autokap 1.3.28 → 1.3.30

package/dist/browser.d.ts

@@ -99,6 +99,8 @@ export declare class Browser {
     private elementMap;
     private akNodeIndex;
     private poolContext;
+    private persistentContext;
+    private ownedChromiumProfileDir;
     /**
      * Xvfb instance backing the headed Chromium used by clip capture on Cloud
      * Run with NVIDIA L4. Set when forClipCapture spawns Xvfb; null otherwise
@@ -134,10 +136,10 @@ export declare class Browser {
      */
     static forClipCapture(options: BrowserOptions, cursorScript: string): Promise<Browser>;
     /**
-     * Close only the browser context (not the browser process).
-     * Used by clip capture to release the context promptly after the CDP loop
-     * has stopped, while keeping the browser process alive for any pending teardown.
-     * Call browser.close() afterwards to shut down the browser process.
+     * Close the active capture context promptly after recording stops. Regular
+     * clip capture leaves the browser process alive for teardown; persistent
+     * Cloud Run contexts close their owned browser process with the context.
+     * Call browser.close() afterwards to stop Xvfb and remove owned temp state.
      */
     closeContext(): Promise<void>;
     launch(): Promise<void>;

package/dist/browser.js

@@ -1,6 +1,8 @@
 import { chromium } from 'playwright';
 import sharp from 'sharp';
 import { createHash } from 'crypto';
+import { cp, mkdir, readFile, rm, writeFile } from 'fs/promises';
+import { join } from 'path';
 import { buildAKNodeRuntimeIndex, deriveInteractiveElementsFromAKTree, disambiguateFingerprint, focusAKTree, fingerprintAKNode, serializeAKTree, } from './ak-tree.js';
 /**
  * Set-of-Marks (SoM) annotation: overlays colored [N] badges on each visible
@@ -100,6 +102,110 @@ import { CAPTURE_HIDE_STYLE_ID, dismissCookiesAndWidgets, ensureCaptureHideStyle
 import { CHROMIUM_ARGS, browserPool } from './browser-pool.js';
 import { isDebugEnabled, logger } from './logger.js';
 import { XvfbProcess } from './xvfb-process.js';
+const CLOUD_CHROMIUM_PROFILE_TEMPLATE_DIR = '/opt/chromium-profile';
+const CLOUD_CHROMIUM_PROFILE_ROOT = '/tmp/autokap-chromium-profiles';
+const CLOUD_CHROMIUM_TRANSLATE_BLOCKED_LANGUAGES = [
+    'fr',
+    'en',
+    'de',
+    'es',
+    'it',
+    'pt',
+    'nl',
+    'ja',
+    'zh',
+    'ko',
+    'ar',
+    'ru',
+];
+const CLOUD_CHROMIUM_ACCEPT_LANGUAGES = [
+    'fr-FR',
+    'fr',
+    'en-US',
+    'en',
+    'en-GB',
+    'es-ES',
+    'es',
+    'de-DE',
+    'de',
+    'it-IT',
+    'it',
+    'pt-PT',
+    'pt-BR',
+    'pt',
+    'nl-NL',
+    'nl',
+    'ja-JP',
+    'ja',
+    'zh-CN',
+    'zh-TW',
+    'zh',
+    'ko-KR',
+    'ko',
+    'ar',
+    'ru-RU',
+    'ru',
+    'he',
+    'th',
+    'tr',
+    'vi',
+    'sv',
+    'no',
+    'da',
+    'fi',
+    'pl',
+    'cs',
+];
+function isPlainRecord(value) {
+    return typeof value === 'object' && value !== null && !Array.isArray(value);
+}
+async function prepareCloudChromiumProfile() {
+    const rootDir = process.env.AUTOKAP_CHROMIUM_PROFILE_ROOT || CLOUD_CHROMIUM_PROFILE_ROOT;
+    const templateDir = process.env.AUTOKAP_CHROMIUM_PROFILE_TEMPLATE_DIR || CLOUD_CHROMIUM_PROFILE_TEMPLATE_DIR;
+    const userDataDir = join(rootDir, `${process.pid}-${Date.now()}-${Math.random().toString(36).slice(2, 8)}`);
