messages = [
    {"role": "system",    "content": "人设"},
    {"role": "user",      "content": "问题a"},
    {"role": "assistant", "content": "回答a"},
    {"role": "user",      "content": '问题b'},
]

messages = [
    {"role": "system",    "content": "人设"},
    {"role": "user",      "content": "问题"},
    {"role": "assistant", "content": "", "tool_calls": [{"function": {"name": "function_name", "arguments": "function_args"}, "id": "", "type": "function"}]},
    {"role": "tool",      "content": "函数调用结果", "name": "function_name"},
    {"role": "assistant", "content": "回答"},
]

prompt_template = ("任务指令：作为线上百货商城的专业客服，为用户{}（用户ID：{}，性别：{}）提供全方位咨询服务，当前时间为{}。在利用内置工具函数处理查询请求时，如信息不足，请主动引导用户提供详细信息。"
                   "回答策略："
                   "1. **主动信息索取**：在需要调用如订单查询等工具功能时，若必要参数（如订单号）缺失，采用友好且明确的语言主动询问：“尊敬的张三，为了快速查询您的信息，请提供一下订单号好吗？”"
                   "2. **精准解答**：基于用户提出的问题，严格参照公司政策与操作流程，结合最新文档内容给予精确解答，避免无关扩展，确保用户问题得到有效解决。"
                   "3. **透明化操作**：在处理用户请求过程中，如需使用特定工具函数，简要告知用户将采取的步骤，增加服务透明度，例如：“我将通过我们的订单查询系统来获取您的订单详情，请稍候。”"
                   "4. **确认与跟进**：解答完毕后，确认用户是否满意解答，并主动询问是否有其他可以帮助的地方，如：“张三先生/女士，您的问题已解答完毕，请问还有其他方面需要我的协助吗？”"
                   "注意事项："
                   "- 维持专业且亲切的交流风格，确保每一次互动都能提升用户满意度。"
                   "- 对于所有工具函数调用，务必确保在获取足够且准确的参数后再执行，避免因信息不全导致处理错误。 "
                   "- 记录重要咨询细节，以便后续跟踪服务或内部评估使用。")

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_orders",
            "description": "当你想通过用户id查询一段时间内的订单时非常有用，返回结果包括订单号和商品名称",
            "parameters": {
                "type": "object",
                "properties": {
                    "user_id": {
                        "type": "string",
                        "description": "用户id"
                    },
                    "start_time": {
                        "type": "string",
                        "description": "开始时间，格式为%Y-%m-%d %H:%M:%S"
                    },
                    "end_time": {
                        "type": "string",
                        "description": "结束时间，格式为%Y-%m-%d %H:%M:%S"
                    }
                },
                "required": ["user_id", "start_time", "end_time"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "get_order",
            "description": "当你想通过订单号查询订单信息时非常有用，返回结果包含物流信息、下单时间以及商品id、名称、类型、价格、文档id等信息",
            "parameters": {
                "type": "object",
                "properties": {
                    "number": {
                        "type": "string",
                        "description": "订单号"
                    }
                },
                "required": ["number"]
            }
        }
    },
    {
        "type": "function",
        "function": {
            "name": "get_product",
            "description": "当你想通过商品名称查询商品信息时非常有用，返回结果包括商品id、商品名称、商品类型、商品价格、商品库存、商品文档id",
            "parameters": {
                "type": "object",
                "properties": {
                    "name": {
                        "type": "string",
                        "description": "商品名称"
                    }
                },
                "required": ["name"]
            }
        }
    },
]
def get_orders(user_id: str, start_time: datetime, end_time: datetime) -> str:
    pass
def get_order(number: str) -> str:
    pass
def get_product(name: str) -> str:
    pass

import os
from llama_index.core import Settings
from llama_index.indices.managed.dashscope import DashScopeCloudIndex
from llama_index.llms.dashscope import DashScope
from llama_index.embeddings.dashscope import DashScopeEmbedding
class Rag:
    _instance = None
    @classmethod
    def get_instance(cls):
        if cls._instance is None:
            cls._instance = cls()
        return cls._instance
    def __init__(self):
        llm = DashScope(model_name="qwen-max")
        embed_model = DashScopeEmbedding(model_name="text-embedding-v2", text_type="query")
        Settings.llm = llm
        Settings.embed_model = embed_model
    def query(self, index_name: str, question: str) -> str:
        index = DashScopeCloudIndex(name=index_name)
        query_engine = index.as_query_engine()
        answer = query_engine.query(question)
        return str(answer)

tools = [
    {
        "type": "function",
        "function": {
            "name": "retrieving_documents",
            "description": "当你想通过商品文档id查询商品参数或功能使用说明时非常有用，返回结果为问题查询得到的答案",
            "parameters": {
                "type": "object",
                "properties": {
                    "document_id": {
                        "type": "string",
                        "description": "文档id，该值只能通过其他function查询得到"
                    },
                    "question": {
                        "type": "string",
                        "description": "问题描述，比如：该产品电池容量多大？"
                    }
                },
                "required": ["document_id", "question"]
            }
        }
    }
]
def retrieving_documents(document_id: str, question: str) -> str:
    return Rag.get_instance().query(index_name=document_id, question=question)