@melaya/runner 1.1.12 → 1.1.14

package/dist/assistantHost.py

@@ -621,7 +621,11 @@ def _build_agent():
         toolkit=toolkit,
         model_name=model,
         provider=provider,
-        api_key="",  # model.py resolves the provider's own creds (OAuth off disk / localhost)
+        # Cloud providers on the runner path (browser control turns pinned to the
+        # runner even for a cloud model) get their key passed down from the server
+        # as MEL_ASSISTANT_PROVIDER_API_KEY. Empty for local providers (claude_code
+        # OAuth / ollama / lmstudio), where model.py resolves on-disk/localhost auth.
+        api_key=os.environ.get("MEL_ASSISTANT_PROVIDER_API_KEY", "") or "",
         # Phone tasks that act on N items ("like + comment on 10 posts") take
         # ~10-15 tool calls EACH (screen_tree→read→tap→type→post_comment→verify→
         # swipe to the next), so a flat 40 capped a 10-post run at ~3 done and

package/dist/browserBridge.js

@@ -59,7 +59,7 @@ const MELAYA_BRAND_SCRIPT = `(() => {
     if (document.getElementById(ID + "_border")) return;
     const style = document.createElement("style");
     style.id = ID + "_style";
-    style.textContent = "@keyframes melBrandPulse{0%,100%{box-shadow:inset 0 0 9px 2px rgba(34,211,238,.5),inset 0 0 30px 7px rgba(124,111,240,.18)}50%{box-shadow:inset 0 0 16px 4px rgba(34,211,238,.8),inset 0 0 54px 13px rgba(124,111,240,.32)}}@keyframes melBrandGlow{0%,100%{opacity:.78}50%{opacity:1}}#" + ID + "_border{position:fixed;inset:0;pointer-events:none;z-index:2147483647;border-radius:7px;animation:melBrandPulse 2.6s ease-in-out infinite}#" + ID + "_badge{position:fixed;top:10px;left:50%;transform:translateX(-50%);pointer-events:none;z-index:2147483647;font-family:-apple-system,Segoe UI,Roboto,sans-serif;font-weight:600;font-size:12px;line-height:1;white-space:nowrap;max-width:calc(100vw - 24px);box-sizing:border-box;color:#E2E8F0;background:linear-gradient(180deg,rgba(15,23,42,.94),rgba(15,23,42,.82));padding:5px 13px 5px 7px;border-radius:999px;letter-spacing:.2px;border:1px solid rgba(34,211,238,.45);box-shadow:0 0 16px 1px rgba(34,211,238,.4),0 4px 14px rgba(0,0,0,.45);display:flex;align-items:center;gap:8px;backdrop-filter:blur(6px);animation:melBrandGlow 2.6s ease-in-out infinite}#" + ID + "_logo{height:16px;width:auto;display:block;filter:drop-shadow(0 0 4px rgba(34,211,238,.55))}@media (max-width:600px){#" + ID + "_badge{font-size:10px;padding:4px 10px 4px 6px;gap:6px;top:6px}#" + ID + "_logo{height:13px}}";
+    style.textContent = "@keyframes melBrandPulse{0%,100%{box-shadow:inset 0 0 9px 2px rgba(34,211,238,.5),inset 0 0 30px 7px rgba(124,111,240,.18)}50%{box-shadow:inset 0 0 16px 4px rgba(34,211,238,.8),inset 0 0 54px 13px rgba(124,111,240,.32)}}@keyframes melBrandGlow{0%,100%{opacity:.78}50%{opacity:1}}#" + ID + "_border{position:fixed;inset:0;pointer-events:none;z-index:2147483647;border-radius:7px;animation:melBrandPulse 2.6s ease-in-out infinite}#" + ID + "_badge{position:fixed;top:10px;left:50%;transform:translateX(-50%);pointer-events:none;z-index:2147483647;font-family:-apple-system,Segoe UI,Roboto,sans-serif;font-weight:600;font-size:12px;line-height:16px;white-space:nowrap;max-width:calc(100vw - 24px);box-sizing:border-box;color:#E2E8F0;background:linear-gradient(180deg,rgba(15,23,42,.94),rgba(15,23,42,.82));padding:5px 13px 5px 7px;border-radius:999px;letter-spacing:.2px;border:1px solid rgba(34,211,238,.45);box-shadow:0 0 16px 1px rgba(34,211,238,.4),0 4px 14px rgba(0,0,0,.45);display:flex;align-items:center;gap:8px;backdrop-filter:blur(6px);animation:melBrandGlow 2.6s ease-in-out infinite}#" + ID + "_logo{height:16px;width:auto;display:block;filter:drop-shadow(0 0 4px rgba(34,211,238,.55))}@media (max-width:600px){#" + ID + "_badge{font-size:10px;line-height:13px;padding:4px 10px 4px 6px;gap:6px;top:6px}#" + ID + "_logo{height:13px}}";
