opc-agent 4.0.44 → 4.1.1

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 (250) hide show
  1. package/.github/ISSUE_TEMPLATE/bug_report.md +20 -20
  2. package/.github/ISSUE_TEMPLATE/feature_request.md +14 -14
  3. package/.github/PULL_REQUEST_TEMPLATE.md +13 -13
  4. package/CHANGELOG.md +48 -48
  5. package/CONTRIBUTING.md +36 -36
  6. package/README.zh-CN.md +497 -497
  7. package/dist/channels/wechat.js +6 -6
  8. package/dist/cli.js +2 -2
  9. package/dist/core/runtime.js +18 -0
  10. package/dist/deploy/index.js +56 -56
  11. package/dist/providers/index.js +39 -13
  12. package/dist/studio/server.js +211 -20
  13. package/dist/studio-ui/index.html +279 -24
  14. package/dist/ui/components.js +105 -105
  15. package/examples/README.md +22 -22
  16. package/examples/basic-agent.ts +90 -90
  17. package/examples/brain-integration.ts +71 -71
  18. package/examples/multi-channel.ts +74 -74
  19. package/fix-sidebar.mjs +188 -188
  20. package/install.ps1 +154 -154
  21. package/install.sh +164 -164
  22. package/package.json +1 -1
  23. package/scripts/install.ps1 +31 -31
  24. package/scripts/install.sh +40 -40
  25. package/serve-studio.js +13 -13
  26. package/serve-test.js +25 -25
  27. package/src/channels/dingtalk.ts +46 -46
  28. package/src/channels/email.ts +351 -351
  29. package/src/channels/feishu.ts +349 -349
  30. package/src/channels/googlechat.ts +42 -42
  31. package/src/channels/imessage.ts +31 -31
  32. package/src/channels/irc.ts +82 -82
  33. package/src/channels/line.ts +32 -32
  34. package/src/channels/matrix.ts +33 -33
  35. package/src/channels/mattermost.ts +57 -57
  36. package/src/channels/msteams.ts +32 -32
  37. package/src/channels/nostr.ts +32 -32
  38. package/src/channels/qq.ts +33 -33
  39. package/src/channels/signal.ts +32 -32
  40. package/src/channels/sms.ts +33 -33
  41. package/src/channels/telegram.ts +616 -616
  42. package/src/channels/twitch.ts +65 -65
  43. package/src/channels/voice-call.ts +100 -100
  44. package/src/channels/websocket.ts +399 -399
  45. package/src/channels/wechat.ts +329 -329
  46. package/src/channels/whatsapp.ts +32 -32
  47. package/src/cli/chat.ts +99 -99
  48. package/src/cli/setup.ts +314 -314
  49. package/src/cli.ts +2 -2
  50. package/src/core/agent.ts +476 -476
  51. package/src/core/api-server.ts +277 -277
  52. package/src/core/audio.ts +98 -98
  53. package/src/core/collaboration.ts +275 -275
  54. package/src/core/context-discovery.ts +85 -85
  55. package/src/core/context-refs.ts +140 -140
  56. package/src/core/gateway.ts +106 -106
  57. package/src/core/heartbeat.ts +51 -51
  58. package/src/core/hooks.ts +105 -105
  59. package/src/core/ide-bridge.ts +133 -133
  60. package/src/core/node-network.ts +86 -86
  61. package/src/core/profiles.ts +122 -122
  62. package/src/core/runtime.ts +18 -0
  63. package/src/core/scheduler.ts +187 -187
  64. package/src/core/session-manager.ts +137 -137
  65. package/src/core/subagent.ts +98 -98
  66. package/src/core/vision.ts +180 -180
  67. package/src/core/workflow-graph.ts +365 -365
  68. package/src/daemon.ts +96 -96
  69. package/src/deploy/index.ts +255 -255
  70. package/src/doctor.ts +156 -156
  71. package/src/eval/index.ts +211 -211
  72. package/src/eval/suites/basic.json +16 -16
  73. package/src/eval/suites/memory.json +12 -12
  74. package/src/eval/suites/safety.json +14 -14
  75. package/src/hub/brain-seed.ts +54 -54
  76. package/src/hub/client.ts +60 -60
  77. package/src/mcp/servers/calculator-mcp.ts +65 -65
  78. package/src/mcp/servers/crypto-mcp.ts +73 -73
  79. package/src/mcp/servers/database-mcp.ts +72 -72
  80. package/src/mcp/servers/datetime-mcp.ts +69 -69
  81. package/src/mcp/servers/filesystem.ts +66 -66
  82. package/src/mcp/servers/github-mcp.ts +58 -58
  83. package/src/mcp/servers/index.ts +63 -63
  84. package/src/mcp/servers/json-mcp.ts +102 -102
  85. package/src/mcp/servers/memory-mcp.ts +56 -56
  86. package/src/mcp/servers/regex-mcp.ts +53 -53
  87. package/src/mcp/servers/web-mcp.ts +49 -49
  88. package/src/memory/context-compressor.ts +189 -189
  89. package/src/memory/seed-loader.ts +212 -212
  90. package/src/memory/user-profiler.ts +215 -215
  91. package/src/plugins/content-filter.ts +23 -23
  92. package/src/plugins/logger.ts +18 -18
  93. package/src/plugins/rate-limiter.ts +38 -38
