@detiq/crawler 0.1.3 → 0.1.5

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.
package/Dockerfile CHANGED
@@ -40,6 +40,10 @@ RUN npx prisma generate --schema packages/database/prisma/schema.prisma && \
40
40
  # Screenshot cache dir
41
41
  RUN mkdir -p /app/.screenshots-cache
42
42
 
43
+ RUN addgroup --system appgroup && adduser --system --ingroup appgroup appuser
44
+ RUN chown -R appuser:appgroup /app
45
+ USER appuser
46
+
43
47
  EXPOSE 4001
44
48
  # index.js is a side-effect-free library barrel (also imported by @detiq/api) — it holds
45
49
  # no listener/timer, so running it directly starts the process, finishes its imports,
package/dist/ai-login.js CHANGED
@@ -1,5 +1,5 @@
1
1
  import { z } from "zod";
2
- import { AppBrain } from "@qip/app-brain";
2
+ import { AppBrain } from "@detiq/app-brain";
3
3
  import { solveCaptcha } from "./captcha-solver.js";
4
4
  import { waitForEmailContent } from "./email-otp.js";
5
5
  import { logger } from "./logger.js";
@@ -87,12 +87,20 @@ async function loginWithStagehand(stagehand, creds, startUrl, captchaOpts, email
87
87
  break;
88
88
  }
89
89
  }
90
+ const MAX_LOGIN_INSTRUCTIONS = 2e3;
91
+ if (loginInstructions && loginInstructions.length > MAX_LOGIN_INSTRUCTIONS) {
92
+ throw new Error(`loginInstructions exceeds ${MAX_LOGIN_INSTRUCTIONS} character limit`);
93
+ }
90
94
  if (loginInstructions) {
91
95
  await logStep("custom_recipe", () => actWithMouseSim(stagehand, loginInstructions));
92
96
  await pg.waitForLoadState("networkidle").catch(() => pg.waitForTimeout(1500));
93
97
  }
94
98
  await logStep("fill_username", () => actWithMouseSim(stagehand, `Type "${creds.username}" into the username or email input field`));
95
- await logStep("fill_password", () => actWithMouseSim(stagehand, `Type "${creds.password}" into the password input field`));
99
+ await logStep("fill_password", async () => {
100
+ const observations = await stagehand.observe("Find the password input field");
101
+ const selector = observations[0]?.selector || 'input[type="password"]';
102
+ await pg.locator(selector).fill(creds.password);
103
+ });
96
104
  await logStep("click_submit", () => actWithMouseSim(stagehand, "Click the sign in, log in, login, or continue button to submit credentials"));
97
105
  await pg.waitForLoadState("networkidle").catch(() => pg.waitForTimeout(2e3));
98
106
  const twoFA = await stagehand.extract(
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "version": 3,
3
3
  "sources": ["../src/ai-login.ts"],
4
- "sourcesContent": ["import { z } from 'zod';\nimport type { Stagehand } from '@browserbasehq/stagehand';\nimport { AppBrain } from '@qip/app-brain';\nimport { solveCaptcha } from './captcha-solver.js';\nimport { waitForEmailContent } from './email-otp.js';\nimport { logger } from './logger.js';\nimport type { Credentials } from './auth.js';\nimport { simulateMouseMovement } from './stealth/human-behavior.js';\n\nconst CLICK_INTENT = /\\b(click|tap|press|submit|select|choose|toggle)\\b/i;\n\n/** Simulate human mouse movement before click-type actions, then call stagehand.act(). */\nasync function actWithMouseSim(stagehand: Stagehand, instruction: string): Promise<any> {\n if (CLICK_INTENT.test(instruction)) {\n try {\n const pg = await (stagehand as any).resolvePage();\n const viewport = pg.viewportSize() ?? { width: 1280, height: 720 };\n // Move mouse toward center of page with some randomness before Stagehand clicks\n const targetX = viewport.width * 0.3 + Math.random() * viewport.width * 0.4;\n const targetY = viewport.height * 0.2 + Math.random() * viewport.height * 0.5;\n await simulateMouseMovement(pg, targetX, targetY);\n } catch { /* non-fatal \u2014 don't block the action */ }\n }\n return stagehand.act(instruction);\n}\n\n/**\n * Recorded login step from the Chrome extension's step recorder.\n * {username} and {password} in `value` are substituted at replay time.