     (document.head || root).appendChild(style);
     const border = document.createElement("div"); border.id = ID + "_border";
     const badge = document.createElement("div"); badge.id = ID + "_badge";
@@ -70,6 +70,120 @@ const MELAYA_BRAND_SCRIPT = `(() => {
   if (document.readyState === "loading") document.addEventListener("DOMContentLoaded", inject); else inject();
   try { new MutationObserver(() => { if (!document.getElementById(ID + "_border")) inject(); }).observe(document.documentElement, { childList: true }); } catch (e) {}
 })();`;
+// ── Stealth init script ────────────────────────────────────────────────────
+// Runs in EVERY frame BEFORE page scripts (context.addInitScript). A headed
+// Brave/Chrome already carries a genuine fingerprint, so this only patches the
+// handful of vectors a CDP-driven session can still leak (webdriver flag, an
+// inconsistent Notification permission, a missing chrome stub, empty plugins,
+// masked WebGL vendor). It cannot hide the CDP transport itself — that is why
+// behaviour is also humanised below — but it defeats the common JS bot checks.
+const STEALTH_SCRIPT = `(() => {
+  const def = (o, k, get) => { try { Object.defineProperty(o, k, { get, configurable: true }); } catch (e) {} };
+  // navigator.webdriver → false (belt-and-suspenders; the launch flag already
+  // disables it, but some builds still expose it truthy under CDP).
+  def(Object.getPrototypeOf(navigator) || navigator, "webdriver", () => false);
+  try { delete Navigator.prototype.webdriver; } catch (e) {}
+  // Minimal chrome runtime stub when absent (headless-style gap).
+  try { window.chrome = window.chrome || {}; window.chrome.runtime = window.chrome.runtime || {}; } catch (e) {}
+  def(navigator, "languages", () => ["en-US", "en"]);
+  // Non-empty plugins/mimeTypes so length checks pass.
+  try { if (!navigator.plugins || navigator.plugins.length === 0) def(navigator, "plugins", () => [1, 2, 3, 4, 5]); } catch (e) {}
+  // Notification permission must agree between Notification.permission and
+  // permissions.query — a classic mismatch CDP introduces.
+  try {
+    const q = navigator.permissions && navigator.permissions.query ? navigator.permissions.query.bind(navigator.permissions) : null;
+    if (q) navigator.permissions.query = (p) => (p && p.name === "notifications")
+      ? Promise.resolve({ state: (typeof Notification !== "undefined" ? Notification.permission : "default"), onchange: null })
+      : q(p);
+  } catch (e) {}
+  // Mask WebGL UNMASKED_VENDOR/RENDERER (37445/37446) to a plausible GPU.
+  try {
+    const patch = (proto) => { if (!proto) return; const gp = proto.getParameter; if (!gp) return; proto.getParameter = function (p) { if (p === 37445) return "Intel Inc."; if (p === 37446) return "Intel Iris OpenGL Engine"; return gp.call(this, p); }; };
+    patch(window.WebGLRenderingContext && WebGLRenderingContext.prototype);
+    patch(window.WebGL2RenderingContext && WebGL2RenderingContext.prototype);
+  } catch (e) {}
+  def(navigator, "hardwareConcurrency", () => 8);
+  def(navigator, "deviceMemory", () => 8);
+})();`;
+// ── Humanized input helpers ────────────────────────────────────────────────
+// Instant teleport-clicks and zero-delay typing are a behavioural bot tell.
+// These move the pointer along an eased, jittered Bezier path, dwell before
+// pressing, type with variable per-key cadence, and scroll in small steps.
+// (PWPage is already aliased above to import("playwright").Page.)
+const _rand = (min, max) => min + Math.random() * (max - min);
+const _randInt = (min, max) => Math.floor(_rand(min, max + 1));
+const _lastMouse = new WeakMap();
+async function humanPause(page, min = 40, max = 140) {
+    await page.waitForTimeout(_randInt(min, max));
+}
+// Eased, jittered pointer travel from the last known position to (tx,ty).