  94. package/src/protocols/a2a/client.ts +132 -132
  95. package/src/protocols/a2a/index.ts +8 -8
  96. package/src/protocols/a2a/server.ts +333 -333
  97. package/src/protocols/a2a/types.ts +88 -88
  98. package/src/protocols/a2a/utils.ts +50 -50
  99. package/src/protocols/agui/client.ts +83 -83
  100. package/src/protocols/agui/index.ts +4 -4
  101. package/src/protocols/agui/server.ts +218 -218
  102. package/src/protocols/agui/types.ts +153 -153
  103. package/src/protocols/index.ts +2 -2
  104. package/src/protocols/mcp/agent-tools.ts +134 -134
  105. package/src/protocols/mcp/index.ts +8 -8
  106. package/src/protocols/mcp/server.ts +262 -262
  107. package/src/protocols/mcp/types.ts +69 -69
  108. package/src/providers/index.ts +632 -608
  109. package/src/publish/index.ts +376 -376
  110. package/src/scheduler/cron-engine.ts +191 -191
  111. package/src/scheduler/index.ts +2 -2
  112. package/src/schema/oad.ts +217 -217
  113. package/src/security/approval.ts +131 -131
  114. package/src/security/approvals.ts +143 -143
  115. package/src/security/elevated.ts +105 -105
  116. package/src/security/guardrails.ts +248 -248
  117. package/src/security/index.ts +9 -9
  118. package/src/security/keys.ts +87 -87
  119. package/src/security/secrets.ts +129 -129
  120. package/src/skills/builtin/index.ts +408 -408
  121. package/src/skills/marketplace.ts +113 -113
  122. package/src/skills/types.ts +42 -42
  123. package/src/studio/server.ts +209 -22
  124. package/src/studio/templates-data.ts +178 -178
  125. package/src/studio-ui/index.html +279 -24
  126. package/src/telemetry/index.ts +324 -324
  127. package/src/tools/builtin/browser.ts +299 -299
  128. package/src/tools/builtin/datetime.ts +41 -41
  129. package/src/tools/builtin/file.ts +107 -107
  130. package/src/tools/builtin/home-assistant.ts +116 -116
  131. package/src/tools/builtin/rl-tools.ts +243 -243
  132. package/src/tools/builtin/shell.ts +43 -43
  133. package/src/tools/builtin/vision.ts +64 -64
  134. package/src/tools/builtin/web-search.ts +126 -126
  135. package/src/tools/builtin/web.ts +35 -35
  136. package/src/tools/document-processor.ts +213 -213
  137. package/src/tools/image-generator.ts +150 -150
  138. package/src/tools/integrations/calendar.ts +73 -73
  139. package/src/tools/integrations/code-exec.ts +39 -39
  140. package/src/tools/integrations/csv-analyzer.ts +92 -92
  141. package/src/tools/integrations/database.ts +44 -44
  142. package/src/tools/integrations/email-send.ts +76 -76
  143. package/src/tools/integrations/git-tool.ts +42 -42
  144. package/src/tools/integrations/github-tool.ts +76 -76
  145. package/src/tools/integrations/image-gen.ts +56 -56
  146. package/src/tools/integrations/index.ts +92 -92
  147. package/src/tools/integrations/jira.ts +83 -83
  148. package/src/tools/integrations/notion.ts +71 -71
  149. package/src/tools/integrations/npm-tool.ts +48 -48
  150. package/src/tools/integrations/pdf-reader.ts +58 -58
  151. package/src/tools/integrations/slack.ts +65 -65
  152. package/src/tools/integrations/summarizer.ts +49 -49
  153. package/src/tools/integrations/translator.ts +48 -48
  154. package/src/tools/integrations/trello.ts +60 -60
  155. package/src/tools/integrations/vector-search.ts +42 -42
  156. package/src/tools/integrations/web-scraper.ts +47 -47
  157. package/src/tools/integrations/web-search.ts +58 -58
  158. package/src/tools/integrations/webhook.ts +38 -38
  159. package/src/tools/mcp-client.ts +131 -131
  160. package/src/tools/web-scraper.ts +179 -179
  161. package/src/tools/web-search.ts +180 -180
  162. package/src/ui/components.ts +127 -127
  163. package/srv-out.txt +1 -1
  164. package/templates/ecommerce-assistant/README.md +45 -45
  165. package/templates/ecommerce-assistant/oad.yaml +47 -47
  166. package/templates/tech-support/README.md +43 -43
  167. package/templates/tech-support/oad.yaml +45 -45
  168. package/test-agent/Dockerfile +9 -9
  169. package/test-agent/README.md +50 -50
  170. package/test-agent/agent.yaml +23 -23
  171. package/test-agent/docker-compose.yml +11 -11
  172. package/test-agent/oad.yaml +31 -31
  173. package/test-agent/package-lock.json +1492 -1492
  174. package/test-agent/package.json +17 -17
  175. package/test-agent/src/index.ts +24 -24