\n */\nexport interface LoginStep {\n action: 'fill' | 'click' | 'wait';\n selector: string;\n value?: string;\n ms?: number;\n}\n\nconst TwoFASchema = z.object({\n has2FA: z.boolean(),\n fieldLabel: z.string().optional(),\n});\n\nconst CAPTCHA_SELECTORS = [\n 'iframe[src*=\"recaptcha\"]',\n 'iframe[src*=\"hcaptcha\"]',\n 'iframe[src*=\"challenges.cloudflare.com\"]',\n '.cf-turnstile',\n '[data-sitekey]',\n '#captcha',\n '.captcha',\n '[id*=\"captcha\"]',\n '[class*=\"captcha\"]',\n];\n\n/**\n * AI-guided login using Stagehand act/extract.\n *\n * @param loginInstructions - Optional natural language description of the login flow.\n * When provided, an initial stagehand.act() is called with these instructions to\n * navigate the UI (e.g. \"click Sign In in the nav, fill email, click Continue\").\n * The standard 3-step credential fill follows immediately after.\n * @param opts - Optional tenant/project/job IDs for writing AUTH_STEP CrawlEvidence.\n */\nexport async function loginWithStagehand(\n stagehand: Stagehand,\n creds: Credentials,\n startUrl: string,\n captchaOpts?: { apiKey?: string; provider?: '2captcha' | 'capsolver' },\n emailOtpOpts?: { apiKey?: string; inboxId?: string },\n loginInstructions?: string,\n opts?: { tenantId?: string; projectId?: string; jobId?: string },\n): Promise<boolean> {\n async function logStep(step: string, fn: () => Promise<any>): Promise<{ success: boolean; detail: string }> {\n const t0 = Date.now();\n try {\n const r = await fn();\n const result = { success: true, detail: r?.description ?? r?.message ?? step };\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step, ...result, ms: Date.now() - t0 },\n }).catch(() => {});\n }\n return result;\n } catch (e: any) {\n const result = { success: false, detail: e.message ?? 'failed' };\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step, ...result, ms: Date.now() - t0 },\n }).catch(() => {});\n }\n throw e;\n }\n }\n\n try {\n const pg = await (stagehand as any).resolvePage();\n\n // CAPTCHA detection\n for (const sel of CAPTCHA_SELECTORS) {\n const count = await pg.locator(sel).count().catch(() => 0);\n if (count > 0) {\n logger.warn({ selector: sel }, '[crawler] CAPTCHA detected on login page');\n if (captchaOpts?.apiKey) {\n logger.info({ provider: captchaOpts.provider ?? '2captcha' }, '[crawler] Attempting auto-solve');\n try {\n const result = await solveCaptcha(pg, startUrl, captchaOpts.apiKey, captchaOpts.provider ?? '2captcha');\n if (result) {\n logger.info({ provider: result.provider }, '[crawler] CAPTCHA solved \u2014 proceeding with login');\n break;\n }\n } catch (solveErr) {\n logger.warn({ err: solveErr }, '[crawler] CAPTCHA solve failed \u2014 falling back to manual error');\n throw new Error(`CAPTCHA_DETECTED: Auto-solve failed (${String(solveErr)}). Use session capture instead.`);\n }\n } else {\n throw new Error(\n 'CAPTCHA_DETECTED: Login page has a CAPTCHA. Use \"Session Capture \u2192 Launch Browser\" to log in manually, then capture the session. Alternatively, disable CAPTCHA for the ZeTa crawler IP.'\n );\n }\n break;\n }\n }\n\n // If caller provided natural language login instructions, use them to navigate to the login form\n if (loginInstructions) {\n await logStep('custom_recipe', () => actWithMouseSim(stagehand, loginInstructions));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(1500));\n }\n\n await logStep('fill_username', () => actWithMouseSim(stagehand, `Type \"${creds.username}\" into the username or email input field`));\n await logStep('fill_password', () => actWithMouseSim(stagehand, `Type \"${creds.password}\" into the password input field`));\n await logStep('click_submit', () => actWithMouseSim(stagehand, 'Click the sign in, log in, login, or continue button to submit credentials'));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(2000));\n\n const twoFA = await stagehand.extract(\n 'Is there a two-factor authentication, OTP, or verification code input visible right now?',\n TwoFASchema as any\n );\n\n if (twoFA.has2FA) {\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step: '2fa_detect', success: true, detail: twoFA.fieldLabel ?? '2FA detected', ms: 0 },\n }).catch(() => {});\n }\n\n let code = creds.oneTimeCode ?? await creds.getOneTimeCode?.