agent.py

from op_llm_client import OllamaClient


class CustomerServiceAgent:
    def __init__(self, client, config):
        self.client = client
        self.config = config
        self.messages = []
        self.system_prompt = """
        You are a Intelligent customer service assistant for e-commerce platform. It is necessary to answer the user's consultation about the product in a timely manner. If it has nothing to do with the specific product, you can answer it directly.
        output it as Answer: [Your answer here].
       
        Example :
        Answer: Is there anything else I can help you with
        
        If specific information about the product is involved, You run in a loop of Thought, Action, Observation.
        Use Thought to describe your analysis process.
        Use Action to run one of the available tools - then wait for an Observation.
        When you have a final answer, output it as Answer: [Your answer here].
        
        Available tools:
        1. query_by_product_name: Query the database to retrieve a list of products that match or contain the specified product name. This function can be used to assist customers in finding products by name via an online platform or customer support interface
        2. read_store_promotions: Read the store's promotion document to find specific promotions related to the provided product name. This function scans a text document for any promotional entries that include the product name.
        3. calculate: Calculate the final transaction price by combining the selling price and preferential information of the product


        When using an Action, always format it as:
        Action: tool_name: argument1, argument2, ...

        Example :
        Human: Do you sell football in your shop? If you sell soccer balls, what are the preferential policies now? If I buy it now, how much will I get in the end?
        Thought: To answer this question, I need to check the database of the background first.
        Action: query_by_product_name: football

        Observation: At present, I have checked that the ball is in stock, and I know its price is 120 yuan.

        Thought: I need to further inquire about the preferential policy of football
        Action: read_store_promotions: football

        Observation: The current promotional policy for this ball is: 10% discount upon purchase

        Thought: Now I need to combine the selling price and preferential policies of the ball to calculate the final transaction price
        Action: calculate: 120 * 0.9

        Observation: The final price of the ball was 108.0 yuan

        Thought: I now have all the information needed to answer the question.
        Answer:  According to your enquiry, we do sell soccer balls in our store, the current price is 120 yuan. At present, we offer a 10% discount on the purchase of football. Therefore, if you buy now, the final transaction price will be 108 yuan.

        Note: You must reply to the final result in Chinese
        
        Now it's your turn:
        """.strip()
        self.messages.append({"role": "system", "content": self.system_prompt})

    # __call__ 方法可以使得一个类的实例可以被像函数那样调用，提供了类实例的“可调用”能力。
    # 当使用类实例后面跟着括号并传递参数时，就会触发 __call__ 方法。
    def __call__(self, message):
        self.messages.append({"role": "user", "content": message})
        response = self.execute()
        if not isinstance(response, str):
            raise TypeError(f"Expected string response from execute, got {type(response)}")
        self.messages.append({"role": "assistant", "content": response})
        return response

    def execute(self):

        # 检查 self.client 是否是 OllamaClient 类的一个实例。这是类型安全的一种做法，确保 self.client 具有执行接下来代码所需的方法和属性。
        if isinstance(self.client, OllamaClient):
            completion = self.client.chat_completions_create(
                model=self.config["ollama"]['model_name'],
                messages=self.messages,
                temperature=self.config["ollama"]['temperature']
            )

            # 如果 completion 是一个字典并且包含一个键为 message 的项，则尝试从 message 中提取 content 键对应的值。如果没有 content，则返回一个空字符串。
            if isinstance(completion, dict) and 'message' in completion:
                return completion['message'].get('content', '')
            # 如果 completion 直接是一个字符串，则直接返回这个字符串。
            elif isinstance(completion, str):
                return completion
            else:
                raise ValueError(f"Unexpected response structure from OllamaClient: {completion}")
        else:
            # 使用 OpenAI 的 GPT 系列模型
            completion = self.client.chat.completions.create(
                model=self.config['openai']['model_name'],
                messages=self.messages,
            )
            response = completion.choices[0].message.content
            if response != None:
                return completion.choices[0].message.content
            else:
                return "当前没有正常的生成回复，请重新思考当前的问题，并再次进行尝试"


if __name__ == '__main__':
    import json
    from openai import OpenAI

    # 加载全局的配置信息
    with open('config.json', 'r') as f:
        config = json.load(f)

    # 测试当前环境下是否可以连通OpenAI的GPT模型，如果能正常返回回复，则说明当前环境的网络正常
    client = OpenAI()
    completion = client.chat.completions.create(
        model=config['openai']['model_name'],
        messages=[{"role": "user", "content": "你好，测试OpenAI模型在当前环境下的连通性"}],
        temperature=config['openai']['temperature'],
    )
    print(completion.choices[0].message.content)