fix: resolve UTF-8 encoding issues in HTTP request bodies

- Add charset=utf-8 to Content-Type and Accept headers
- Serialize JSON with ensure_ascii=False for proper UTF-8 encoding
- Fix body size calculation to prevent Content-Length mismatches
- Apply changes to sync HTTP, async HTTP, WebSocket, and form data paths
- Add comprehensive unit and integration tests

This fix reduces request body size by 57.5% for non-ASCII content
(e.g., 12.41MB → 5.27MB for websearch.json test data) and ensures
compatibility with servers that enforce strict body size limits.

Closes issue with large UTF-8 payloads exceeding server limits.

Author: zhansheng.lzs

dashscope/api_entities/aiohttp_request.py

@@ -57,7 +57,7 @@ def __init__(
         self.async_request = async_request
         self._external_aio_session = session
         self.headers = {
-            "Accept": "application/json",
+            "Accept": "application/json; charset=utf-8",
             "Authorization": f"Bearer {api_key}",
             "Cache-Control": "no-cache",
             **self.headers,  # type: ignore[has-type]
@@ -70,7 +70,7 @@ def __init__(
             }
         self.method = http_method
         if self.method == HTTPMethod.POST:
-            self.headers["Content-Type"] = "application/json"
+            self.headers["Content-Type"] = "application/json; charset=utf-8"
 
         self.stream = stream
         if self.stream:
@@ -286,10 +286,13 @@ async def _handle_request(self):
                             timeout=request_timeout,
                         )
                     else:
+                        body = json.dumps(obj, ensure_ascii=False).encode(
+                            "utf-8",
+                        )
                         response = await session.request(
                             "POST",
                             url=self.url,
-                            json=obj,
+                            data=body,
                             headers=self.headers,
                             timeout=request_timeout,
                         )

dashscope/api_entities/api_request_data.py

@@ -83,8 +83,14 @@ def get_aiohttp_payload(self):
                 form.add_field(key, value)
             form.add_field("model", data["model"])
             if "input" in data:
-                form.add_field("input", json.dumps(data["input"]))
-            form.add_field("parameters", json.dumps(data["parameters"]))
+                form.add_field(
+                    "input",
+                    json.dumps(data["input"], ensure_ascii=False),
+                )
+            form.add_field(
+                "parameters",
+                json.dumps(data["parameters"], ensure_ascii=False),
+            )
             return True, form()
             # pylint: disable=unreachable,pointless-string-statement
             """

dashscope/api_entities/http_request.py

@@ -90,7 +90,7 @@ def __init__(
             self._external_session = None
             self._external_aio_session = None
         self.headers: Dict = {
-            "Accept": "application/json",
+            "Accept": "application/json; charset=utf-8",
             "Authorization": f"Bearer {api_key}",
             **self.headers,
         }
@@ -115,7 +115,7 @@ def __init__(
             }
         self.method = http_method
         if self.method == HTTPMethod.POST:
-            self.headers["Content-Type"] = "application/json"
+            self.headers["Content-Type"] = "application/json; charset=utf-8"
 
         self.stream = stream
         if self.stream:
@@ -194,10 +194,13 @@ async def _handle_aio_request(self):  # pylint: disable=too-many-branches
                             timeout=request_timeout,
                         )
                     else:
+                        body = json.dumps(obj, ensure_ascii=False).encode(
+                            "utf-8",
+                        )
                         response = await session.request(
                             "POST",
                             url=self.url,
-                            json=obj,
+                            data=body,
                             headers=self.headers,
                             timeout=request_timeout,
                         )
@@ -483,10 +486,13 @@ def _handle_request(self):  # pylint: disable=too-many-branches
                         )
                     else:
                         logger.debug("Request body: %s", obj)
+                        body = json.dumps(obj, ensure_ascii=False).encode(
+                            "utf-8",
+                        )
                         response = session.post(
                             url=self.url,
                             stream=self.stream,
-                            json=obj,
+                            data=body,
                             headers={**self.headers},
                             timeout=self.timeout,
                         )

