AutoGen+Langchain can't read my local files.

[TS10armourer]

I made some modifications to your code sample, replacing it with my own model instead of GPT-4. Initially, everything went smoothly. However, when I used the following code, the Langchain_AutoGen() function couldn't access and read local files. On the other hand, the agentchat_groupchat() function worked perfectly, delivering the same quality as GPT-4.
I didn't use Docker; instead, I installed the package directly in a Conda environment using pip.

import autogen
from accelerate import Accelerator
accelerator = Accelerator()
config_list_baichuan2=[
    {
        "model": "Baichuan2-13b-base",
        "api_base": "http://localhost:8000/v1",
        "api_type": "open_ai",
        "api_key": "NULL",
    }
]

def agentchat_groupchat():
    # 定义LLM配置
    llm_config = {
        "seed": 42,  # 更改种子以进行不同的尝试
        "temperature": 0,
        "config_list": config_list_baichuan2,
        "request_timeout": 120,
    }

    # 创建用户代理
    user_proxy = autogen.UserProxyAgent(
        name="User_proxy",
        system_message="人类管理员。",
        code_execution_config={"last_n_messages": 2, "work_dir": "groupchat"},
        human_input_mode="TERMINATE"
    )

    # 创建编码代理
    coder = autogen.AssistantAgent(
        name="Coder",
        llm_config=llm_config,
    )

    # 创建产品经理代理
    pm = autogen.AssistantAgent(
        name="Product_manager",
        system_message="在软件产品创意中具有创造性。",
        llm_config=llm_config,
    )

    # 定义群聊
    groupchat = autogen.GroupChat(agents=[user_proxy, coder, pm], messages=[], max_round=12)
    
    # 创建群聊管理器
    manager = autogen.GroupChatManager(groupchat=groupchat, llm_config=llm_config)

    # 用户代理启动与管理器的聊天，寻找关于arxiv上的gpt-4的最新论文，并找到其在软件中的潜在应用
    user_proxy.initiate_chat(manager, message="在arxiv上找到关于gpt-4的最新论文，并找到其在软件中的潜在应用。")
    
    # 输入exit来终止聊天


def Langchain_AutoGen():
    # 通用导入
    from langchain.pydantic_v1 import BaseModel, Field
    from langchain.tools import BaseTool
    from typing import Optional, Type
    import math

    # 定义计算圆周长的工具的输入模型
    class CircumferenceToolInput(BaseModel):
        radius: float = Field()

    # 定义计算圆周长的工具
    class CircumferenceTool(BaseTool):
        name = "circumference_calculator"
        description = "当你需要使用圆的半径计算圆周长时使用此工具"
        args_schema: Type[BaseModel] = CircumferenceToolInput

        # 定义如何计算圆周长
        def _run(self, radius: float):
            return float(radius) * 2.0 * math.pi

    # 导入文件读取工具
    from langchain.tools.file_management.read import ReadFileTool

    # 定义一个从LangChain工具生成llm_config的函数
    def generate_llm_config(tool):
        # 基于工具的args_schema定义函数模式
        function_schema = {
            "name": tool.name.lower().replace(' ', '_'),
            "description": tool.description,
            "parameters": {
                "type": "object",
                "properties": {},
                "required": [],
            },
        }

        if tool.args is not None:
            function_schema["parameters"]["properties"] = tool.args

        return function_schema

    # 实例化读取文件的工具
    read_file_tool = ReadFileTool()
    custom_tool = CircumferenceTool()

    # 构造llm_config
    llm_config = {
        # 为工具生成函数配置
        "functions": [
            generate_llm_config(custom_tool),
            generate_llm_config(read_file_tool),
        ],
        "config_list": config_list_baichuan2,  # 假设你在其他地方已经定义了这个
        "request_timeout": 120,
    }

    print(llm_config)

    # 创建用户代理
    user_proxy = autogen.UserProxyAgent(
        name="user_proxy",
        is_termination_msg=lambda x: x.get("content", "") and x.get("content", "").rstrip().endswith("TERMINATE"),
        human_input_mode="NEVER",
        max_consecutive_auto_reply=10,
        code_execution_config={"work_dir": "coding"},
    )

    # 在用户代理中注册工具并开始对话
    user_proxy.register_function(
        function_map={
            custom_tool.name: custom_tool._run,
            read_file_tool.name: read_file_tool._run,
        }
    )

    # 创建助手代理
    chatbot = autogen.AssistantAgent(
        name="chatbot",
        system_message="对于编码任务，只使用你所提供的函数。当任务完成时回复TERMINATE。",
        llm_config=llm_config,
    )

    # 初始化与助手代理的对话
    user_proxy.initiate_chat(
        chatbot,
        message="读取文件路径为'Test.txt'的文件，然后计算文件内容作为半径的圆的周长。",  # 文件中为"7.81mm"
        llm_config=llm_config,
    )

if __name__ == '__main__':
    Langchain_AutoGen()

[sonichi]

What error message did you get? @ElliotWood is the author of the LangChain tooling.