  176. package/test-agent/src/skills/echo.ts +15 -15
  177. package/test-agent/tsconfig.json +24 -24
  178. package/test-full.js +43 -43
  179. package/test-sidebar.js +22 -22
  180. package/test-studio3.js +75 -75
  181. package/test-studio4.js +41 -41
  182. package/tests/a2a-protocol.test.ts +285 -285
  183. package/tests/agui-protocol.test.ts +246 -246
  184. package/tests/api-server.test.ts +148 -148
  185. package/tests/approvals.test.ts +89 -89
  186. package/tests/audio.test.ts +40 -40
  187. package/tests/brain-seed-extended.test.ts +490 -490
  188. package/tests/brain-seed.test.ts +239 -239
  189. package/tests/browser.test.ts +179 -179
  190. package/tests/channels/discord.test.ts +79 -79
  191. package/tests/channels/email.test.ts +148 -148
  192. package/tests/channels/feishu.test.ts +123 -123
  193. package/tests/channels/telegram.test.ts +129 -129
  194. package/tests/channels/websocket.test.ts +53 -53
  195. package/tests/channels/wechat.test.ts +170 -170
  196. package/tests/channels-extra.test.ts +45 -45
  197. package/tests/chat-cli.test.ts +160 -160
  198. package/tests/cli.test.ts +46 -46
  199. package/tests/context-compressor.test.ts +172 -172
  200. package/tests/context-refs.test.ts +121 -121
  201. package/tests/cron-engine.test.ts +101 -101
  202. package/tests/daemon.test.ts +135 -135
  203. package/tests/deepbrain-wire.test.ts +234 -234
  204. package/tests/deploy-and-dag.test.ts +196 -196
  205. package/tests/doctor.test.ts +38 -38
  206. package/tests/document-processor.test.ts +69 -69
  207. package/tests/e2e-nocode.test.ts +442 -442
  208. package/tests/elevated.test.ts +69 -69
  209. package/tests/eval.test.ts +173 -173
  210. package/tests/gateway.test.ts +63 -63
  211. package/tests/guardrails.test.ts +177 -177
  212. package/tests/home-assistant.test.ts +40 -40
  213. package/tests/hooks.test.ts +79 -79
  214. package/tests/ide-bridge.test.ts +38 -38
  215. package/tests/image-generator.test.ts +84 -84
  216. package/tests/init-role.test.ts +124 -124
  217. package/tests/integrations.test.ts +249 -249
  218. package/tests/mcp-client.test.ts +92 -92
  219. package/tests/mcp-server.test.ts +178 -178
  220. package/tests/mcp-servers.test.ts +260 -260
  221. package/tests/node-network.test.ts +74 -74
  222. package/tests/plugin-a2a-enhanced.test.ts +230 -230
  223. package/tests/profiles.test.ts +61 -61
  224. package/tests/publish.test.ts +231 -231
  225. package/tests/rl-tools.test.ts +93 -93
  226. package/tests/sandbox-manager.test.ts +46 -46
  227. package/tests/scheduler.test.ts +200 -200
  228. package/tests/secrets.test.ts +107 -107
  229. package/tests/security-enhanced.test.ts +233 -233
  230. package/tests/settings-api.test.ts +148 -148
  231. package/tests/setup.test.ts +73 -73
  232. package/tests/subagent.test.ts +193 -193
  233. package/tests/telegram-discord.test.ts +60 -60
  234. package/tests/telemetry.test.ts +186 -186
  235. package/tests/user-profiler.test.ts +169 -169
  236. package/tests/v090-features.test.ts +254 -254
  237. package/tests/vision.test.ts +61 -61
  238. package/tests/voice-call.test.ts +47 -47
  239. package/tests/voice-enhanced.test.ts +169 -169
  240. package/tests/voice-interaction.test.ts +38 -38
  241. package/tests/web-search.test.ts +155 -155
  242. package/tests/workflow-graph.test.ts +279 -279
  243. package/tutorial/customer-service-agent/README.md +612 -612
  244. package/tutorial/customer-service-agent/SOUL.md +26 -26
  245. package/tutorial/customer-service-agent/agent.yaml +63 -63
  246. package/tutorial/customer-service-agent/package.json +19 -19
  247. package/tutorial/customer-service-agent/src/index.ts +69 -69
  248. package/tutorial/customer-service-agent/src/skills/faq.ts +27 -27
  249. package/tutorial/customer-service-agent/src/skills/ticket.ts +22 -22
  250. package/tutorial/customer-service-agent/tsconfig.json +14 -14
@@ -1,299 +1,299 @@
1
- import type { MCPTool, MCPToolResult } from '../mcp';
2
-
3
- const IDLE_TIMEOUT_MS = 5 * 60 * 1000;
4
- const MAX_TEXT_LENGTH = 5000;
5
-
6
- export class BrowserManager {
7
- private browser: any = null;
8
- private page: any = null;
9
- private lastActivity: number = 0;
10
- private idleTimer: ReturnType<typeof setTimeout> | null = null;
11
- private playwrightFactory: (() => any) | null;
12
-
13
- constructor(playwrightFactory?: () => any) {
