release: v0.3.3 — Windows x64, macOS v145, auto-spoof docs

Bump wrapper to 0.3.3. Update README fingerprint section to
document auto-spoof behavior (zero-config stealth). Improve
reCAPTCHA test with wait_for_selector instead of blind sleep.

CHANGELOG.md

@@ -6,6 +6,21 @@ Changes are tagged: **[wrapper]** for Python/JS wrapper, **[binary]** for Chromi
 
 ---
 
+## [0.3.3] — 2026-03-03
+
+All platforms now run Chromium 145 v2 with 25 patches. Windows x64 added.
+
+- **[binary]** Auto-spoof by default — binary is stealthy with zero flags. Random fingerprint seed auto-generated at startup, no wrapper or configuration required
+- **[binary]** Platform-aware auto-detection — GPU, screen dimensions, and User-Agent automatically match the real OS (macOS, Linux, Windows) without explicit flags
+- **[binary]** Expanded GPU model database for realistic per-session diversity
+- **[binary]** First macOS v145 builds (arm64 + x64) — 25 patches, up from 16 on v142
+- **[binary]** First Windows x64 v145 build — 25 patches
+- **[wrapper]** Add Windows x64 platform support — auto-download, binary path resolution, and platform detection
+- **[wrapper]** Upgrade macOS (arm64 + x64) from Chromium 142 to 145 — all platforms now ship the same 25-patch build
+- **[wrapper]** Add explicit Mac GPU flags (`Apple M3 Metal` renderer) to default stealth args for consistent WebGL fingerprints
+- **[wrapper]** Improve reCAPTCHA stealth test — wait for score element instead of blind sleep
+- **[wrapper]** JS: add `win32-x64` platform mapping, Windows binary path (`chrome.exe`)
+
 ## [0.3.1] — 2026-03-03
 
 - **[wrapper]** Auto-check for wrapper updates on startup (PyPI/npm). Notifies users when a newer wrapper version is available. Runs once per process, respects `CLOAKBROWSER_AUTO_UPDATE=false`.

README.md

@@ -342,7 +342,17 @@ clearCache();
 
 ## Fingerprint Management
 
-Every launch automatically generates a **unique fingerprint**. A random seed (10000–99999) drives all seed-based patches — canvas, WebGL, audio, fonts, and client rects all produce consistent, correlated values derived from that single seed.
+The binary is **stealthy by default** — no flags needed. It auto-generates a random fingerprint seed at startup and spoofs all detectable values (GPU, hardware specs, screen dimensions, canvas, WebGL, audio, fonts). Every launch produces a fresh, coherent identity.
+
+**How fingerprinting works:**
+
+| Scenario | What happens |
+|----------|-------------|
+| **No flags** | Random seed auto-generated at startup. GPU, screen, hardware specs, and all noise patches are spoofed automatically. Fresh identity each launch. |
+| **`--fingerprint=seed`** | Deterministic identity from the seed. Same seed = same fingerprint across launches. Use this for session persistence (returning visitor). |
+| **`--fingerprint=seed` + explicit flags** | Explicit flags override individual auto-generated values. The seed fills in everything else. |
+
+The binary detects its platform at compile time — a macOS binary reports Apple GPU and macOS screen (1440x900), a Linux binary reports NVIDIA GPU and 1080p screen. Override with `--fingerprint-platform` for cross-platform spoofing (e.g. Linux binary appearing as Windows).
 
 > **Tip: Use a fixed seed when revisiting the same site.** A random seed makes every session look like a different device — which can be suspicious when hitting the same site repeatedly from the same IP. For reCAPTCHA v3 Enterprise and similar scoring systems, a fixed seed produces a consistent fingerprint across sessions, making you look like a returning visitor:
 > ```python
@@ -368,9 +378,9 @@ Every `launch()` call sets these automatically. Defaults are **platform-aware**
 | `--fingerprint-screen-height` | `1080` | *(not set)* | Screen height reporting |
 | `--window-size` | `1920,1080` | *(not set)* | Browser window dimensions |
 
-> **Important:** `--fingerprint-platform` should always be set. Without it, platform-specific patches (GPU, UA, screen, taskbar) won't activate. The wrapper handles this automatically.
+> **Using the binary directly?** It works out of the box with zero flags — the binary auto-spoofs everything. Pass `--fingerprint=seed` for a persistent identity, or use explicit flags like `--fingerprint-gpu-renderer` to override any auto-generated value.
 
-> **⚠️ Using the binary directly (without the wrapper)?** The binary does not auto-spoof without flags. You must pass `--fingerprint`, `--fingerprint-platform`, and GPU flags explicitly. Without these, real device values pass through unmodified. See the table above for the full list of flags the wrapper sets automatically.
+> **Production tip:** For better stealth at scale, pass your own GPU, screen, and hardware values instead of relying on defaults. Custom parameters make your sessions harder to cluster by anti-bot systems that look for uniform fingerprint profiles.
 
 ### Additional Flags
 

cloakbrowser/_version.py

@@ -1 +1 @@
-__version__ = "0.3.1"
+__version__ = "0.3.3"

examples/stealth_test.py

@@ -146,8 +146,12 @@ def test_recaptcha(page):
         wait_until="domcontentloaded",
         timeout=30000,
     )
-    # Page auto-submits via grecaptcha.execute() — wait for backend response
+    # Wait for backend response (step3 element appears when score arrives)
-    time.sleep(8)
+    try:
+        page.wait_for_selector("li.step3", timeout=20000)
+        time.sleep(1)
+    except Exception:
+        time.sleep(10)  # fallback
 
     results = page.evaluate("""() => {
         const text = document.body.innerText;

js/package.json

@@ -1,6 +1,6 @@
 {
   "name": "cloakbrowser",
-  "version": "0.3.1",
+  "version": "0.3.3",
   "description": "Stealth Chromium that passes every bot detection test. Drop-in Playwright/Puppeteer replacement with source-level fingerprint patches.",
   "type": "module",
   "main": "dist/index.js",