fix(sdk): preserve JSON document downloads (#19367)

Fix SDK Document.download() — check Content-Disposition: attachment before JSON parsing. Arrays/scalars/plain files now work.

sdk/python/ragflow_sdk/modules/document.py

@@ -14,11 +14,13 @@
 #  limitations under the License.
 #
 
-import json
+import logging
 
 from .base import Base
 from .chunk import Chunk
 
+logger = logging.getLogger(__name__)
+
 
 class Document(Base):
     class ParserConfig(Base):
@@ -64,16 +66,19 @@ class Document(Base):
 
     def download(self):
         res = self.get(f"/datasets/{self.dataset_id}/documents/{self.id}")
-        error_keys = set(["code", "message"])
+        content_disposition = res.headers.get("Content-Disposition", "")
+        if content_disposition.lstrip().lower().startswith("attachment"):
+            logger.debug("Document.download returning attachment content dataset_id=%s document_id=%s", self.dataset_id, self.id)
+            return res.content
+
         try:
             response = res.json()
-            actual_keys = set(response.keys())
-            if actual_keys == error_keys:
+            if isinstance(response, dict) and "code" in response and response["code"] != 0:
                 raise Exception(response.get("message"))
-            else:
-                return res.content
-        except json.JSONDecodeError:
-            return res.content
+        except ValueError:
+            pass
+        logger.debug("Document.download returning non-attachment content dataset_id=%s document_id=%s", self.dataset_id, self.id)
+        return res.content
 
     def list_chunks(self, page=1, page_size=30, keywords="", id=""):
         data = {"keywords": keywords, "page": page, "page_size": page_size, "id": id}