+    await mkdir(rootDir, { recursive: true });
+    try {
+        await cp(templateDir, userDataDir, { recursive: true, force: true });
+    }
+    catch (err) {
+        if (err.code !== 'ENOENT') {
+            throw err;
+        }
+        await mkdir(userDataDir, { recursive: true });
+    }
+    await seedCloudChromiumPreferences(userDataDir);
+    return userDataDir;
+}
+async function seedCloudChromiumPreferences(userDataDir) {
+    const defaultDir = join(userDataDir, 'Default');
+    const preferencesPath = join(defaultDir, 'Preferences');
+    await mkdir(defaultDir, { recursive: true });
+    let preferences = {};
+    try {
+        const raw = await readFile(preferencesPath, 'utf8');
+        const parsed = JSON.parse(raw);
+        if (isPlainRecord(parsed)) {
+            preferences = parsed;
+        }
+    }
+    catch (err) {
+        if (err.code !== 'ENOENT') {
+            logger.warn(`[capture] Cloud clip capture: replacing unreadable Chromium Preferences: ${err.message}`);
+        }
+    }
+    const translate = isPlainRecord(preferences.translate) ? preferences.translate : {};
+    preferences.translate = {
+        ...translate,
+        enabled: false,
+    };
+    preferences.translate_blocked_languages = CLOUD_CHROMIUM_TRANSLATE_BLOCKED_LANGUAGES;
+    const intl = isPlainRecord(preferences.intl) ? preferences.intl : {};
+    preferences.intl = {
+        ...intl,
+        accept_languages: CLOUD_CHROMIUM_ACCEPT_LANGUAGES.join(','),
+    };
+    await writeFile(preferencesPath, `${JSON.stringify(preferences)}\n`, 'utf8');
+}
 async function withHelperTimeout(label, timeoutMs, work) {
     if (!timeoutMs || timeoutMs <= 0) {
         return work();
@@ -776,6 +882,8 @@ export class Browser {
     elementMap = new Map();
     akNodeIndex = new Map();
     poolContext = false;
+    persistentContext = false;
+    ownedChromiumProfileDir = null;
     /**
      * Xvfb instance backing the headed Chromium used by clip capture on Cloud
      * Run with NVIDIA L4. Set when forClipCapture spawns Xvfb; null otherwise
@@ -883,37 +991,27 @@ export class Browser {
             process.env.DISPLAY = instance.xvfb.display;
             logger.info(`[capture] Cloud clip capture: Chromium → Xvfb ${instance.xvfb.display} → ffmpeg x11grab + h264_nvenc path enabled`);
         }
+        const cloudChromiumProfileDir = isLinuxWithGpu ? await prepareCloudChromiumProfile() : null;
+        instance.ownedChromiumProfileDir = cloudChromiumProfileDir;
         // Kiosk + zero-position anchor for Xvfb: Chromium normally renders its
         // own toolbar/tabbar in headed mode, which would appear at the top of
         // every clip. `--kiosk` removes the address bar + tab strip;
         // `--window-position=0,0` and `--window-size` make the page fill the
-        // Xvfb screen exactly. The `--disable-features` block kills the
-        // separate "infobar" surfaces (translate suggestion, save-password
-        // prompt, autofill banner, "Chrome is being controlled by automated
-        // software" warning) — these render OUTSIDE kiosk's chrome and would
-        // otherwise show up at the top of every clip captured via x11grab.
-        // CDP screenshot capture (Mac/Win/local Linux) hits the page surface
-        // directly so it never sees these; only ffmpeg x11grab does.
