Troubles on using an third-party API

[eccstartup]

I was trying to use third-party API for workforce and get stuck many times.

Firstly, I don't know that I have to write model_platform=ModelPlatformType.OPENAI_COMPATIBILITY_MODEL. I thought it was model_platform=ModelPlatformType.OPENAI.

model = ModelFactory.create(
    model_platform=ModelPlatformType.OPENAI_COMPATIBILITY_MODEL,
    model_type='deepseek-ai/DeepSeek-V2-Chat',
    url='https://api.siliconflow.cn/v1',
    api_key='sk-xxxxxx',
    model_config_dict={"temperature": 0.4, 'max_tokens': 4096},
)

Secondly, in a workforce, I have to set task_agent_kwargs and coordinator_agent_kwargs because these parameters are OpenAI models by default.

workforce = Workforce("A Simple Workforce",
                      task_agent_kwargs={"model": model},
                      coordinator_agent_kwargs={"model": model},)

self.model_backend: BaseModelBackend = (
            model
            if model is not None
            else ModelFactory.create(
                model_platform=ModelPlatformType.OPENAI,
                model_type=ModelType.GPT_4O_MINI,
                model_config_dict=ChatGPTConfig().as_dict(),
            )
        )

Thirdly, a newly fixed bug stops me again which has not be pushed to pypi.

    if output_schema is not None and self.model_type.supports_tool_calling:
                                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'str' object has no attribute 'supports_tool_calling'

[Asher-hss]

The issue will be resolved after the refactoring of the modeltype.
#982

[eccstartup]

Thanks for replying.

[eccstartup]

A few more troubles:

• Third-party API does not support model_config_dict["tool_choice"] = "required"

• I cannot read API error message, this is the code.

/Users/eccstartup/miniforge3/envs/camelai/bin/python /Users/eccstartup/code/llm4dummies/camel_examples/workforce_example.py 
Traceback (most recent call last):
  File "/Users/eccstartup/code/llm4dummies/camel_examples/workforce_example.py", line 87, in <module>
    task = workforce.process_task(task)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/workforce/utils.py", line 63, in wrapper
    return func(self, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/workforce/workforce.py", line 149, in process_task
    subtasks = self._decompose_task(task)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/workforce/workforce.py", line 124, in _decompose_task
    subtasks = task.decompose(self.task_agent, decompose_prompt)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/tasks/task.py", line 226, in decompose
    response = agent.step(msg)
               ^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/agents/chat_agent.py", line 388, in step
    ) = self._step_model_response(openai_messages, num_tokens)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/agents/chat_agent.py", line 676, in _step_model_response
    response = self.model_backend.run(openai_messages)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/camel/models/openai_compatibility_model.py", line 77, in run
    response = self._client.chat.completions.create(
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/openai/_utils/_utils.py", line 274, in wrapper
    return func(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/openai/resources/chat/completions.py", line 704, in create
    return self._post(
           ^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/openai/_base_client.py", line 1270, in post
    return cast(ResponseT, self.request(cast_to, opts, stream=stream, stream_cls=stream_cls))
                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/openai/_base_client.py", line 947, in request
    return self._request(
           ^^^^^^^^^^^^^^
  File "/Users/eccstartup/miniforge3/envs/camelai/lib/python3.11/site-packages/openai/_base_client.py", line 1051, in _request
    raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'code': 20015, 'message': 'The parameter is invalid. Please check again.', 'data': None}

Process finished with exit code 1

[eccstartup]

It is difficult to guess how to use tools configuration.