Initial VRC OSC commit

src/screen/screen_actions.py

@@ -0,0 +1,167 @@
+import pyautogui
+import pygetwindow as gw
+
+from common.config_loader import loadOcrConfig
+from common.project_paths import RUNTIME_DIR
+
+SCREENSHOT_PATH = RUNTIME_DIR / "latest_screenshot.png"
+OCR_SCREENSHOT_PATH = RUNTIME_DIR / "latest_ocr.png"
+
+EXCLUDED_WINDOW_KEYWORDS = [
+    "Google Chrome",
+    "Microsoft Edge",
+    "Mozilla Firefox",
+    "Visual Studio Code",
+    "PowerShell",
+    "Windows Terminal",
+    "ChatGPT",
+]
+
+
+def log(level, message):
+    from datetime import datetime
+
+    now = datetime.now().strftime("%H:%M:%S")
+    print(f"[{level}] {now} {message}", flush=True)
+
+
+def findVrchatWindow():
+    windows = gw.getAllWindows()
+    candidates = []
+
+    for window in windows:
+        title = window.title or ""
+        normalized = title.strip().lower()
+        if not normalized:
+            continue
+        if any(keyword.lower() in normalized for keyword in EXCLUDED_WINDOW_KEYWORDS):
+            continue
+
+        if normalized == "vrchat":
+            candidates.append((0, window))
+            continue
+        if normalized.startswith("vrchat "):
+            candidates.append((1, window))
+            continue
+        if normalized.startswith("vrchat"):
+            candidates.append((2, window))
+
+    if not candidates:
+        visible_titles = []
+        for window in windows:
+            title = (window.title or "").strip()
+            if title:
+                visible_titles.append(title)
+
+        log("ERROR", "VRChatウィンドウが見つかりません")
+        log("CHECK", "現在検出できるウィンドウ")
+        for title in visible_titles[:30]:
+            print(title, flush=True)
+
+        raise RuntimeError("VRChatウィンドウが見つかりません")
+
+    candidates = sorted(
+        candidates,
+        key=lambda item: (item[0], -(item[1].width * item[1].height)),
+    )
+    window = candidates[0][1]
+
+    if window.width <= 0 or window.height <= 0:
+        raise RuntimeError("VRChatウィンドウサイズが不正です")
+
+    return window
+
+
+def getVrchatRegion():
+    window = findVrchatWindow()
+
+    try:
+        if window.isMinimized:
+            window.restore()
+    except Exception:
+        pass
+
+    left = int(window.left)
+    top = int(window.top)
+    width = int(window.width)
+    height = int(window.height)
+
+    if width <= 0 or height <= 0:
+        raise RuntimeError("VRChatウィンドウ範囲が不正です")
+
+    log("INFO", f"VRChat window title={window.title}")
+    log("INFO", f"VRChat region left={left} top={top} width={width} height={height}")
+    return left, top, width, height
+
+
+def clampRatio(value):
+    return max(0.0, min(1.0, value))
+
+
+def getOcrRegion():
+    left, top, width, height = getVrchatRegion()
+    config = loadOcrConfig()
+
+    # Default to the lower central area. It is much faster than OCRing the
+    # whole VRChat window and usually contains chat/subtitle text.
+    crop_left = clampRatio(config["crop_left"])
+    crop_top = clampRatio(config["crop_top"])
+    crop_right = clampRatio(config["crop_right"])
+    crop_bottom = clampRatio(config["crop_bottom"])
+
+    if crop_right <= crop_left:
+        crop_left = 0.0
+        crop_right = 1.0
+
+    if crop_bottom <= crop_top:
+        crop_top = 0.0
+        crop_bottom = 1.0
+
+    region_left = left + int(width * crop_left)
+    region_top = top + int(height * crop_top)
+    region_width = max(1, int(width * (crop_right - crop_left)))
+    region_height = max(1, int(height * (crop_bottom - crop_top)))
+
+    log(
+        "INFO",
+        f"OCR crop left={region_left} top={region_top} width={region_width} height={region_height}",
+    )
+    return region_left, region_top, region_width, region_height
+
+
+def captureRegionTo(image_path, region):
+    image_path.parent.mkdir(parents=True, exist_ok=True)
+
+    left, top, width, height = region
+
+    image = pyautogui.screenshot(region=(left, top, width, height))
+    image.save(image_path)
+
+    log("ACTION", f"captureVrchatWindow saved={image_path}")
+    return image_path
+
+
+def captureVrchatWindowTo(image_path):
+    return captureRegionTo(
+        image_path,
+        getVrchatRegion(),
+    )
+
+
+def captureVrchatWindow():
+    return captureVrchatWindowTo(SCREENSHOT_PATH)
+
+
+def captureOcrRegion():
+    return captureRegionTo(
+        OCR_SCREENSHOT_PATH,
+        getOcrRegion(),
+    )
+
+
+def captureScreen():
+    return captureVrchatWindow()
+
+
+def takeScreenshot():
+    return captureVrchatWindow()