();\n\n if (!code && emailOtpOpts?.apiKey && emailOtpOpts?.inboxId) {\n const since = new Date().toISOString();\n logger.info('[crawler] 2FA detected \u2014 waiting for email OTP or magic link via MailSlurp');\n const emailContent = await waitForEmailContent(\n { provider: 'mailslurp', apiKey: emailOtpOpts.apiKey, inboxId: emailOtpOpts.inboxId },\n since\n );\n if (emailContent?.magicLink) {\n logger.info({ url: emailContent.magicLink }, '[crawler] Magic link received \u2014 navigating');\n await pg.goto(emailContent.magicLink, { waitUntil: 'domcontentloaded' });\n await pg.waitForLoadState('networkidle').catch(() => {});\n return true; // magic link navigation = login complete\n }\n code = emailContent?.otp ?? undefined;\n if (code) {\n logger.info('[crawler] Email OTP received');\n } else {\n logger.warn('[crawler] Email OTP timed out after 60s');\n }\n }\n\n if (!code) {\n console.warn('[crawler] 2FA detected but no oneTimeCode provided');\n return false;\n }\n await logStep('2fa_inject', () => actWithMouseSim(stagehand, `Enter \"${code}\" into the ${twoFA.fieldLabel ?? 'verification code'} field`));\n await logStep('2fa_submit', () => actWithMouseSim(stagehand, 'Click verify or confirm to submit the verification code'));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(2000));\n }\n\n // URL-based success check\n const currentUrl = pg.url();\n const loginPathPattern = /\\/(login|signin|sign-in|auth|authenticate|log-in|account\\/login)(\\?|$|\\/)/i;\n try {\n const { pathname } = new URL(currentUrl);\n const wasRedirected = currentUrl !== startUrl;\n const stillOnLoginPath = loginPathPattern.test(pathname);\n if (wasRedirected && !stillOnLoginPath) {\n logger.info({ url: currentUrl }, 'Login confirmed via URL change');\n return true;\n }\n } catch { /* invalid URL \u2014 fall through */ }\n\n const passwordCount = await pg.locator('input[type=\"password\"]').count().catch(() => 1);\n const bodyText = await pg.evaluate(() => (document as any).body?.innerText?.trim()?.length ?? 0).catch(() => 0);\n if (passwordCount === 0 && bodyText > 30) {\n logger.info('Login confirmed: password field no longer visible');\n return true;\n }\n\n logger.warn({ url: currentUrl }, 'Login appears failed');\n try {\n const screenshotPath = `/tmp/zeta-auth-failure-${Date.now()}.jpg`;\n await pg.screenshot({ path: screenshotPath, type: 'jpeg', quality: 70, fullPage: false });\n logger.warn({ screenshotPath }, '[crawler] Auth failure screenshot saved');\n } catch { /* non-fatal */ }\n return false;\n } catch (err) {\n const msg = String((err as Error).message ?? err);\n if (msg.includes('CAPTCHA_DETECTED')) throw err;\n logger.error({ err }, 'Login error');\n return false;\n }\n}\n"],
5
- "mappings": "AAAA,SAAS,SAAS;AAElB,SAAS,gBAAgB;AACzB,SAAS,oBAAoB;AAC7B,SAAS,2BAA2B;AACpC,SAAS,cAAc;AAEvB,SAAS,6BAA6B;AAEtC,MAAM,eAAe;AAGrB,eAAe,gBAAgB,WAAsB,aAAmC;AACtF,MAAI,aAAa,KAAK,WAAW,GAAG;AAClC,QAAI;AACF,YAAM,KAAK,MAAO,UAAkB,YAAY;AAChD,YAAM,WAAW,GAAG,aAAa,KAAK,EAAE,OAAO,MAAM,QAAQ,IAAI;AAEjE,YAAM,UAAU,SAAS,QAAQ,MAAM,KAAK,OAAO,IAAI,SAAS,QAAQ;AACxE,YAAM,UAAU,SAAS,SAAS,MAAM,KAAK,OAAO,IAAI,SAAS,SAAS;AAC1E,YAAM,sBAAsB,IAAI,SAAS,OAAO;AAAA,IAClD,QAAQ;AAAA,IAA2C;AAAA,EACrD;AACA,SAAO,UAAU,IAAI,WAAW;AAClC;AAaA,MAAM,cAAc,EAAE,OAAO;AAAA,EAC3B,QAAQ,EAAE,QAAQ;AAAA,EAClB,YAAY,EAAE,OAAO,EAAE,SAAS;AAClC,CAAC;AAED,MAAM,oBAAoB;AAAA,EACxB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAWA,eAAsB,mBACpB,WACA,OACA,UACA,aACA,cACA,mBACA,MACkB;AAClB,iBAAe,QAAQ,MAAc,IAAuE;AAC1G,UAAM,KAAK,KAAK,IAAI;AACpB,QAAI;AACF,YAAM,IAAI,MAAM,GAAG;AACnB,YAAM,SAAS,EAAE,SAAS,MAAM,QAAQ,GAAG,eAAe,GAAG,WAAW,KAAK;AAC7E,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,GAAG,QAAQ,IAAI,KAAK,IAAI,IAAI,GAAG;AAAA,QAClD,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AACA,aAAO;AAAA,IACT,SAAS,GAAQ;AACf,YAAM,SAAS,EAAE,SAAS,OAAO,QAAQ,EAAE,WAAW,SAAS;AAC/D,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,GAAG,QAAQ,IAAI,KAAK,IAAI,IAAI,GAAG;AAAA,QAClD,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