Merge pull request 'Add Node/TS automation twin sharing the extension core' (#1) from feat/automation-js into master

.gitea/workflows/ci.yml

@@ -56,6 +56,10 @@ jobs:
           test -L automation/context_extractor/js/prompt.js
           diff -q automation/context_extractor/js/dom.js extension/core/dom.js
           diff -q automation/context_extractor/js/prompt.js extension/core/prompt.js
+          test -L automation-js/src/js/dom.js
+          test -L automation-js/src/js/prompt.js
+          diff -q automation-js/src/js/dom.js extension/core/dom.js
+          diff -q automation-js/src/js/prompt.js extension/core/prompt.js
 
       - name: Install package + dev deps
         run: |
@@ -73,3 +77,34 @@ jobs:
         run: |
           python3 scripts/package_extension.py
           test -f dist/context-extractor-extension.zip
+
+  quality-js:
+    name: automation-js (vitest + real Chromium)
+    runs-on: ubuntu-latest
+    defaults:
+      run:
+        working-directory: automation-js
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: "20"
+
+      - name: Install deps
+        run: npm install
+
+      - name: Install Playwright Chromium
+        run: npx playwright install --with-deps chromium
+
+      - name: Typecheck
+        run: npm run typecheck
+
+      - name: Vitest
+        run: npm test
+
+      - name: Build (verifies dist/js copy step)
+        run: npm run build

Makefile

@@ -1,4 +1,4 @@
-.PHONY: help test lint package ci install
+.PHONY: help test test-js lint package ci install install-js
 
 help: ## Show targets
 	@grep -E '^[a-zA-Z_-]+:.*?## ' $(MAKEFILE_LIST) | awk 'BEGIN {FS = ":.*?## "}; {printf "  %-12s %s\n", $$1, $$2}'
@@ -8,9 +8,15 @@ install: ## Create venv and install automation package + Playwright Chromium
 	cd automation && .venv/bin/pip install -U pip && .venv/bin/pip install -e ".[dev]"
 	cd automation && .venv/bin/playwright install chromium
 
+install-js: ## npm install automation-js (uses the Chromium cache npx playwright already has)
+	cd automation-js && npm install
+
 test: ## Run pytest
 	cd automation && .venv/bin/pytest -q
 
+test-js: ## Run automation-js vitest suite (real headless Chromium)
+	cd automation-js && npm test
+
 lint: ## Syntax-check extension JS + validate manifest JSON
 	@node --check extension/core/dom.js
 	@node --check extension/core/prompt.js
@@ -24,4 +30,4 @@ lint: ## Syntax-check extension JS + validate manifest JSON
 package: ## Zip extension/ into dist/
 	python3 scripts/package_extension.py
 
-ci: lint test package ## Local stand-in for GitHub Actions
+ci: lint test test-js package ## Local stand-in for GitHub Actions

README.md

@@ -4,12 +4,16 @@
 **Clone:** `gitea@git.levkin.ca:ilia/context-extractor.git`
 
 Capture console logs, network activity, JS errors, and clean markdown content
-from a webpage, formatted as an AI-ready prompt. Ships two ways from one
+from a webpage, formatted as an AI-ready prompt. Ships three ways from one
 shared core:
 