14
- this.playwrightFactory = playwrightFactory || null;
15
- }
16
-
17
- private resetIdleTimer(): void {
18
- this.lastActivity = Date.now();
19
- if (this.idleTimer) clearTimeout(this.idleTimer);
20
- this.idleTimer = setTimeout(() => this.close(), IDLE_TIMEOUT_MS);
21
- }
22
-
23
- async ensureBrowser(): Promise<any> {
24
- if (!this.browser) {
25
- let playwright: any;
26
- if (this.playwrightFactory) {
27
- playwright = this.playwrightFactory();
28
- } else {
29
- try {
30
- playwright = require('playwright');
31
- } catch {
32
- throw new Error('Install playwright: npm i playwright');
33
- }
34
- }
35
- this.browser = await playwright.chromium.launch({ headless: true });
36
- const context = await this.browser.newContext();
37
- this.page = await context.newPage();
38
- }
39
- this.resetIdleTimer();
40
- return this.page;
41
- }
42
-
43
- async navigate(url: string): Promise<{ title: string; text: string; url: string }> {
44
- const page = await this.ensureBrowser();
45
- await page.goto(url, { waitUntil: 'domcontentloaded', timeout: 30000 });
46
- const title = await page.title();
47
- const text = (await page.innerText('body')).slice(0, MAX_TEXT_LENGTH);
48
- return { title, text, url: page.url() };
49
- }
50
-
51
- async click(selector: string): Promise<void> {
52
- const page = await this.ensureBrowser();
53
- await page.click(selector, { timeout: 10000 });
54
- }
55
-
56
- async type(selector: string, text: string): Promise<void> {
57
- const page = await this.ensureBrowser();
58
- await page.fill(selector, text, { timeout: 10000 });
59
- }
60
-
61
- async screenshot(): Promise<string> {
62
- const page = await this.ensureBrowser();
63
- const buffer = await page.screenshot({ type: 'png' });
64
- return buffer.toString('base64');
65
- }
66
-
67
- async extract(): Promise<{ text: string; links: string[]; images: string[] }> {
68
- const page = await this.ensureBrowser();
69
- const text = (await page.innerText('body')).slice(0, MAX_TEXT_LENGTH);
70
- const links: string[] = await page.$$eval('a[href]', (els: any[]) => els.map((e: any) => e.href).slice(0, 100));
71
- const images: string[] = await page.$$eval('img[src]', (els: any[]) => els.map((e: any) => e.src).slice(0, 100));
72
- return { text, links, images };
73
- }
74
-
75
- async scroll(direction: 'up' | 'down', amount?: number): Promise<void> {
76
- const page = await this.ensureBrowser();
77
- const delta = amount || 500;
78
- const scrollScript = direction === 'down'
79
- ? `window.scrollBy(0, ${delta})`
80
- : `window.scrollBy(0, -${delta})`;
81
- await page.evaluate(scrollScript);
82
- }
83
-
84
- async back(): Promise<void> {
85
- const page = await this.ensureBrowser();
86
- await page.goBack({ timeout: 10000 });
87
- }
88
-
89
- async evaluate(script: string): Promise<any> {
90
- const page = await this.ensureBrowser();
91
- return await page.evaluate(script);
92
- }
93
-
94
- async getImages(): Promise<Array<{ src: string; alt: string }>> {
95
- const page = await this.ensureBrowser();
96
- return await page.$$eval('img', (els: any[]) =>
97
- els.map((e: any) => ({ src: e.src, alt: e.alt || '' })).slice(0, 200)
98
- );
99
- }
100
-
101
- async waitFor(selector: string, timeout?: number): Promise<boolean> {
102
- const page = await this.ensureBrowser();
103
- try {
104
- await page.waitForSelector(selector, { timeout: timeout || 5000 });
105
- return true;
106
- } catch {
107
- return false;
108
- }
109
- }
110
-
111
- async close(): Promise<void> {
112
- if (this.idleTimer) {
113
- clearTimeout(this.idleTimer);
114
- this.idleTimer = null;
115
- }
116
- if (this.browser) {
117
- try { await this.browser.close(); } catch {}
118
- this.browser = null;
119
- this.page = null;
120
- }
121
- }
122
- }
123
-
124
- // Singleton
125
- const browserManager = new BrowserManager();
126
-
127
- function wrapTool(fn: () => Promise<MCPToolResult>): Promise<MCPToolResult> {
128
- return fn().catch((err: any) => ({
129
- content: `Browser error: ${err instanceof Error ? err.message : String(err)}`,
130
- isError: true,
131
- }));
132
- }
133
-
134
- export const browserNavigateTool: MCPTool = {
135
- name: 'browser_navigate',
136
- description: 'Navigate to a URL and return page title + text content (truncated)',
137
- inputSchema: {
138
- type: 'object',
139
- properties: {