AACA,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI;AACF,UAAM,KAAK,MAAO,UAAkB,YAAY;AAGhD,eAAW,OAAO,mBAAmB;AACnC,YAAM,QAAQ,MAAM,GAAG,QAAQ,GAAG,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AACzD,UAAI,QAAQ,GAAG;AACb,eAAO,KAAK,EAAE,UAAU,IAAI,GAAG,0CAA0C;AACzE,YAAI,aAAa,QAAQ;AACvB,iBAAO,KAAK,EAAE,UAAU,YAAY,YAAY,WAAW,GAAG,iCAAiC;AAC/F,cAAI;AACF,kBAAM,SAAS,MAAM,aAAa,IAAI,UAAU,YAAY,QAAQ,YAAY,YAAY,UAAU;AACtG,gBAAI,QAAQ;AACV,qBAAO,KAAK,EAAE,UAAU,OAAO,SAAS,GAAG,uDAAkD;AAC7F;AAAA,YACF;AAAA,UACF,SAAS,UAAU;AACjB,mBAAO,KAAK,EAAE,KAAK,SAAS,GAAG,oEAA+D;AAC9F,kBAAM,IAAI,MAAM,wCAAwC,OAAO,QAAQ,CAAC,iCAAiC;AAAA,UAC3G;AAAA,QACF,OAAO;AACL,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA;AAAA,MACF;AAAA,IACF;AAGA,QAAI,mBAAmB;AACrB,YAAM,QAAQ,iBAAiB,MAAM,gBAAgB,WAAW,iBAAiB,CAAC;AAClF,YAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,IAAI,CAAC;AAAA,IAC9E;AAEA,UAAM,QAAQ,iBAAiB,MAAM,gBAAgB,WAAW,SAAS,MAAM,QAAQ,0CAA0C,CAAC;AAClI,UAAM,QAAQ,iBAAiB,MAAM,gBAAgB,WAAW,SAAS,MAAM,QAAQ,iCAAiC,CAAC;AACzH,UAAM,QAAQ,gBAAgB,MAAM,gBAAgB,WAAW,4EAA4E,CAAC;AAC5I,UAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,GAAI,CAAC;AAE5E,UAAM,QAAQ,MAAM,UAAU;AAAA,MAC5B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,MAAM,QAAQ;AAChB,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,cAAc,SAAS,MAAM,QAAQ,MAAM,cAAc,gBAAgB,IAAI,EAAE;AAAA,QAClG,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AAEA,UAAI,OAAO,MAAM,eAAe,MAAM,MAAM,iBAAiB;AAE7D,UAAI,CAAC,QAAQ,cAAc,UAAU,cAAc,SAAS;AAC1D,cAAM,SAAQ,oBAAI,KAAK,GAAE,YAAY;AACrC,eAAO,KAAK,iFAA4E;AACxF,cAAM,eAAe,MAAM;AAAA,UACzB,EAAE,UAAU,aAAa,QAAQ,aAAa,QAAQ,SAAS,aAAa,QAAQ;AAAA,UACpF;AAAA,QACF;AACA,YAAI,cAAc,WAAW;AAC3B,iBAAO,KAAK,EAAE,KAAK,aAAa,UAAU,GAAG,iDAA4C;AACzF,gBAAM,GAAG,KAAK,aAAa,WAAW,EAAE,WAAW,mBAAmB,CAAC;AACvE,gBAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM;AAAA,UAAC,CAAC;AACvD,iBAAO;AAAA,QACT;AACA,eAAO,cAAc,OAAO;AAC5B,YAAI,MAAM;AACR,iBAAO,KAAK,8BAA8B;AAAA,QAC5C,OAAO;AACL,iBAAO,KAAK,yCAAyC;AAAA,QACvD;AAAA,MACF;AAEA,UAAI,CAAC,MAAM;AACT,gBAAQ,KAAK,oDAAoD;AACjE,eAAO;AAAA,MACT;AACA,YAAM,QAAQ,cAAc,MAAM,gBAAgB,WAAW,UAAU,IAAI,cAAc,MAAM,cAAc,mBAAmB,QAAQ,CAAC;AACzI,YAAM,QAAQ,cAAc,MAAM,gBAAgB,WAAW,yDAAyD,CAAC;AACvH,YAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,GAAI,CAAC;AAAA,IAC9E;AAGA,UAAM,aAAa,GAAG,IAAI;AAC1B,UAAM,mBAAmB;AACzB,QAAI;AACF,YAAM,EAAE,SAAS,IAAI,IAAI,IAAI,UAAU;AACvC,YAAM,gBAAgB,eAAe;AACrC,YAAM,mBAAmB,iBAAiB,KAAK,QAAQ;AACvD,UAAI,iBAAiB,CAAC,kBAAkB;AACtC,eAAO,KAAK,EAAE,KAAK,WAAW,GAAG,gCAAgC;AACjE,eAAO;AAAA,MACT;AAAA,IACF,QAAQ;AAAA,IAAmC;AAE3C,UAAM,gBAAgB,MAAM,GAAG,QAAQ,wBAAwB,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AACtF,UAAM,WAAW,MAAM,GAAG,SAAS,MAAO,SAAiB,MAAM,WAAW,KAAK,GAAG,UAAU,CAAC,EAAE,MAAM,MAAM,CAAC;AAC9G,QAAI,kBAAkB,KAAK,WAAW,IAAI;AACxC,aAAO,KAAK,mDAAmD;AAC/D,aAAO;AAAA,IACT;AAEA,WAAO,KAAK,EAAE,KAAK,WAAW,GAAG,sBAAsB;AACvD,QAAI;AACF,YAAM,iBAAiB,0BAA0B,KAAK,IAAI,CAAC;AAC3D,YAAM,GAAG,WAAW,EAAE,MAAM,gBAAgB,MAAM,QAAQ,SAAS,IAAI,UAAU,MAAM,CAAC;AACxF,aAAO,KAAK,EAAE,eAAe,GAAG,yCAAyC;AAAA,IAC3E,QAAQ;AAAA,IAAkB;AAC1B,WAAO;AAAA,EACT,SAAS,KAAK;AACZ,UAAM,MAAM,OAAQ,IAAc,WAAW,GAAG;AAChD,QAAI,IAAI,SAAS,kBAAkB,EAAG,OAAM;AAC5C,WAAO,MAAM,EAAE,IAAI,GAAG,aAAa;AACnC,WAAO;AAAA,EACT;AACF;",
4
+ "sourcesContent": ["import { z } from 'zod';\nimport type { Stagehand } from '@browserbasehq/stagehand';\nimport { AppBrain } from '@detiq/app-brain';\nimport { solveCaptcha } from './captcha-solver.js';\nimport { waitForEmailContent } from './email-otp.js';\nimport { logger } from './logger.js';\nimport type { Credentials } from './auth.js';\nimport { simulateMouseMovement } from './stealth/human-behavior.js';\n\nconst CLICK_INTENT = /\\b(click|tap|press|submit|select|choose|toggle)\\b/i;\n\n/** Simulate human mouse movement before click-type actions, then call stagehand.act(). */\nasync function actWithMouseSim(stagehand: Stagehand, instruction: string): Promise<any> {\n if (CLICK_INTENT.test(instruction)) {\n try {\n const pg = await (stagehand as any).resolvePage();\n const viewport = pg.viewportSize() ?? { width: 1280, height: 720 };\n // Move mouse toward center of page with some randomness before Stagehand clicks\n const targetX = viewport.width * 0.3 + Math.random() * viewport.width * 0.4;\n const targetY = viewport.height * 0.2 + Math.random() * viewport.height * 0.5;\n await simulateMouseMovement(pg, targetX, targetY);\n } catch { /* non-fatal \u2014 don't block the action */ }\n }\n return stagehand.act(instruction);\n}\n\n/**\n * Recorded login step from the Chrome extension's step recorder.