\r\n tool(\r\n createToolFunction('click'),\r\n {\r\n name: EBrowserTools.CLICK,\r\n description: `Click an element on the current web page by its index number.\r\nThe element list shows clickable items like: [0]<button>Submit</button> [1]<a href=\"/home\">Home</a>\r\nUse the index number in brackets to click that element.\r\nAfter clicking, you receive an updated element list showing the new page state.`,\r\n schema: BrowserClickSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_type\r\n tools.push(\r\n tool(\r\n createToolFunction('type'),\r\n {\r\n name: EBrowserTools.TYPE,\r\n description: `Type text into an input element on the page.\r\nFind the input element in the list by its index (e.g., [5]<input placeholder=\"Search\">).\r\nSet pressEnter: true to submit forms after typing.\r\nAfter typing, you receive an updated element list.`,\r\n schema: BrowserTypeSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_navigate\r\n tools.push(\r\n tool(\r\n createToolFunction('navigate'),\r\n {\r\n name: EBrowserTools.NAVIGATE,\r\n description: `Navigate to a URL. Always include the full URL with https://.\r\nAfter navigation, you receive the new page's element list.`,\r\n schema: BrowserNavigateSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_scroll\r\n tools.push(\r\n tool(\r\n createToolFunction('scroll'),\r\n {\r\n name: EBrowserTools.SCROLL,\r\n description: `Scroll the page to reveal more content.\r\nUse 'down' to scroll down, 'up' to scroll up.\r\nAfter scrolling, you receive an updated element list with newly visible elements.`,\r\n schema: BrowserScrollSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_extract\r\n tools.push(\r\n tool(\r\n createToolFunction('extract'),\r\n {\r\n name: EBrowserTools.EXTRACT,\r\n description: `Extract content from the current page.\r\nReturns page URL, title, and element list.`,\r\n schema: BrowserExtractSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_hover\r\n tools.push(\r\n tool(\r\n createToolFunction('hover'),\r\n {\r\n name: EBrowserTools.HOVER,\r\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\r\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\r\n schema: BrowserHoverSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_wait\r\n tools.push(\r\n tool(\r\n createToolFunction('wait'),\r\n {\r\n name: EBrowserTools.WAIT,\r\n description: `Wait for a specified duration for page content to load.\r\nUse this after actions that trigger async content loading.\r\nAfter waiting, you receive an updated element list.`,\r\n schema: BrowserWaitSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_back\r\n tools.push(\r\n tool(\r\n createToolFunction('back'),\r\n {\r\n name: EBrowserTools.BACK,\r\n description: `Go back to the previous page in browser history.\r\nAfter going back, you receive the previous page's element list.`,\r\n schema: BrowserBackSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_screenshot\r\n tools.push(\r\n tool(\r\n createToolFunction('screenshot'),\r\n {\r\n name: EBrowserTools.SCREENSHOT,\r\n description: `Capture a screenshot of the current page.\r\nReturns the page state with a note that screenshot was displayed to the user.\r\nUse browser_get_page_state to get the element list for automation.`,\r\n schema: BrowserScreenshotSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_get_page_state\r\n tools.push(\r\n tool(\r\n createToolFunction('get_page_state'),\r\n {\r\n name: EBrowserTools.GET_PAGE_STATE,\r\n description: `Get the current page state including URL, title, and all interactive elements.\r\nUse this at the start of a task to see what elements are available.\r\nReturns a text list of elements with their index numbers for interaction.`,\r\n schema: BrowserGetPageStateSchema,\r\n }\r\n )\r\n );\r\n\r\n return tools;\r\n}\r\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;;AAiC1C;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAAiE,EAAA;AACpG,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;IAClC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,0FAA0F,CAAC;AACvH,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAGA,KAAC,CAAC,MAAM,CAAC;IACjC,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,oDAAoD,CAAC;IAChF,IAAI,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAEA,KAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,+DAA+D,CAAC;AAC7G,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAGA,KAAC,CAAC,MAAM,CAAC;IACrC,GAAG,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAChF,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAEA,KAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,QAAQ,CAAC,qBAAqB,CAAC;AAClF,IAAA,MAAM,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,iDAAiD,CAAC;AAC1F,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAGA,KAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAEA,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAC7F,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,KAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,sCAAsC,CAAC;AACjF,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;AAyB9C;;AAEG;AACH,SAAS,kBAAkB,CAAC,MAA2B,EAAE,MAAc,EAAA;IACrE,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,OAAO,mBAAmB,MAAM,CAAA,UAAA,EAAa,MAAM,CAAC,KAAK,EAAE;;IAG7D,MAAM,KAAK,GAAa,EAAE;AAE1B,IAAA,IAAI,MAAM,CAAC,GAAG,EAAE;QACd,KAAK,CAAC,IAAI,CAAC,CAAA,iBAAA,EAAoB,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;;AAE9C,IAAA,IAAI,MAAM,CAAC,KAAK,EAAE;QAChB,KAAK,CAAC,IAAI,CAAC,CAAA,gBAAA,EAAmB,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;;AAE/C,IAAA,IAAI,MAAM,CAAC,WAAW,EAAE;QACtB,KAAK,CAAC,IAAI,CAAC,CAAA,6BAAA,EAAgC,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;;AAElE,IAAA,IAAI,MAAM,CAAC,UAAU,EAAE;AACrB,QAAA,KAAK,CAAC,IAAI,CAAC,CAAA,6CAAA,CAA+C,CAAC;;AAG7D,IAAA,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE;QACtB,OAAO,CAAA,gBAAA,EAAmB,MAAM,CAAA,yBAAA,CAA2B;;AAG7D,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAAC,OAAmC,EAAA;AACpE,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,OAAO,IAA6B,EAAE,MAAuC,KAAqB;AACvG,YAAA,MAAM,UAAU,GAAG,MAAM,EAAE,QAAQ,EAAE,EAAE,IAAI,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,CAAA,CAAA,EAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAGtG,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;AAC5D,gBAAA,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,CAAC;;YACzC,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAAG,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;AAC3E,gBAAA,OAAO,CAAmB,gBAAA,EAAA,MAAM,CAAa,UAAA,EAAA,YAAY,EAAE;;AAE/D,SAAC;AACH,KAAC;;IAGDA,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,OAAO,CAAC,EAC3B;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;AAG2D,+EAAA,CAAA;AACxE,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;AAG8B,kDAAA,CAAA;AAC3C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,UAAU,CAAC,EAC9B;QACE,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;AACsC,0DAAA,CAAA;AACnD,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,QAAQ,CAAC,EAC5B;QACE,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAAA;;AAE6D,iFAAA,CAAA;AAC1E,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,SAAS,CAAC,EAC7B;QACE,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACsB,0CAAA,CAAA;AACnC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,OAAO,CAAC,EAC3B;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACiE,qFAAA,CAAA;AAC9E,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAE+B,mDAAA,CAAA;AAC5C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC2C,+DAAA,CAAA;AACxD,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,YAAY,CAAC,EAChC;QACE,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAA;;AAE8C,kEAAA,CAAA;AAC3D,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CACF,CACF;;IAGDD,OAAK,CAAC,IAAI,CACRC,UAAI,CACF,kBAAkB,CAAC,gBAAgB,CAAC,EACpC;QACE,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAA;;AAEqD,yEAAA,CAAA;AAClE,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CACF,CACF;AAED,IAAA,OAAOD,OAAK;AACd;;;;;;"}
