omnius 1.0.31 → 1.0.32

package/dist/scripts/web_scrape.py

@@ -140,6 +140,26 @@ def log_message(msg: str, level: str = "INFO") -> None:
     print(f"[{ts}] [{level.upper()}] {msg}")
 
 
+def _bounded_int(value, default: int, minimum: int, maximum: int) -> int:
+    try:
+        n = int(float(value))
+    except Exception:
+        n = default
+    return max(minimum, min(maximum, n))
+
+
+def _bounded_float(value, default: float, minimum: float, maximum: float) -> float:
+    try:
+        n = float(value)
+    except Exception:
+        n = default
+    return max(minimum, min(maximum, n))
+
+
+def _truthy(value) -> bool:
+    return str(value).lower() in ("1", "true", "yes", "on")
+
+
 class Tools:
     _driver: Optional[webdriver.Chrome] = None
 
@@ -163,7 +183,13 @@ class Tools:
         return None
 
     @staticmethod
-    def open_browser(headless: bool = False, force_new: bool = False) -> str:
+    def open_browser(
+        headless: bool = False,
+        force_new: bool = False,
+        width: int = 1280,
+        height: int = 720,
+        device_scale_factor: float = 1.0,
+    ) -> str:
         if force_new and Tools._driver:
             try:
                 Tools._driver.quit()
@@ -184,12 +210,16 @@ class Tools:
             or "/usr/bin/chromium"
         )
 
+        width = _bounded_int(width, 1280, 320, 3840)
+        height = _bounded_int(height, 720, 240, 2160)
+        device_scale_factor = _bounded_float(device_scale_factor, 1.0, 0.25, 3.0)
+
         opts = Options()
         if chrome_bin:
             opts.binary_location = chrome_bin
         if headless:
             opts.add_argument("--headless=new")
-        opts.add_argument("--window-size=1920,1080")
+        opts.add_argument(f"--window-size={width},{height}")
         opts.add_argument("--disable-gpu")
         opts.add_argument("--no-sandbox")
         opts.add_argument("--disable-dev-shm-usage")
@@ -199,6 +229,7 @@ class Tools:
         try:
             log_message("[open_browser] Trying Selenium-Manager…", "DEBUG")
             Tools._driver = webdriver.Chrome(options=opts)
+            Tools.set_viewport(width, height, device_scale_factor)
             log_message("[open_browser] Launched via Selenium-Manager.", "SUCCESS")
             return "Browser launched (selenium-manager)"
         except WebDriverException as e:
@@ -209,6 +240,7 @@ class Tools:
             try:
                 log_message(f"[open_browser] Using snap chromedriver at {snap_drv}", "DEBUG")
                 Tools._driver = webdriver.Chrome(service=Service(snap_drv), options=opts)
+                Tools.set_viewport(width, height, device_scale_factor)
                 log_message("[open_browser] Launched via snap chromedriver.", "SUCCESS")
                 return "Browser launched (snap chromedriver)"
             except WebDriverException as e:
@@ -219,6 +251,7 @@ class Tools:
             try:
                 log_message(f"[open_browser] Trying system chromedriver at {sys_drv}", "DEBUG")
                 Tools._driver = webdriver.Chrome(service=Service(sys_drv), options=opts)
+                Tools.set_viewport(width, height, device_scale_factor)
                 log_message("[open_browser] Launched via system chromedriver.", "SUCCESS")
                 return "Browser launched (system chromedriver)"
             except WebDriverException as e:
@@ -242,6 +275,7 @@ class Tools:
                 drv = shutil.which("chromedriver")
                 log_message(f"[open_browser] Installed ARM64 driver at {drv}", "DEBUG")
                 Tools._driver = webdriver.Chrome(service=Service(drv), options=opts)
+                Tools.set_viewport(width, height, device_scale_factor)
                 log_message("[open_browser] Launched via downloaded ARM64 chromedriver.", "SUCCESS")
                 return "Browser launched (downloaded ARM64 chromedriver)"
             except Exception as e:
@@ -257,6 +291,7 @@ class Tools:
                 log_message(f"[open_browser] Installing ChromeDriver {browser_major} via webdriver-manager", "DEBUG")
                 drv_path = ChromeDriverManager(driver_version=browser_major).install()
                 Tools._driver = webdriver.Chrome(service=Service(drv_path), options=opts)
+                Tools.set_viewport(width, height, device_scale_factor)
                 log_message("[open_browser] Launched via webdriver-manager.", "SUCCESS")
                 return "Browser launched (webdriver-manager)"
             except Exception as e:
@@ -266,6 +301,7 @@ class Tools:
             log_message("[open_browser] Attempting `sudo snap install chromium`…", "DEBUG")
             subprocess.check_call(["sudo", "snap", "install", "chromium"])
             Tools._driver = webdriver.Chrome(service=Service(snap_drv), options=opts)
+            Tools.set_viewport(width, height, device_scale_factor)
             log_message("[open_browser] Launched via newly-installed snap chromium.", "SUCCESS")
             return "Browser launched (snap install fallback)"
         except Exception as e:
@@ -292,6 +328,46 @@ class Tools:
     def is_browser_open() -> bool:
         return Tools._driver is not None
 