\n * {username} and {password} in `value` are substituted at replay time.\n */\nexport interface LoginStep {\n action: 'fill' | 'click' | 'wait';\n selector: string;\n value?: string;\n ms?: number;\n}\n\nconst TwoFASchema = z.object({\n has2FA: z.boolean(),\n fieldLabel: z.string().optional(),\n});\n\nconst CAPTCHA_SELECTORS = [\n 'iframe[src*=\"recaptcha\"]',\n 'iframe[src*=\"hcaptcha\"]',\n 'iframe[src*=\"challenges.cloudflare.com\"]',\n '.cf-turnstile',\n '[data-sitekey]',\n '#captcha',\n '.captcha',\n '[id*=\"captcha\"]',\n '[class*=\"captcha\"]',\n];\n\n/**\n * AI-guided login using Stagehand act/extract.\n *\n * @param loginInstructions - Optional natural language description of the login flow.\n * When provided, an initial stagehand.act() is called with these instructions to\n * navigate the UI (e.g. \"click Sign In in the nav, fill email, click Continue\").\n * The standard 3-step credential fill follows immediately after.\n * @param opts - Optional tenant/project/job IDs for writing AUTH_STEP CrawlEvidence.\n */\nexport async function loginWithStagehand(\n stagehand: Stagehand,\n creds: Credentials,\n startUrl: string,\n captchaOpts?: { apiKey?: string; provider?: '2captcha' | 'capsolver' },\n emailOtpOpts?: { apiKey?: string; inboxId?: string },\n loginInstructions?: string,\n opts?: { tenantId?: string; projectId?: string; jobId?: string },\n): Promise<boolean> {\n async function logStep(step: string, fn: () => Promise<any>): Promise<{ success: boolean; detail: string }> {\n const t0 = Date.now();\n try {\n const r = await fn();\n const result = { success: true, detail: r?.description ?? r?.message ?? step };\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step, ...result, ms: Date.now() - t0 },\n }).catch(() => {});\n }\n return result;\n } catch (e: any) {\n const result = { success: false, detail: e.message ?? 'failed' };\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step, ...result, ms: Date.now() - t0 },\n }).catch(() => {});\n }\n throw e;\n }\n }\n\n try {\n const pg = await (stagehand as any).resolvePage();\n\n // CAPTCHA detection\n for (const sel of CAPTCHA_SELECTORS) {\n const count = await pg.locator(sel).count().catch(() => 0);\n if (count > 0) {\n logger.warn({ selector: sel }, '[crawler] CAPTCHA detected on login page');\n if (captchaOpts?.apiKey) {\n logger.info({ provider: captchaOpts.provider ?? '2captcha' }, '[crawler] Attempting auto-solve');\n try {\n const result = await solveCaptcha(pg, startUrl, captchaOpts.apiKey, captchaOpts.provider ?? '2captcha');\n if (result) {\n logger.info({ provider: result.provider }, '[crawler] CAPTCHA solved \u2014 proceeding with login');\n break;\n }\n } catch (solveErr) {\n logger.warn({ err: solveErr }, '[crawler] CAPTCHA solve failed \u2014 falling back to manual error');\n throw new Error(`CAPTCHA_DETECTED: Auto-solve failed (${String(solveErr)}). Use session capture instead.`);\n }\n } else {\n throw new Error(\n 'CAPTCHA_DETECTED: Login page has a CAPTCHA. Use \"Session Capture \u2192 Launch Browser\" to log in manually, then capture the session. Alternatively, disable CAPTCHA for the ZeTa crawler IP.'\n );\n }\n break;\n }\n }\n\n // If caller provided natural language login instructions, use them to navigate to the login form\n const MAX_LOGIN_INSTRUCTIONS = 2000;\n if (loginInstructions && loginInstructions.length > MAX_LOGIN_INSTRUCTIONS) {\n throw new Error(`loginInstructions exceeds ${MAX_LOGIN_INSTRUCTIONS} character limit`);\n }\n if (loginInstructions) {\n await logStep('custom_recipe', () => actWithMouseSim(stagehand, loginInstructions));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(1500));\n }\n\n await logStep('fill_username', () => actWithMouseSim(stagehand, `Type \"${creds.username}\" into the username or email input field`));\n // Don't send password to LLM \u2014 use direct Playwright fill to avoid\n // leaking credentials into the Stagehand/LLM provider's request logs.