@monoes/monobrowse 1.0.0

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.
Files changed (208) hide show
  1. package/dist/src/browser/action-builder/analyzer.d.ts +11 -0
  2. package/dist/src/browser/action-builder/analyzer.d.ts.map +1 -0
  3. package/dist/src/browser/action-builder/analyzer.js +93 -0
  4. package/dist/src/browser/action-builder/analyzer.js.map +1 -0
  5. package/dist/src/browser/action-builder/types.d.ts +46 -0
  6. package/dist/src/browser/action-builder/types.d.ts.map +1 -0
  7. package/dist/src/browser/action-builder/types.js +2 -0
  8. package/dist/src/browser/action-builder/types.js.map +1 -0
  9. package/dist/src/browser/actions.d.ts +29 -0
  10. package/dist/src/browser/actions.d.ts.map +1 -0
  11. package/dist/src/browser/actions.js +398 -0
  12. package/dist/src/browser/actions.js.map +1 -0
  13. package/dist/src/browser/adapters/gemini.d.ts +3 -0
  14. package/dist/src/browser/adapters/gemini.d.ts.map +1 -0
  15. package/dist/src/browser/adapters/gemini.js +13 -0
  16. package/dist/src/browser/adapters/gemini.js.map +1 -0
  17. package/dist/src/browser/adapters/google.d.ts +3 -0
  18. package/dist/src/browser/adapters/google.d.ts.map +1 -0
  19. package/dist/src/browser/adapters/google.js +17 -0
  20. package/dist/src/browser/adapters/google.js.map +1 -0
  21. package/dist/src/browser/adapters/index.d.ts +15 -0
  22. package/dist/src/browser/adapters/index.d.ts.map +1 -0
  23. package/dist/src/browser/adapters/index.js +18 -0
  24. package/dist/src/browser/adapters/index.js.map +1 -0
  25. package/dist/src/browser/adapters/instagram.d.ts +3 -0
  26. package/dist/src/browser/adapters/instagram.d.ts.map +1 -0
  27. package/dist/src/browser/adapters/instagram.js +14 -0
  28. package/dist/src/browser/adapters/instagram.js.map +1 -0
  29. package/dist/src/browser/adapters/linkedin.d.ts +3 -0
  30. package/dist/src/browser/adapters/linkedin.d.ts.map +1 -0
  31. package/dist/src/browser/adapters/linkedin.js +13 -0
  32. package/dist/src/browser/adapters/linkedin.js.map +1 -0
  33. package/dist/src/browser/adapters/microsoft.d.ts +3 -0
  34. package/dist/src/browser/adapters/microsoft.d.ts.map +1 -0
  35. package/dist/src/browser/adapters/microsoft.js +16 -0
  36. package/dist/src/browser/adapters/microsoft.js.map +1 -0
  37. package/dist/src/browser/adapters/x.d.ts +3 -0
  38. package/dist/src/browser/adapters/x.d.ts.map +1 -0
  39. package/dist/src/browser/adapters/x.js +13 -0
  40. package/dist/src/browser/adapters/x.js.map +1 -0
  41. package/dist/src/browser/batch.d.ts +5 -0
  42. package/dist/src/browser/batch.d.ts.map +1 -0
  43. package/dist/src/browser/batch.js +2 -0
  44. package/dist/src/browser/batch.js.map +1 -0
  45. package/dist/src/browser/browser.d.ts +15 -0
  46. package/dist/src/browser/browser.d.ts.map +1 -0
  47. package/dist/src/browser/browser.js +256 -0
  48. package/dist/src/browser/browser.js.map +1 -0
  49. package/dist/src/browser/cdp.d.ts +18 -0
  50. package/dist/src/browser/cdp.d.ts.map +1 -0
  51. package/dist/src/browser/cdp.js +153 -0
  52. package/dist/src/browser/cdp.js.map +1 -0
  53. package/dist/src/browser/console-log.d.ts +23 -0
  54. package/dist/src/browser/console-log.d.ts.map +1 -0
  55. package/dist/src/browser/console-log.js +104 -0
  56. package/dist/src/browser/console-log.js.map +1 -0
  57. package/dist/src/browser/dashboard/api-types.d.ts +50 -0
  58. package/dist/src/browser/dashboard/api-types.d.ts.map +1 -0
  59. package/dist/src/browser/dashboard/api-types.js +14 -0
  60. package/dist/src/browser/dashboard/api-types.js.map +1 -0
  61. package/dist/src/browser/dashboard/server.d.ts +11 -0
  62. package/dist/src/browser/dashboard/server.d.ts.map +1 -0
  63. package/dist/src/browser/dashboard/server.js +191 -0
  64. package/dist/src/browser/dashboard/server.js.map +1 -0
  65. package/dist/src/browser/dialog.d.ts +12 -0
  66. package/dist/src/browser/dialog.d.ts.map +1 -0
  67. package/dist/src/browser/dialog.js +53 -0
  68. package/dist/src/browser/dialog.js.map +1 -0
  69. package/dist/src/browser/emulation.d.ts +15 -0
  70. package/dist/src/browser/emulation.d.ts.map +1 -0
  71. package/dist/src/browser/emulation.js +62 -0
  72. package/dist/src/browser/emulation.js.map +1 -0
  73. package/dist/src/browser/find.d.ts +22 -0
  74. package/dist/src/browser/find.d.ts.map +1 -0
  75. package/dist/src/browser/find.js +169 -0
  76. package/dist/src/browser/find.js.map +1 -0
  77. package/dist/src/browser/har.d.ts +27 -0
  78. package/dist/src/browser/har.d.ts.map +1 -0
  79. package/dist/src/browser/har.js +161 -0
  80. package/dist/src/browser/har.js.map +1 -0
  81. package/dist/src/browser/index.d.ts +23 -0
  82. package/dist/src/browser/index.d.ts.map +1 -0
  83. package/dist/src/browser/index.js +23 -0
  84. package/dist/src/browser/index.js.map +1 -0
  85. package/dist/src/browser/network.d.ts +32 -0
