@shadowcoderr/context-graph 0.3.2 → 0.3.4

package/README.md

@@ -1,49 +1,108 @@
+<div align="center">
+
+<br />
+
+<img width="80" height="80" src="https://raw.githubusercontent.com/shadowcoderr/ContextGraph/main/.github/assets/logo.svg" alt="ContextGraph logo" onerror="this.style.display='none'" />
+
 # ContextGraph
 
-> **A deterministic "Flight Data Recorder" for web applications** — captures rich, AI-ready context from browser sessions in restricted enterprise environments where MCP is unavailable.
+**The deterministic "Flight Data Recorder" for web applications**
+
+*Transform manual browser sessions into structured, AI-ready intelligence — 100% offline, zero telemetry, enterprise-safe*
+
+<br />
 
-[![Version](https://img.shields.io/badge/version-0.3.0-blue)](package.json)
-[![Node](https://img.shields.io/badge/node-%3E%3D18.0.0-green)](package.json)
-[![TypeScript](https://img.shields.io/badge/TypeScript-5.x-blue)](tsconfig.json)
-[![Playwright](https://img.shields.io/badge/Playwright-1.40%2B-orange)](package.json)
-[![License](https://img.shields.io/badge/license-MIT-lightgrey)](LICENSE)
+[![npm version](https://img.shields.io/npm/v/@shadowcoderr/context-graph?style=flat-square&color=0066cc&label=npm)](https://www.npmjs.com/package/@shadowcoderr/context-graph)
+[![Node.js](https://img.shields.io/badge/node-%3E%3D18.0.0-brightgreen?style=flat-square&logo=node.js&logoColor=white)](https://nodejs.org)
+[![TypeScript](https://img.shields.io/badge/TypeScript-5.x-3178c6?style=flat-square&logo=typescript&logoColor=white)](https://www.typescriptlang.org)
+[![Playwright](https://img.shields.io/badge/Playwright-1.40%2B-2ead33?style=flat-square&logo=playwright&logoColor=white)](https://playwright.dev)
+[![License: MIT](https://img.shields.io/badge/license-MIT-yellow?style=flat-square)](LICENSE)
+[![Zero Telemetry](https://img.shields.io/badge/telemetry-none-red?style=flat-square)](README.md#security-model)
+
+<br />
+
+```
+  IDE AI Agent          ContextGraph          Your Browser
+  (Copilot/Cursor)   ←  (Playwright)    ←    (You Navigate)
+   generates tests       captures facts        visits pages
+```
+
+<br />
+
+[**Quick Start**](#-quick-start) • [**Features**](#-key-features) • [**Output**](#-what-you-get) • [**AI Integration**](#-ai-integration) • [**Config**](#-configuration) • [**Docs**](docs/USER_GUIDE.md)
+
+<br />
+
+</div>
 
 ---
 
 ## Why ContextGraph?
 
-IDE-based AI agents (Copilot, Cursor, etc.) are blind to your running browser. Without MCP, they cannot see the live DOM, accessibility tree, or real API traffic of the app you are testing. In enterprise environments, MCP and cloud-based recording tools are often blocked by security policy.
+IDE-based AI agents (GitHub Copilot, Cursor, Claude) are **blind to your running browser**. Without MCP, they cannot see the live DOM, accessibility tree, or real API traffic of the application you're testing.
 
-ContextGraph solves this by acting as a **pre-processor** — it bridges the gap between manual exploration and AI-powered test generation using only Playwright, which is already whitelisted on most QA teams' machines.
+In enterprise environments, MCP and cloud-based recording tools are frequently blocked by security policy.
+
+**ContextGraph solves this** by acting as a pre-processor — bridging the gap between manual exploration and AI-powered test generation using only Playwright, which is already whitelisted on most QA teams' machines.
 
 ```
-Your Browser Session
-        │
-        ▼
-  ContextGraph          ← You are here
-  (Playwright-native)
-        │
-        ▼
-  Structured Output       ← DOM, Locators, A11y Tree, Network
-  (Local Filesystem)
-        │
-        ▼
-  AI Agent / LLM          ← Receives clean, noise-free context
-  (Copilot / Cursor / Claude)
-        │
-        ▼
-  Generated Tests         ← Playwright .spec.ts files
+Your Browser Session  →  ContextGraph  →  Structured Output  →  AI Agent  →  Generated Tests
+                         (Playwright)      (Local Filesystem)    (Copilot)     (.spec.ts files)
 ```
 