140
- url: { type: 'string', description: 'URL to navigate to' },
141
- },
142
- required: ['url'],
143
- },
144
- execute: (input) => wrapTool(async () => {
145
- const url = input.url as string;
146
- if (!url) return { content: 'Missing required parameter: url', isError: true };
147
- const result = await browserManager.navigate(url);
148
- return { content: `Title: ${result.title}\nURL: ${result.url}\n\n${result.text}`, isError: false };
149
- }),
150
- };
151
-
152
- export const browserClickTool: MCPTool = {
153
- name: 'browser_click',
154
- description: 'Click an element by CSS selector',
155
- inputSchema: {
156
- type: 'object',
157
- properties: {
158
- selector: { type: 'string', description: 'CSS selector of element to click' },
159
- },
160
- required: ['selector'],
161
- },
162
- execute: (input) => wrapTool(async () => {
163
- const selector = input.selector as string;
164
- if (!selector) return { content: 'Missing required parameter: selector', isError: true };
165
- await browserManager.click(selector);
166
- return { content: `Clicked: ${selector}`, isError: false };
167
- }),
168
- };
169
-
170
- export const browserTypeTool: MCPTool = {
171
- name: 'browser_type',
172
- description: 'Type text into an element by CSS selector',
173
- inputSchema: {
174
- type: 'object',
175
- properties: {
176
- selector: { type: 'string', description: 'CSS selector of input element' },
177
- text: { type: 'string', description: 'Text to type' },
178
- },
179
- required: ['selector', 'text'],
180
- },
181
- execute: (input) => wrapTool(async () => {
182
- const selector = input.selector as string;
183
- const text = input.text as string;
184
- if (!selector || text === undefined) return { content: 'Missing required parameters: selector, text', isError: true };
185
- await browserManager.type(selector, text);
186
- return { content: `Typed into ${selector}`, isError: false };
187
- }),
188
- };
189
-
190
- export const browserScreenshotTool: MCPTool = {
191
- name: 'browser_screenshot',
192
- description: 'Take a screenshot of the current page, returned as base64 PNG',
193
- inputSchema: { type: 'object', properties: {} },
194
- execute: () => wrapTool(async () => {
195
- const base64 = await browserManager.screenshot();
196
- return { content: base64, isError: false, metadata: { encoding: 'base64', mimeType: 'image/png' } };
197
- }),
198
- };
199
-
200
- export const browserExtractTool: MCPTool = {
201
- name: 'browser_extract',
202
- description: 'Extract text, links, and images from the current page',
203
- inputSchema: { type: 'object', properties: {} },
204
- execute: () => wrapTool(async () => {
205
- const data = await browserManager.extract();
206
- return { content: JSON.stringify(data, null, 2), isError: false };
207
- }),
208
- };
209
-
210
- export const browserScrollTool: MCPTool = {
211
- name: 'browser_scroll',
212
- description: 'Scroll the page up or down',
213
- inputSchema: {
214
- type: 'object',
215
- properties: {
216
- direction: { type: 'string', enum: ['up', 'down'], description: 'Scroll direction' },
217
- amount: { type: 'number', description: 'Pixels to scroll (default 500)' },
218
- },
219
- required: ['direction'],
220
- },
221
- execute: (input) => wrapTool(async () => {
222
- const direction = input.direction as 'up' | 'down';
223
- if (!direction) return { content: 'Missing required parameter: direction', isError: true };
224
- await browserManager.scroll(direction, input.amount as number | undefined);
225
- return { content: `Scrolled ${direction}`, isError: false };
226
- }),
227
- };
228
-
229
- export const browserBackTool: MCPTool = {
230
- name: 'browser_back',
231
- description: 'Navigate back in browser history',
232
- inputSchema: { type: 'object', properties: {} },
233
- execute: () => wrapTool(async () => {
234
- await browserManager.back();
235
- return { content: 'Navigated back', isError: false };
236
- }),
237
- };
238
-
239
- export const browserEvalTool: MCPTool = {
240
- name: 'browser_eval',
241
- description: 'Execute JavaScript in the page context',
242
- inputSchema: {
243
- type: 'object',
244
- properties: {
245
- script: { type: 'string', description: 'JavaScript to execute' },
246
- },
247
- required: ['script'],
248
- },
249
- execute: (input) => wrapTool(async () => {
250