  86. package/dist/src/browser/network.d.ts.map +1 -0
  87. package/dist/src/browser/network.js +214 -0
  88. package/dist/src/browser/network.js.map +1 -0
  89. package/dist/src/browser/pdf.d.ts +15 -0
  90. package/dist/src/browser/pdf.d.ts.map +1 -0
  91. package/dist/src/browser/pdf.js +27 -0
  92. package/dist/src/browser/pdf.js.map +1 -0
  93. package/dist/src/browser/profiler.d.ts +10 -0
  94. package/dist/src/browser/profiler.d.ts.map +1 -0
  95. package/dist/src/browser/profiler.js +77 -0
  96. package/dist/src/browser/profiler.js.map +1 -0
  97. package/dist/src/browser/record.d.ts +21 -0
  98. package/dist/src/browser/record.d.ts.map +1 -0
  99. package/dist/src/browser/record.js +59 -0
  100. package/dist/src/browser/record.js.map +1 -0
  101. package/dist/src/browser/screenshot.d.ts +15 -0
  102. package/dist/src/browser/screenshot.d.ts.map +1 -0
  103. package/dist/src/browser/screenshot.js +38 -0
  104. package/dist/src/browser/screenshot.js.map +1 -0
  105. package/dist/src/browser/session.d.ts +8 -0
  106. package/dist/src/browser/session.d.ts.map +1 -0
  107. package/dist/src/browser/session.js +91 -0
  108. package/dist/src/browser/session.js.map +1 -0
  109. package/dist/src/browser/snapshot.d.ts +12 -0
  110. package/dist/src/browser/snapshot.d.ts.map +1 -0
  111. package/dist/src/browser/snapshot.js +179 -0
  112. package/dist/src/browser/snapshot.js.map +1 -0
  113. package/dist/src/browser/storage.d.ts +12 -0
  114. package/dist/src/browser/storage.d.ts.map +1 -0
  115. package/dist/src/browser/storage.js +46 -0
  116. package/dist/src/browser/storage.js.map +1 -0
  117. package/dist/src/browser/tabs.d.ts +8 -0
  118. package/dist/src/browser/tabs.d.ts.map +1 -0
  119. package/dist/src/browser/tabs.js +25 -0
  120. package/dist/src/browser/tabs.js.map +1 -0
  121. package/dist/src/browser/trace.d.ts +10 -0
  122. package/dist/src/browser/trace.d.ts.map +1 -0
  123. package/dist/src/browser/trace.js +87 -0
  124. package/dist/src/browser/trace.js.map +1 -0
  125. package/dist/src/browser/types.d.ts +111 -0
  126. package/dist/src/browser/types.d.ts.map +1 -0
  127. package/dist/src/browser/types.js +16 -0
  128. package/dist/src/browser/types.js.map +1 -0
  129. package/dist/src/browser/vitals.d.ts +15 -0
  130. package/dist/src/browser/vitals.d.ts.map +1 -0
  131. package/dist/src/browser/vitals.js +116 -0
  132. package/dist/src/browser/vitals.js.map +1 -0
  133. package/dist/src/browser/wait.d.ts +4 -0
  134. package/dist/src/browser/wait.d.ts.map +1 -0
  135. package/dist/src/browser/wait.js +146 -0
  136. package/dist/src/browser/wait.js.map +1 -0
  137. package/dist/src/browser/workflow/builtin-handlers.d.ts +3 -0
  138. package/dist/src/browser/workflow/builtin-handlers.d.ts.map +1 -0
  139. package/dist/src/browser/workflow/builtin-handlers.js +343 -0
  140. package/dist/src/browser/workflow/builtin-handlers.js.map +1 -0
  141. package/dist/src/browser/workflow/engine.d.ts +10 -0
  142. package/dist/src/browser/workflow/engine.d.ts.map +1 -0
  143. package/dist/src/browser/workflow/engine.js +174 -0
  144. package/dist/src/browser/workflow/engine.js.map +1 -0
  145. package/dist/src/browser/workflow/expression.d.ts +4 -0
  146. package/dist/src/browser/workflow/expression.d.ts.map +1 -0
  147. package/dist/src/browser/workflow/expression.js +89 -0
  148. package/dist/src/browser/workflow/expression.js.map +1 -0
  149. package/dist/src/browser/workflow/index.d.ts +6 -0
  150. package/dist/src/browser/workflow/index.d.ts.map +1 -0
  151. package/dist/src/browser/workflow/index.js +6 -0
  152. package/dist/src/browser/workflow/index.js.map +1 -0
  153. package/dist/src/browser/workflow/store.d.ts +8 -0
  154. package/dist/src/browser/workflow/store.d.ts.map +1 -0
  155. package/dist/src/browser/workflow/store.js +87 -0
  156. package/dist/src/browser/workflow/store.js.map +1 -0
  157. package/dist/src/browser/workflow/types.d.ts +62 -0
  158. package/dist/src/browser/workflow/types.d.ts.map +1 -0
  159. package/dist/src/browser/workflow/types.js +2 -0
  160. package/dist/src/browser/workflow/types.js.map +1 -0
  161. package/dist/src/index.d.ts +7 -0
  162. package/dist/src/index.d.ts.map +1 -0
  163. package/dist/src/index.js +7 -0
  164. package/dist/src/index.js.map +1 -0
  165. package/dist/tsconfig.tsbuildinfo +1 -0
  166. package/package.json +56 -0
  167. package/src/browser/action-builder/analyzer.ts +127 -0
  168. package/src/browser/action-builder/types.ts +17 -0
  169. package/src/browser/actions.ts +518 -0
  170. package/src/browser/adapters/gemini.ts +15 -0
  171. package/src/browser/adapters/google.ts +20 -0
  172. package/src/browser/adapters/index.ts +33 -0
  173. package/src/browser/adapters/instagram.ts +16 -0
  174. package/src/browser/adapters/linkedin.ts +15 -0
  175. package/src/browser/adapters/microsoft.ts +19 -0
  176. package/src/browser/adapters/x.ts +15 -0