 - **`extension/`** — a browser extension (Manifest V3) for interactive use.
 - **`automation/`** — a Python package for scripted/headless use with
   Playwright or [Camoufox](https://camoufox.com).
+- **`automation-js/`** — a Node/TS twin of `automation/` for Playwright
+  (`playwright-core`) or [camoufox-js](https://www.npmjs.com/package/camoufox-js)
+  automation that's already JS, so it can `import` the session directly
+  instead of shelling out to Python.
 
 ```
 context-extractor/
@@ -24,6 +28,10 @@ context-extractor/
 │   ├── context_extractor/       session + CLI
 │   │   └── js/                  symlinks → extension/core/*.js
 │   └── tests/                   pytest + fixture page
+├── automation-js/                Node/TS package for Playwright / camoufox-js
+│   ├── src/session.ts           ← ExtractorSession (async, one class)
+│   │   └── js/                  symlinks → extension/core/*.js
+│   └── tests/                   vitest + real Chromium, shared fixture
 ├── scripts/package_extension.py optional zip for distribution
 ├── .gitea/workflows/ci.yml      Gitea Actions: JS lint + pytest + package smoke
 ├── CURSOR_PROMPT.md             paste into other Cursor chats
@@ -32,9 +40,11 @@ context-extractor/
 
 `extension/core/dom.js` and `extension/core/prompt.js` are the single source
 of truth for DOM/markdown/prompt logic — the extension loads them directly as
-content-script files, and the Python package reads the exact same files
-(via symlinks) and runs them with `page.evaluate()`. There's only one place
-to fix a markdown-formatting bug.
+content-script files, and **both** automation packages read the exact same
+files (via symlinks) and run them with `page.evaluate()`. There's only one
+place to fix a markdown-formatting bug, and the Python/JS test suites both
+run against the same fixture page (`automation/tests/fixtures/sample.html`)
+so a regression can't silently diverge between the two.
 
 ## Use from another Cursor project
 
@@ -126,6 +136,27 @@ Headless CLI, handy in CI:
 context-extractor https://example.com --selector "#main" --engine camoufox --out prompt.md
 ```
 
+### JS/TS twin (`automation-js/`) — for Node/Playwright or camoufox-js automation
+
+Same idea, no Python subprocess:
+
+```bash
+cd automation-js
+npm install && npm run build
+```
+
+```ts
+import { chromium } from 'playwright-core';
+import { ExtractorSession } from 'context-extractor'; // automation-js
+
+const page = await (await chromium.launch()).newPage();
+const session = new ExtractorSession(page); // attach BEFORE navigating
+await page.goto('https://example.com', { waitUntil: 'networkidle' });
+console.log(await session.buildAiPrompt()); // or session.extractMarkdown('#main')
+```
+
+See `automation-js/README.md`.
+
 ### Why automation doesn't reuse the extension's patcher
 
 Camoufox executes Playwright's own JS (`page.add_init_script()`,

automation-js/README.md

@@ -0,0 +1,85 @@
+# context-extractor (JS/TS automation)
+
+Node/Playwright twin of the Python `automation/` package — same
+`extension/core/dom.js` + `extension/core/prompt.js` (via symlinks in
+`src/js/`), same output shape, same test fixture. Use this one from Node
+tooling (Playwright, `playwright-core`, or [camoufox-js](https://www.npmjs.com/package/camoufox-js))
+instead of shelling out to Python.
+
+One `ExtractorSession` class covers both sync/async use cases — Node's
+Playwright API is async-only to begin with, so there's no sync/async split
+like the Python package's `ExtractorSession` / `AsyncExtractorSession`.
+
+## Install
+
+```bash
+cd automation-js
+npm install
+npm run build
+```
+
+Not published to a registry yet — consume it from this checkout via a
+relative `file:` dependency, or copy `dist/` + `dist/js/` into a consumer.
+
+## Use
+
+```ts
+import { chromium } from 'playwright-core'; // or 'playwright', or camoufox-js
+import { ExtractorSession } from 'context-extractor'; // this package
+
+const browser = await chromium.launch();
+const page = await browser.newPage();
+const session = new ExtractorSession(page); // attach BEFORE navigating
+
+await page.goto('https://example.com', { waitUntil: 'networkidle' });
+
+console.log(await session.buildAiPrompt()); // or session.extractMarkdown('#main')
+console.log(session.getStore());            // { console, errors, network }
+```
+
+With [camoufox-js](https://www.npmjs.com/package/camoufox-js) — same idea,
+`launchServer`/`launchOptions` differ but the returned `page` is a normal
+Playwright `Page`:
+
+```ts
+import { launchOptions } from 'camoufox-js';
+import { firefox } from 'playwright-core';
+import { ExtractorSession } from 'context-extractor';
+
+const browser = await firefox.launch(await launchOptions({ headless: false }));
+const page = await browser.newPage();
+const session = new ExtractorSession(page);
+```
+
+## Why this exists (vs. shelling out to the Python package)
+
+If your automation is already Node/Playwright (Cursor-driven scripts, a
+`camoufox-js` session, a Playwright test suite), the Python package meant
+spawning a subprocess and shipping JSON across a process boundary just to get
+a markdown dump. This package removes that boundary — same core JS, same
+output, but a normal `import` + `await` in the same process/event loop as the
+rest of your automation.
+
+## Tests
+
+```bash
+npm test
+```
+
+Runs against a real headless Chromium (`playwright-core`) loading the shared
+fixture at `../automation/tests/fixtures/sample.html` — the exact same file
+the Python `automation/tests/test_session.py` suite uses, so a
+markdown-formatting regression can't silently diverge between the two
+automation layers.
+
+## Notes
+
+- `extractMarkdown(selector?)` returns `{ url, title, ts, selector, markdown }`.
+- `buildAiPrompt(selector?, maxChars?)` wraps that in the same
+  `# Page Context` / `## Page Content` / errors / console / network prompt
+  format as the extension's Copy button and the Python package.
+- `getStore()` / `clear(which?)` expose the same bounded ring buffers
+  (`console`, `errors`, `network`, capped at 200 entries each) as the Python
+  `_Store`.
+- Call `session.detach()` if you keep a page open long after you're done
+  capturing and want to stop accumulating entries.