@@ -60,25 +60,79 @@ const BrowserBackSchema = z.object({});
60
60
  const BrowserScreenshotSchema = z.object({});
61
61
  const BrowserGetPageStateSchema = z.object({});
62
62
  /**
63
- * Create placeholder browser tools that signal to the extension to execute locally
64
- * The server creates these tools to give the LLM the schema, but actual execution
65
- * happens in the browser extension which intercepts tool calls
63
+ * Format browser action result for LLM consumption
64
+ */
65
+ function formatResultForLLM(result, action) {
66
+ if (!result.success && result.error) {
67
+ return `Browser action "${action}" failed: ${result.error}`;
68
+ }
69
+ const parts = [];
70
+ if (result.url) {
71
+ parts.push(`**Current URL:** ${result.url}`);
72
+ }
73
+ if (result.title) {
74
+ parts.push(`**Page Title:** ${result.title}`);
75
+ }
76
+ if (result.elementList) {
77
+ parts.push(`\n**Interactive Elements:**\n${result.elementList}`);
78
+ }
79
+ if (result.screenshot) {
80
+ parts.push(`\n[Screenshot captured and displayed to user]`);
81
+ }
82
+ if (parts.length === 0) {
83
+ return `Browser action "${action}" completed successfully.`;
84
+ }
85
+ return parts.join('\n');
86
+ }
87
+ /**
88
+ * Create browser tools with optional callback for waiting on results
89
+ *
90
+ * When waitForResult callback is provided:
91
+ * 1. Tool returns marker that triggers extension
92
+ * 2. Tool then awaits callback to get actual results
93
+ * 3. Returns real page state to LLM
94
+ *
95
+ * When no callback:
96
+ * 1. Tool returns marker only (for non-server contexts)
66
97
  *
67
98
  * NOTE: These tools use TEXT-BASED element lists, NOT screenshots
68
99
  * Screenshots would be 100K+ tokens each - element lists are ~100 tokens
69
100
  */
70
- function createBrowserTools() {
101
+ function createBrowserTools(options) {
102
+ const { waitForResult } = options || {};
71
103
  const tools = [];
72
- // browser_click
73
- tools.push(tool(async (args) => {
74
- // This returns a marker that tells the extension to execute locally
75
- const response = {
76
- requiresBrowserExecution: true,
77
- action: 'click',
78
- args,
104
+ /**
105
+ * Helper to create tool function that optionally waits for results
106
+ * The toolCallId is extracted from the RunnableConfig passed by LangChain
107
+ */
108
+ const createToolFunction = (action) => {
109
+ return async (args, config) => {
110
+ const toolCallId = config?.toolCall?.id || `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;
111
+ // Create marker for extension
112
+ const marker = {
113
+ requiresBrowserExecution: true,
114
+ action,
115
+ args,
116
+ toolCallId,
117
+ };
118
+ // If no callback, return marker immediately (extension handles via SSE interception)
119
+ if (!waitForResult) {
120
+ return JSON.stringify(marker);
121
+ }
122
+ // With callback: wait for actual results from extension
123
+ // The marker is still returned initially via SSE, but we wait for the callback
124
+ try {
125
+ const result = await waitForResult(action, args, toolCallId);
126
+ return formatResultForLLM(result, action);
127
+ }
128
+ catch (error) {
129
+ const errorMessage = error instanceof Error ? error.message : String(error);
130
+ return `Browser action "${action}" failed: ${errorMessage}`;
131
+ }
79
132
  };
80
- return JSON.stringify(response);
81
- }, {
133
+ };
134
+ // browser_click
135
+ tools.push(tool(createToolFunction('click'), {
82
136
  name: EBrowserTools.CLICK,
83
137
  description: `Click an element on the current web page by its index number.
84
138
  The element list shows clickable items like: [0]<button>Submit</button> [1]<a href="/home">Home</a>
@@ -87,14 +141,7 @@ After clicking, you receive an updated element list showing the new page state.`
87
141
  schema: BrowserClickSchema,
88
142
  }));
89
143
  // browser_type
90
- tools.push(tool(async (args) => {
91
- const response = {
92
- requiresBrowserExecution: true,
93
- action: 'type',
94
- args,
95
- };
96
- return JSON.stringify(response);
97
- }, {
144
+ tools.push(tool(createToolFunction('type'), {
98
145
  name: EBrowserTools.TYPE,
99
146
  description: `Type text into an input element on the page.
100
147
  Find the input element in the list by its index (e.g., [5]<input placeholder="Search">).
@@ -103,28 +150,14 @@ After typing, you receive an updated element list.`,
103
150
  schema: BrowserTypeSchema,
104
151
  }));
105
152
  // browser_navigate
106
- tools.push(tool(async (args) => {
107
- const response = {
108
- requiresBrowserExecution: true,
109
- action: 'navigate',
110
- args,
111
- };
112
- return JSON.stringify(response);
113
- }, {
153
+ tools.push(tool(createToolFunction('navigate'), {
114
154
  name: EBrowserTools.NAVIGATE,
115
155
  description: `Navigate to a URL. Always include the full URL with https://.
116
156
  After navigation, you receive the new page's element list.`,
117
157
  schema: BrowserNavigateSchema,
118
158
  }));
119
159
  // browser_scroll
120
- tools.push(tool(async (args) => {
121
- const response = {
122
- requiresBrowserExecution: true,
123
- action: 'scroll',
124
- args,
125
- };
126
- return JSON.stringify(response);
127
- }, {
160
+ tools.push(tool(createToolFunction('scroll'), {
128
161
  name: EBrowserTools.SCROLL,
129
162
  description: `Scroll the page to reveal more content.
130
163
  Use 'down' to scroll down, 'up' to scroll up.
@@ -132,42 +165,21 @@ After scrolling, you receive an updated element list with newly visible elements
132
165
  schema: BrowserScrollSchema,
133
166
  }));
134
167
  // browser_extract
135
- tools.push(tool(async (args) => {
136
- const response = {
137
- requiresBrowserExecution: true,
138
- action: 'extract',
139
- args,
140
- };
141
- return JSON.stringify(response);
142
- }, {
168
+ tools.push(tool(createToolFunction('extract'), {
143
169
  name: EBrowserTools.EXTRACT,
144
170
  description: `Extract content from the current page.
145
171
  Returns page URL, title, and element list.`,
146
172
  schema: BrowserExtractSchema,
147
173
  }));
148
174
  // browser_hover
149
- tools.push(tool(async (args) => {
150
- const response = {
151
- requiresBrowserExecution: true,
152
- action: 'hover',
153
- args,
154
- };
155
- return JSON.stringify(response);
156
- }, {
175
+ tools.push(tool(createToolFunction('hover'), {
157
176
  name: EBrowserTools.HOVER,
158
177
  description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.
159
178
  After hovering, you receive an updated element list with any newly revealed elements.`,
160
179
  schema: BrowserHoverSchema,
161
180
  }));
162
181
  // browser_wait
163
- tools.push(tool(async (args) => {
164
- const response = {
165
- requiresBrowserExecution: true,
166
- action: 'wait',
167
- args,
168
- };
169
- return JSON.stringify(response);
170
- }, {
182
+ tools.push(tool(createToolFunction('wait'), {
171
183
  name: EBrowserTools.WAIT,
172
184
  description: `Wait for a specified duration for page content to load.
173
185
  Use this after actions that trigger async content loading.