  177. package/src/browser/batch.ts +4 -0
  178. package/src/browser/browser.ts +265 -0
  179. package/src/browser/cdp.ts +156 -0
  180. package/src/browser/console-log.ts +120 -0
  181. package/src/browser/dashboard/api-types.ts +71 -0
  182. package/src/browser/dashboard/server.ts +206 -0
  183. package/src/browser/dashboard/ui.html +444 -0
  184. package/src/browser/dialog.ts +63 -0
  185. package/src/browser/emulation.ts +94 -0
  186. package/src/browser/find.ts +225 -0
  187. package/src/browser/har.ts +202 -0
  188. package/src/browser/index.ts +22 -0
  189. package/src/browser/network.ts +244 -0
  190. package/src/browser/pdf.ts +41 -0
  191. package/src/browser/profiler.ts +96 -0
  192. package/src/browser/record.ts +92 -0
  193. package/src/browser/screenshot.ts +67 -0
  194. package/src/browser/session.ts +118 -0
  195. package/src/browser/snapshot.ts +248 -0
  196. package/src/browser/storage.ts +54 -0
  197. package/src/browser/tabs.ts +37 -0
  198. package/src/browser/trace.ts +106 -0
  199. package/src/browser/types.ts +135 -0
  200. package/src/browser/vitals.ts +122 -0
  201. package/src/browser/wait.ts +178 -0
  202. package/src/browser/workflow/builtin-handlers.ts +379 -0
  203. package/src/browser/workflow/engine.ts +202 -0
  204. package/src/browser/workflow/expression.ts +107 -0
  205. package/src/browser/workflow/index.ts +5 -0
  206. package/src/browser/workflow/store.ts +87 -0
  207. package/src/browser/workflow/types.ts +62 -0
  208. package/src/index.ts +6 -0
@@ -0,0 +1,127 @@
1
+ // src/browser/action-builder/analyzer.ts
2
+ import Anthropic from '@anthropic-ai/sdk';
3
+ import type { ActionDef } from './types.js';
4
+ const MODEL_DEFAULTS = { sonnet: 'claude-sonnet-4-6' } as const;
5
+
6
+ export interface AnalyzerPage {
7
+ evaluate<T>(expression: string): Promise<T>;
8
+ url(): Promise<string>;
9
+ }
10
+
11
+ export interface AnalyzerOptions {
12
+ apiKey?: string;
13
+ model?: string;
14
+ }
15
+
16
+ interface DomElement {
17
+ tag: string;
18
+ text: string;
19
+ attrs: {
20
+ id?: string;
21
+ name?: string | null;
22
+ ariaLabel?: string | null;
23
+ type?: string | null;
24
+ href?: string | null;
25
+ };
26
+ }
27
+
28
+ const DOM_CAPTURE_EXPR = `
29
+ JSON.stringify(
30
+ Array.from(document.querySelectorAll('button,input,a,textarea,[role="button"]'))
31
+ .slice(0, 100)
32
+ .map(el => ({
33
+ tag: el.tagName.toLowerCase(),
34
+ text: el.textContent?.trim().slice(0, 80) ?? '',
35
+ attrs: {
36
+ id: el.id || undefined,
37
+ name: el.getAttribute('name') || undefined,
38
+ ariaLabel: el.getAttribute('aria-label') || undefined,
39
+ type: el.getAttribute('type') || undefined,
40
+ href: el.tagName === 'A' ? (el.getAttribute('href') || undefined) : undefined
41
+ }
42
+ }))
43
+ )`;
44
+
45
+ const SYSTEM_PROMPT = `You are an expert browser automation engineer. Given a webpage's interactive elements and a task description, generate a browser automation ActionDef JSON.
46
+
47
+ Rules:
48
+ - Output ONLY valid JSON matching the ActionDef schema, no markdown, no explanation
49
+ - Use multi-selector find steps: provide CSS selector, aria-label selector, and XPath as fallbacks
50
+ - Use {{params.param_name}} for dynamic values the caller will provide
51
+ - Steps must be executable in order
52
+
53
+ ActionDef schema:
54
+ {
55
+ "id": "<platform>:<action_name>",
56
+ "platform": "<platform>",
57
+ "name": "<human readable>",
58
+ "params": ["param1", "param2"],
59
+ "steps": [
60
+ { "type": "navigate", "url": "{{params.url}}" },
61
+ { "type": "find", "selectors": ["css-selector", "[aria-label='...']", "//xpath"], "as": "element_name" },
62
+ { "type": "click", "target": "{{element_name}}" },
63
+ { "type": "type", "target": "{{element_name}}", "text": "{{params.text}}", "humanDelay": true },
64
+ { "type": "wait", "condition": "network_idle", "timeout": 3000 },
65
+ { "type": "extract", "target": "{{element_name}}", "as": "result" },
66
+ { "type": "condition", "expression": "{{extracted_value}}", "then": [...steps], "else": [...steps] }
67
+ ]
68
+ }`;
69
+
70
+ export async function analyzePageForAction(
71
+ page: AnalyzerPage,
72
+ task: string,
73
+ options: AnalyzerOptions = {},
74
+ ): Promise<ActionDef> {
75
+ const url = await page.url();
76
+ const title = await page.evaluate<string>('document.title');
77
+ const elementsJson = await page.evaluate<string>(DOM_CAPTURE_EXPR);
78
+
79
+ let elements: DomElement[] = [];
80
+ try {
81
+ elements = JSON.parse(elementsJson) as DomElement[];
82
+ } catch {
83
+ elements = [];
84
+ }
85
+
86
+ const domContext = `URL: ${url}
87
+ Title: ${title}
88
+ Interactive elements (${elements.length}):
89
+ ${elements.map((el, i) => {
90
+ const attrs = Object.entries(el.attrs)
91
+ .filter(([, v]) => v)
92
+ .map(([k, v]) => `${k}="${v}"`)
93
+ .join(' ');
94