automation-js/package.json

@@ -0,0 +1,54 @@
+{
+  "name": "context-extractor",
+  "version": "1.0.0",
+  "description": "Capture console/network/errors and extract AI-ready markdown from a Playwright (Node) page — JS/TS twin of the automation/ Python package, sharing the same extension/core/*.js.",
+  "license": "MIT",
+  "type": "module",
+  "main": "./dist/index.js",
+  "module": "./dist/index.js",
+  "types": "./dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js"
+    }
+  },
+  "files": [
+    "dist",
+    "README.md",
+    "LICENSE"
+  ],
+  "scripts": {
+    "build": "tsup && npm run copy-js",
+    "copy-js": "mkdir -p dist/js && cp -L src/js/dom.js src/js/prompt.js dist/js/",
+    "typecheck": "tsc --noEmit",
+    "test": "vitest run",
+    "test:watch": "vitest"
+  },
+  "peerDependencies": {
+    "playwright-core": ">=1.40.0"
+  },
+  "devDependencies": {
+    "@types/node": "^22.15.0",
+    "playwright": "^1.55.0",
+    "playwright-core": "^1.55.0",
+    "tsup": "^8.4.0",
+    "typescript": "^5.8.0",
+    "vitest": "^3.1.0"
+  },
+  "engines": {
+    "node": ">=20"
+  },
+  "repository": {
+    "type": "git",
+    "url": "https://git.levkin.ca/ilia/context-extractor.git"
+  },
+  "keywords": [
+    "playwright",
+    "camoufox",
+    "scraping",
+    "llm",
+    "markdown",
+    "browser-extension"
+  ]
+}

automation-js/src/index.ts

@@ -0,0 +1,9 @@
+export {
+  ExtractorSession,
+  type CaptureStore,
+  type ConsoleEntry,
+  type ErrorEntry,
+  type NetworkEntry,
+  type ExtractedMarkdown,
+} from './session.js';
+export { readSharedJs } from './loadJs.js';