+> **Zero cloud. Zero telemetry. Zero MCP required.** Everything stays inside your enterprise perimeter.
+
 ---
 
-## Quick Start
+## ✨ Key Features
+
+<table>
+<tr>
+<td width="50%">
+
+**🎯 Guided Capture**
+Captures only what you navigate — no autonomous crawling, no surprises.
 
-### Prerequisites
+**🧹 Clean DOM Output**
+Scripts, event handlers, and noise stripped before saving. Shadow DOM included.
 
-- **Node.js** 18.0.0 or higher
-- **Microsoft Edge** (installed — used as the capture browser)
-- **Windows 10/11** (primary support; Linux/macOS in beta)
+**🔑 Smart Locators**
+5+ selector strategies per element, ranked by resilience (TestID → Role → Label → Text → CSS).
+
+**♿ Accessibility Tree**
+Three-strategy extraction: Playwright API → CDP → DOM fallback. Full ARIA/role hierarchy.
+
+</td>
+<td width="50%">
+
+**🌐 Network Intelligence**
+All HTTP traffic captured. Concurrent request timing fixed via WeakMap. Failed requests tracked.
+
+**🔒 Security by Design**
+JWT, Bearer tokens, credit cards, SSNs, API keys, URL query params — all redacted before any disk write.
+
+**🤖 AI Context Bundle**
+Export entire session as a single Markdown file, ready to paste into any AI agent.
+
+**📊 API Inventory**
+Auto-generates structured API surface documentation with path-parameter normalisation.
+
+</td>
+</tr>
+</table>
+
+| Feature | Description |
+|---|---|
+| 📸 **Full Page Screenshots** | Automatic per-page + optional per-element captures |
+| 🧩 **Component Registry** | Tracks reusable UI patterns across all captured pages |
+| 🔄 **Change Detection** | Structural content hashing — skips unchanged pages automatically |
+| 📦 **100% Offline** | Zero cloud dependencies, zero telemetry, runs inside enterprise perimeter |
+| 🎭 **Recorder Mode** | Playwright Codegen integration — convert manual walkthroughs to `.spec.ts` |
+| 🩺 **Console Logging** | JS errors, warnings, and uncaught exceptions captured per page |
+
+---
+
+## ⚡ Quick Start
 
 ### Install
 
@@ -55,40 +114,137 @@ Verify:
 
 ```bash
 context-graph --version
-# Context Graph v0.3.0
 ```
 
-### Capture Your First Page (2 minutes)
+### Capture Your First Page
 
 ```bash
-# 1. Start a capture session
+# Start a capture session (browser will open)
 context-graph --mode browser --url https://your-app.com
 
-# 2. Navigate through the app — every page you visit is captured
-# 3. Press Ctrl+C when done
+# Navigate freely — every page you visit is captured automatically
+# Press Ctrl+C when done
+```
+
+### Generate an AI Context Bundle
 
-# 4. Inspect results
-cat context-graph-output/global_manifest.json
+```bash
+# After capture, bundle everything into one LLM-ready file
+context-graph --export-bundle
+
+# → context-graph-output/bundles/ai_context_bundle.md
 ```
 
+Paste `ai_context_bundle.md` into Copilot Chat, Cursor, or Claude to immediately generate tests.
+
 ---
 