@@ -175,47 +187,26 @@ After waiting, you receive an updated element list.`,
175
187
  schema: BrowserWaitSchema,
176
188
  }));
177
189
  // browser_back
178
- tools.push(tool(async (args) => {
179
- const response = {
180
- requiresBrowserExecution: true,
181
- action: 'back',
182
- args,
183
- };
184
- return JSON.stringify(response);
185
- }, {
190
+ tools.push(tool(createToolFunction('back'), {
186
191
  name: EBrowserTools.BACK,
187
192
  description: `Go back to the previous page in browser history.
188
193
  After going back, you receive the previous page's element list.`,
189
194
  schema: BrowserBackSchema,
190
195
  }));
191
196
  // browser_screenshot
192
- tools.push(tool(async (args) => {
193
- const response = {
194
- requiresBrowserExecution: true,
195
- action: 'screenshot',
196
- args,
197
- };
198
- return JSON.stringify(response);
199
- }, {
197
+ tools.push(tool(createToolFunction('screenshot'), {
200
198
  name: EBrowserTools.SCREENSHOT,
201
199
  description: `Capture a screenshot of the current page.
202
- NOTE: Screenshot is displayed to the USER only, not returned to you.
203
- Use browser_get_page_state instead to get the element list.`,
200
+ Returns the page state with a note that screenshot was displayed to the user.
201
+ Use browser_get_page_state to get the element list for automation.`,
204
202
  schema: BrowserScreenshotSchema,
205
203
  }));
206
204
  // browser_get_page_state
207
- tools.push(tool(async (args) => {
208
- const response = {
209
- requiresBrowserExecution: true,
210
- action: 'get_page_state',
211
- args,
212
- };
213
- return JSON.stringify(response);
214
- }, {
205
+ tools.push(tool(createToolFunction('get_page_state'), {
215
206
  name: EBrowserTools.GET_PAGE_STATE,
216
207
  description: `Get the current page state including URL, title, and all interactive elements.
217
208
  Use this at the start of a task to see what elements are available.
218
- Returns a text list of elements with their index numbers.`,
209
+ Returns a text list of elements with their index numbers for interaction.`,
219
210
  schema: BrowserGetPageStateSchema,
220
211
  }));
221
212
  return tools;
@@ -1 +1 @@
1
- {"version":3,"file":"BrowserTools.mjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\r\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\r\nimport type * as t from '@/types';\r\n\r\n/**\r\n * Browser tool names - keep in sync with ranger-browser extension\r\n * These tools execute locally in the browser extension, NOT on the server\r\n */\r\nexport const EBrowserTools = {\r\n CLICK: 'browser_click',\r\n TYPE: 'browser_type',\r\n NAVIGATE: 'browser_navigate',\r\n SCROLL: 'browser_scroll',\r\n EXTRACT: 'browser_extract',\r\n HOVER: 'browser_hover',\r\n WAIT: 'browser_wait',\r\n BACK: 'browser_back',\r\n SCREENSHOT: 'browser_screenshot',\r\n GET_PAGE_STATE: 'browser_get_page_state',\r\n} as const;\r\n\r\nexport type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];\r\n\r\n/**\r\n * Check if browser capability is available based on request headers or context\r\n * The browser extension sets these headers when connected:\r\n * - X-Ranger-Browser-Extension: true\r\n * - X-Ranger-Browser-Capable: true\r\n */\r\nexport function hasBrowserCapability(req?: { headers?: Record<string, string | string[] | undefined> }): boolean {\r\n if (!req?.headers) {\r\n return false;\r\n }\r\n \r\n const browserExtension = req.headers['x-ranger-browser-extension'];\r\n const browserCapable = req.headers['x-ranger-browser-capable'];\r\n \r\n return browserExtension === 'true' || browserCapable === 'true';\r\n}\r\n\r\n// Tool schemas\r\nconst BrowserClickSchema = z.object({\r\n index: z.number().describe('The index number [0], [1], etc. of the element to click from the page state element list'),\r\n});\r\n\r\nconst BrowserTypeSchema = z.object({\r\n index: z.number().describe('The index number of the input element to type into'),\r\n text: z.string().describe('The text to type into the element'),\r\n pressEnter: z.boolean().optional().describe('Whether to press Enter after typing (useful for search forms)'),\r\n});\r\n\r\nconst BrowserNavigateSchema = z.object({\r\n url: z.string().describe('The full URL to navigate to (must include https://)'),\r\n});\r\n\r\nconst BrowserScrollSchema = z.object({\r\n direction: z.enum(['up', 'down', 'left', 'right']).describe('Direction to scroll'),\r\n amount: z.number().optional().describe('Pixels to scroll (default: one viewport height)'),\r\n});\r\n\r\nconst BrowserExtractSchema = z.object({\r\n query: z.string().optional().describe('Optional: specific content to extract from the page'),\r\n});\r\n\r\nconst BrowserHoverSchema = z.object({\r\n index: z.number().describe('The index number of the element to hover over'),\r\n});\r\n\r\nconst BrowserWaitSchema = z.object({\r\n duration: z.number().optional().describe('Milliseconds to wait (default: 1000)'),\r\n});\r\n\r\nconst BrowserBackSchema = z.object({});\r\n\r\nconst BrowserScreenshotSchema = z.object({});\r\n\r\nconst BrowserGetPageStateSchema = z.object({});\r\n\r\n/**\r\n * Browser tool response interface\r\n * This is what the extension returns after executing the action\r\n */\r\nexport interface BrowserToolResponse {\r\n requiresBrowserExecution: true;\r\n action: string;\r\n args: Record<string, unknown>;\r\n}\r\n\r\n/**\r\n * Create placeholder browser tools that signal to the extension to execute locally\r\n * The server creates these tools to give the LLM the schema, but actual execution\r\n * happens in the browser extension which intercepts tool calls\r\n * \r\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\r\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\r\n */\r\nexport function createBrowserTools(): DynamicStructuredTool[] {\r\n const tools: DynamicStructuredTool[] = [];\r\n\r\n // browser_click\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n // This returns a marker that tells the extension to execute locally\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'click',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.CLICK,\r\n description: `Click an element on the current web page by its index number.\r\nThe element list shows clickable items like: [0]<button>Submit</button> [1]<a href=\"/home\">Home</a>\r\nUse the index number in brackets to click that element.\r\nAfter clicking, you receive an updated element list showing the new page state.`,\r\n schema: BrowserClickSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_type\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'type',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.TYPE,\r\n description: `Type text into an input element on the page.\r\nFind the input element in the list by its index (e.g., [5]<input placeholder=\"Search\">).\r\nSet pressEnter: true to submit forms after typing.\r\nAfter typing, you receive an updated element list.`,\r\n schema: BrowserTypeSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_navigate\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'navigate',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.NAVIGATE,\r\n description: `Navigate to a URL. Always include the full URL with https://.