+ return `${i + 1}. <${el.tag}${attrs ? ' ' + attrs : ''}>${el.text}</${el.tag}>`;
95
+ }).join('\n')}`;
96
+
97
+ const client = new Anthropic({ apiKey: options.apiKey ?? process.env['ANTHROPIC_API_KEY'] });
98
+ const model = options.model ?? MODEL_DEFAULTS.sonnet;
99
+
100
+ const message = await client.messages.create({
101
+ model,
102
+ max_tokens: 2048,
103
+ system: SYSTEM_PROMPT,
104
+ messages: [
105
+ {
106
+ role: 'user',
107
+ content: `Task: ${task}\n\nPage context:\n${domContext}`,
108
+ },
109
+ ],
110
+ });
111
+
112
+ const content = message.content[0];
113
+ if (content.type !== 'text') throw new Error('Claude returned non-text response');
114
+
115
+ let actionDef: ActionDef;
116
+ try {
117
+ actionDef = JSON.parse(content.text) as ActionDef;
118
+ } catch {
119
+ throw new Error(`Claude returned invalid JSON:\n${content.text.slice(0, 500)}`);
120
+ }
121
+
122
+ if (!actionDef.id || !actionDef.steps || !Array.isArray(actionDef.steps)) {
123
+ throw new Error(`Claude returned invalid ActionDef: missing id or steps\n${content.text.slice(0, 500)}`);
124
+ }
125
+
126
+ return actionDef;
127
+ }
@@ -0,0 +1,17 @@
1
+ export interface NavigateStep { type: 'navigate'; url: string; }
2
+ export interface FindStep { type: 'find'; selectors: string[]; as: string; }
3
+ export interface ClickStep { type: 'click'; target: string; }
4
+ export interface TypeStep { type: 'type'; target: string; text: string; humanDelay?: boolean; }
5
+ export interface WaitStep { type: 'wait'; condition: 'network_idle' | 'selector' | 'duration'; timeout?: number; selector?: string; }
6
+ export interface ExtractStep { type: 'extract'; target: string; as: string; attribute?: string; }
7
+ export interface ConditionStep { type: 'condition'; expression: string; then: StepDef[]; else?: StepDef[]; }
8
+
9
+ export type StepDef = NavigateStep | FindStep | ClickStep | TypeStep | WaitStep | ExtractStep | ConditionStep;
10
+
11
+ export interface ActionDef {
12
+ id: string;
13
+ platform: string;
14
+ name: string;
15
+ params: string[];
16
+ steps: StepDef[];
17
+ }
@@ -0,0 +1,518 @@
1
+ import { resolve } from 'path';
2
+ import type { CdpClient } from './cdp.js';
3
+ import type { ElementRef, ClickOptions } from './types.js';
4
+ import { getObjectIdForRef, getElementBox } from './snapshot.js';
5
+
6
+ export async function clickElement(
7
+ client: CdpClient,
8
+ sessionId: string,
9
+ ref: ElementRef,
10
+ options: ClickOptions = {}
11
+ ): Promise<void> {
12
+ const box = await getElementBox(client, sessionId, ref);
13
+
14
+ if (box) {
15
+ await clickPoint(client, sessionId, box.x, box.y, options);
16
+ return;
17
+ }
18
+
19
+ // Fallback: use JS click via objectId
20
+ const objectId = await getObjectIdForRef(client, sessionId, ref);
21
+ if (objectId) {
22
+ await client.send('Runtime.callFunctionOn', {
23
+ functionDeclaration: 'function() { this.click(); }',
24
+ objectId,
25
+ returnByValue: true,
26
+ }, sessionId);
27
+ return;
28
+ }
29
+
30
+ throw new Error(`Cannot click ref @${ref.ref}: element not found in DOM`);
31
+ }
32
+
33
+ export async function clickPoint(
34
+ client: CdpClient,
35
+ sessionId: string,
36
+ x: number,
37
+ y: number,
38
+ options: ClickOptions = {}
39
+ ): Promise<void> {
40
+ const button = options.button ?? 'left';
41
+ // Cap clickCount to prevent unbounded loop DoS
42
+ const clickCount = Math.min(Math.max(1, Math.floor(options.clickCount ?? 1)), 100);
43
+ const modifiers = options.modifiers ?? 0;
44
+
45
+ const shared = { x, y, button, modifiers };
46
+ const buttonsMask = button === 'right' ? 2 : button === 'middle' ? 4 : 1;
47
+
48
+ for (let i = 0; i < clickCount; i++) {
49
+ const count = i + 1;
50
+ await client.send('Input.dispatchMouseEvent', { ...shared, type: 'mousePressed', buttons: buttonsMask, clickCount: count }, sessionId);
51
+ await client.send('Input.dispatchMouseEvent', { ...shared, type: 'mouseReleased', buttons: i < clickCount - 1 ? buttonsMask : 0, clickCount: count }, sessionId);
52
+ }
53
+ }
54
+
55
+ export async function fillElement(
56
+ client: CdpClient,
57
+ sessionId: string,
58
+ ref: ElementRef,
59
+ value: string
60
+ ): Promise<void> {
61
+ const box = await getElementBox(client, sessionId, ref);
62
+
63
+ if (box) {
64
+ // Click to focus
65
+ await clickPoint(client, sessionId, box.x, box.y);
66
+ }
67
+
68
+ // Select all and replace
69
+ const objectId = await getObjectIdForRef(client, sessionId, ref);
70
+ if (objectId) {
71
+ const fillSelectResult = await client.send<{ result: unknown; exceptionDetails?: { text: string; exception?: { description?: string } } }>('Runtime.callFunctionOn', {
72
+ functionDeclaration: `function() {
73
+ this.focus();
74
+ if (this.tagName === 'INPUT' || this.tagName === 'TEXTAREA') {
75
+ this.select();
76
+ } else if (this.isContentEditable) {
77
+ const range = document.createRange();
78
+ range.selectNodeContents(this);