-## Key Features
+## 🗺️ Architecture
 
-| Feature | Description |
+```
+CLI (Commander.js + Inquirer.js)
+  └── RuntimeController
+        ├── BrowserAdapter          Playwright lifecycle, event hooks
+        ├── CaptureEngine
+        │     ├── DOMAnalyzer       Sanitise + Shadow DOM serialisation
+        │     ├── A11yExtractor     3-strategy: Playwright → CDP → DOM
+        │     ├── LocatorGenerator  Multi-strategy selector ranking
+        │     ├── NetworkLogger     HTTP interception (WeakMap timing)
+        │     ├── ScreenshotCapturer Full-page + element screenshots
+        │     └── SecurityRedactor  PII / credential / URL param redaction
+        ├── NetworkPatternAnalyzer  API surface documentation
+        ├── AIContextBundler        Single-file LLM export
+        ├── ComponentsRegistry      Cross-page UI pattern tracking
+        └── StorageEngine           Filesystem persistence + statistics
+```
+
+---
+
+## 📁 What You Get
+
+For each captured page, ContextGraph saves a self-contained directory:
+
+```
+context-graph-output/
+├── global_manifest.json                ← Master index of all captures
+├── bundles/
+│   └── ai_context_bundle.md            ← Ready-to-paste LLM context
+├── <domain>/
+│   ├── api_inventory.json              ← Structured API surface docs
+│   ├── components_registry.json        ← Cross-page UI patterns
+│   ├── network/
+│   │   └── traffic_log.jsonl           ← HTTP traffic (redacted)
+│   ├── user_interactions.json          ← Recorder mode interactions
+│   └── pages/
+│       └── <page-name>/
+│           ├── metadata.json           ← URL, title, timing, viewport
+│           ├── DOM                     ← Beautified, sanitised HTML
+│           ├── a11y_tree.json          ← Full ARIA hierarchy
+│           ├── locators.json           ← Ranked Playwright selectors ⭐
+│           ├── frames.json             ← iframe + Shadow DOM structure
+│           ├── console_errors.json     ← JS errors & warnings
+│           ├── screenshot_manifest.json← Screenshot paths + metadata
+│           └── screenshots/
+│               └── screenshot.png      ← Full-page capture
+└── scripts/
+    └── <domain>.spec.ts                ← Recorder mode output
+```
+
+### File Reference
+
+| File | Purpose |
 |---|---|
-| 🎯 **Guided Capture** | Captures only what you navigate — no autonomous crawling |
-| 🧹 **Clean DOM Output** | Scripts, event handlers, and noise stripped before saving |
-| 🔑 **Smart Locators** | 5+ selector strategies per element, ranked by resilience |
-| ♿ **Accessibility Tree** | Full ARIA/role hierarchy for semantic understanding |
-| 🌐 **Network Logging** | All HTTP traffic captured with automatic PII redaction |
-| 🔒 **Security by Design** | Credentials, tokens, PII never touch the filesystem |
-| 🧩 **Component Registry** | Tracks reusable UI patterns across captured pages |
-| 📦 **100% Offline** | Zero cloud dependencies, zero telemetry, runs inside enterprise perimeter |
+| `metadata.json` | URL, page title, timing metrics, performance counters |
+| `DOM` | Sanitised HTML — scripts removed, Shadow DOM inlined |
+| `a11y_tree.json` | Full ARIA tree — the most semantically dense representation |
+| `locators.json` | 5+ selector strategies per element, ranked by resilience |
+| `api_inventory.json` | All observed API endpoints with normalised path params |
+| `ai_context_bundle.md` | Single-file LLM export of the entire session |
 
 ---
 
-## Operating Modes
+## 🤖 AI Integration
+
+### The 3-Step Workflow
+
+```bash
+# 1. Capture your application (you drive, ContextGraph records)
+context-graph --mode browser --url https://app.example.com
+
+# 2. Bundle everything for your AI agent
+context-graph --export-bundle
+
+# 3. Open your AI IDE, paste the bundle, and ask for tests
+```
+
+### Example Prompt (after pasting the bundle)
+
+```
+I am testing the app shown in this session context.
+
+Using ONLY the locators from the "Interactive Elements" tables above,
+generate Playwright TypeScript tests that:
+  1. Navigate to the login page and log in with valid credentials
+  2. Assert the dashboard renders correctly
+  3. Test the shopping cart add/remove flow
+  4. Cover the checkout form submission
+
+Use test.describe() groups per page. Include assertions for
+visible elements and navigation outcomes.
+```
+
+### What to Feed Your AI Agent
+
+| File | When to Include |
+|---|---|
+| `ai_context_bundle.md` | **Always** — the complete single-file export |
+| `locators.json` | When you need precise selector details |
+| `api_inventory.json` | When generating API integration tests |
+| `a11y_tree.json` | When generating accessibility tests |
+
+---
+
+## 🚀 Operating Modes
 