[TS10armourer]

I change the prompts and find different answer, It seems that there comes an error:

(modelscope) star@PowerEdge-C4140:~/Code/DataLinguistic/AutoGen$ /home/star/miniconda3/envs/modelscope/bin/python /home/star/Code/DataLinguistic/AutoGen/main.py
[2023-10-18 08:51:59,824] [INFO] [real_accelerator.py:158:get_accelerator] Setting ds_accelerator to cuda (auto detect)
2023-10-18 08:52:00.914915: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2023-10-18 08:52:02.111359: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
{'functions': [{'name': 'circumference_calculator', 'description': '当你需要使用圆的半径计算圆周长时使用此工具', 'parameters': {'type': 'object', 'properties': {'radius': {'title': 'Radius', 'type': 'number'}}, 'required': []}}, {'name': 'read_file', 'description': 'Read file from disk', 'parameters': {'type': 'object', 'properties': {'file_path': {'title': 'File Path', 'description': 'name of file', 'type': 'string'}}, 'required': []}}], 'config_list': [{'model': 'Baichuan2-13b-base', 'api_base': 'http://localhost:8000/v1', 'api_type': 'open_ai', 'api_key': 'NULL'}], 'request_timeout': 120}
user_proxy (to coder):

读取文件路径为'/home/star/Code/DataLinguistic/AutoGen/Tutorial/Test.txt'的文件，然后计算文件内容作为半径的圆的周长。

--------------------------------------------------------------------------------
coder (to user_proxy):

 好的，我将读取文件并计算周长。请稍等片刻。
```python
import math

def calculate_circle_circumference(radius):
    return 2 * math.pi * radius

file_path = "/home/star/Code/DataLinguistic/AutoGen/Tutorial/Test.txt"
with open(file_path, "r") as file:
    file_content = file.read()

radius = len(file_content)
circumference = calculate_circle_circumference(radius)
print(f"The circumference of the file is: {circumference}")

---

EXECUTING CODE BLOCK 0 (inferred language is python)...
execute_code was called with use_docker evaluating to True, but the python docker package is not available. Falling back to native code execution. Note: this fallback behavior is subject to change
user_proxy (to coder):

exitcode: 0 (execution succeeded)
Code output:
The circumference of the file is: 37.69911184307752

---

coder (to user_proxy):

TERMINATE

---

[GXKIM]

Test.txt 填写的内容
使用百川但是还是无法读取里面的半径，请问是不是我填写错误？

[sonichi]

what's the system_message used for the last run?

[TS10armourer]

I find if I use the local LLM, the AutoGen won't use these func anymore:
read_file_tool = ReadFileTool() custom_tool = CircumferenceTool()
and the dialogue is:
`
user_proxy (to chatbot):

Read the file with the path '/home/star/Code/DataLinguistic/AutoGen/Tutorial/Test.txt', then calculate the circumference of a circle that has a radius of that files contents.

---

chatbot (to user_proxy):

Sure, here's the code to read the file and calculate the circumference of a circle with the radius equal to the file contents:

import os
import math

# Read the file contents
file_path = "/home/star/Code/DataLinguistic/AutoGen/Tutorial/Test.txt"
with open(file_path, "r") as f:
    radius = float(f.read())

# Calculate the circumference
circumference = 2 * math.pi * radius

# Print the result
print("The circumference of the circle is:", circumference)

...

---

EXECUTING CODE BLOCK 0 (inferred language is python)...
Output is truncated. View as a scrollable element or open in a text editor. Adjust cell output settings...
execute_code was called with use_docker evaluating to True, but the python docker package is not available. Falling back to native code execution. Note: this fallback behavior is subject to change
`

[sonichi]

That's a property of the OSS model used. It's a known issue that has been discussed a lot on Discord.

[luohao123]

Am also very confused by how does it defines tools,

these decorators not work anymore?

@user_proxy.register_function()
@assistant.register_for_llm(description="Currency exchange calculator.")
def currency_calculator(
base_amount: Annotated[float, "Amount of currency in base_currency"],
base_currency: Annotated[CurrencySymbol, "Base currency"] = "USD",
quote_currency: Annotated[CurrencySymbol, "Quote currency"] = "EUR",
) -> str:
quote_amount = exchange_rate(base_currency, quote_currency) * base_amount
return f"{quote_amount} {quote_currency}"

[rickyloynd-microsoft]

To use the new function decorators, you can clone the repo then install it with pip install -e . from the autogen dir.