79
+ const sel = window.getSelection();
80
+ if (sel) { sel.removeAllRanges(); sel.addRange(range); }
81
+ }
82
+ }`,
83
+ objectId,
84
+ returnByValue: true,
85
+ }, sessionId);
86
+ if (fillSelectResult.exceptionDetails) {
87
+ throw new Error(`fillElement select-all failed: ${fillSelectResult.exceptionDetails.exception?.description ?? fillSelectResult.exceptionDetails.text}`);
88
+ }
89
+ } else if (box) {
90
+ // Fallback for elements without a resolvable objectId: keyboard select-all clears existing content
91
+ const mod = process.platform === 'darwin' ? 4 : 2;
92
+ await pressKeyCombo(client, sessionId, 'a', mod);
93
+ }
94
+
95
+ if (!objectId && !box) throw new Error(`Cannot fill ref @${ref.ref}: element not found in DOM`);
96
+
97
+ // Type the value character by character for natural input
98
+ await typeText(client, sessionId, value);
99
+ }
100
+
101
+ export async function typeText(client: CdpClient, sessionId: string, text: string): Promise<void> {
102
+ // Cap to 100 KB to prevent OOM in CDP message serializer
103
+ const safeText = text.length > 102_400 ? text.slice(0, 102_400) : text;
104
+ await client.send('Input.insertText', { text: safeText }, sessionId);
105
+ }
106
+
107
+ export async function pressKeyCombo(
108
+ client: CdpClient,
109
+ sessionId: string,
110
+ key: string,
111
+ modifiers: number
112
+ ): Promise<void> {
113
+ const { text: _text, ...keyInfo } = resolveKey(key);
114
+ // rawKeyDown prevents Chrome from inserting the character text; only the shortcut fires
115
+ await client.send('Input.dispatchKeyEvent', { type: 'rawKeyDown', ...keyInfo, modifiers }, sessionId);
116
+ await client.send('Input.dispatchKeyEvent', { type: 'keyUp', ...keyInfo, modifiers }, sessionId);
117
+ }
118
+
119
+ export async function pressKey(
120
+ client: CdpClient,
121
+ sessionId: string,
122
+ key: string
123
+ ): Promise<void> {
124
+ const { text, ...keyInfo } = resolveKey(key);
125
+
126
+ // rawKeyDown does not insert text; the explicit char event handles insertion
127
+ await client.send('Input.dispatchKeyEvent', {
128
+ type: 'rawKeyDown',
129
+ ...keyInfo,
130
+ }, sessionId);
131
+
132
+ if (text) {
133
+ await client.send('Input.dispatchKeyEvent', {
134
+ type: 'char',
135
+ text,
136
+ }, sessionId);
137
+ }
138
+
139
+ await client.send('Input.dispatchKeyEvent', {
140
+ type: 'keyUp',
141
+ ...keyInfo,
142
+ }, sessionId);
143
+ }
144
+
145
+ function resolveKey(key: string): { key: string; code: string; text?: string; windowsVirtualKeyCode?: number } {
146
+ const keyMap: Record<string, { key: string; code: string; text?: string; windowsVirtualKeyCode?: number }> = {
147
+ Enter: { key: 'Enter', code: 'Enter', text: '\r', windowsVirtualKeyCode: 13 },
148
+ Tab: { key: 'Tab', code: 'Tab', windowsVirtualKeyCode: 9 },
149
+ Escape: { key: 'Escape', code: 'Escape', windowsVirtualKeyCode: 27 },
150
+ Backspace: { key: 'Backspace', code: 'Backspace', windowsVirtualKeyCode: 8 },
151
+ Delete: { key: 'Delete', code: 'Delete', windowsVirtualKeyCode: 46 },
152
+ ArrowUp: { key: 'ArrowUp', code: 'ArrowUp', windowsVirtualKeyCode: 38 },
153
+ ArrowDown: { key: 'ArrowDown', code: 'ArrowDown', windowsVirtualKeyCode: 40 },
154
+ ArrowLeft: { key: 'ArrowLeft', code: 'ArrowLeft', windowsVirtualKeyCode: 37 },
155
+ ArrowRight: { key: 'ArrowRight', code: 'ArrowRight', windowsVirtualKeyCode: 39 },
156
+ Home: { key: 'Home', code: 'Home', windowsVirtualKeyCode: 36 },
157
+ End: { key: 'End', code: 'End', windowsVirtualKeyCode: 35 },
158
+ PageUp: { key: 'PageUp', code: 'PageUp', windowsVirtualKeyCode: 33 },
159
+ PageDown: { key: 'PageDown', code: 'PageDown', windowsVirtualKeyCode: 34 },
160
+ Space: { key: ' ', code: 'Space', text: ' ', windowsVirtualKeyCode: 32 },
161
+ ' ': { key: ' ', code: 'Space', text: ' ', windowsVirtualKeyCode: 32 },
162
+ Shift: { key: 'Shift', code: 'ShiftLeft', windowsVirtualKeyCode: 16 },
163
+ Control: { key: 'Control', code: 'ControlLeft', windowsVirtualKeyCode: 17 },
164
+ Alt: { key: 'Alt', code: 'AltLeft', windowsVirtualKeyCode: 18 },
165
+ Meta: { key: 'Meta', code: 'MetaLeft', windowsVirtualKeyCode: 91 },
166
+ F1: { key: 'F1', code: 'F1', windowsVirtualKeyCode: 112 },
167
+ F2: { key: 'F2', code: 'F2', windowsVirtualKeyCode: 113 },
168
+ F3: { key: 'F3', code: 'F3', windowsVirtualKeyCode: 114 },
169
+ F4: { key: 'F4', code: 'F4', windowsVirtualKeyCode: 115 },
170
+ F5: { key: 'F5', code: 'F5', windowsVirtualKeyCode: 116 },
171
+ F6: { key: 'F6', code: 'F6', windowsVirtualKeyCode: 117 },
172
+ F7: { key: 'F7', code: 'F7', windowsVirtualKeyCode: 118 },
173
+ F8: { key: 'F8', code: 'F8', windowsVirtualKeyCode: 119 },
174
+ F9: { key: 'F9', code: 'F9', windowsVirtualKeyCode: 120 },
175
+ F10: { key: 'F10', code: 'F10', windowsVirtualKeyCode: 121 },
176
+ F11: { key: 'F11', code: 'F11', windowsVirtualKeyCode: 122 },
177