- const script = input.script as string;
251
- if (!script) return { content: 'Missing required parameter: script', isError: true };
252
- const result = await browserManager.evaluate(script);
253
- return { content: typeof result === 'string' ? result : JSON.stringify(result, null, 2), isError: false };
254
- }),
255
- };
256
-
257
- export const browserGetImagesTool: MCPTool = {
258
- name: 'browser_get_images',
259
- description: 'List all images on the current page with src and alt attributes',
260
- inputSchema: { type: 'object', properties: {} },
261
- execute: () => wrapTool(async () => {
262
- const images = await browserManager.getImages();
263
- return { content: JSON.stringify(images, null, 2), isError: false };
264
- }),
265
- };
266
-
267
- export const browserWaitTool: MCPTool = {
268
- name: 'browser_wait',
269
- description: 'Wait for a CSS selector to appear on the page',
270
- inputSchema: {
271
- type: 'object',
272
- properties: {
273
- selector: { type: 'string', description: 'CSS selector to wait for' },
274
- timeout: { type: 'number', description: 'Max wait time in ms (default 5000)' },
275
- },
276
- required: ['selector'],
277
- },
278
- execute: (input) => wrapTool(async () => {
279
- const selector = input.selector as string;
280
- if (!selector) return { content: 'Missing required parameter: selector', isError: true };
281
- const found = await browserManager.waitFor(selector, input.timeout as number | undefined);
282
- return { content: found ? `Found: ${selector}` : `Timeout waiting for: ${selector}`, isError: !found };
283
- }),
284
- };
285
-
286
- export const browserTools: MCPTool[] = [
287
- browserNavigateTool,
288
- browserClickTool,
289
- browserTypeTool,
290
- browserScreenshotTool,
291
- browserExtractTool,
292
- browserScrollTool,
293
- browserBackTool,
294
- browserEvalTool,
295
- browserGetImagesTool,
296
- browserWaitTool,
297
- ];
298
-
299
- export { browserManager };
1
+ import type { MCPTool, MCPToolResult } from '../mcp';
2
+
3
+ const IDLE_TIMEOUT_MS = 5 * 60 * 1000;
4
+ const MAX_TEXT_LENGTH = 5000;
5
+
6
+ export class BrowserManager {
7
+ private browser: any = null;
8
+ private page: any = null;
9
+ private lastActivity: number = 0;
10
+ private idleTimer: ReturnType<typeof setTimeout> | null = null;
11
+ private playwrightFactory: (() => any) | null;
12
+
13
+ constructor(playwrightFactory?: () => any) {
14
+ this.playwrightFactory = playwrightFactory || null;
15
+ }
16
+
17
+ private resetIdleTimer(): void {
18
+ this.lastActivity = Date.now();
19
+ if (this.idleTimer) clearTimeout(this.idleTimer);
20
+ this.idleTimer = setTimeout(() => this.close(), IDLE_TIMEOUT_MS);
21
+ }
22
+
23
+ async ensureBrowser(): Promise<any> {
24
+ if (!this.browser) {
25
+ let playwright: any;
26
+ if (this.playwrightFactory) {
27
+ playwright = this.playwrightFactory();
28
+ } else {
29
+ try {
30
+ playwright = require('playwright');
31
+ } catch {
32
+ throw new Error('Install playwright: npm i playwright');
33
+ }
34
+ }
35
+ this.browser = await playwright.chromium.launch({ headless: true });
36
+ const context = await this.browser.newContext();
37
+ this.page = await context.newPage();
38
+ }
39
+ this.resetIdleTimer();
40
+ return this.page;
41
+ }
42
+
43
+ async navigate(url: string): Promise<{ title: string; text: string; url: string }> {
44
+ const page = await this.ensureBrowser();
45
+ await page.goto(url, { waitUntil: 'domcontentloaded', timeout: 30000 });
46
+ const title = await page.title();
47
+ const text = (await page.innerText('body')).slice(0, MAX_TEXT_LENGTH);
48
+ return { title, text, url: page.url() };
49
+ }
50
+
51
+ async click(selector: string): Promise<void> {
52
+ const page = await this.ensureBrowser();
53
+ await page.click(selector, { timeout: 10000 });
54
+ }
55
+
56
+ async type(selector: string, text: string): Promise<void> {
57
+ const page = await this.ensureBrowser();
58
+ await page.fill(selector, text, { timeout: 10000 });
59
+ }
60
+
61
+ async screenshot(): Promise<string> {
62
+ const page = await this.ensureBrowser();
63
+ const buffer = await page.screenshot({ type: 'png' });
64
+ return buffer.toString('base64');
65
+ }
66
+
67
+ async extract(): Promise<{ text: string; links: string[]; images: string[] }> {
68
+ const page = await this.ensureBrowser();
69
