install: show wait progress, shorten OCR health wait, log to file.

deploy/ocr-worker/app.py

@@ -110,27 +110,14 @@ def run_ocr_on_bytes(content: bytes) -> dict:
     }
 
 
-@app.on_event("startup")
-def warmup():
-    """延迟预热，不阻塞 HTTP 服务启动。"""
-    import threading
-
-    def _run():
-        buf = __import__("io").BytesIO()
-        Image.new("RGB", (120, 40), color=(255, 255, 255)).save(buf, format="JPEG")
-        try:
-            run_ocr_on_bytes(buf.getvalue())
-        except Exception:
-            pass
-
-    threading.Thread(target=_run, daemon=True, name="ocr-warmup").start()
-
-
 @app.get("/health")
 def health():
     return {"status": "ok", "gpu": OCR_USE_GPU}
 
 
+# 首次 /api/ocr/regions 请求时再加载模型（/health 立即响应，避免安装脚本长时间等待）
+
+
 @app.post("/api/ocr/regions")
 async def ocr_regions(
     file: UploadFile = File(...),