+    @staticmethod
+    def set_viewport(width: int = 1280, height: int = 720, device_scale_factor: float = 1.0) -> Dict[str, float]:
+        if not Tools._driver:
+            raise RuntimeError("browser not open")
+        width = _bounded_int(width, 1280, 320, 3840)
+        height = _bounded_int(height, 720, 240, 2160)
+        device_scale_factor = _bounded_float(device_scale_factor, 1.0, 0.25, 3.0)
+        try:
+            Tools._driver.set_window_size(width, height)
+        except Exception as exc:
+            log_message(f"[viewport] set_window_size failed: {exc}", "WARNING")
+        try:
+            Tools._driver.execute_cdp_cmd("Emulation.setDeviceMetricsOverride", {
+                "mobile": False,
+                "width": width,
+                "height": height,
+                "deviceScaleFactor": device_scale_factor,
+            })
+        except Exception as exc:
+            log_message(f"[viewport] CDP override failed: {exc}", "DEBUG")
+        return Tools.viewport()
+
+    @staticmethod
+    def viewport() -> Dict[str, float]:
+        if not Tools._driver:
+            return {"width": 0, "height": 0, "deviceScaleFactor": 1}
+        try:
+            data = Tools._driver.execute_script(
+                "return {width: window.innerWidth, height: window.innerHeight, deviceScaleFactor: window.devicePixelRatio || 1};"
+            )
+            if isinstance(data, dict):
+                return {
+                    "width": int(data.get("width") or 0),
+                    "height": int(data.get("height") or 0),
+                    "deviceScaleFactor": float(data.get("deviceScaleFactor") or 1),
+                }
+        except Exception as exc:
+            log_message(f"[viewport] read failed: {exc}", "DEBUG")
+        return {"width": 0, "height": 0, "deviceScaleFactor": 1}
+
     @staticmethod
     def navigate(url: str) -> str:
         if not Tools._driver:
@@ -357,9 +433,36 @@ class Tools:
             return f"Error scrolling: {exc}"
 
     @staticmethod
-    def screenshot(filename: str = "screenshot.png") -> str:
+    def screenshot(filename: str = "screenshot.png", full_page: bool = False) -> str:
         if not Tools._driver:
             return "Error: browser not open"
+        # Prefer Chrome DevTools Protocol capture. It renders the browser
+        # surface directly, so this path cannot accidentally fall back to an
+        # OS desktop screenshot tool.
+        try:
+            drv = Tools._driver
+            layout = drv.execute_cdp_cmd("Page.getLayoutMetrics", {})
+            if full_page:
+                content = layout.get("contentSize") or {}
+                width = int(content.get("width") or 1280)
+                height = int(content.get("height") or 720)
+                clip = {"x": 0, "y": 0, "width": width, "height": height, "scale": 1}
+            else:
+                viewport = Tools.viewport()
+                width = int(viewport.get("width") or 1280)
+                height = int(viewport.get("height") or 720)
+                clip = {"x": 0, "y": 0, "width": width, "height": height, "scale": 1}
+            data = drv.execute_cdp_cmd("Page.captureScreenshot", {
+                "format": "png",
+                "fromSurface": True,
+                "captureBeyondViewport": bool(full_page),
+                "clip": clip,
+            })
+            raw = base64.b64decode(data.get("data") or "")
+            Path(filename).write_bytes(raw)
+            return filename
+        except Exception as exc:
+            log_message(f"[screenshot] CDP capture failed, falling back to webdriver screenshot: {exc}", "WARNING")
         Tools._driver.save_screenshot(filename)
         return filename
 