automation-js/src/js/dom.js

@@ -0,0 +1 @@
+../../../extension/core/dom.js

automation-js/src/js/prompt.js

@@ -0,0 +1 @@
+../../../extension/core/prompt.js

automation-js/src/loadJs.ts

@@ -0,0 +1,26 @@
+import { existsSync, readFileSync } from 'node:fs';
+import { dirname, join } from 'node:path';
+import { fileURLToPath } from 'node:url';
+
+const here = dirname(fileURLToPath(import.meta.url));
+
+/**
+ * Read one of the shared `extension/core/*.js` files from disk, the same way
+ * the Python `automation/` package's `_read_js()` does — so both automation
+ * layers (and the browser extension) run byte-identical DOM/prompt logic.
+ *
+ * Resolves `js/<name>` next to this module: `src/js/*.js` (symlinks into
+ * `../../extension/core/`) when running from TS source, or `dist/js/*.js`
+ * (copied by `npm run build`) once built.
+ */
+export function readSharedJs(name: 'dom.js' | 'prompt.js'): string {
+  const path = join(here, 'js', name);
+  if (!existsSync(path)) {
+    throw new Error(
+      `Missing shared JS file: ${path}. This package expects to run from a checkout of the ` +
+        'context-extractor repo where automation-js/src/js/*.js are symlinks into extension/core/ ' +
+        '(and, once built, dist/js/*.js — see the copy-js build step).',
+    );
+  }
+  return readFileSync(path, 'utf-8');
+}

