Classify malformed-multimodal rejects as invalid_request (#27451)

Co-authored-by: cctry <cctry@meta.com>
Co-authored-by: cctry <cctry@fb.com>

python/sglang/srt/entrypoints/http_server.py

@@ -726,7 +726,21 @@ async def generate_request(obj: GenerateReqInput, request: Request):
                 ):
                     yield b"data: " + dumps_json(out) + b"\n\n"
             except ValueError as e:
-                out = {"error": {"message": str(e)}}
+                # A client disconnect also surfaces here. It's a client-side
+                # cancellation, not a server error or bad input -- log it and
+                # stop (the request was already aborted upstream) instead of
+                # emitting a 400.
+                if request is not None and await request.is_disconnected():
+                    logger.info(f"[http_server] Client disconnected: {e}")
+                    return
+                out = {
+                    "error": {
+                        "message": str(e),
+                        "type": "invalid_request_error",
+                        "code": 400,
+                        "retryable": False,
+                    }
+                }
                 logger.error(f"[http_server] Error: {e}")
                 yield b"data: " + dumps_json(out) + b"\n\n"
             yield b"data: [DONE]\n\n"

python/sglang/srt/multimodal/processors/base_processor.py

@@ -549,8 +549,17 @@ class BaseMultimodalProcessor(ABC):
             elif modality == Modality.AUDIO:
                 return load_audio(data, audio_sample_rate)
 
+        except ValueError as e:
+            # Bad input (e.g. invalid base64) -> 400, not 500.
+            data_str = str(data)
+            if len(data_str) > 100:
+                data_str = data_str[:100] + "..."
+            raise ValueError(f"Error while loading data {data_str}: {e}") from e
         except Exception as e:
-            raise RuntimeError(f"Error while loading data {data}: {e}")
+            data_str = str(data)
+            if len(data_str) > 100:
+                data_str = data_str[:100] + "..."
+            raise RuntimeError(f"Error while loading data {data_str}: {e}") from e
 
     @staticmethod
     def _get_preprocessed_input_format(data):