\r\nAfter navigation, you receive the new page's element list.`,\r\n schema: BrowserNavigateSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_scroll\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'scroll',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.SCROLL,\r\n description: `Scroll the page to reveal more content.\r\nUse 'down' to scroll down, 'up' to scroll up.\r\nAfter scrolling, you receive an updated element list with newly visible elements.`,\r\n schema: BrowserScrollSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_extract\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'extract',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.EXTRACT,\r\n description: `Extract content from the current page.\r\nReturns page URL, title, and element list.`,\r\n schema: BrowserExtractSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_hover\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'hover',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.HOVER,\r\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\r\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\r\n schema: BrowserHoverSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_wait\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'wait',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.WAIT,\r\n description: `Wait for a specified duration for page content to load.\r\nUse this after actions that trigger async content loading.\r\nAfter waiting, you receive an updated element list.`,\r\n schema: BrowserWaitSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_back\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'back',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.BACK,\r\n description: `Go back to the previous page in browser history.\r\nAfter going back, you receive the previous page's element list.`,\r\n schema: BrowserBackSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_screenshot\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'screenshot',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.SCREENSHOT,\r\n description: `Capture a screenshot of the current page.\r\nNOTE: Screenshot is displayed to the USER only, not returned to you.\r\nUse browser_get_page_state instead to get the element list.`,\r\n schema: BrowserScreenshotSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_get_page_state\r\n tools.push(\r\n tool(\r\n async (args): Promise<string> => {\r\n const response: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action: 'get_page_state',\r\n args,\r\n };\r\n return JSON.stringify(response);\r\n },\r\n {\r\n name: EBrowserTools.GET_PAGE_STATE,\r\n description: `Get the current page state including URL, title, and all interactive elements.\r\nUse this at the start of a task to see what elements are available.\r\nReturns a text list of elements with their index numbers.`,\r\n schema: BrowserGetPageStateSchema,\r\n }\r\n )\r\n );\r\n\r\n return tools;\r\n}\r\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;;AAK1C;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAAiE,EAAA;AACpG,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;IAClC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,0FAA0F,CAAC;AACvH,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,oDAAoD,CAAC;IAChF,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,+DAA+D,CAAC;AAC7G,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;IACrC,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAChF,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,QAAQ,CAAC,qBAAqB,CAAC;AAClF,IAAA,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,iDAAiD,CAAC;AAC1F,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAC7F,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,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,sCAAsC,CAAC;AACjF,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;AAY9C;;;;;;;AAOG;SACa,kBAAkB,GAAA;IAChC,MAAM,KAAK,GAA4B,EAAE;;IAGzC,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;;AAE9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,OAAO;YACf,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;AAG2D,+EAAA,CAAA;AACxE,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;AAG8B,kDAAA,CAAA;AAC3C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,UAAU;YAClB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;AACsC,0DAAA,CAAA;AACnD,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,QAAQ;YAChB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAAA;;AAE6D,iFAAA,CAAA;AAC1E,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,SAAS;YACjB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACsB,0CAAA,CAAA;AACnC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,OAAO;YACf,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACiE,qFAAA,CAAA;AAC9E,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAE+B,mDAAA,CAAA;AAC5C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,MAAM;YACd,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC2C,+DAAA,CAAA;AACxD,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,YAAY;YACpB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAA;;AAEuC,2DAAA,CAAA;AACpD,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,OAAO,IAAI,KAAqB;AAC9B,QAAA,MAAM,QAAQ,GAAwB;AACpC,YAAA,wBAAwB,EAAE,IAAI;AAC9B,YAAA,MAAM,EAAE,gBAAgB;YACxB,IAAI;SACL;AACD,QAAA,OAAO,IAAI,CAAC,SAAS,CAAC,QAAQ,CAAC;AACjC,KAAC,EACD;QACE,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAA;;AAEqC,yDAAA,CAAA;AAClD,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CACF,CACF;AAED,IAAA,OAAO,KAAK;AACd;;;;"}
1
+ {"version":3,"file":"BrowserTools.mjs","sources":["../../../src/tools/BrowserTools.ts"],"sourcesContent":["import { z } from 'zod';\r\nimport { tool, DynamicStructuredTool } from '@langchain/core/tools';\r\nimport type * as t from '@/types';\r\n\r\n/**\r\n * Browser tool names - keep in sync with ranger-browser extension\r\n * These tools execute locally in the browser extension, NOT on the server\r\n */\r\nexport const EBrowserTools = {\r\n CLICK: 'browser_click',\r\n TYPE: 'browser_type',\r\n NAVIGATE: 'browser_navigate',\r\n SCROLL: 'browser_scroll',\r\n EXTRACT: 'browser_extract',\r\n HOVER: 'browser_hover',\r\n WAIT: 'browser_wait',\r\n BACK: 'browser_back',\r\n SCREENSHOT: 'browser_screenshot',\r\n GET_PAGE_STATE: 'browser_get_page_state',\r\n} as const;\r\n\r\nexport type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];\r\n\r\n/**\r\n * Callback function type for waiting on browser action results\r\n * This allows the server (Ranger) to provide a callback that waits for the extension\r\n * to POST results back to the server before returning to the LLM.\r\n * \r\n * @param action - The browser action (click, type, navigate, etc.)\r\n * @param args - Arguments for the action\r\n * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)\r\n * @returns Promise that resolves with the actual browser result (page state, etc.)