automation-js/src/session.ts

@@ -0,0 +1,265 @@
+/**
+ * Console/network/error capture + markdown extraction for a Playwright
+ * (Node) page — the JS/TS twin of the Python `automation/` package's
+ * `ExtractorSession` / `AsyncExtractorSession`. One class here covers both,
+ * since Node's Playwright API is async-only to begin with.
+ *
+ * Works with anything that hands you a real Playwright `Page`: `playwright`,
+ * `playwright-core`, `@playwright/test`, and `camoufox-js` (which wraps
+ * `playwright-core` directly).
+ *
+ * Design notes (see the repo README for the full rationale — same as the
+ * Python package):
+ *   - Console/network/page-error capture uses Playwright's *native* event
+ *     hooks (`page.on('console'/'request'/'requestfinished'/'requestfailed'/'pageerror')`)
+ *     rather than an injected JS patcher, because:
+ *       1. It's more robust — no reliance on `page.addInitScript()`, which is
+ *          known to be unreliable under Camoufox's isolated-world execution
+ *          model (https://github.com/daijro/camoufox/issues/48).
+ *       2. It captures more than the extension does (any resource type, not
+ *          just fetch/XHR), and can't be blocked by a page's CSP.
+ *   - Markdown extraction and CSS-selector helpers *do* need to run inside
+ *     the page (DOM traversal). Those live in `extension/core/*.js` and are
+ *     read from disk here, then run via `page.evaluate()`. That's safe under
+ *     Camoufox's default isolated world because that code only reads the DOM
+ *     — it never writes to the live page — so no `mainWorld`/init-script
+ *     workaround is required.
+ */
+
+import type { ConsoleMessage, Page, Request } from 'playwright-core';
+import { readSharedJs } from './loadJs.js';
+
+const MAX_ENTRIES = 200;
+
+export interface ConsoleEntry {
+  level: string;
+  ts: number;
+  msg: string;
+}
+
+export interface ErrorEntry {
+  type: string;
+  ts: number;
+  msg: string;
+  source: string;
+  line: number;
+  col: number;
+  stack: string;
+}
+
+export interface NetworkEntry {
+  type: string;
+  method: string;
+  url: string;
+  status: number;
+  ts: number;
+  duration: number;
+  error: string | null;
+}
+
+export interface CaptureStore {
+  console: ConsoleEntry[];
+  errors: ErrorEntry[];
+  network: NetworkEntry[];
+}
+
+export interface ExtractedMarkdown {
+  url: string;
+  title: string;
+  ts: number;
+  selector: string;
+  markdown: string;
+}
+
+const CONSOLE_LEVEL_MAP: Record<string, string> = { warning: 'warn' };
+
+function now(): number {
+  return Date.now();
+}
+
+class RingStore implements CaptureStore {
+  console: ConsoleEntry[] = [];
+  errors: ErrorEntry[] = [];
+  network: NetworkEntry[] = [];
+
+  push<K extends keyof CaptureStore>(which: K, entry: CaptureStore[K][number]): void {
+    const arr = this[which];
+    (arr as Array<CaptureStore[K][number]>).push(entry);
+    while (arr.length > MAX_ENTRIES) arr.shift();
+  }
+
+  clear(which?: keyof CaptureStore): void {
+    if (which) {
+      this[which].length = 0;
+    } else {
+      this.console.length = 0;
+      this.errors.length = 0;
+      this.network.length = 0;
+    }
+  }
+
+  asDict(): CaptureStore {
+    return { console: this.console, errors: this.errors, network: this.network };
+  }
+}
+
+function buildExtractScript(selector: string | null | undefined): string {
+  const domJs = readSharedJs('dom.js');
+  const selJson = JSON.stringify(selector || '');
+  return `
+(() => {
+${domJs}
+let el = null;
+const sel = ${selJson};
+if (sel) { try { el = document.querySelector(sel); } catch (_) { el = null; } }
+if (!el) el = document.body;
+return extractMarkdown(el);
+})()
+`;
+}
+
+function buildPromptScript(
+  meta: Pick<ExtractedMarkdown, 'url' | 'title' | 'ts' | 'selector'>,
+  markdown: string,
+  store: CaptureStore,
+  maxChars: number | undefined,
+): string {
+  const promptJs = readSharedJs('prompt.js');
+  const maxCharsJs = maxChars === undefined ? 'undefined' : JSON.stringify(maxChars);
+  return `
+(() => {
+${promptJs}
+return buildAIPrompt(${JSON.stringify(meta)}, ${JSON.stringify(markdown)}, ${JSON.stringify(store)}, ${maxCharsJs});
