feat: support openai structured output as objecct

camel/agents/chat_agent.py

@@ -1090,6 +1090,7 @@ def handle_batch_response(
                 role_type=self.role_type,
                 meta_dict=dict(),
                 content=choice.message.content or "",
+                parsed=getattr(choice.message, 'parsed', None),
             )
             # Process log probabilities and append to the message meta information
             if choice.logprobs is not None:

camel/messages/base.py

@@ -15,10 +15,11 @@
 import io
 import re
 from dataclasses import dataclass
-from typing import Any, Dict, List, Literal, Optional, Tuple, Union
+from typing import Any, Dict, List, Literal, Optional, Tuple, Type, Union
 
 import numpy as np
 from PIL import Image
+from pydantic import BaseModel
 
 from camel.messages import (
     FunctionCallFormatter,
@@ -51,13 +52,15 @@ class BaseMessage:
             for the message.
         content (str): The content of the message.
         video_bytes (Optional[bytes]): Optional bytes of a video associated
-            with the message. Default is None.
+            with the message. (default::obj:`None`)
         image_list (Optional[List[Image.Image]]): Optional list of PIL Image
-            objects associated with the message. Default is None.
+            objects associated with the message. (default::obj:`None`)
         image_detail (Literal["auto", "low", "high"]): Detail level of the
-            images associated with the message. Default is "auto".
+            images associated with the message. (default::obj:`auto`)
         video_detail (Literal["auto", "low", "high"]): Detail level of the
-            videos associated with the message. Default is "low".
+            videos associated with the message. (default::obj:`low`)
+        parsed: Optional[Union[Type[BaseModel], dict]]: Optional object which
+            is parsed from the content. (default::obj:`None`)
     """
 
     role_name: str
@@ -69,6 +72,7 @@ class BaseMessage:
     image_list: Optional[List[Image.Image]] = None
     image_detail: Literal["auto", "low", "high"] = "auto"
     video_detail: Literal["auto", "low", "high"] = "low"
+    parsed: Optional[Union[Type[BaseModel], dict]] = None
 
     @classmethod
     def make_user_message(
@@ -419,7 +423,6 @@ def to_openai_user_message(self) -> OpenAIUserMessage:
                 "text": self.content,
             }
         )
-
         if self.image_list and len(self.image_list) > 0:
             for image in self.image_list:
                 if image.format is None:

camel/models/openai_model.py

@@ -158,6 +158,7 @@ def _to_chat_completion(
                 "role": response.choices[0].message.role,
                 "content": response.choices[0].message.content,
                 "tool_calls": response.choices[0].message.tool_calls,
+                "parsed": response.choices[0].message.parsed,
             },
             finish_reason=response.choices[0].finish_reason,
         )

examples/models/openai_structured_output_example.py

@@ -25,7 +25,7 @@ class Student(BaseModel):
 
 
 class StudentList(BaseModel):
-    studentLlst: list[Student]
+    studentList: list[Student]
 
 
 openai_model = ModelFactory.create(
@@ -45,10 +45,15 @@ class StudentList(BaseModel):
 # Get response information
 response = camel_agent.step(user_msg)
 print(response.msgs[0].content)
+print(response.msgs[0].parsed)
 '''
 ===============================================================================
 {"studentLlst":[{"name":"Alice Johnson","age":"20"},{"name":"Brian Smith",
 "age":"22"},{"name":"Catherine Lee","age":"19"},{"name":"David Brown",
 "age":"21"},{"name":"Eva White","age":"20"}]}
+
+studentList=[Student(name='Alice Johnson', age='20'), Student(name=
+'Brian Smith', age='22'), Student(name='Catherine Lee', age='19'),
+Student(name='David Brown', age='21'), Student(name='Eva White', age='23')]
 ===============================================================================
 '''