\r\n */\r\nexport type BrowserToolCallback = (\r\n action: string,\r\n args: Record<string, unknown>,\r\n toolCallId: string\r\n) => Promise<BrowserActionResult>;\r\n\r\n/**\r\n * Result returned from browser action execution\r\n */\r\nexport interface BrowserActionResult {\r\n success: boolean;\r\n url?: string;\r\n title?: string;\r\n elementList?: string; // Text-based element list\r\n error?: string;\r\n screenshot?: string; // Base64 screenshot (if requested)\r\n}\r\n\r\n/**\r\n * Check if browser capability is available based on request headers or context\r\n * The browser extension sets these headers when connected:\r\n * - X-Ranger-Browser-Extension: true\r\n * - X-Ranger-Browser-Capable: true\r\n */\r\nexport function hasBrowserCapability(req?: { headers?: Record<string, string | string[] | undefined> }): boolean {\r\n if (!req?.headers) {\r\n return false;\r\n }\r\n \r\n const browserExtension = req.headers['x-ranger-browser-extension'];\r\n const browserCapable = req.headers['x-ranger-browser-capable'];\r\n \r\n return browserExtension === 'true' || browserCapable === 'true';\r\n}\r\n\r\n// Tool schemas\r\nconst BrowserClickSchema = z.object({\r\n index: z.number().describe('The index number [0], [1], etc. of the element to click from the page state element list'),\r\n});\r\n\r\nconst BrowserTypeSchema = z.object({\r\n index: z.number().describe('The index number of the input element to type into'),\r\n text: z.string().describe('The text to type into the element'),\r\n pressEnter: z.boolean().optional().describe('Whether to press Enter after typing (useful for search forms)'),\r\n});\r\n\r\nconst BrowserNavigateSchema = z.object({\r\n url: z.string().describe('The full URL to navigate to (must include https://)'),\r\n});\r\n\r\nconst BrowserScrollSchema = z.object({\r\n direction: z.enum(['up', 'down', 'left', 'right']).describe('Direction to scroll'),\r\n amount: z.number().optional().describe('Pixels to scroll (default: one viewport height)'),\r\n});\r\n\r\nconst BrowserExtractSchema = z.object({\r\n query: z.string().optional().describe('Optional: specific content to extract from the page'),\r\n});\r\n\r\nconst BrowserHoverSchema = z.object({\r\n index: z.number().describe('The index number of the element to hover over'),\r\n});\r\n\r\nconst BrowserWaitSchema = z.object({\r\n duration: z.number().optional().describe('Milliseconds to wait (default: 1000)'),\r\n});\r\n\r\nconst BrowserBackSchema = z.object({});\r\n\r\nconst BrowserScreenshotSchema = z.object({});\r\n\r\nconst BrowserGetPageStateSchema = z.object({});\r\n\r\n/**\r\n * Browser tool response interface\r\n * This is what the extension returns after executing the action\r\n */\r\nexport interface BrowserToolResponse {\r\n requiresBrowserExecution: true;\r\n action: string;\r\n args: Record<string, unknown>;\r\n toolCallId?: string; // Added to help extension correlate with callback\r\n}\r\n\r\n/**\r\n * Options for creating browser tools\r\n */\r\nexport interface CreateBrowserToolsOptions {\r\n /**\r\n * Optional callback that waits for browser action results.\r\n * When provided, tools will await this callback to get actual results from the extension.\r\n * When not provided, tools return markers immediately (for non-server contexts).\r\n */\r\n waitForResult?: BrowserToolCallback;\r\n}\r\n\r\n/**\r\n * Format browser action result for LLM consumption\r\n */\r\nfunction formatResultForLLM(result: BrowserActionResult, action: string): string {\r\n if (!result.success && result.error) {\r\n return `Browser action \"${action}\" failed: ${result.error}`;\r\n }\r\n\r\n const parts: string[] = [];\r\n \r\n if (result.url) {\r\n parts.push(`**Current URL:** ${result.url}`);\r\n }\r\n if (result.title) {\r\n parts.push(`**Page Title:** ${result.title}`);\r\n }\r\n if (result.elementList) {\r\n parts.push(`\\n**Interactive Elements:**\\n${result.elementList}`);\r\n }\r\n if (result.screenshot) {\r\n parts.push(`\\n[Screenshot captured and displayed to user]`);\r\n }\r\n \r\n if (parts.length === 0) {\r\n return `Browser action \"${action}\" completed successfully.`;\r\n }\r\n \r\n return parts.join('\\n');\r\n}\r\n\r\n/**\r\n * Create browser tools with optional callback for waiting on results\r\n * \r\n * When waitForResult callback is provided:\r\n * 1. Tool returns marker that triggers extension\r\n * 2. Tool then awaits callback to get actual results\r\n * 3. Returns real page state to LLM\r\n * \r\n * When no callback:\r\n * 1. Tool returns marker only (for non-server contexts)\r\n * \r\n * NOTE: These tools use TEXT-BASED element lists, NOT screenshots\r\n * Screenshots would be 100K+ tokens each - element lists are ~100 tokens\r\n */\r\nexport function createBrowserTools(options?: CreateBrowserToolsOptions): DynamicStructuredTool[] {\r\n const { waitForResult } = options || {};\r\n const tools: DynamicStructuredTool[] = [];\r\n\r\n /**\r\n * Helper to create tool function that optionally waits for results\r\n * The toolCallId is extracted from the RunnableConfig passed by LangChain\r\n */\r\n const createToolFunction = (action: string) => {\r\n return async (args: Record<string, unknown>, config?: { toolCall?: { id?: string } }): Promise<string> => {\r\n const toolCallId = config?.toolCall?.id || `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;\r\n \r\n // Create marker for extension\r\n const marker: BrowserToolResponse = {\r\n requiresBrowserExecution: true,\r\n action,\r\n args,\r\n toolCallId,\r\n };\r\n \r\n // If no callback, return marker immediately (extension handles via SSE interception)\r\n if (!waitForResult) {\r\n return JSON.stringify(marker);\r\n }\r\n \r\n // With callback: wait for actual results from extension\r\n // The marker is still returned initially via SSE, but we wait for the callback\r\n try {\r\n const result = await waitForResult(action, args, toolCallId);\r\n return formatResultForLLM(result, action);\r\n } catch (error) {\r\n const errorMessage = error instanceof Error ? error.message : String(error);\r\n return `Browser action \"${action}\" failed: ${errorMessage}`;\r\n }\r\n };\r\n };\r\n\r\n // browser_click\r\n tools.push(\r\n tool(\r\n createToolFunction('click'),\r\n {\r\n name: EBrowserTools.CLICK,\r\n description: `Click an element on the current web page by its index number.\r\nThe element list shows clickable items like: [0]<button>Submit</button> [1]<a href=\"/home\">Home</a>\r\nUse the index number in brackets to click that element.\r\nAfter clicking, you receive an updated element list showing the new page state.`,\r\n schema: BrowserClickSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_type\r\n tools.push(\r\n tool(\r\n createToolFunction('type'),\r\n {\r\n name: EBrowserTools.TYPE,\r\n description: `Type text into an input element on the page.\r\nFind the input element in the list by its index (e.g., [5]<input placeholder=\"Search\">).\r\nSet pressEnter: true to submit forms after typing.\r\nAfter typing, you receive an updated element list.`,\r\n schema: BrowserTypeSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_navigate\r\n tools.push(\r\n tool(\r\n createToolFunction('navigate'),\r\n {\r\n name: EBrowserTools.NAVIGATE,\r\n description: `Navigate to a URL. Always include the full URL with https://.\r\nAfter navigation, you receive the new page's element list.`,\r\n schema: BrowserNavigateSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_scroll\r\n tools.push(\r\n tool(\r\n createToolFunction('scroll'),\r\n {\r\n name: EBrowserTools.SCROLL,\r\n description: `Scroll the page to reveal more content.