\n await logStep('fill_password', async () => {\n const observations = await stagehand.observe('Find the password input field');\n const selector = observations[0]?.selector || 'input[type=\"password\"]';\n await pg.locator(selector).fill(creds.password);\n });\n await logStep('click_submit', () => actWithMouseSim(stagehand, 'Click the sign in, log in, login, or continue button to submit credentials'));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(2000));\n\n const twoFA = await stagehand.extract(\n 'Is there a two-factor authentication, OTP, or verification code input visible right now?',\n TwoFASchema as any\n );\n\n if (twoFA.has2FA) {\n if (opts?.tenantId && opts?.projectId) {\n AppBrain.saveCrawlEvidence(opts.tenantId, opts.projectId, {\n jobId: opts?.jobId,\n type: 'AUTH_STEP',\n payload: { step: '2fa_detect', success: true, detail: twoFA.fieldLabel ?? '2FA detected', ms: 0 },\n }).catch(() => {});\n }\n\n let code = creds.oneTimeCode ?? await creds.getOneTimeCode?.();\n\n if (!code && emailOtpOpts?.apiKey && emailOtpOpts?.inboxId) {\n const since = new Date().toISOString();\n logger.info('[crawler] 2FA detected \u2014 waiting for email OTP or magic link via MailSlurp');\n const emailContent = await waitForEmailContent(\n { provider: 'mailslurp', apiKey: emailOtpOpts.apiKey, inboxId: emailOtpOpts.inboxId },\n since\n );\n if (emailContent?.magicLink) {\n logger.info({ url: emailContent.magicLink }, '[crawler] Magic link received \u2014 navigating');\n await pg.goto(emailContent.magicLink, { waitUntil: 'domcontentloaded' });\n await pg.waitForLoadState('networkidle').catch(() => {});\n return true; // magic link navigation = login complete\n }\n code = emailContent?.otp ?? undefined;\n if (code) {\n logger.info('[crawler] Email OTP received');\n } else {\n logger.warn('[crawler] Email OTP timed out after 60s');\n }\n }\n\n if (!code) {\n console.warn('[crawler] 2FA detected but no oneTimeCode provided');\n return false;\n }\n await logStep('2fa_inject', () => actWithMouseSim(stagehand, `Enter \"${code}\" into the ${twoFA.fieldLabel ?? 'verification code'} field`));\n await logStep('2fa_submit', () => actWithMouseSim(stagehand, 'Click verify or confirm to submit the verification code'));\n await pg.waitForLoadState('networkidle').catch(() => pg.waitForTimeout(2000));\n }\n\n // URL-based success check\n const currentUrl = pg.url();\n const loginPathPattern = /\\/(login|signin|sign-in|auth|authenticate|log-in|account\\/login)(\\?|$|\\/)/i;\n try {\n const { pathname } = new URL(currentUrl);\n const wasRedirected = currentUrl !== startUrl;\n const stillOnLoginPath = loginPathPattern.test(pathname);\n if (wasRedirected && !stillOnLoginPath) {\n logger.info({ url: currentUrl }, 'Login confirmed via URL change');\n return true;\n }\n } catch { /* invalid URL \u2014 fall through */ }\n\n const passwordCount = await pg.locator('input[type=\"password\"]').count().catch(() => 1);\n const bodyText = await pg.evaluate(() => (document as any).body?.innerText?.trim()?.length ?? 0).catch(() => 0);\n if (passwordCount === 0 && bodyText > 30) {\n logger.info('Login confirmed: password field no longer visible');\n return true;\n }\n\n logger.warn({ url: currentUrl }, 'Login appears failed');\n try {\n const screenshotPath = `/tmp/zeta-auth-failure-${Date.now()}.jpg`;\n await pg.screenshot({ path: screenshotPath, type: 'jpeg', quality: 70, fullPage: false });\n logger.warn({ screenshotPath }, '[crawler] Auth failure screenshot saved');\n } catch { /* non-fatal */ }\n return false;\n } catch (err) {\n const msg = String((err as Error).message ?? err);\n if (msg.includes('CAPTCHA_DETECTED')) throw err;\n logger.error({ err }, 'Login error');\n return false;\n }\n}\n"],