+        // Xvfb screen exactly. Chrome UI surfaces render OUTSIDE kiosk's chrome
+        // and would otherwise show up in clips captured via x11grab. Translate is
+        // controlled by the seeded Chromium preferences below; avoid passing our
+        // own --disable-features list here because it replaces Playwright's
+        // default disabled-feature map instead of merging with it.
         const xvfbWindowArgs = isLinuxWithGpu ? [
             '--kiosk',
             '--window-position=0,0',
-            '--disable-features=Translate,TranslateUI,AutofillServerCommunication,InfoBars',
-            '--disable-infobars',
             '--disable-blink-features=AutomationControlled',
-            '--disable-translate',
             '--no-default-browser-check',
             '--no-first-run',
             '--noerrdialogs',
-            // Chrome's translate prompt only appears when the page language is NOT
-            // in the user's preferred-languages list. By listing every language we
-            // might capture (covers our marketing demos: fr, en, es, de, it, pt, nl,
-            // ja, zh, ko, ar, ru, plus regional variants), Chrome treats every page
-            // as "user already speaks this" and never offers translation. Belt-and-
-            // suspenders alongside the Dockerfile policy file (which Playwright's
-            // bundled Chromium might not honor per issue #32324) and the
-            // notranslate meta init script. accept-lang affects HTTP requests AND
-            // Chrome's translate decision; it's the only Chrome-side knob that
-            // works regardless of policy path detection or feature flag drift.
+            // Belt-and-suspenders with the seeded translate prefs: keep the browser's
+            // preferred-language list broad enough that Chrome has no reason to offer
+            // translation on the marketing/demo languages we capture most often.
             '--accept-lang=fr-FR,fr,en-US,en,en-GB,es-ES,es,de-DE,de,it-IT,it,pt-PT,pt-BR,pt,nl-NL,nl,ja-JP,ja,zh-CN,zh-TW,zh,ko-KR,ko,ar,ru-RU,ru,he,th,tr,vi,sv,no,da,fi,pl,cs',
         ] : [];
         const clipArgs = [
@@ -924,14 +1022,6 @@ export class Browser {
             ...cloudGpuArgs,
             ...xvfbWindowArgs,
         ];
-        // Dedicated browser process for clip capture. Not pooled because clip
-        // capture installs context-level init scripts (cursor overlay).
-        // Cloud Run with Xvfb: launch headed (headless: false) so Chromium
-        // renders to the Xvfb framebuffer that ffmpeg captures.
-        instance.browser = await chromium.launch({
-            headless: isLinuxWithGpu ? false : !options.headed,
-            args: clipArgs,
-        });
         const contextOptions = {
             viewport: options.viewport,
             deviceScaleFactor,
@@ -939,7 +1029,43 @@ export class Browser {
             colorScheme: options.colorScheme ?? 'light',
             storageState: options.storageState,
         };
-        instance.context = await instance.browser.newContext(contextOptions);
+        // Dedicated browser process for clip capture. Not pooled because clip
+        // capture installs context-level init scripts (cursor overlay). Cloud Run
+        // uses a seeded persistent profile so Chromium reads translate.enabled=false
+        // from Default/Preferences. Policy files and flags are not reliable against
+        // the Chromium 127+ TranslateUI2024 bubble; the user-data-dir pref is.
+        if (isLinuxWithGpu && cloudChromiumProfileDir) {
+            instance.context = await chromium.launchPersistentContext(cloudChromiumProfileDir, {
+                ...contextOptions,
+                headless: false,
+                args: clipArgs,
+            });
+            instance.browser = instance.context.browser();
+            instance.persistentContext = true;
+            // launchPersistentContext opens an initial about:blank page automatically.
+            // Reuse it instead of closing it + opening a new one: Chromium's --kiosk
+            // mode blocks `Target.createTarget` (one-window-only enforcement), so
+            // `context.newPage()` would fail with "Failed to open a new tab" — which
+            // is exactly what bricked 1.3.29. The init scripts registered below run
+            // on the next navigation regardless of when the page was created.
+            const existingPages = instance.context.pages();
+            instance.page = existingPages[0] ?? await instance.context.newPage();
+            // Defensive: close any extra pages Chromium might have opened (unlikely
+            // under --kiosk but cheap insurance). Keep index 0.
+            for (let i = 1; i < existingPages.length; i++) {
+                await existingPages[i].close().catch(() => undefined);
+            }
+            logger.info(`[capture] Cloud clip capture: persistent Chromium profile seeded at ${cloudChromiumProfileDir} (reusing initial page, kiosk blocks Target.createTarget)`);
+        }
+        else {
+            // Non-cloud clip capture keeps the regular browser + incognito context
+            // model used by local/macOS/Windows frame capture.