\r\nUse 'down' to scroll down, 'up' to scroll up.\r\nAfter scrolling, you receive an updated element list with newly visible elements.`,\r\n schema: BrowserScrollSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_extract\r\n tools.push(\r\n tool(\r\n createToolFunction('extract'),\r\n {\r\n name: EBrowserTools.EXTRACT,\r\n description: `Extract content from the current page.\r\nReturns page URL, title, and element list.`,\r\n schema: BrowserExtractSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_hover\r\n tools.push(\r\n tool(\r\n createToolFunction('hover'),\r\n {\r\n name: EBrowserTools.HOVER,\r\n description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.\r\nAfter hovering, you receive an updated element list with any newly revealed elements.`,\r\n schema: BrowserHoverSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_wait\r\n tools.push(\r\n tool(\r\n createToolFunction('wait'),\r\n {\r\n name: EBrowserTools.WAIT,\r\n description: `Wait for a specified duration for page content to load.\r\nUse this after actions that trigger async content loading.\r\nAfter waiting, you receive an updated element list.`,\r\n schema: BrowserWaitSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_back\r\n tools.push(\r\n tool(\r\n createToolFunction('back'),\r\n {\r\n name: EBrowserTools.BACK,\r\n description: `Go back to the previous page in browser history.\r\nAfter going back, you receive the previous page's element list.`,\r\n schema: BrowserBackSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_screenshot\r\n tools.push(\r\n tool(\r\n createToolFunction('screenshot'),\r\n {\r\n name: EBrowserTools.SCREENSHOT,\r\n description: `Capture a screenshot of the current page.\r\nReturns the page state with a note that screenshot was displayed to the user.\r\nUse browser_get_page_state to get the element list for automation.`,\r\n schema: BrowserScreenshotSchema,\r\n }\r\n )\r\n );\r\n\r\n // browser_get_page_state\r\n tools.push(\r\n tool(\r\n createToolFunction('get_page_state'),\r\n {\r\n name: EBrowserTools.GET_PAGE_STATE,\r\n description: `Get the current page state including URL, title, and all interactive elements.\r\nUse this at the start of a task to see what elements are available.\r\nReturns a text list of elements with their index numbers for interaction.`,\r\n schema: BrowserGetPageStateSchema,\r\n }\r\n )\r\n );\r\n\r\n return tools;\r\n}\r\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;;AAiC1C;;;;;AAKG;AACG,SAAU,oBAAoB,CAAC,GAAiE,EAAA;AACpG,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;IAClC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,0FAA0F,CAAC;AACvH,CAAA,CAAC;AAEF,MAAM,iBAAiB,GAAG,CAAC,CAAC,MAAM,CAAC;IACjC,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,oDAAoD,CAAC;IAChF,IAAI,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,mCAAmC,CAAC;AAC9D,IAAA,UAAU,EAAE,CAAC,CAAC,OAAO,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,+DAA+D,CAAC;AAC7G,CAAA,CAAC;AAEF,MAAM,qBAAqB,GAAG,CAAC,CAAC,MAAM,CAAC;IACrC,GAAG,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAChF,CAAA,CAAC;AAEF,MAAM,mBAAmB,GAAG,CAAC,CAAC,MAAM,CAAC;AACnC,IAAA,SAAS,EAAE,CAAC,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,QAAQ,CAAC,qBAAqB,CAAC;AAClF,IAAA,MAAM,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,iDAAiD,CAAC;AAC1F,CAAA,CAAC;AAEF,MAAM,oBAAoB,GAAG,CAAC,CAAC,MAAM,CAAC;AACpC,IAAA,KAAK,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,qDAAqD,CAAC;AAC7F,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,CAAC,CAAC,MAAM,EAAE,CAAC,QAAQ,EAAE,CAAC,QAAQ,CAAC,sCAAsC,CAAC;AACjF,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;AAyB9C;;AAEG;AACH,SAAS,kBAAkB,CAAC,MAA2B,EAAE,MAAc,EAAA;IACrE,IAAI,CAAC,MAAM,CAAC,OAAO,IAAI,MAAM,CAAC,KAAK,EAAE;AACnC,QAAA,OAAO,mBAAmB,MAAM,CAAA,UAAA,EAAa,MAAM,CAAC,KAAK,EAAE;;IAG7D,MAAM,KAAK,GAAa,EAAE;AAE1B,IAAA,IAAI,MAAM,CAAC,GAAG,EAAE;QACd,KAAK,CAAC,IAAI,CAAC,CAAA,iBAAA,EAAoB,MAAM,CAAC,GAAG,CAAE,CAAA,CAAC;;AAE9C,IAAA,IAAI,MAAM,CAAC,KAAK,EAAE;QAChB,KAAK,CAAC,IAAI,CAAC,CAAA,gBAAA,EAAmB,MAAM,CAAC,KAAK,CAAE,CAAA,CAAC;;AAE/C,IAAA,IAAI,MAAM,CAAC,WAAW,EAAE;QACtB,KAAK,CAAC,IAAI,CAAC,CAAA,6BAAA,EAAgC,MAAM,CAAC,WAAW,CAAE,CAAA,CAAC;;AAElE,IAAA,IAAI,MAAM,CAAC,UAAU,EAAE;AACrB,QAAA,KAAK,CAAC,IAAI,CAAC,CAAA,6CAAA,CAA+C,CAAC;;AAG7D,IAAA,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE;QACtB,OAAO,CAAA,gBAAA,EAAmB,MAAM,CAAA,yBAAA,CAA2B;;AAG7D,IAAA,OAAO,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC;AACzB;AAEA;;;;;;;;;;;;;AAaG;AACG,SAAU,kBAAkB,CAAC,OAAmC,EAAA;AACpE,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,OAAO,IAA6B,EAAE,MAAuC,KAAqB;AACvG,YAAA,MAAM,UAAU,GAAG,MAAM,EAAE,QAAQ,EAAE,EAAE,IAAI,CAAQ,KAAA,EAAA,IAAI,CAAC,GAAG,EAAE,CAAA,CAAA,EAAI,IAAI,CAAC,MAAM,EAAE,CAAC,QAAQ,CAAC,EAAE,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE;;AAGtG,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;AAC5D,gBAAA,OAAO,kBAAkB,CAAC,MAAM,EAAE,MAAM,CAAC;;YACzC,OAAO,KAAK,EAAE;AACd,gBAAA,MAAM,YAAY,GAAG,KAAK,YAAY,KAAK,GAAG,KAAK,CAAC,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC;AAC3E,gBAAA,OAAO,CAAmB,gBAAA,EAAA,MAAM,CAAa,UAAA,EAAA,YAAY,EAAE;;AAE/D,SAAC;AACH,KAAC;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,OAAO,CAAC,EAC3B;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;;;AAG2D,+EAAA,CAAA;AACxE,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;;AAG8B,kDAAA,CAAA;AAC3C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,UAAU,CAAC,EAC9B;QACE,IAAI,EAAE,aAAa,CAAC,QAAQ;AAC5B,QAAA,WAAW,EAAE,CAAA;AACsC,0DAAA,CAAA;AACnD,QAAA,MAAM,EAAE,qBAAqB;AAC9B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,QAAQ,CAAC,EAC5B;QACE,IAAI,EAAE,aAAa,CAAC,MAAM;AAC1B,QAAA,WAAW,EAAE,CAAA;;AAE6D,iFAAA,CAAA;AAC1E,QAAA,MAAM,EAAE,mBAAmB;AAC5B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,SAAS,CAAC,EAC7B;QACE,IAAI,EAAE,aAAa,CAAC,OAAO;AAC3B,QAAA,WAAW,EAAE,CAAA;AACsB,0CAAA,CAAA;AACnC,QAAA,MAAM,EAAE,oBAAoB;AAC7B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,OAAO,CAAC,EAC3B;QACE,IAAI,EAAE,aAAa,CAAC,KAAK;AACzB,QAAA,WAAW,EAAE,CAAA;AACiE,qFAAA,CAAA;AAC9E,QAAA,MAAM,EAAE,kBAAkB;AAC3B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;;AAE+B,mDAAA,CAAA;AAC5C,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,MAAM,CAAC,EAC1B;QACE,IAAI,EAAE,aAAa,CAAC,IAAI;AACxB,QAAA,WAAW,EAAE,CAAA;AAC2C,+DAAA,CAAA;AACxD,QAAA,MAAM,EAAE,iBAAiB;AAC1B,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,YAAY,CAAC,EAChC;QACE,IAAI,EAAE,aAAa,CAAC,UAAU;AAC9B,QAAA,WAAW,EAAE,CAAA;;AAE8C,kEAAA,CAAA;AAC3D,QAAA,MAAM,EAAE,uBAAuB;AAChC,KAAA,CACF,CACF;;IAGD,KAAK,CAAC,IAAI,CACR,IAAI,CACF,kBAAkB,CAAC,gBAAgB,CAAC,EACpC;QACE,IAAI,EAAE,aAAa,CAAC,cAAc;AAClC,QAAA,WAAW,EAAE,CAAA;;AAEqD,yEAAA,CAAA;AAClE,QAAA,MAAM,EAAE,yBAAyB;AAClC,KAAA,CACF,CACF;AAED,IAAA,OAAO,KAAK;AACd;;;;"}
@@ -16,6 +16,28 @@ export declare const EBrowserTools: {
16
16
  readonly GET_PAGE_STATE: "browser_get_page_state";
17
17
  };
18
18
  export type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];
19
+ /**
20
+ * Callback function type for waiting on browser action results
21
+ * This allows the server (Ranger) to provide a callback that waits for the extension
22
+ * to POST results back to the server before returning to the LLM.