5
+ "mappings": "AAAA,SAAS,SAAS;AAElB,SAAS,gBAAgB;AACzB,SAAS,oBAAoB;AAC7B,SAAS,2BAA2B;AACpC,SAAS,cAAc;AAEvB,SAAS,6BAA6B;AAEtC,MAAM,eAAe;AAGrB,eAAe,gBAAgB,WAAsB,aAAmC;AACtF,MAAI,aAAa,KAAK,WAAW,GAAG;AAClC,QAAI;AACF,YAAM,KAAK,MAAO,UAAkB,YAAY;AAChD,YAAM,WAAW,GAAG,aAAa,KAAK,EAAE,OAAO,MAAM,QAAQ,IAAI;AAEjE,YAAM,UAAU,SAAS,QAAQ,MAAM,KAAK,OAAO,IAAI,SAAS,QAAQ;AACxE,YAAM,UAAU,SAAS,SAAS,MAAM,KAAK,OAAO,IAAI,SAAS,SAAS;AAC1E,YAAM,sBAAsB,IAAI,SAAS,OAAO;AAAA,IAClD,QAAQ;AAAA,IAA2C;AAAA,EACrD;AACA,SAAO,UAAU,IAAI,WAAW;AAClC;AAaA,MAAM,cAAc,EAAE,OAAO;AAAA,EAC3B,QAAQ,EAAE,QAAQ;AAAA,EAClB,YAAY,EAAE,OAAO,EAAE,SAAS;AAClC,CAAC;AAED,MAAM,oBAAoB;AAAA,EACxB;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAWA,eAAsB,mBACpB,WACA,OACA,UACA,aACA,cACA,mBACA,MACkB;AAClB,iBAAe,QAAQ,MAAc,IAAuE;AAC1G,UAAM,KAAK,KAAK,IAAI;AACpB,QAAI;AACF,YAAM,IAAI,MAAM,GAAG;AACnB,YAAM,SAAS,EAAE,SAAS,MAAM,QAAQ,GAAG,eAAe,GAAG,WAAW,KAAK;AAC7E,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,GAAG,QAAQ,IAAI,KAAK,IAAI,IAAI,GAAG;AAAA,QAClD,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AACA,aAAO;AAAA,IACT,SAAS,GAAQ;AACf,YAAM,SAAS,EAAE,SAAS,OAAO,QAAQ,EAAE,WAAW,SAAS;AAC/D,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,GAAG,QAAQ,IAAI,KAAK,IAAI,IAAI,GAAG;AAAA,QAClD,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AACA,YAAM;AAAA,IACR;AAAA,EACF;AAEA,MAAI;AACF,UAAM,KAAK,MAAO,UAAkB,YAAY;AAGhD,eAAW,OAAO,mBAAmB;AACnC,YAAM,QAAQ,MAAM,GAAG,QAAQ,GAAG,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AACzD,UAAI,QAAQ,GAAG;AACb,eAAO,KAAK,EAAE,UAAU,IAAI,GAAG,0CAA0C;AACzE,YAAI,aAAa,QAAQ;AACvB,iBAAO,KAAK,EAAE,UAAU,YAAY,YAAY,WAAW,GAAG,iCAAiC;AAC/F,cAAI;AACF,kBAAM,SAAS,MAAM,aAAa,IAAI,UAAU,YAAY,QAAQ,YAAY,YAAY,UAAU;AACtG,gBAAI,QAAQ;AACV,qBAAO,KAAK,EAAE,UAAU,OAAO,SAAS,GAAG,uDAAkD;AAC7F;AAAA,YACF;AAAA,UACF,SAAS,UAAU;AACjB,mBAAO,KAAK,EAAE,KAAK,SAAS,GAAG,oEAA+D;AAC9F,kBAAM,IAAI,MAAM,wCAAwC,OAAO,QAAQ,CAAC,iCAAiC;AAAA,UAC3G;AAAA,QACF,OAAO;AACL,gBAAM,IAAI;AAAA,YACR;AAAA,UACF;AAAA,QACF;AACA;AAAA,MACF;AAAA,IACF;AAGA,UAAM,yBAAyB;AAC/B,QAAI,qBAAqB,kBAAkB,SAAS,wBAAwB;AAC1E,YAAM,IAAI,MAAM,6BAA6B,sBAAsB,kBAAkB;AAAA,IACvF;AACA,QAAI,mBAAmB;AACrB,YAAM,QAAQ,iBAAiB,MAAM,gBAAgB,WAAW,iBAAiB,CAAC;AAClF,YAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,IAAI,CAAC;AAAA,IAC9E;AAEA,UAAM,QAAQ,iBAAiB,MAAM,gBAAgB,WAAW,SAAS,MAAM,QAAQ,0CAA0C,CAAC;AAGlI,UAAM,QAAQ,iBAAiB,YAAY;AACzC,YAAM,eAAe,MAAM,UAAU,QAAQ,+BAA+B;AAC5E,YAAM,WAAW,aAAa,CAAC,GAAG,YAAY;AAC9C,YAAM,GAAG,QAAQ,QAAQ,EAAE,KAAK,MAAM,QAAQ;AAAA,IAChD,CAAC;AACD,UAAM,QAAQ,gBAAgB,MAAM,gBAAgB,WAAW,4EAA4E,CAAC;AAC5I,UAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,GAAI,CAAC;AAE5E,UAAM,QAAQ,MAAM,UAAU;AAAA,MAC5B;AAAA,MACA;AAAA,IACF;AAEA,QAAI,MAAM,QAAQ;AAChB,UAAI,MAAM,YAAY,MAAM,WAAW;AACrC,iBAAS,kBAAkB,KAAK,UAAU,KAAK,WAAW;AAAA,UACxD,OAAO,MAAM;AAAA,UACb,MAAM;AAAA,UACN,SAAS,EAAE,MAAM,cAAc,SAAS,MAAM,QAAQ,MAAM,cAAc,gBAAgB,IAAI,EAAE;AAAA,QAClG,CAAC,EAAE,MAAM,MAAM;AAAA,QAAC,CAAC;AAAA,MACnB;AAEA,UAAI,OAAO,MAAM,eAAe,MAAM,MAAM,iBAAiB;AAE7D,UAAI,CAAC,QAAQ,cAAc,UAAU,cAAc,SAAS;AAC1D,cAAM,SAAQ,oBAAI,KAAK,GAAE,YAAY;AACrC,eAAO,KAAK,iFAA4E;AACxF,cAAM,eAAe,MAAM;AAAA,UACzB,EAAE,UAAU,aAAa,QAAQ,aAAa,QAAQ,SAAS,aAAa,QAAQ;AAAA,UACpF;AAAA,QACF;AACA,YAAI,cAAc,WAAW;AAC3B,iBAAO,KAAK,EAAE,KAAK,aAAa,UAAU,GAAG,iDAA4C;AACzF,gBAAM,GAAG,KAAK,aAAa,WAAW,EAAE,WAAW,mBAAmB,CAAC;AACvE,gBAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM;AAAA,UAAC,CAAC;AACvD,iBAAO;AAAA,QACT;AACA,eAAO,cAAc,OAAO;AAC5B,YAAI,MAAM;AACR,iBAAO,KAAK,8BAA8B;AAAA,QAC5C,OAAO;AACL,iBAAO,KAAK,yCAAyC;AAAA,QACvD;AAAA,MACF;AAEA,UAAI,CAAC,MAAM;AACT,gBAAQ,KAAK,oDAAoD;AACjE,eAAO;AAAA,MACT;AACA,YAAM,QAAQ,cAAc,MAAM,gBAAgB,WAAW,UAAU,IAAI,cAAc,MAAM,cAAc,mBAAmB,QAAQ,CAAC;AACzI,YAAM,QAAQ,cAAc,MAAM,gBAAgB,WAAW,yDAAyD,CAAC;AACvH,YAAM,GAAG,iBAAiB,aAAa,EAAE,MAAM,MAAM,GAAG,eAAe,GAAI,CAAC;AAAA,IAC9E;AAGA,UAAM,aAAa,GAAG,IAAI;AAC1B,UAAM,mBAAmB;AACzB,QAAI;AACF,YAAM,EAAE,SAAS,IAAI,IAAI,IAAI,UAAU;AACvC,YAAM,gBAAgB,eAAe;AACrC,YAAM,mBAAmB,iBAAiB,KAAK,QAAQ;AACvD,UAAI,iBAAiB,CAAC,kBAAkB;AACtC,eAAO,KAAK,EAAE,KAAK,WAAW,GAAG,gCAAgC;AACjE,eAAO;AAAA,MACT;AAAA,IACF,QAAQ;AAAA,IAAmC;AAE3C,UAAM,gBAAgB,MAAM,GAAG,QAAQ,wBAAwB,EAAE,MAAM,EAAE,MAAM,MAAM,CAAC;AACtF,UAAM,WAAW,MAAM,GAAG,SAAS,MAAO,SAAiB,MAAM,WAAW,KAAK,GAAG,UAAU,CAAC,EAAE,MAAM,MAAM,CAAC;AAC9G,QAAI,kBAAkB,KAAK,WAAW,IAAI;AACxC,aAAO,KAAK,mDAAmD;AAC/D,aAAO;AAAA,IACT;AAEA,WAAO,KAAK,EAAE,KAAK,WAAW,GAAG,sBAAsB;AACvD,QAAI;AACF,YAAM,iBAAiB,0BAA0B,KAAK,IAAI,CAAC;AAC3D,YAAM,GAAG,WAAW,EAAE,MAAM,gBAAgB,MAAM,QAAQ,SAAS,IAAI,UAAU,MAAM,CAAC;AACxF,aAAO,KAAK,EAAE,eAAe,GAAG,yCAAyC;AAAA,IAC3E,QAAQ;AAAA,IAAkB;AAC1B,WAAO;AAAA,EACT,SAAS,KAAK;AACZ,UAAM,MAAM,OAAQ,IAAc,WAAW,GAAG;AAChD,QAAI,IAAI,SAAS,kBAAkB,EAAG,OAAM;AAC5C,WAAO,MAAM,EAAE,IAAI,GAAG,aAAa;AACnC,WAAO;AAAA,EACT;AACF;",
6
6
  "names": []
7
7
  }
package/dist/crawler.js CHANGED
@@ -9,9 +9,9 @@ stealthChromium.use(StealthPlugin());
9
9
  import { z } from "zod";
10
10
  import fs from "node:fs/promises";
11
11
  import path from "node:path";
12
- import { AppBrain, recordLLMCall } from "@qip/app-brain";
13
- import { aiVisualCompare, askTracked, estimateCost, getTaskAgentLLMConfig } from "@qip/agents";
14
- import { config } from "@qip/core";
12
+ import { AppBrain, recordLLMCall } from "@detiq/app-brain";
13
+ import { aiVisualCompare, askTracked, estimateCost, getTaskAgentLLMConfig } from "@detiq/agents";
14
+ import { config } from "@detiq/core";
15
15
  import { pHashVerdict, mergeAiVisualVerdict } from "./visual-diff-verdict.js";
16
16
  import { skyvernFallback, skyvernNavigate } from "./skyvern.js";
17
17
  import { loginWithStagehand } from "./ai-login.js";
@@ -1002,6 +1002,7 @@ async function crawlProject(opts, onProgress) {
1002
1002
  }
1003
1003
  let checkpointSuccessCount = 0;
1004
1004
  let maxSeenTotal = 0;
1005
+ let lastEmittedPct = 20;
1005
1006
  const startUrl = page.url();
1006
1007
  const resumeUrls = opts.resumeUrls ?? [];
1007
1008
  const gqlIntrospectionPromise = (async () => {
@@ -1045,8 +1046,9 @@ async function crawlProject(opts, onProgress) {
1045
1046
  const u = s.url;
1046
1047
  if (!u) return false;
1047
1048
  try {
1048
- const { origin: origin2, pathname } = new URL(u);
1049
- if (origin2 !== new URL(startUrl).origin) return false;
1049
+ const { pathname } = new URL(u);
1050
+ const stripWww = (h) => h.replace(/^www\./i, "").toLowerCase();
1051
+ if (stripWww(new URL(u).hostname) !== stripWww(new URL(startUrl).hostname)) return false;
1050
1052
  if (isGeneratedDynamicPath(pathname)) return false;
1051
1053
  return true;
1052
1054
  } catch {
@@ -1937,7 +1939,7 @@ ${snippet}`,
1937
1939
  if (changeType !== "NONE") {
1938
1940
  let newScreenshotCloudUrl;
1939
1941
  try {
1940
- const { getTenantStorageConfig, uploadScreenshotToTenantStorage, buildArtifactKey } = await import("@qip/app-brain");
1942
+ const { getTenantStorageConfig, uploadScreenshotToTenantStorage, buildArtifactKey } = await import("@detiq/app-brain");
1941
1943
  const storageCfg = await getTenantStorageConfig(opts.tenantId).catch(() => null);
1942
1944
  if (storageCfg) {
1943
1945
  const key = buildArtifactKey(opts.tenantId, opts.projectId, "visual-diffs", `${screen.id}-v${screen.version}.png`, storageCfg.prefix);
@@ -2157,8 +2159,10 @@ ${snippet}`,
2157
2159
  }
2158
2160
  discovered.push({ url, screenId: screen.id, elements: elements.length, elementFacts: elements, screenshotPath: shotPath, version: screen.version });
2159
2161
  maxSeenTotal = Math.max(maxSeenTotal, visited.size + queue.length);
2160
- const screenPct = Math.min(60, visited.size / Math.max(1, maxSeenTotal) * 60);
2161
- await onProgress?.(20 + Math.round(screenPct), `Crawling screens (${visited.size} mapped, ${queue.length} pending)`);
2162
+ const rawPct = 20 + Math.round(Math.min(60, visited.size / Math.max(1, maxSeenTotal) * 60));
2163
+ const clampedPct = Math.max(lastEmittedPct, rawPct);
2164
+ lastEmittedPct = clampedPct;
2165
+ await onProgress?.(clampedPct, `Crawling screens (${visited.size} mapped, ${queue.length} pending)`);
2162
2166
  logger.info({ name, elements: elements.length }, "Screen crawled");
2163
2167
  if (++checkpointSuccessCount % 10 === 0) {
2164
2168
  await opts.onCheckpoint?.(