+async function humanMove(page, tx, ty) {
+    const vp = page.viewportSize() ?? { width: 1280, height: 800 };
+    const from = _lastMouse.get(page) ?? { x: _rand(0, vp.width), y: _rand(0, vp.height) };
+    const dist = Math.hypot(tx - from.x, ty - from.y);
+    const steps = Math.max(6, Math.min(42, Math.round(dist / _rand(18, 34))));
+    const curve = Math.min(90, dist * 0.28);
+    const midX = (from.x + tx) / 2 + _rand(-1, 1) * curve;
+    const midY = (from.y + ty) / 2 + _rand(-1, 1) * curve;
+    for (let i = 1; i <= steps; i++) {
+        const t = i / steps;
+        const e = t < 0.5 ? 2 * t * t : -1 + (4 - 2 * t) * t; // easeInOutQuad
+        const mt = 1 - e;
+        let x = mt * mt * from.x + 2 * mt * e * midX + e * e * tx;
+        let y = mt * mt * from.y + 2 * mt * e * midY + e * e * ty;
+        if (i < steps) {
+            x += _rand(-1.2, 1.2);
+            y += _rand(-1.2, 1.2);
+        } // hand tremor
+        await page.mouse.move(x, y);
+        if (i % _randInt(3, 5) === 0)
+            await page.waitForTimeout(_randInt(4, 16));
+    }
+    _lastMouse.set(page, { x: tx, y: ty });
+}
+// Move to a point a few px off the exact center, dwell, then press/release.
+async function humanClick(page, x, y, clickCount = 1) {
+    const jx = x + _rand(-3, 3), jy = y + _rand(-3, 3);
+    await humanMove(page, jx, jy);
+    await humanPause(page, 30, 120);
+    for (let c = 0; c < clickCount; c++) {
+        await page.mouse.down();
+        await page.waitForTimeout(_randInt(40, 110)); // press dwell
+        await page.mouse.up();
+        if (c + 1 < clickCount)
+            await page.waitForTimeout(_randInt(60, 130));
+    }
+    _lastMouse.set(page, { x: jx, y: jy });
+}
+// Type with variable cadence: slower on spaces, longer at punctuation, and an
+// occasional "thinking" pause.
+async function humanType(page, text) {
+    for (const ch of text) {
+        await page.keyboard.type(ch);
+        let d = _rand(45, 150);
+        if (ch === " ")
+            d += _rand(10, 60);
+        if (/[.,!?;:]/.test(ch))
+            d += _rand(80, 220);
+        if (Math.random() < 0.04)
+            d += _rand(180, 450);
+        await page.waitForTimeout(Math.round(d));
+    }
+}
+// Break a wheel scroll into a few smaller, jittered steps with pauses.
+async function humanScroll(page, dx, dy) {
+    const steps = _randInt(3, 6);
+    let sx = 0, sy = 0;
+    for (let i = 0; i < steps; i++) {
+        const px = Math.round(dx / steps + _rand(-8, 8));
+        const py = Math.round(dy / steps + _rand(-14, 14));
+        await page.mouse.wheel(px, py);
+        sx += px;
+        sy += py;
+        await page.waitForTimeout(_randInt(40, 130));
+    }
+    return { dx: sx, dy: sy };
+}
 const MAX_BODY_BYTES = 512 * 1024; // /browser/run scripts included
 const NODE_CAP = 240; // snapshot node cap (Section 8)
 const SCREENSHOT_MAX_B64 = 1_400_000; // ~1 MB binary; ladder recompresses
@@ -260,17 +374,24 @@ export async function startBrowserBridge(opts) {
                     headless: reg.spec.headless === true,
                     viewport: { width: 1280, height: 800 },
                     acceptDownloads: false, // downloads default-denied (Section 10)
+                    ignoreDefaultArgs: ["--enable-automation"],
                     args: [
                         "--no-first-run",
                         "--no-default-browser-check",
                         "--disable-background-networking",
                         "--disable-sync",
+                        "--disable-blink-features=AutomationControlled",
                     ],
                 });
                 sessions.attachHandles(rec, {
                     context,
                     ephemeralUserDataDir: ephemeral ? userDataDir : null,
                 });
+                // Stealth patches before any page script (see STEALTH_SCRIPT).