 ### Browser Mode — Passive Observer
 
@@ -98,88 +254,292 @@ Best for: understanding app structure, feeding AI with full UI context.
 context-graph --mode browser --url https://app.example.com
 ```
 
+Automatically captures on every page navigation: DOM, A11y tree, smart locators, network traffic, screenshots, and console errors.
+
 ### Recorder Mode — Action Capture
 
 Best for: converting manual test walkthroughs into replayable scripts.
 
 ```bash
 context-graph --mode recorder --url https://app.example.com
+
+# Optionally replay the recording to capture full artifacts too:
+context-graph --mode recorder --url https://app.example.com --recorder-capture
 ```
 
 ---
 
-## What You Get
+## ⚙️ Configuration
+
+Create a `context-graph.config.json` file in your project root:
+
+```json
+{
+  "browser": {
+    "channel": "msedge",
+    "headless": false,
+    "viewport": { "width": 1920, "height": 1080 },
+    "slowMo": 0
+  },
+  "capture": {
+    "screenshots": {
+      "enabled": true,
+      "fullPage": true,
+      "elementTargeting": false
+    },
+    "network": {
+      "enabled": true,
+      "captureHeaders": true,
+      "captureBody": false
+    },
+    "accessibility": {
+      "enabled": true,
+      "includeHidden": false
+    },
+    "components": {
+      "enabled": true,
+      "minOccurrences": 1
+    },
+    "forceCapture": false
+  },
+  "security": {
+    "redactPatterns": ["jwt", "bearer", "creditcard", "ssn", "email", "api_key"],
+    "redactHeaders": ["authorization", "cookie", "set-cookie", "x-api-key"],
+    "customPatterns": [
+      {
+        "name": "employee_id",
+        "pattern": "EMP-\\d{6}",
+        "replacement": "[REDACTED:EMP_ID]",
+        "severity": "high"
+      }
+    ]
+  },
+  "storage": {
+    "outputDir": "./context-graph-output",
+    "prettyJson": true
+  }
+}
+```
+
+Run with config:
+
+```bash
+context-graph --config ./context-graph.config.json --mode browser
+```
 
-For each captured page, ContextGraph generates:
+### Configuration Reference
 
-- **Sanitized DOM** - Clean HTML with scripts removed
-- **Accessibility Tree** - Full ARIA hierarchy and semantic structure  
-- **Smart Locators** - Ranked Playwright selectors (Role → TestID → Label → Text)
-- **Network Log** - HTTP traffic with automatic PII redaction
-- **Screenshots** - Full-page visual captures
-- **Component Registry** - Cross-page UI pattern tracking
+<details>
+<summary><strong>Browser options</strong></summary>
 
----
+| Option | Type | Default | Description |
+|---|---|---|---|
+| `channel` | `msedge \| chromium \| firefox` | `msedge` | Browser engine |
+| `headless` | `boolean` | `false` | Run without visible UI |
+| `viewport.width` | `number` | `1920` | Browser width |
+| `viewport.height` | `number` | `1080` | Browser height |
+| `slowMo` | `number` | `0` | Slow motion delay (ms) |
 
-## Feeding Output to an AI Agent
+</details>
 
-### What to Give the AI
+<details>
+<summary><strong>Capture options</strong></summary>
 
-For test generation, provide the AI these files from a single captured page:
+| Option | Type | Default | Description |
+|---|---|---|---|
+| `screenshots.enabled` | `boolean` | `true` | Capture full-page screenshots |
+| `screenshots.fullPage` | `boolean` | `true` | Include off-viewport content |
+| `screenshots.elementTargeting` | `boolean` | `false` | Also capture key element screenshots |
+| `network.captureBody` | `boolean` | `false` | Include response body (JSON/text only) |
+| `accessibility.includeHidden` | `boolean` | `false` | Include aria-hidden elements |
+| `forceCapture` | `boolean` | `false` | Re-capture even if content hash matches |
 
-| File | Why it Matters |
-|---|---|
-| `metadata.json` | URL, page title, timing context |
-| `a11y_tree.json` | Semantic structure — what the page *means* |
-| `locators.json` | Ready-to-use, ranked Playwright selectors |
-| `dom_snapshot.html` | Full structural HTML if deep context is needed |
-| `network/traffic_log.jsonl` | What API calls the page makes |
+</details>
+
+<details>
+<summary><strong>Security options</strong></summary>
+
+**Built-in redaction patterns:** `jwt`, `bearer`, `creditcard`, `ssn`, `email`, `phone_number`, `api_key`, `aws_key`
+
+**Always-redacted headers:** `authorization`, `cookie`, `set-cookie`, `x-api-key`, `x-auth-token`, `x-csrf-token`, `x-access-token`, `x-refresh-token`, `proxy-authorization`
+
+**URL query parameters:** `token`, `api_key`, `auth`, `secret`, `password`, `session`, `code`, `client_secret`, and more — values are redacted before any disk write.
 
-### Example Prompt Structure
+</details>
 
+---
+
+## 🖥️ CLI Reference
+
+```
+context-graph [startUrl] [options]
+
+Options:
+  -m, --mode <type>         browser | recorder            (required)
+  -u, --url <url>           Starting URL
+  -o, --output <path>       Output directory               [default: ./context-graph-output]
+  -c, --config <path>       Config file path
+  -v, --viewport <WxH>      Viewport size                  [default: 1920x1080]
+  --headless                Run headless (no visible browser)
+  --slow-mo <ms>            Slow motion delay in ms
+  --no-screenshots          Disable screenshot capture
+  --no-network              Disable network logging
+  --recorder-capture        (Recorder) Replay script to capture full artifacts
+  --export-bundle           Generate AI context bundle after capture
+  --force-capture           Re-capture already-visited pages
+  --verbose                 Enable debug logging
+  --version                 Show version
+  --help                    Show help
 ```
-I am testing the following web page:
-URL: https://app.example.com/login
 