+})()
+`;
+}
+
+/**
+ * Attach to a page BEFORE navigating, so no console/network/error activity
+ * from the initial load is missed.
+ *
+ * @example
+ * ```ts
+ * import { chromium } from 'playwright-core';
+ * import { ExtractorSession } from 'context-extractor';
+ *
+ * const page = await (await chromium.launch()).newPage();
+ * const session = new ExtractorSession(page);
+ * await page.goto('https://example.com', { waitUntil: 'networkidle' });
+ * console.log(await session.buildAiPrompt());       // or session.extractMarkdown('#main')
+ * ```
+ */
+export class ExtractorSession {
+  readonly page: Page;
+  private readonly store = new RingStore();
+  private readonly requestStarts = new WeakMap<Request, number>();
+
+  constructor(page: Page) {
+    this.page = page;
+    page.on('console', this.onConsole);
+    page.on('pageerror', this.onPageError);
+    page.on('request', this.onRequest);
+    page.on('requestfinished', this.onRequestFinished);
+    page.on('requestfailed', this.onRequestFailed);
+  }
+
+  /** Stop listening. Call this if you keep a page open long after you're done capturing. */
+  detach(): void {
+    this.page.off('console', this.onConsole);
+    this.page.off('pageerror', this.onPageError);
+    this.page.off('request', this.onRequest);
+    this.page.off('requestfinished', this.onRequestFinished);
+    this.page.off('requestfailed', this.onRequestFailed);
+  }
+
+  getStore(): CaptureStore {
+    return this.store.asDict();
+  }
+
+  clear(which?: keyof CaptureStore): void {
+    this.store.clear(which);
+  }
+
+  async extractMarkdown(selector?: string | null): Promise<ExtractedMarkdown> {
+    const markdown = (await this.page.evaluate(buildExtractScript(selector))) as string;
+    return {
+      url: this.page.url(),
+      title: await this.page.title(),
+      ts: now(),
+      selector: selector || 'body',
+      markdown: markdown || '',
+    };
+  }
+
+  async buildAiPrompt(selector?: string | null, maxChars?: number): Promise<string> {
+    const extracted = await this.extractMarkdown(selector);
+    const meta = {
+      url: extracted.url,
+      title: extracted.title,
+      ts: extracted.ts,
+      selector: extracted.selector,
+    };
+    const script = buildPromptScript(meta, extracted.markdown, this.getStore(), maxChars);
+    return (await this.page.evaluate(script)) as string;
+  }
+
+  private onConsole = (msg: ConsoleMessage): void => {
+    const type = msg.type();
+    const level = CONSOLE_LEVEL_MAP[type] ?? type;
+    this.store.push('console', { level, ts: now(), msg: msg.text() });
+  };
+
+  private onPageError = (error: Error): void => {
+    this.store.push('errors', {
+      type: 'error',
+      ts: now(),
+      msg: error?.message ?? String(error),
+      source: '',
+      line: 0,
+      col: 0,
+      stack: error?.stack ?? '',
+    });
+  };
+
+  private onRequest = (request: Request): void => {
+    this.requestStarts.set(request, Date.now());
+  };
+
+  private onRequestFinished = async (request: Request): Promise<void> => {
+    const start = this.requestStarts.get(request);
+    this.requestStarts.delete(request);
+    const duration = start ? Date.now() - start : 0;
+    let status = 0;
+    try {
+      const response = await request.response();
+      if (response) status = response.status();
+    } catch {
+      /* response unavailable — leave status 0 */
+    }
+    this.store.push('network', {
+      type: request.resourceType(),
+      method: request.method(),
+      url: request.url(),
+      status,
+      ts: now(),
+      duration,
+      error: null,
+    });
+  };
+
+  private onRequestFailed = (request: Request): void => {
+    const start = this.requestStarts.get(request);
+    this.requestStarts.delete(request);
+    const duration = start ? Date.now() - start : 0;
+    const failure = request.failure();
+    this.store.push('network', {
+      type: request.resourceType(),
+      method: request.method(),
+      url: request.url(),
+      status: 0,
+      ts: now(),
+      duration,
+      error: failure?.errorText || 'Network error',
+    });
+  };
+}

