DT走过的路

前面是通过 提示词 的形式，将工具箱带过去。

这种方式有什么问题？

1. 繁琐：大段大段的提示词，仅仅是为了约束大模型的输出
2. 不标准：每个开发者的提示词的描述千差万别
3. 约束力不高：即便使用了语气最重的提示词，大模型的底层原理决定了它总会有不按照要求回复的情况

• 工具箱的提供

// 工具箱
const tools = [{
    type: "function",
    name: "get_weather",
    description: "Get current temperature for provided coordinates in celsius.",
    parameters: {
        type: "object",
        properties: {
            latitude: { type: "number" },
            longitude: { type: "number" }
        },
        required: ["latitude", "longitude"],
        additionalProperties: false
    },
    strict: true
}];

• 返回的调用工具请求

[{
    "type": "function_call",
    "id": "fc_12345xyz",
    "call_id": "call_12345xyz",
    "name": "get_weather",
    "arguments": "{\"latitude\":48.8566,\"longitude\":2.3522}"
}]

不同的模型，Function Calling的格式不一致。

deepseek

tools = [{
  "type": "function",
  "function": {
      "name": "get_weather",
      "description": "Get weather of an location",
      "parameters": {
          "type": "object",
          "properties": {
              "location": {
                  "type": "string",
                  "description": "The city and state, e.g. San Francisco, CA",
              }
          },
          "required": ["location"]
      },
  }
}]

GPT

const tools = [{
    type: "function",
    name: "get_weather",
    description: "Get current temperature for provided coordinates in celsius.",
    parameters: {
        type: "object",
        properties: {
            latitude: { type: "number" },
            longitude: { type: "number" }
        },
        required: ["latitude", "longitude"],
        additionalProperties: false
    },
    strict: true
}];

claude

"tools": [{
  "name": "get_weather",
  "description": "Get the current weather in a given location",
  "input_schema": {
    "type": "object",
    "properties": {
      "location": {
        "type": "string",
        "description": "The city and state, e.g. San Francisco, CA"
      }
    },
    "required": ["location"]
  }
}],