-Here is the accessibility tree:
-[paste a11y_tree.json]
+---
 
-Here are the available locators:
-[paste locators.json]
+## 🔒 Security Model
 
-Here is the API traffic observed:
-[paste relevant entries from traffic_log.jsonl]
+ContextGraph enforces a multi-layer redaction pipeline — **all redaction happens before any disk write**. There is no "raw then redact" staging.
 
-Generate a Playwright test that:
-1. Navigates to the login page
-2. Enters valid credentials
-3. Asserts the dashboard is shown
 ```
+Raw Data
+  ↓ 1. Header stripping     (authorization, cookie, x-api-key, etc.)
+  ↓ 2. URL sanitisation     (sensitive query params: ?token=, ?api_key=)
+  ↓ 3. Pattern matching     (JWT, Bearer, credit cards, SSNs, emails, AWS keys)
+  ↓ 4. Key-based redaction  (object properties with sensitive names)
+  ↓ 5. Custom rules         (user-defined patterns from config)
+  ↓
+Redacted Data → Audit Log → Filesystem Write
+```
+
+**Security invariants (hardcoded, cannot be disabled):**
+- `<script>` content is always stripped from DOM snapshots
+- `type="password"` input values are always replaced with `[REDACTED]`
+- `authorization`, `cookie`, and `set-cookie` headers are always removed
+- Zero outbound connections, zero telemetry, zero update checks
+
+Every redaction event is logged to `logs/redaction_audit.jsonl` for compliance review.
 
 ---
 
-## Documentation
+## 🏢 Enterprise STLC Patterns
+
+### Pattern A — AI-Assisted Test Generation
+
+```bash
+context-graph --mode browser --url https://app.example.com
+context-graph --export-bundle
+# → Paste ai_context_bundle.md into Copilot Chat → get .spec.ts tests
+```
+
+### Pattern B — Regression Detection Between Releases
+
+```bash
+# Capture baseline before release
+context-graph --mode browser --url https://staging.example.com
+mv context-graph-output context-graph-baseline
+
+# Deploy release candidate, then capture again
+context-graph --mode browser --url https://staging.example.com
+
+# Compare the two captures to find locator drift
+# Elements whose best locator strategy changed = tests likely to break
+```
+
+### Pattern C — Undocumented App Onboarding
+
+```bash
+# Capture 1–2 hours of navigation through an unfamiliar enterprise app
+context-graph --mode browser --url https://internal-erp.corp.com
+
+# Generate API inventory — no Swagger/OpenAPI docs needed
+# api_inventory.json shows every endpoint, method, and status code observed
+```
+
+### Pattern D — Manual Test Script Conversion
+
+```bash
+# Follow your manual test script step-by-step
+context-graph --mode recorder --url https://app.example.com --recorder-capture
 
-📖 **[Complete User Guide](docs/USER_GUIDE.md)** - Advanced configuration, CLI reference, troubleshooting, and architecture details.
+# Output: <domain>.spec.ts with your steps as Playwright actions
+# Add assertions, or ask AI to add them given the a11y tree context
+```
 
 ---
 
-## Installation
+## 🛠️ Development
 