automation-js/tests/session.test.ts

@@ -0,0 +1,117 @@
+import { afterAll, beforeAll, describe, expect, it } from 'vitest';
+import { chromium, type Browser, type Page } from 'playwright-core';
+import { fileURLToPath } from 'node:url';
+import { dirname, join } from 'node:path';
+import { pathToFileURL } from 'node:url';
+import { ExtractorSession } from '../src/session.js';
+
+// Same golden fixture the Python `automation/tests/` suite uses — one file,
+// two automation layers, so a markdown-formatting bug can't silently diverge
+// between them.
+const here = dirname(fileURLToPath(import.meta.url));
+const FIXTURE_URL = pathToFileURL(
+  join(here, '..', '..', 'automation', 'tests', 'fixtures', 'sample.html'),
+).toString();
+
+let browser: Browser;
+
+beforeAll(async () => {
+  browser = await chromium.launch({ headless: true });
+});
+
+afterAll(async () => {
+  await browser.close();
+});
+
+async function withPage<T>(fn: (page: Page) => Promise<T>): Promise<T> {
+  const p = await browser.newPage();
+  try {
+    return await fn(p);
+  } finally {
+    await p.close();
+  }
+}
+
+describe('ExtractorSession', () => {
+  it('captures console messages, page errors, and network activity', async () => {
+    await withPage(async (page) => {
+      const session = new ExtractorSession(page);
+      await page.goto(FIXTURE_URL, { waitUntil: 'domcontentloaded' });
+      await page.waitForTimeout(150);
+
+      const store = session.getStore();
+      const levels = new Set(store.console.map((e) => e.level));
+      expect(levels.has('warn')).toBe(true);
+      expect(
+        levels.has('error') ||
+          store.console.some((e) => (e.msg || '').toLowerCase().includes('error')),
+      ).toBe(true);
+
+      expect(store.errors.some((e) => (e.msg || '').includes('fixture boom'))).toBe(true);
+      expect(store.network.length).toBeGreaterThan(0);
+    });
+  });
+
+  it('extracts markdown from the full body and from a selector', async () => {
+    await withPage(async (page) => {
+      const session = new ExtractorSession(page);
+      await page.goto(FIXTURE_URL, { waitUntil: 'domcontentloaded' });
+
+      const full = await session.extractMarkdown();
+      expect(full.title).toBe('Context Extractor Fixture');
+      expect(full.markdown).toContain('Fixture Page');
+      expect(full.markdown).toContain('**world**');
+      expect(full.markdown).toContain('[docs](/docs)');
+      expect(full.markdown).toContain('- Alpha');
+
+      const scoped = await session.extractMarkdown('#main-content');
+      expect(scoped.selector).toBe('#main-content');
+      expect(scoped.markdown).toContain('Section');
+      expect(scoped.markdown).not.toContain('Skip this noise');
+    });
+  });
+
+  it('excludes hidden nodes (inline style, stylesheet class, [hidden], aria-hidden) from markdown', async () => {
+    await withPage(async (page) => {
+      const session = new ExtractorSession(page);
+      await page.goto(FIXTURE_URL, { waitUntil: 'domcontentloaded' });
+
+      const { markdown } = await session.extractMarkdown('#main-content');
+      expect(markdown).not.toContain('should-not-appear');
+      expect(markdown).not.toContain('secretConfig');
+      expect(markdown).not.toContain('lixTracking');
+      expect(markdown).toContain('Alpha');
+      expect(markdown).toContain('Beta');
+    });
+  });
+
+  it('builds an AI prompt with the expected sections', async () => {
+    await withPage(async (page) => {
+      const session = new ExtractorSession(page);
+      await page.goto(FIXTURE_URL, { waitUntil: 'domcontentloaded' });
+      await page.waitForTimeout(150);
+
+      const prompt = await session.buildAiPrompt('#main-content');
+      expect(prompt.startsWith('# Page Context')).toBe(true);
+      expect(prompt).toContain('## Page Content');
+      expect(prompt).toContain('Hello **world**');
+      expect(/JavaScript Errors|Console Errors/.test(prompt)).toBe(true);
+      expect(prompt).toContain('_Extracted by Context Extractor_');
+    });
+  });
+
+  it('clears the capture store, selectively and fully', async () => {
+    await withPage(async (page) => {
+      const session = new ExtractorSession(page);
+      await page.goto(FIXTURE_URL, { waitUntil: 'domcontentloaded' });
+      await page.waitForTimeout(100);
+
+      expect(session.getStore().console.length).toBeGreaterThan(0);
+      session.clear('console');
+      expect(session.getStore().console).toEqual([]);
+
+      session.clear();
+      expect(session.getStore()).toEqual({ console: [], errors: [], network: [] });
+    });
+  });
+});

automation-js/tsconfig.json

@@ -0,0 +1,21 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "module": "ESNext",
+    "moduleResolution": "Bundler",
+    "lib": ["ES2022"],
+    "strict": true,
+    "declaration": true,
+    "declarationMap": true,
+    "sourceMap": true,
+    "skipLibCheck": true,
+    "esModuleInterop": true,
+    "forceConsistentCasingInFileNames": true,
+    "resolveJsonModule": true,
+    "outDir": "dist",
+    "rootDir": "src",
+    "types": ["node"]
+  },
+  "include": ["src/**/*.ts"],
+  "exclude": ["node_modules", "dist", "tests"]
+}

automation-js/tsup.config.ts

@@ -0,0 +1,12 @@
+import { defineConfig } from 'tsup';
+
+export default defineConfig({
+  entry: { index: 'src/index.ts' },
+  format: ['esm'],
+  dts: true,
+  sourcemap: true,
+  splitting: false,
+  treeshake: true,
+  clean: true,
+  external: ['playwright-core', '@playwright/test', 'playwright'],
+});

automation-js/vitest.config.ts

@@ -0,0 +1,8 @@
+import { defineConfig } from 'vitest/config';
+
+export default defineConfig({
+  test: {
+    include: ['tests/**/*.test.ts'],
+    testTimeout: 20_000,
+  },
+});