+ const text = (await page.innerText('body')).slice(0, MAX_TEXT_LENGTH);
70
+ const links: string[] = await page.$$eval('a[href]', (els: any[]) => els.map((e: any) => e.href).slice(0, 100));
71
+ const images: string[] = await page.$$eval('img[src]', (els: any[]) => els.map((e: any) => e.src).slice(0, 100));
72
+ return { text, links, images };
73
+ }
74
+
75
+ async scroll(direction: 'up' | 'down', amount?: number): Promise<void> {
76
+ const page = await this.ensureBrowser();
77
+ const delta = amount || 500;
78
+ const scrollScript = direction === 'down'
79
+ ? `window.scrollBy(0, ${delta})`
80
+ : `window.scrollBy(0, -${delta})`;
81
+ await page.evaluate(scrollScript);
82
+ }
83
+
84
+ async back(): Promise<void> {
85
+ const page = await this.ensureBrowser();
86
+ await page.goBack({ timeout: 10000 });
87
+ }
88
+
89
+ async evaluate(script: string): Promise<any> {
90
+ const page = await this.ensureBrowser();
91
+ return await page.evaluate(script);
92
+ }
93
+
94
+ async getImages(): Promise<Array<{ src: string; alt: string }>> {
95
+ const page = await this.ensureBrowser();
96
+ return await page.$$eval('img', (els: any[]) =>
97
+ els.map((e: any) => ({ src: e.src, alt: e.alt || '' })).slice(0, 200)
98
+ );
99
+ }
100
+
101
+ async waitFor(selector: string, timeout?: number): Promise<boolean> {
102
+ const page = await this.ensureBrowser();
103
+ try {
104
+ await page.waitForSelector(selector, { timeout: timeout || 5000 });
105
+ return true;
106
+ } catch {
107
+ return false;
108
+ }
109
+ }
110
+
111
+ async close(): Promise<void> {
112
+ if (this.idleTimer) {
113
+ clearTimeout(this.idleTimer);
114
+ this.idleTimer = null;
115
+ }
116
+ if (this.browser) {
117
+ try { await this.browser.close(); } catch {}
118
+ this.browser = null;
119
+ this.page = null;
120
+ }
121
+ }
122
+ }
123
+
124
+ // Singleton
125
+ const browserManager = new BrowserManager();
126
+
127
+ function wrapTool(fn: () => Promise<MCPToolResult>): Promise<MCPToolResult> {
128
+ return fn().catch((err: any) => ({
129
+ content: `Browser error: ${err instanceof Error ? err.message : String(err)}`,
130
+ isError: true,
131
+ }));
132
+ }
133
+
134
+ export const browserNavigateTool: MCPTool = {
135
+ name: 'browser_navigate',
136
+ description: 'Navigate to a URL and return page title + text content (truncated)',
137
+ inputSchema: {
138
+ type: 'object',
139
+ properties: {
140
+ url: { type: 'string', description: 'URL to navigate to' },
141
+ },
142
+ required: ['url'],
143
+ },
144
+ execute: (input) => wrapTool(async () => {
145
+ const url = input.url as string;
146
+ if (!url) return { content: 'Missing required parameter: url', isError: true };
147
+ const result = await browserManager.navigate(url);
148
+ return { content: `Title: ${result.title}\nURL: ${result.url}\n\n${result.text}`, isError: false };
149
+ }),
150
+ };
151
+
152
+ export const browserClickTool: MCPTool = {
153
+ name: 'browser_click',
154
+ description: 'Click an element by CSS selector',
155
+ inputSchema: {
156
+ type: 'object',
157
+ properties: {
158
+ selector: { type: 'string', description: 'CSS selector of element to click' },
159
+ },
160
+ required: ['selector'],
161
+ },
162
+ execute: (input) => wrapTool(async () => {
163
+ const selector = input.selector as string;
164
+ if (!selector) return { content: 'Missing required parameter: selector', isError: true };
165
+ await browserManager.click(selector);
166
+ return { content: `Clicked: ${selector}`, isError: false };
167
+ }),
168
+ };
169
+
170
+ export const browserTypeTool: MCPTool = {
171
+ name: 'browser_type',
172
+ description: 'Type text into an element by CSS selector',
173
+ inputSchema: {
174
+ type: 'object',
175
+ properties: {
176
+ selector: { type: 'string', description: 'CSS selector of input element' },
177
+ text: { type: 'string', description: 'Text to type' },
178
+ },
179
+ required: ['selector', 'text'],
180
+ },
181
+ execute: (input) => wrapTool(async () => {
182
+ const selector = input.selector as string;
183
+ const text = input.text as string;
184
+ if (!selector || text === undefined) return { content: 'Missing required parameters: selector, text', isError: true };
185
+ await browserManager.type(selector, text);
186
+ return { content: `Typed into ${selector}`, isError: false };
187
+ }),
188
+ };
189
+