@@ -827,8 +930,17 @@ def session_start():
         return _error("unauthorized", 401)
     payload = request.get_json(silent=True) or {}
     headless = bool(payload.get("headless", HEADLESS_DEFAULT))
+    width = _bounded_int(payload.get("width"), 1280, 320, 3840)
+    height = _bounded_int(payload.get("height"), 720, 240, 2160)
+    device_scale_factor = _bounded_float(payload.get("deviceScaleFactor") or payload.get("device_scale_factor"), 1.0, 0.25, 3.0)
     with _slot():
-        msg = Tools.open_browser(headless=headless, force_new=True)
+        msg = Tools.open_browser(
+            headless=headless,
+            force_new=True,
+            width=width,
+            height=height,
+            device_scale_factor=device_scale_factor,
+        )
     if not _result_ok(msg):
         return _error(msg, 500)
     sid = uuid.uuid4().hex
@@ -842,7 +954,7 @@ def session_start():
             "frames": {},
         }
     _queue_event(sid, {"type": "status", "msg": "browser_started", "detail": msg, "sid": sid, "ts": int(time.time() * 1000)})
-    return _ok(session_id=sid, message=msg, headless=headless)
+    return _ok(session_id=sid, message=msg, headless=headless, viewport=Tools.viewport())
 
 
 @app.post("/session/close")
@@ -855,6 +967,29 @@ def session_close():
     return _ok(message=msg)
 
 
+@app.route("/viewport", methods=["GET", "POST"])
+def viewport():
+    if not _auth_ok(request):
+        return _error("unauthorized", 401)
+    if not Tools.is_browser_open():
+        return _error("browser not open", 409)
+    if request.method == "POST":
+        payload = request.get_json(silent=True) or {}
+        width = _bounded_int(payload.get("width"), 1280, 320, 3840)
+        height = _bounded_int(payload.get("height"), 720, 240, 2160)
+        device_scale_factor = _bounded_float(
+            payload.get("deviceScaleFactor") or payload.get("device_scale_factor"),
+            1.0,
+            0.25,
+            3.0,
+        )
+        with _slot():
+            vp = Tools.set_viewport(width, height, device_scale_factor)
+    else:
+        vp = Tools.viewport()
+    return _ok(viewport=vp, width=vp.get("width"), height=vp.get("height"), deviceScaleFactor=vp.get("deviceScaleFactor"))
+
+
 @app.post("/navigate")
 def navigate():
     if not _auth_ok(request):
@@ -1200,10 +1335,20 @@ def screenshot():
     if not _auth_ok(request):
         return _error("unauthorized", 401)
     sid = request.args.get("sid") or next(iter(_SESSIONS), "")
+    width_raw = request.args.get("width")
+    height_raw = request.args.get("height")
+    scale_raw = request.args.get("deviceScaleFactor") or request.args.get("device_scale_factor")
+    if width_raw or height_raw or scale_raw:
+        width = _bounded_int(width_raw, 1280, 320, 3840)
+        height = _bounded_int(height_raw, 720, 240, 2160)
+        device_scale_factor = _bounded_float(scale_raw, 1.0, 0.25, 3.0)
+        with _slot():
+            Tools.set_viewport(width, height, device_scale_factor)
+    full_page = _truthy(request.args.get("full_page") or request.args.get("fullPage"))
     fname = f"{uuid.uuid4().hex}.png"
     fpath = OUT_DIR / fname
     with _slot():
-        msg = Tools.screenshot(str(fpath))
+        msg = Tools.screenshot(str(fpath), full_page=full_page)
     if not _result_ok(msg):
         return _error(msg, 500)
     try:

package/npm-shrinkwrap.json

@@ -1,12 +1,12 @@
 {
   "name": "omnius",
-  "version": "1.0.31",
+  "version": "1.0.32",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "omnius",
-      "version": "1.0.31",
+      "version": "1.0.32",
       "bundleDependencies": [
         "image-to-ascii"
       ],

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "omnius",
-  "version": "1.0.31",
+  "version": "1.0.32",
   "description": "AI coding agent powered by open-source models (Ollama/vLLM) — interactive TUI with agentic tool-calling loop",
   "type": "module",
   "main": "./dist/index.js",