fix: change agent tool

src/codeinterpreterapi/agents/config/coding_agent/config.yaml

@@ -1,6 +1,6 @@
 agent_definition:
   agent_name: "main_function_create_agent"
-  agent_type: structured_chat
+  agent_type: tool_calling
   agent_role: |
     pythonプログラムのmain関数を作成する作業をお願いします。
     別のagentはプログラムの分割やテストの作成を実施することになります。

src/codeinterpreterapi/agents/config/split_agent/config.yaml

@@ -1,6 +1,6 @@
 agent_definition:
   agent_name: "code_split_agent"
-  agent_type: structured_chat
+  agent_type: tool_calling
   agent_role: |
     プログラムを分割する作業をお願いします。
     別のagentはプログラムの作成やテストの作成を実施することになります。

src/codeinterpreterapi/agents/structured_chat/agent.py

@@ -172,7 +172,7 @@ def create_structured_chat_agent(
         | llm_with_stop
     )
     agent = assign_runnable_history(agent, runnable_config)
-    agent = agent | output_parser
+    # agent = agent | output_parser
     return agent
 
 

src/codeinterpreterapi/crew/custom_agent.py

@@ -4,6 +4,8 @@
 from langchain_core.tools import BaseTool
 from pydantic import Field
 
+from codeinterpreterapi.utils.multi_converter import MultiConverter
+
 
 class CustomAgent(BaseAgent):
     agent_executor: Any = Field(default=None, description="Verbose mode for the Agent Execution")
@@ -28,7 +30,11 @@ def interpolate_inputs(self, inputs: Dict[str, Any]) -> None:
         print("interpolate_inputs inputs=", inputs)
         super().interpolate_inputs(inputs)
 
-    def execute_task(self, task: Any, context: Optional[str] = None, tools: Optional[List[Any]] = None):
+    def execute_task(self, task: Any, context: Optional[str] = None, tools: Optional[List[Any]] = None) -> str:
+        # Notice: ValidationError  - 1 validation error for TaskOutput | raw: Input should be a valid string
+        # crewaiのTaskOutputのrawに入るのでstrで返す必要がある。
+        # TODO: 直接dictを返せるようにcrewaiを直す？
+
         # AgentExecutorを使用してタスクを実行
         # print("execute_task task=", task)
         print("execute_task context=", context)
@@ -38,9 +44,11 @@ def execute_task(self, task: Any, context: Optional[str] = None, tools: Optional
         input_dict["question"] = task.prompt_context
         input_dict["message"] = "タスクを実行してください。\n" + task.expected_output
         result = self.agent_executor.invoke(input=input_dict)
-        print("execute_task result=", result)
+        result_str = MultiConverter.to_str(result)
+        print("execute_task result=", result_str)
+
         # TODO: return full dict when crewai is updated
-        return result["output"]
+        return result_str
 
     def create_agent_executor(self, tools=None) -> None:
         pass

src/codeinterpreterapi/session.py

@@ -25,6 +25,7 @@
 from codeinterpreterapi.config import settings
 from codeinterpreterapi.llm.llm import CodeInterpreterLlm
 from codeinterpreterapi.schema import CodeInterpreterResponse, File, SessionStatus, UserRequest
+from codeinterpreterapi.utils.multi_converter import MultiConverter
 
 
 def _handle_deprecated_kwargs(kwargs: dict) -> None:
@@ -49,6 +50,7 @@ def on_chain_start(self, serialized: Dict[str, Any], inputs: Dict[str, Any], **k
     def on_chain_end(self, outputs: Dict[str, Any], **kwargs: Any) -> Any:
         """Run when chain ends running."""
         print("AgentCallbackHandler on_chain_end type(outputs)=", type(outputs))
+        print("AgentCallbackHandler on_chain_end type(outputs)=", outputs)
         self.agent_callback_func(outputs)
 
     def on_chat_model_start(
@@ -228,9 +230,10 @@ async def _ainput_handler(self, request: UserRequest) -> None:
 
     def _output_handler_pre(self, response: Any) -> str:
         print("_output_handler_pre response(type)=", type(response))
-        if isinstance(response, str):
-            output_str = response
-        elif isinstance(response, dict):
+        output_str = MultiConverter.to_str(response)
+
+        # TODO: MultiConverterに共通化
+        if isinstance(response, dict):
             output_str = ""
             code_log_item = {}
             if "output" in response:
@@ -242,8 +245,6 @@ def _output_handler_pre(self, response: Any) -> str:
             if "log" in response:
                 code_log_item["log"] = str(response["log"])
             self.output_code_log = code_log_item
-        else:
-            output_str = "response=" + str(response)
         return output_str
 
     def _output_handler_post(self, final_response: str) -> CodeInterpreterResponse:

src/codeinterpreterapi/tools/code_checker.py

@@ -5,7 +5,7 @@
 from codeinterpreterapi.brain.params import CodeInterpreterParams
 from codeinterpreterapi.config import settings
 from codeinterpreterapi.llm.llm import prepare_test_llm
-from codeinterpreterapi.schema import CodeInput
+from codeinterpreterapi.schema import FileInput
 from codeinterpreterapi.utils.file_util import FileUtil
 
 
@@ -27,13 +27,13 @@ def get_tools_code_checker(cls, ci_params: CodeInterpreterParams) -> None:
                 "Start from this latest code first.\n",
                 func=tools_instance._get_latest_code,
                 coroutine=tools_instance._aget_latest_code,
-                args_schema=CodeInput,  # type: ignore
+                args_schema=FileInput,  # type: ignore
             ),
         ]
 
         return tools
 
-    def _get__get_latest_code_common(self, filename=""):
+    def _get_latest_code_common(self, filename=""):
         target_dir = "./"
         target_filename = "main.py"
 
@@ -51,10 +51,10 @@ def _get__get_latest_code_common(self, filename=""):
             return ""
 
     def _get_latest_code(self, filename=""):
-        return self._get__get_latest_code_common(filename)
+        return self._get_latest_code_common(filename)
 
     async def _aget_latest_code(self, filename=""):
-        return self._get__get_latest_code_common(filename)
+        return self._get_latest_code_common(filename)
 
 
 def test():