dashscope/api_entities/websocket_request.py

@@ -73,7 +73,7 @@ def __init__(
 
         self.headers = {
             "Authorization": f"bearer {api_key}",
-            **self.headers,
+            **self.headers,  # type: ignore[has-type]
         }
 
         self.task_headers = {
@@ -422,4 +422,4 @@ async def _check_websocket_unexpected_message(self, msg):
 
     def _build_up_message(self, headers, payload):
         message = {"header": headers, "payload": payload}
-        return json.dumps(message)
+        return json.dumps(message, ensure_ascii=False)

dashscope/client/base_api.py

@@ -2,6 +2,7 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 import asyncio
 import collections
+import json
 import time
 from http import HTTPStatus
 from typing import Any, Dict, Iterator, List, Union
@@ -1266,11 +1267,13 @@ def call(
         flattened_output = kwargs.pop("flattened_output", False)
         with requests.Session() as session:
             logger.debug("Starting request: %s", url)
+            body = json.dumps(data, ensure_ascii=False).encode("utf-8")
             response = session.post(
                 url,
-                json=data,
+                data=body,
                 stream=stream,
                 headers={
+                    "Content-Type": "application/json; charset=utf-8",
                     **_workspace_header(workspace),
                     **default_headers(api_key),
                     **kwargs.pop("headers", {}),
@@ -1295,7 +1298,7 @@ class UpdateMixin:
     def update(
         cls,
         target: str,
-        json: object,
+        json_body: object,
         api_key: str = None,
         path: str = None,
         workspace: str = None,
@@ -1306,7 +1309,7 @@ def update(
 
         Args:
             target (str): The target to update.
-            json (object): The create request json body.
+            json_body (object): The create request json body.
             api_key (str, optional): The api api_key, if not present,
                 will get by default rule(TODO: api key doc). Defaults to None.
 
@@ -1329,11 +1332,13 @@ def update(
         flattened_output = kwargs.pop("flattened_output", False)
         with requests.Session() as session:
             logger.debug("Starting request: %s", url)
+            body = json.dumps(json_body, ensure_ascii=False).encode("utf-8")
             if method == "post":
                 response = session.post(
                     url,
-                    json=json,
+                    data=body,
                     headers={
+                        "Content-Type": "application/json; charset=utf-8",
                         **_workspace_header(workspace),
                         **default_headers(api_key),
                         **kwargs.pop("headers", {}),
@@ -1343,8 +1348,9 @@ def update(
             else:
                 response = session.patch(
                     url,
-                    json=json,
+                    data=body,
                     headers={
+                        "Content-Type": "application/json; charset=utf-8",
                         **_workspace_header(workspace),
                         **default_headers(api_key),
                         **kwargs.pop("headers", {}),
@@ -1360,7 +1366,7 @@ class PutMixin:
     def put(
         cls,
         target: str,
-        json: object,
+        json_body: object,
         path: str = None,
         api_key: str = None,
         workspace: str = None,
@@ -1370,7 +1376,7 @@ def put(
 
         Args:
             target (str): The target to update.
-            json (object): The create request json body.
+            json_body (object): The create request json body.
             api_key (str, optional): The api api_key, if not present,
                 will get by default rule(TODO: api key doc). Defaults to None.
 
@@ -1392,10 +1398,12 @@ def put(
         )
         with requests.Session() as session:
             logger.debug("Starting request: %s", url)
+            body = json.dumps(json_body, ensure_ascii=False).encode("utf-8")
             response = session.put(
                 url,
-                json=json,
+                data=body,
                 headers={
+                    "Content-Type": "application/json; charset=utf-8",
                     **_workspace_header(workspace),
                     **default_headers(api_key),
                     **kwargs.pop("headers", {}),

dashscope/common/utils.py

@@ -166,7 +166,7 @@ def default_headers(api_key: str = None) -> Dict[str, str]:
     if api_key is None:
         api_key = get_default_api_key()
     headers["Authorization"] = f"Bearer {api_key}"
-    headers["Accept"] = "application/json"
+    headers["Accept"] = "application/json; charset=utf-8"
     return headers