+                try {
+                    await context.addInitScript(STEALTH_SCRIPT);
+                }
+                catch { /* non-fatal */ }
                 await enforceOnContext(context, reg.policy, hooks);
                 const page = context.pages()[0] ?? (await context.newPage());
                 sessions.leaseTarget(rec, reg.spec.grant.target.ref, page);
@@ -401,6 +522,14 @@ export async function startBrowserBridge(opts) {
             // all navigations and new pages. The one-shot evaluate covers the
             // already-open page (typically about:blank) immediately.
             // Both are wrapped so a branding failure cannot fail the launch.
+            // Stealth FIRST (runs before page + brand scripts in every frame) so the
+            // navigator/webgl patches are in place before any site fingerprinting.
+            try {
+                await context.addInitScript(STEALTH_SCRIPT);
+            }
+            catch (stealthErr) {
+                log(`[browser-launch] stealth init-script install failed (non-fatal): ${stealthErr?.message || stealthErr}`);
+            }
             try {
                 await context.addInitScript(MELAYA_BRAND_SCRIPT);
             }
@@ -852,16 +981,16 @@ export async function startBrowserBridge(opts) {
                         }
                         const pt = { x: box.x + box.width / 2, y: box.y + box.height / 2 };
                         if (kind === "hover")
-                            await page.mouse.move(pt.x, pt.y);
+                            await humanMove(page, pt.x, pt.y);
                         else
-                            await page.mouse.click(pt.x, pt.y, { clickCount: kind === "dblclick" ? 2 : 1 });
+                            await humanClick(page, pt.x, pt.y, kind === "dblclick" ? 2 : 1);
                         return { at: { x: Math.round(pt.x), y: Math.round(pt.y) }, matched_text: label };
                     }
                     const pt = await resolveActionPoint(reg, rec, lease, args);
                     if (kind === "hover")
-                        await page.mouse.move(pt.x, pt.y);
+                        await humanMove(page, pt.x, pt.y);
                     else
-                        await page.mouse.click(pt.x, pt.y, { clickCount: kind === "dblclick" ? 2 : 1 });
+                        await humanClick(page, pt.x, pt.y, kind === "dblclick" ? 2 : 1);
                     return { at: { x: Math.round(pt.x), y: Math.round(pt.y) } };
                 }
                 case "tap": {
@@ -869,7 +998,7 @@ export async function startBrowserBridge(opts) {
                     // browser.py sends {x, y} as 0..1 fractions or CSS px.
                     // resolveActionPoint already handles the fraction/px heuristic.
                     const pt = await resolveActionPoint(reg, rec, lease, args);
-                    await page.mouse.click(pt.x, pt.y);
+                    await humanClick(page, pt.x, pt.y);
                     return { at: { x: Math.round(pt.x), y: Math.round(pt.y) } };
                 }
                 case "input_text": {
@@ -877,10 +1006,11 @@ export async function startBrowserBridge(opts) {
                     if (text.length > 20_000)
                         throw new BridgeError("act_args_invalid", "text exceeds 20000 chars");
                     const pt = await resolveActionPoint(reg, rec, lease, args);
-                    await page.mouse.click(pt.x, pt.y);
-                    // Clear-then-type: select-all + type preserves site key handlers.
+                    await humanClick(page, pt.x, pt.y);
+                    // Clear-then-type: select-all + humanized type preserves site key handlers.
                     await page.keyboard.press(process.platform === "darwin" ? "Meta+A" : "Control+A").catch(() => { });
-                    await page.keyboard.type(text, { delay: 8 });
+                    await humanPause(page, 40, 120);
+                    await humanType(page, text);
                     return { typed: text.length };
                 }
                 case "press_key": {
@@ -920,8 +1050,8 @@ export async function startBrowserBridge(opts) {
                     else {
                         dy = Math.max(-4000, Math.min(4000, Number(args.dy ?? 600)));
                     }
-                    await page.mouse.wheel(dx, dy);
-                    return { scrolled: { dx, dy } };
+                    const scrolled = await humanScroll(page, dx, dy);
+                    return { scrolled };
                 }
                 case "get_text": {
                     // Fix 1: read-only page-text extraction.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@melaya/runner",
-  "version": "1.1.12",
+  "version": "1.1.14",
   "description": "Run Melaya AI pipelines locally with your own LM Studio or Ollama models",
   "license": "UNLICENSED",
   "private": false,