+ F12: { key: 'F12', code: 'F12', windowsVirtualKeyCode: 123 },
178
+ };
179
+
180
+ if (keyMap[key]) return keyMap[key];
181
+
182
+ // Single printable character — windowsVirtualKeyCode required for rawKeyDown shortcut dispatch
183
+ if (key.length === 1) {
184
+ const charCode = key.charCodeAt(0);
185
+ if (charCode >= 48 && charCode <= 57) {
186
+ return { key, code: `Digit${key}`, text: key, windowsVirtualKeyCode: charCode };
187
+ }
188
+ if ((charCode >= 65 && charCode <= 90) || (charCode >= 97 && charCode <= 122)) {
189
+ return { key, code: `Key${key.toUpperCase()}`, text: key, windowsVirtualKeyCode: key.toUpperCase().charCodeAt(0) };
190
+ }
191
+ // Symbol keys: map to the physical key's code and Windows virtual key code
192
+ const symbolMap: Record<string, { code: string; windowsVirtualKeyCode: number }> = {
193
+ '!': { code: 'Digit1', windowsVirtualKeyCode: 49 },
194
+ '@': { code: 'Digit2', windowsVirtualKeyCode: 50 },
195
+ '#': { code: 'Digit3', windowsVirtualKeyCode: 51 },
196
+ '$': { code: 'Digit4', windowsVirtualKeyCode: 52 },
197
+ '%': { code: 'Digit5', windowsVirtualKeyCode: 53 },
198
+ '^': { code: 'Digit6', windowsVirtualKeyCode: 54 },
199
+ '&': { code: 'Digit7', windowsVirtualKeyCode: 55 },
200
+ '*': { code: 'Digit8', windowsVirtualKeyCode: 56 },
201
+ '(': { code: 'Digit9', windowsVirtualKeyCode: 57 },
202
+ ')': { code: 'Digit0', windowsVirtualKeyCode: 48 },
203
+ '-': { code: 'Minus', windowsVirtualKeyCode: 189 },
204
+ '_': { code: 'Minus', windowsVirtualKeyCode: 189 },
205
+ '=': { code: 'Equal', windowsVirtualKeyCode: 187 },
206
+ '+': { code: 'Equal', windowsVirtualKeyCode: 187 },
207
+ '[': { code: 'BracketLeft', windowsVirtualKeyCode: 219 },
208
+ '{': { code: 'BracketLeft', windowsVirtualKeyCode: 219 },
209
+ ']': { code: 'BracketRight', windowsVirtualKeyCode: 221 },
210
+ '}': { code: 'BracketRight', windowsVirtualKeyCode: 221 },
211
+ '\\': { code: 'Backslash', windowsVirtualKeyCode: 220 },
212
+ '|': { code: 'Backslash', windowsVirtualKeyCode: 220 },
213
+ ';': { code: 'Semicolon', windowsVirtualKeyCode: 186 },
214
+ ':': { code: 'Semicolon', windowsVirtualKeyCode: 186 },
215
+ "'": { code: 'Quote', windowsVirtualKeyCode: 222 },
216
+ '"': { code: 'Quote', windowsVirtualKeyCode: 222 },
217
+ '`': { code: 'Backquote', windowsVirtualKeyCode: 192 },
218
+ '~': { code: 'Backquote', windowsVirtualKeyCode: 192 },
219
+ ',': { code: 'Comma', windowsVirtualKeyCode: 188 },
220
+ '<': { code: 'Comma', windowsVirtualKeyCode: 188 },
221
+ '.': { code: 'Period', windowsVirtualKeyCode: 190 },
222
+ '>': { code: 'Period', windowsVirtualKeyCode: 190 },
223
+ '/': { code: 'Slash', windowsVirtualKeyCode: 191 },
224
+ '?': { code: 'Slash', windowsVirtualKeyCode: 191 },
225
+ };
226
+ const sym = symbolMap[key];
227
+ if (sym) return { key, code: sym.code, text: key, windowsVirtualKeyCode: sym.windowsVirtualKeyCode };
228
+ // Unknown/non-ASCII character — omit code since no valid DOM KeyboardEvent.code exists
229
+ return { key, code: '', text: key };
230
+ }
231
+
232
+ return { key, code: key };
233
+ }
234
+
235
+ export async function scrollElement(
236
+ client: CdpClient,
237
+ sessionId: string,
238
+ direction: 'up' | 'down' | 'left' | 'right',
239
+ amount = 300,
240
+ ref?: ElementRef
241
+ ): Promise<void> {
242
+ // Cap scroll amount to prevent extreme delta values
243
+ amount = Math.min(Math.max(1, Math.floor(amount)), 100_000);
244
+ let x = 0;
245
+ let y = 0;
246
+ let deltaX = 0;
247
+ let deltaY = 0;
248
+
249
+ if (ref) {
250
+ const box = await getElementBox(client, sessionId, ref);
251
+ if (!box) throw new Error(`Cannot scroll: ref @${ref.ref} not found in DOM`);
252
+ x = box.x; y = box.y;
253
+ } else {
254
+ // Center of viewport
255
+ const vp = await client.send<{ result: { value: string } }>('Runtime.evaluate', {
256
+ expression: 'JSON.stringify({w: window.innerWidth, h: window.innerHeight})',
257
+ returnByValue: true,
258
+ }, sessionId);
259
+ const dims = JSON.parse(vp.result?.value ?? '{"w":1280,"h":720}');
260
+ x = dims.w / 2;
261
+ y = dims.h / 2;
262
+ }
263
+
264
+ switch (direction) {
265
+ case 'down': deltaY = amount; break;
266
+ case 'up': deltaY = -amount; break;
267
+ case 'right': deltaX = amount; break;
268
+ case 'left': deltaX = -amount; break;
269
+ }
270
+
271
+ await client.send('Input.dispatchMouseEvent', {
272
+ type: 'mouseWheel',
273
+ x,
274
+ y,
275
+ deltaX,
276
+ deltaY,
277
+ }, sessionId);
278
+ }
279
+
280
+ export async function hoverElement(
281
+ client: CdpClient,
282
+ sessionId: string,
283
+ ref: ElementRef
284
+ ): Promise<void> {
285
+ const box = await getElementBox(client, sessionId, ref);
286
+ if (!box) throw new Error(`Cannot hover ref @${ref.ref}: element not found in DOM`);
287
+
288
+ await client.send('Input.dispatchMouseEvent', {
289
+ type: 'mouseMoved',
290
+ x: box.x,
291
+ y: box.y,
292