+            instance.browser = await chromium.launch({
+                headless: !options.headed,
+                args: clipArgs,
+            });
+            instance.context = await instance.browser.newContext(contextOptions);
+        }
         // Cloud Run only: inject the notranslate meta on every navigation so
         // Chromium's translate UI never prompts. The --disable-features=Translate*
         // launch flags are unreliable across Chromium versions (some translate
@@ -982,7 +1108,12 @@ export class Browser {
                 document.addEventListener('DOMContentLoaded', install, { once: true });
             }
         }, { styleId: CAPTURE_HIDE_STYLE_ID, css: getCaptureHideCSS() });
-        instance.page = await instance.context.newPage();
+        // Persistent context (cloud) already set instance.page by reusing the
+        // initial page launchPersistentContext opened. Only call newPage() for
+        // the regular browser+context path (Mac/Win/local Linux).
+        if (!instance.page) {
+            instance.page = await instance.context.newPage();
+        }
         // Cloud Run only: force the page window to fullscreen via CDP. The
         // --kiosk launch flag is silently ignored by Chromium for windows
         // created by Playwright via Target.createTarget (CDP) — kiosk only
@@ -1041,10 +1172,10 @@ export class Browser {
         return instance;
     }
     /**
-     * Close only the browser context (not the browser process).
-     * Used by clip capture to release the context promptly after the CDP loop
-     * has stopped, while keeping the browser process alive for any pending teardown.
-     * Call browser.close() afterwards to shut down the browser process.
+     * Close the active capture context promptly after recording stops. Regular
+     * clip capture leaves the browser process alive for teardown; persistent
+     * Cloud Run contexts close their owned browser process with the context.
+     * Call browser.close() afterwards to stop Xvfb and remove owned temp state.
      */
     async closeContext() {
         if (this.context) {
@@ -1053,6 +1184,10 @@ export class Browser {
             }
             catch { /* ignore */ }
             this.context = null;
+            this.page = null;
+            if (this.persistentContext) {
+                this.browser = null;
+            }
         }
     }
     async launch() {
@@ -1201,13 +1336,32 @@ export class Browser {
             this.poolContext = false;
             return;
         }
-        // Standalone mode (CLI): close the entire browser process
-        if (this.browser) {
+        // Standalone mode (CLI): close the entire browser process. Persistent
+        // contexts own their browser process, so closing the context is enough.
+        if (this.persistentContext && this.context) {
+            try {
+                await this.context.close();
+            }
+            catch { /* ignore */ }
+            this.context = null;
+            this.browser = null;
+            this.page = null;
+        }
+        else if (this.browser) {
             await this.browser.close();
             this.browser = null;
             this.context = null;
             this.page = null;
         }
+        else if (this.context) {
+            try {
+                await this.context.close();
+            }
+            catch { /* ignore */ }
+            this.context = null;
+            this.page = null;
+        }
+        this.persistentContext = false;
         // Tear down Xvfb only after Chromium is fully gone — Chromium needs the
         // X display for its own teardown (releasing GL contexts, X resources).
         if (this.xvfb) {
@@ -1219,6 +1373,15 @@ export class Browser {
             }
             this.xvfb = null;
         }
+        if (this.ownedChromiumProfileDir) {
+            try {
+                await rm(this.ownedChromiumProfileDir, { recursive: true, force: true });
+            }
+            catch (err) {
+                logger.warn(`[capture] Cloud clip capture: failed to remove Chromium profile ${this.ownedChromiumProfileDir}: ${err.message}`);
+            }
+            this.ownedChromiumProfileDir = null;
+        }
     }
     async navigateTo(url) {
         const page = this.ensurePage();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "autokap",
-  "version": "1.3.28",
+  "version": "1.3.30",
   "description": "AI-powered CLI tool for capturing clean screenshots of websites",
   "type": "module",
   "main": "./dist/index.js",