fix: wenxin error not raise when stream mode (#884)

api/core/third_party/langchain/llms/wenxin.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 
 import json
 import logging
+from json import JSONDecodeError
 from typing import (
     Any,
     Dict,
@@ -223,6 +224,8 @@ class Wenxin(LLM):
         for token in self._client.post(request).iter_lines():
             if token:
                 token = token.decode("utf-8")
+
+                if token.startswith('data:'):
                     completion = json.loads(token[5:])
 
                     yield GenerationChunk(text=completion['result'])
@@ -231,3 +234,14 @@ class Wenxin(LLM):
 
                     if completion['is_end']:
                         break
+                else:
+                    try:
+                        json_response = json.loads(token)
+                    except JSONDecodeError:
+                        raise ValueError(f"Wenxin Response Error {token}")
+
+                    raise ValueError(
+                        f"Wenxin API {json_response['error_code']}"
+                        f" error: {json_response['error_msg']}, "
+                        f"please confirm if the model you have chosen is already paid for."
+                    )