23
+ *
24
+ * @param action - The browser action (click, type, navigate, etc.)
25
+ * @param args - Arguments for the action
26
+ * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)
27
+ * @returns Promise that resolves with the actual browser result (page state, etc.)
28
+ */
29
+ export type BrowserToolCallback = (action: string, args: Record<string, unknown>, toolCallId: string) => Promise<BrowserActionResult>;
30
+ /**
31
+ * Result returned from browser action execution
32
+ */
33
+ export interface BrowserActionResult {
34
+ success: boolean;
35
+ url?: string;
36
+ title?: string;
37
+ elementList?: string;
38
+ error?: string;
39
+ screenshot?: string;
40
+ }
19
41
  /**
20
42
  * Check if browser capability is available based on request headers or context
21
43
  * The browser extension sets these headers when connected:
@@ -33,13 +55,31 @@ export interface BrowserToolResponse {
33
55
  requiresBrowserExecution: true;
34
56
  action: string;
35
57
  args: Record<string, unknown>;
58
+ toolCallId?: string;
36
59
  }
37
60
  /**
38
- * Create placeholder browser tools that signal to the extension to execute locally
39
- * The server creates these tools to give the LLM the schema, but actual execution
40
- * happens in the browser extension which intercepts tool calls
61
+ * Options for creating browser tools
62
+ */
63
+ export interface CreateBrowserToolsOptions {
64
+ /**
65
+ * Optional callback that waits for browser action results.
66
+ * When provided, tools will await this callback to get actual results from the extension.
67
+ * When not provided, tools return markers immediately (for non-server contexts).
68
+ */
69
+ waitForResult?: BrowserToolCallback;
70
+ }
71
+ /**
72
+ * Create browser tools with optional callback for waiting on results
73
+ *
74
+ * When waitForResult callback is provided:
75
+ * 1. Tool returns marker that triggers extension
76
+ * 2. Tool then awaits callback to get actual results
77
+ * 3. Returns real page state to LLM
78
+ *
79
+ * When no callback:
80
+ * 1. Tool returns marker only (for non-server contexts)
41
81
  *
42
82
  * NOTE: These tools use TEXT-BASED element lists, NOT screenshots
43
83
  * Screenshots would be 100K+ tokens each - element lists are ~100 tokens
44
84
  */
45
- export declare function createBrowserTools(): DynamicStructuredTool[];
85
+ export declare function createBrowserTools(options?: CreateBrowserToolsOptions): DynamicStructuredTool[];
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "illuma-agents",
3
- "version": "1.0.26",
3
+ "version": "1.0.27",
4
4
  "main": "./dist/cjs/main.cjs",
5
5
  "module": "./dist/esm/main.mjs",
6
6
  "types": "./dist/types/index.d.ts",
@@ -21,6 +21,34 @@ export const EBrowserTools = {
21
21
 
22
22
  export type BrowserToolName = typeof EBrowserTools[keyof typeof EBrowserTools];
23
23
 
24
+ /**
25
+ * Callback function type for waiting on browser action results
26
+ * This allows the server (Ranger) to provide a callback that waits for the extension
27
+ * to POST results back to the server before returning to the LLM.
28
+ *
29
+ * @param action - The browser action (click, type, navigate, etc.)
30
+ * @param args - Arguments for the action
31
+ * @param toolCallId - Unique ID for this tool call (from config.toolCall.id)
32
+ * @returns Promise that resolves with the actual browser result (page state, etc.)
33
+ */
34
+ export type BrowserToolCallback = (
35
+ action: string,
36
+ args: Record<string, unknown>,
37
+ toolCallId: string
38
+ ) => Promise<BrowserActionResult>;
39
+
40
+ /**
41
+ * Result returned from browser action execution
42
+ */
43
+ export interface BrowserActionResult {
44
+ success: boolean;
45
+ url?: string;
46
+ title?: string;
47
+ elementList?: string; // Text-based element list
48
+ error?: string;
49
+ screenshot?: string; // Base64 screenshot (if requested)
50
+ }
51
+
24
52
  /**
25
53
  * Check if browser capability is available based on request headers or context
26
54
  * The browser extension sets these headers when connected:
@@ -84,31 +112,106 @@ export interface BrowserToolResponse {
84
112
  requiresBrowserExecution: true;
85
113
  action: string;
86
114
  args: Record<string, unknown>;
115
+ toolCallId?: string; // Added to help extension correlate with callback
87
116
  }
88
117
 
89
118
  /**
90
- * Create placeholder browser tools that signal to the extension to execute locally
91
- * The server creates these tools to give the LLM the schema, but actual execution
92
- * happens in the browser extension which intercepts tool calls
119
+ * Options for creating browser tools
120
+ */
121
+ export interface CreateBrowserToolsOptions {
122
+ /**
123
+ * Optional callback that waits for browser action results.
124
+ * When provided, tools will await this callback to get actual results from the extension.
125
+ * When not provided, tools return markers immediately (for non-server contexts).
126
+ */
127
+ waitForResult?: BrowserToolCallback;
128
+ }
129
+
130
+ /**
131
+ * Format browser action result for LLM consumption
132
+ */
133
+ function formatResultForLLM(result: BrowserActionResult, action: string): string {
134
+ if (!result.success && result.error) {
135
+ return `Browser action "${action}" failed: ${result.error}`;
136
+ }
137
+
138
+ const parts: string[] = [];
139
+
140
+ if (result.url) {
141
+ parts.push(`**Current URL:** ${result.url}`);
142
+ }
143
+ if (result.title) {
144
+ parts.push(`**Page Title:** ${result.title}`);
145
+ }
146
+ if (result.elementList) {
147
+ parts.push(`\n**Interactive Elements:**\n${result.elementList}`);
148
+ }
149
+ if (result.screenshot) {
150
+ parts.push(`\n[Screenshot captured and displayed to user]`);
151
+ }
152
+
153
+ if (parts.length === 0) {
154
+ return `Browser action "${action}" completed successfully.`;
155
+ }
156
+
157
+ return parts.join('\n');
158
+ }
159
+
160
+ /**
161
+ * Create browser tools with optional callback for waiting on results
162
+ *
163
+ * When waitForResult callback is provided:
164
+ * 1. Tool returns marker that triggers extension
165
+ * 2. Tool then awaits callback to get actual results
166
+ * 3. Returns real page state to LLM
167
+ *
168
+ * When no callback:
169
+ * 1. Tool returns marker only (for non-server contexts)
93
170
  *
94
171
  * NOTE: These tools use TEXT-BASED element lists, NOT screenshots
95
172
  * Screenshots would be 100K+ tokens each - element lists are ~100 tokens
96
173
  */
97
- export function createBrowserTools(): DynamicStructuredTool[] {
174
+ export function createBrowserTools(options?: CreateBrowserToolsOptions): DynamicStructuredTool[] {
175
+ const { waitForResult } = options || {};
98
176
  const tools: DynamicStructuredTool[] = [];
99
177
 
178
+ /**
179
+ * Helper to create tool function that optionally waits for results
180
+ * The toolCallId is extracted from the RunnableConfig passed by LangChain
181
+ */
182
+ const createToolFunction = (action: string) => {
183
+ return async (args: Record<string, unknown>, config?: { toolCall?: { id?: string } }): Promise<string> => {
184
+ const toolCallId = config?.toolCall?.id || `tool_${Date.now()}_${Math.random().toString(36).slice(2)}`;
185
+
186
+ // Create marker for extension
187
+ const marker: BrowserToolResponse = {
188
+ requiresBrowserExecution: true,
189
+ action,
190
+ args,
191
+ toolCallId,
192
+ };
193
+
194
+ // If no callback, return marker immediately (extension handles via SSE interception)
195
+ if (!waitForResult) {
196
+ return JSON.stringify(marker);
197
+ }
198
+
199
+ // With callback: wait for actual results from extension
200
+ // The marker is still returned initially via SSE, but we wait for the callback
201
+ try {
202
+ const result = await waitForResult(action, args, toolCallId);
203
+ return formatResultForLLM(result, action);
204
+ } catch (error) {
205
+ const errorMessage = error instanceof Error ? error.message : String(error);
206
+ return `Browser action "${action}" failed: ${errorMessage}`;
207
+ }
208
+ };
209
+ };
210
+
100
211
  // browser_click
101
212
  tools.push(
102
213
  tool(
103
- async (args): Promise<string> => {
104
- // This returns a marker that tells the extension to execute locally
105
- const response: BrowserToolResponse = {
106
- requiresBrowserExecution: true,
107
- action: 'click',
108
- args,
109
- };
110
- return JSON.stringify(response);
111
- },
214
+ createToolFunction('click'),
112
215
  {
113
216
  name: EBrowserTools.CLICK,
114
217
  description: `Click an element on the current web page by its index number.
