Update app.py

app.py

@@ -17,7 +17,9 @@ logging.basicConfig(level=logging.INFO)
 MODEL_ID = os.getenv("GLMOCR_MODEL_ID", "zai-org/GLM-OCR")
 OCR_API_PORT = int(os.getenv("OCR_API_PORT", "8080"))
 OCR_API_BASE = f"http://127.0.0.1:{OCR_API_PORT}"
-VLLM_MAX_MODEL_LEN = os.getenv("VLLM_MAX_MODEL_LEN", "4096")
+# GLM-OCR self-hosted requests can ask up to 8192 output tokens.
+# Keep server context length aligned unless explicitly overridden by env.
+VLLM_MAX_MODEL_LEN = os.getenv("VLLM_MAX_MODEL_LEN", "8192")
 VLLM_GPU_MEMORY_UTIL = os.getenv("VLLM_GPU_MEMORY_UTIL", "0.75")
 VLLM_EXTRA_ARGS = os.getenv("VLLM_EXTRA_ARGS", "").strip()