 ```bash
-npm install -g @shadowcoderr/context-graph
+# Clone and install
+git clone https://github.com/shadowcoderr/ContextGraph
+cd ContextGraph
+npm install
+
+# Build
+npm run build
+
+# Run in dev mode
+npm run dev -- --mode browser --url https://example.com
+
+# Unit tests (Jest)
+npm test
+
+# Integration tests (Playwright)
+npm run test:playwright
+
+# Lint
+npm run lint
 ```
 
+### Module Status
+
+| Module | Status | Notes |
+|---|---|---|
+| CLI | ✅ Stable | Interactive wizard + full CLI flags |
+| Browser Mode | ✅ Stable | Auto-capture on navigation + SPA history API |
+| Recorder Mode | ✅ Stable | Codegen integration + artifact replay |
+| DOM Analyzer | ✅ Stable | Script removal + Shadow DOM serialisation |
+| A11y Extractor | ✅ Stable | 3-strategy: Playwright → CDP → DOM |
+| Locator Generator | ✅ Stable | 6-strategy ranking + uniqueness checks |
+| Network Logger | ✅ Stable | WeakMap timing, requestfailed, captureBody |
+| Screenshot Capturer | ✅ Stable | Full-page + element-level screenshots |
+| Security Redactor | ✅ Stable | 8 patterns + URL query param redaction |
+| AI Context Bundler | ✅ Stable | Single-file Markdown/JSON LLM export |
+| Network Pattern Analyzer | ✅ Stable | API inventory with path normalisation |
+| Components Registry | ✅ Stable | Cross-page pattern indexing |
+| Shadow DOM Support | 🔄 Partial | Top-level serialisation included |
+| Multi-browser | 🔜 Planned | Edge + Chromium; Firefox beta |
+
+---
+
+## 📖 Documentation
+
+Full technical documentation including architecture deep-dive, module reference, all identified issues and their fixes, proposed features, output schema reference, security model, and STLC integration patterns:
+
+📄 **[Complete Technical Documentation](docs/_DOCUMENTATION.md)**
+
+📖 **[User Guide](docs/USER_GUIDE.md)** — Advanced configuration, CLI reference, troubleshooting
+
+---
+
+## 🤝 Contributing & Support
+
+🐛 **[Report Issues](https://github.com/shadowcoderr/ContextGraph/issues)**
+
+💬 **[Discussions](https://github.com/shadowcoderr/ContextGraph/discussions)**
+
+📦 **[npm Package](https://www.npmjs.com/package/@shadowcoderr/context-graph)**
+
 ---
 
-## License
+## 📄 License
 
-MIT — © Shadow Coderr
+MIT © [Shadow Coderr](https://github.com/shadowcoderr)
 
 ---
 
-## Issues & Support
+<div align="center">
+
+**Built for QA engineers who believe AI should work smarter, not just faster.**
 
-🐛 **[Report Issues](https://github.com/shadowcoderr/ContextGraph/issues)** - Bug reports and feature requests
+*ContextGraph gives your AI the eyes it was missing.*
 