@@ -123,14 +226,7 @@ After clicking, you receive an updated element list showing the new page state.`
123
226
  // browser_type
124
227
  tools.push(
125
228
  tool(
126
- async (args): Promise<string> => {
127
- const response: BrowserToolResponse = {
128
- requiresBrowserExecution: true,
129
- action: 'type',
130
- args,
131
- };
132
- return JSON.stringify(response);
133
- },
229
+ createToolFunction('type'),
134
230
  {
135
231
  name: EBrowserTools.TYPE,
136
232
  description: `Type text into an input element on the page.
@@ -145,14 +241,7 @@ After typing, you receive an updated element list.`,
145
241
  // browser_navigate
146
242
  tools.push(
147
243
  tool(
148
- async (args): Promise<string> => {
149
- const response: BrowserToolResponse = {
150
- requiresBrowserExecution: true,
151
- action: 'navigate',
152
- args,
153
- };
154
- return JSON.stringify(response);
155
- },
244
+ createToolFunction('navigate'),
156
245
  {
157
246
  name: EBrowserTools.NAVIGATE,
158
247
  description: `Navigate to a URL. Always include the full URL with https://.
@@ -165,14 +254,7 @@ After navigation, you receive the new page's element list.`,
165
254
  // browser_scroll
166
255
  tools.push(
167
256
  tool(
168
- async (args): Promise<string> => {
169
- const response: BrowserToolResponse = {
170
- requiresBrowserExecution: true,
171
- action: 'scroll',
172
- args,
173
- };
174
- return JSON.stringify(response);
175
- },
257
+ createToolFunction('scroll'),
176
258
  {
177
259
  name: EBrowserTools.SCROLL,
178
260
  description: `Scroll the page to reveal more content.
@@ -186,14 +268,7 @@ After scrolling, you receive an updated element list with newly visible elements
186
268
  // browser_extract
187
269
  tools.push(
188
270
  tool(
189
- async (args): Promise<string> => {
190
- const response: BrowserToolResponse = {
191
- requiresBrowserExecution: true,
192
- action: 'extract',
193
- args,
194
- };
195
- return JSON.stringify(response);
196
- },
271
+ createToolFunction('extract'),
197
272
  {
198
273
  name: EBrowserTools.EXTRACT,
199
274
  description: `Extract content from the current page.
@@ -206,14 +281,7 @@ Returns page URL, title, and element list.`,
206
281
  // browser_hover
207
282
  tools.push(
208
283
  tool(
209
- async (args): Promise<string> => {
210
- const response: BrowserToolResponse = {
211
- requiresBrowserExecution: true,
212
- action: 'hover',
213
- args,
214
- };
215
- return JSON.stringify(response);
216
- },
284
+ createToolFunction('hover'),
217
285
  {
218
286
  name: EBrowserTools.HOVER,
219
287
  description: `Hover over an element to reveal tooltips, dropdowns, or other hover-triggered content.
@@ -226,14 +294,7 @@ After hovering, you receive an updated element list with any newly revealed elem
226
294
  // browser_wait
227
295
  tools.push(
228
296
  tool(
229
- async (args): Promise<string> => {
230
- const response: BrowserToolResponse = {
231
- requiresBrowserExecution: true,
232
- action: 'wait',
233
- args,
234
- };
235
- return JSON.stringify(response);
236
- },
297
+ createToolFunction('wait'),
237
298
  {
238
299
  name: EBrowserTools.WAIT,
239
300
  description: `Wait for a specified duration for page content to load.
@@ -247,14 +308,7 @@ After waiting, you receive an updated element list.`,
247
308
  // browser_back
248
309
  tools.push(
249
310
  tool(
250
- async (args): Promise<string> => {
251
- const response: BrowserToolResponse = {
252
- requiresBrowserExecution: true,
253
- action: 'back',
254
- args,
255
- };
256
- return JSON.stringify(response);
257
- },
311
+ createToolFunction('back'),
258
312
  {
259
313
  name: EBrowserTools.BACK,
260
314
  description: `Go back to the previous page in browser history.
@@ -267,19 +321,12 @@ After going back, you receive the previous page's element list.`,
267
321
  // browser_screenshot
268
322
  tools.push(
269
323
  tool(
270
- async (args): Promise<string> => {
271
- const response: BrowserToolResponse = {
272
- requiresBrowserExecution: true,
273
- action: 'screenshot',
274
- args,
275
- };
276
- return JSON.stringify(response);
277
- },
324
+ createToolFunction('screenshot'),
278
325
  {
279
326
  name: EBrowserTools.SCREENSHOT,
280
327
  description: `Capture a screenshot of the current page.
281
- NOTE: Screenshot is displayed to the USER only, not returned to you.
282
- Use browser_get_page_state instead to get the element list.`,
328
+ Returns the page state with a note that screenshot was displayed to the user.
329
+ Use browser_get_page_state to get the element list for automation.`,
283
330
  schema: BrowserScreenshotSchema,
284
331
  }
285
332
  )
@@ -288,19 +335,12 @@ Use browser_get_page_state instead to get the element list.`,
288
335
  // browser_get_page_state
289
336
  tools.push(
290
337
  tool(
291
- async (args): Promise<string> => {
292
- const response: BrowserToolResponse = {
293
- requiresBrowserExecution: true,
294
- action: 'get_page_state',
295
- args,
296
- };
297
- return JSON.stringify(response);
298
- },
338
+ createToolFunction('get_page_state'),
299
339
  {
300
340
  name: EBrowserTools.GET_PAGE_STATE,
301
341
  description: `Get the current page state including URL, title, and all interactive elements.
302
342
  Use this at the start of a task to see what elements are available.
303
- Returns a text list of elements with their index numbers.`,
343
+ Returns a text list of elements with their index numbers for interaction.`,
304
344
  schema: BrowserGetPageStateSchema,
305
345
  }
306
346
  )