190
+ export const browserScreenshotTool: MCPTool = {
191
+ name: 'browser_screenshot',
192
+ description: 'Take a screenshot of the current page, returned as base64 PNG',
193
+ inputSchema: { type: 'object', properties: {} },
194
+ execute: () => wrapTool(async () => {
195
+ const base64 = await browserManager.screenshot();
196
+ return { content: base64, isError: false, metadata: { encoding: 'base64', mimeType: 'image/png' } };
197
+ }),
198
+ };
199
+
200
+ export const browserExtractTool: MCPTool = {
201
+ name: 'browser_extract',
202
+ description: 'Extract text, links, and images from the current page',
203
+ inputSchema: { type: 'object', properties: {} },
204
+ execute: () => wrapTool(async () => {
205
+ const data = await browserManager.extract();
206
+ return { content: JSON.stringify(data, null, 2), isError: false };
207
+ }),
208
+ };
209
+
210
+ export const browserScrollTool: MCPTool = {
211
+ name: 'browser_scroll',
212
+ description: 'Scroll the page up or down',
213
+ inputSchema: {
214
+ type: 'object',
215
+ properties: {
216
+ direction: { type: 'string', enum: ['up', 'down'], description: 'Scroll direction' },
217
+ amount: { type: 'number', description: 'Pixels to scroll (default 500)' },
218
+ },
219
+ required: ['direction'],
220
+ },
221
+ execute: (input) => wrapTool(async () => {
222
+ const direction = input.direction as 'up' | 'down';
223
+ if (!direction) return { content: 'Missing required parameter: direction', isError: true };
224
+ await browserManager.scroll(direction, input.amount as number | undefined);
225
+ return { content: `Scrolled ${direction}`, isError: false };
226
+ }),
227
+ };
228
+
229
+ export const browserBackTool: MCPTool = {
230
+ name: 'browser_back',
231
+ description: 'Navigate back in browser history',
232
+ inputSchema: { type: 'object', properties: {} },
233
+ execute: () => wrapTool(async () => {
234
+ await browserManager.back();
235
+ return { content: 'Navigated back', isError: false };
236
+ }),
237
+ };
238
+
239
+ export const browserEvalTool: MCPTool = {
240
+ name: 'browser_eval',
241
+ description: 'Execute JavaScript in the page context',
242
+ inputSchema: {
243
+ type: 'object',
244
+ properties: {
245
+ script: { type: 'string', description: 'JavaScript to execute' },
246
+ },
247
+ required: ['script'],
248
+ },
249
+ execute: (input) => wrapTool(async () => {
250
+ const script = input.script as string;
251
+ if (!script) return { content: 'Missing required parameter: script', isError: true };
252
+ const result = await browserManager.evaluate(script);
253
+ return { content: typeof result === 'string' ? result : JSON.stringify(result, null, 2), isError: false };
254
+ }),
255
+ };
256
+
257
+ export const browserGetImagesTool: MCPTool = {
258
+ name: 'browser_get_images',
259
+ description: 'List all images on the current page with src and alt attributes',
260
+ inputSchema: { type: 'object', properties: {} },
261
+ execute: () => wrapTool(async () => {
262
+ const images = await browserManager.getImages();
263
+ return { content: JSON.stringify(images, null, 2), isError: false };
264
+ }),
265
+ };
266
+
267
+ export const browserWaitTool: MCPTool = {
268
+ name: 'browser_wait',
269
+ description: 'Wait for a CSS selector to appear on the page',
270
+ inputSchema: {
271
+ type: 'object',
272
+ properties: {
273
+ selector: { type: 'string', description: 'CSS selector to wait for' },
274
+ timeout: { type: 'number', description: 'Max wait time in ms (default 5000)' },
275
+ },
276
+ required: ['selector'],
277
+ },
278
+ execute: (input) => wrapTool(async () => {
279
+ const selector = input.selector as string;
280
+ if (!selector) return { content: 'Missing required parameter: selector', isError: true };
281
+ const found = await browserManager.waitFor(selector, input.timeout as number | undefined);
282
+ return { content: found ? `Found: ${selector}` : `Timeout waiting for: ${selector}`, isError: !found };
283
+ }),
284
+ };
285
+
286
+ export const browserTools: MCPTool[] = [
287
+ browserNavigateTool,
288
+ browserClickTool,
289
+ browserTypeTool,
290
+ browserScreenshotTool,
291
+ browserExtractTool,
292
+ browserScrollTool,
293
+ browserBackTool,
294
+ browserEvalTool,
295
+ browserGetImagesTool,
296
+ browserWaitTool,
297
+ ];
298
+
299
+ export { browserManager };