+ }, sessionId);
293
+ }
294
+
295
+ export async function selectOption(
296
+ client: CdpClient,
297
+ sessionId: string,
298
+ ref: ElementRef,
299
+ value: string
300
+ ): Promise<void> {
301
+ const objectId = await getObjectIdForRef(client, sessionId, ref);
302
+ if (!objectId) throw new Error(`Cannot select: ref @${ref.ref} not found in DOM`);
303
+
304
+ const selectResult = await client.send<{ result: unknown; exceptionDetails?: { text: string; exception?: { description?: string } } }>('Runtime.callFunctionOn', {
305
+ functionDeclaration: `function(value) {
306
+ if (this.tagName !== 'SELECT') throw new Error('Not a select element');
307
+ for (const opt of this.options) {
308
+ if (opt.value === value || opt.textContent.trim() === value) {
309
+ this.value = opt.value;
310
+ this.dispatchEvent(new Event('change', { bubbles: true }));
311
+ return;
312
+ }
313
+ }
314
+ throw new Error('Option not found: ' + value);
315
+ }`,
316
+ objectId,
317
+ arguments: [{ value }],
318
+ returnByValue: true,
319
+ }, sessionId);
320
+ if (selectResult.exceptionDetails) {
321
+ throw new Error(`selectOption failed: ${selectResult.exceptionDetails.exception?.description ?? selectResult.exceptionDetails.text}`);
322
+ }
323
+ }
324
+
325
+ export async function checkElement(
326
+ client: CdpClient,
327
+ sessionId: string,
328
+ ref: ElementRef,
329
+ checked = true
330
+ ): Promise<void> {
331
+ const objectId = await getObjectIdForRef(client, sessionId, ref);
332
+ if (!objectId) throw new Error(`Cannot check: ref @${ref.ref} not found in DOM`);
333
+
334
+ const checkResult = await client.send<{ result: unknown; exceptionDetails?: { text: string; exception?: { description?: string } } }>('Runtime.callFunctionOn', {
335
+ functionDeclaration: `function(checked) {
336
+ if (this.checked !== checked) {
337
+ this.click();
338
+ }
339
+ }`,
340
+ objectId,
341
+ arguments: [{ value: checked }],
342
+ returnByValue: true,
343
+ }, sessionId);
344
+ if (checkResult.exceptionDetails) {
345
+ throw new Error(`checkElement failed: ${checkResult.exceptionDetails.exception?.description ?? checkResult.exceptionDetails.text}`);
346
+ }
347
+ }
348
+
349
+ export async function focusElement(client: CdpClient, sessionId: string, ref: ElementRef): Promise<void> {
350
+ const objectId = await getObjectIdForRef(client, sessionId, ref);
351
+ if (objectId) {
352
+ await client.send('Runtime.callFunctionOn', {
353
+ functionDeclaration: 'function() { this.focus(); }',
354
+ objectId,
355
+ returnByValue: true,
356
+ }, sessionId);
357
+ return;
358
+ }
359
+ const box = await getElementBox(client, sessionId, ref);
360
+ if (box) {
361
+ await clickPoint(client, sessionId, box.x, box.y);
362
+ } else {
363
+ throw new Error(`Cannot focus ref @${ref.ref}: element not found in DOM`);
364
+ }
365
+ }
366
+
367
+ export async function typeIntoElement(
368
+ client: CdpClient,
369
+ sessionId: string,
370
+ ref: ElementRef,
371
+ text: string
372
+ ): Promise<void> {
373
+ await focusElement(client, sessionId, ref);
374
+ await typeText(client, sessionId, text);
375
+ }
376
+
377
+ export async function keyDown(client: CdpClient, sessionId: string, key: string): Promise<void> {
378
+ const { text: _text, ...keyInfo } = resolveKey(key);
379
+ await client.send('Input.dispatchKeyEvent', { type: 'rawKeyDown', ...keyInfo }, sessionId);
380
+ }
381
+
382
+ export async function keyUp(client: CdpClient, sessionId: string, key: string): Promise<void> {
383
+ const { text: _text, ...keyInfo } = resolveKey(key);
384
+ await client.send('Input.dispatchKeyEvent', { type: 'keyUp', ...keyInfo }, sessionId);
385
+ }
386
+
387
+ export async function mouseMove(client: CdpClient, sessionId: string, x: number, y: number): Promise<void> {
388
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseMoved', x, y }, sessionId);
389
+ }
390
+
391
+ export async function mouseDown(
392
+ client: CdpClient,
393
+ sessionId: string,
394
+ x: number,
395
+ y: number,
396
+ button: 'left' | 'right' | 'middle' = 'left'
397
+ ): Promise<void> {
398
+ const buttonsMask = button === 'right' ? 2 : button === 'middle' ? 4 : 1;
399
+ await client.send('Input.dispatchMouseEvent', { type: 'mousePressed', x, y, button, buttons: buttonsMask, clickCount: 1 }, sessionId);
400
+ }
401
+
402
+ export async function mouseUp(
403
+ client: CdpClient,
404
+ sessionId: string,
405
+ x: number,
406
+ y: number,
407
+ button: 'left' | 'right' | 'middle' = 'left'
408
+ ): Promise<void> {
409
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseReleased', x, y, button, buttons: 0, clickCount: 1 }, sessionId);
410
+ }
411
+
412
+ export async function mouseWheel(
413
+ client: CdpClient,
414
+ sessionId: string,
415
+ x: number,
416
+ y: number,
417
+ deltaY: number,
418
+ deltaX = 0
419
+ ): Promise<void> {
420
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseWheel', x, y, deltaX, deltaY }, sessionId);