-💬 **[Discussions](https://github.com/shadowcoderr/ContextGraph/discussions)** - Community support and questions
+</div>

package/dist/analyzers/a11y-extractor.d.ts

@@ -1,13 +1,27 @@
 import { Page } from '@playwright/test';
 import { AccessibilityTree } from '../types/capture';
+/**
+ * AccessibilityExtractor — captures the full accessibility tree using three
+ * strategies in order of reliability:
+ *
+ * 1. Playwright's built-in accessibility.snapshot() — the most reliable,
+ *    cross-browser approach.  Configured with interestingOnly: false so every
+ *    node is returned, not just "interesting" leaf nodes.
+ *
+ * 2. Chrome DevTools Protocol (CDP) Accessibility.getFullAXTree — lower-level
+ *    but gives raw browser data when Playwright's API is unavailable.
+ *
+ * 3. DOM-based construction — pure JavaScript executed inside the page.
+ *    Works in every environment but misses computed ARIA states.
+ */
 export declare class AccessibilityExtractor {
-    extract(page: Page, _includeHidden?: boolean): Promise<AccessibilityTree>;
-    /**
-     * Transform CDP accessibility tree format to our format
-     */
+    extract(page: Page, includeHidden?: boolean): Promise<AccessibilityTree>;
+    private transformPlaywrightSnapshot;
     private transformCDPSnapshot;
     /**
-     * Build accessibility tree from DOM when Playwright accessibility API is unavailable
+     * Build an accessibility tree by traversing the live DOM inside the page.
+     * This approach works universally but misses computed ARIA states and
+     * relationships that only the accessibility engine can resolve.
      */
     private buildFromDOM;
     private transformSnapshot;

package/dist/analyzers/a11y-extractor.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"a11y-extractor.d.ts","sourceRoot":"","sources":["../../src/analyzers/a11y-extractor.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,IAAI,EAAE,MAAM,kBAAkB,CAAC;AACxC,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAErD,qBAAa,sBAAsB;IAC3B,OAAO,CAAC,IAAI,EAAE,IAAI,EAAE,cAAc,GAAE,OAAe,GAAG,OAAO,CAAC,iBAAiB,CAAC;IAwCtF;;OAEG;IACH,OAAO,CAAC,oBAAoB;IA0C5B;;OAEG;YACW,YAAY;IA+C1B,OAAO,CAAC,iBAAiB;CAkB1B"}
+{"version":3,"file":"a11y-extractor.d.ts","sourceRoot":"","sources":["../../src/analyzers/a11y-extractor.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,IAAI,EAAE,MAAM,kBAAkB,CAAC;AACxC,OAAO,EAAE,iBAAiB,EAAE,MAAM,kBAAkB,CAAC;AAGrD;;;;;;;;;;;;;GAaG;AACH,qBAAa,sBAAsB;IAC3B,OAAO,CAAC,IAAI,EAAE,IAAI,EAAE,aAAa,GAAE,OAAe,GAAG,OAAO,CAAC,iBAAiB,CAAC;IAoDrF,OAAO,CAAC,2BAA2B;IAyBnC,OAAO,CAAC,oBAAoB;IAsD5B;;;;OAIG;YACW,YAAY;IA0J1B,OAAO,CAAC,iBAAiB;CAsB1B"}