421
+ }
422
+
423
+ export async function dragAndDrop(
424
+ client: CdpClient,
425
+ sessionId: string,
426
+ src: ElementRef,
427
+ tgt: ElementRef
428
+ ): Promise<void> {
429
+ const srcBox = await getElementBox(client, sessionId, src);
430
+ const tgtBox = await getElementBox(client, sessionId, tgt);
431
+ if (!srcBox || !tgtBox) throw new Error('Cannot drag: one or both elements not found in DOM');
432
+
433
+ await client.send('Input.dispatchMouseEvent', { type: 'mousePressed', x: srcBox.x, y: srcBox.y, button: 'left', buttons: 1, clickCount: 1 }, sessionId);
434
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseMoved', x: srcBox.x, y: srcBox.y, button: 'left', buttons: 1 }, sessionId);
435
+
436
+ // Move in steps for smooth drag
437
+ const steps = 10;
438
+ for (let i = 1; i <= steps; i++) {
439
+ const x = srcBox.x + (tgtBox.x - srcBox.x) * (i / steps);
440
+ const y = srcBox.y + (tgtBox.y - srcBox.y) * (i / steps);
441
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseMoved', x, y, button: 'left', buttons: 1 }, sessionId);
442
+ }
443
+
444
+ await client.send('Input.dispatchMouseEvent', { type: 'mouseReleased', x: tgtBox.x, y: tgtBox.y, button: 'left', buttons: 0, clickCount: 1 }, sessionId);
445
+ }
446
+
447
+ export async function uploadFile(
448
+ client: CdpClient,
449
+ sessionId: string,
450
+ ref: ElementRef,
451
+ filePaths: string[]
452
+ ): Promise<void> {
453
+ if (!ref.backendDOMNodeId) throw new Error(`Cannot upload: ref @${ref.ref} has no DOM node`);
454
+
455
+ await client.send('DOM.setFileInputFiles', {
456
+ files: filePaths.map((f) => resolve(f)),
457
+ backendNodeId: ref.backendDOMNodeId,
458
+ }, sessionId);
459
+ }
460
+
461
+ export async function readClipboard(client: CdpClient, sessionId: string): Promise<string> {
462
+ const result = await evaluateJs(client, sessionId, 'navigator.clipboard.readText()');
463
+ return result as string;
464
+ }
465
+
466
+ export async function writeClipboard(client: CdpClient, sessionId: string, text: string): Promise<void> {
467
+ // Cap to 100 KB to prevent OOM when serializing the CDP expression
468
+ const safeText = text.length > 102_400 ? text.slice(0, 102_400) : text;
469
+ await evaluateJs(client, sessionId, `navigator.clipboard.writeText(${JSON.stringify(safeText)})`);
470
+ }
471
+
472
+ export async function pushState(client: CdpClient, sessionId: string, url: string): Promise<void> {
473
+ // Try Next.js router first, then fallback to history.pushState
474
+ await evaluateJs(client, sessionId, `
475
+ (function() {
476
+ const url = ${JSON.stringify(url)};
477
+ if (window.next && window.next.router) {
478
+ window.next.router.push(url).catch(function() {});
479
+ return;
480
+ } else {
481
+ history.pushState({}, '', url);
482
+ window.dispatchEvent(new PopStateEvent('popstate'));
483
+ }
484
+ })()
485
+ `);
486
+ }
487
+
488
+ export async function addInitScript(client: CdpClient, sessionId: string, script: string): Promise<string> {
489
+ const result = await client.send<{ identifier: string }>('Page.addScriptToEvaluateOnNewDocument', {
490
+ source: script,
491
+ }, sessionId);
492
+ return result.identifier;
493
+ }
494
+
495
+ export async function removeInitScript(client: CdpClient, sessionId: string, identifier: string): Promise<void> {
496
+ await client.send('Page.removeScriptToEvaluateOnNewDocument', { identifier }, sessionId);
497
+ }
498
+
499
+ export async function evaluateJs(
500
+ client: CdpClient,
501
+ sessionId: string,
502
+ expression: string
503
+ ): Promise<unknown> {
504
+ const result = await client.send<{
505
+ result: { value?: unknown; type: string; description?: string };
506
+ exceptionDetails?: { text: string; exception?: { description?: string } };
507
+ }>('Runtime.evaluate', {
508
+ expression,
509
+ returnByValue: true,
510
+ awaitPromise: true,
511
+ }, sessionId);
512
+
513
+ if (result.exceptionDetails) {
514
+ throw new Error(`JS evaluation error: ${result.exceptionDetails.exception?.description ?? result.exceptionDetails.text}`);
515
+ }
516
+
517
+ return result.result?.value;
518
+ }
@@ -0,0 +1,15 @@
1
+ // src/browser/adapters/gemini.ts
2
+ import type { PlatformAdapter, PageInterface } from './index.js';
3
+
4
+ export const geminiAdapter: PlatformAdapter = {
5
+ platform: 'gemini',
6
+ baseURL: 'https://gemini.google.com',
7
+ reservedPaths: ['/app', '/chat'],
8
+ loginURL: () => 'https://accounts.google.com/signin/v2/identifier',
9
+ async isLoggedIn(page: PageInterface): Promise<boolean> {
10
+ return page.evaluate<boolean>("!!(document.querySelector('bard-sidenav') || document.querySelector('.ql-editor'))");
11
+ },
12
+ async extractUsername(page: PageInterface): Promise<string> {
13
+ return page.evaluate<string>("document.querySelector('.gb_A.gb_Sa')?.textContent?.trim() ?? ''");
14
+ },
15
+ };