智能问答接口文档

接口描述

创建一个聊天完成请求，支持流式和非流式响应。该接口接收一系列消息作为输入，并返回模型生成的完成内容。本协议在OpenAI的Chat Completions API基础上扩展，支持了reasoning_content字段。最新的完整的OpenAI Chat Completions API参数可以参考OpenAI Chat Completions API。

请求

HTTP 请求

POST /v1/chat/completions

请求体

参数	类型	必填	描述
model	string	是	要使用的模型 ID
messages	array	是	包含对话历史的消息数组。根据使用的模型不同，支持不同的消息类型（模态），如文本、图像和音频
tools	array	否	模型可以调用的工具列表。目前仅支持函数作为工具。使用此参数提供模型可能生成JSON输入的函数列表。最多支持128个函数
tool_choice	string/object	否	控制模型是否调用工具。 - `"none"`: 模型不会调用任何工具，而是生成消息 - `"auto"`: 模型可以在生成消息或调用一个或多个工具之间选择 - `"required"`: 模型必须调用一个或多个工具 - 也可以通过对象指定特定工具：`{"type": "function", "function": {"name": "my_function"}}` 当没有工具时，默认为`"none"`；有工具时，默认为`"auto"`
temperature	number	否	采样温度，默认为 1。较低的值使输出更加确定性
top_p	number	否	核采样的概率质量，默认为 1
n	integer	否	为每个输入消息生成的聊天完成数量，默认为 1
stream	boolean	否	是否启用流式响应，默认为 false
stream_options	object	否	流式响应选项，仅在 stream=true 时设置
stop	string/array	否	最多 4 个序列，API 将在生成这些序列时停止
max_tokens	integer	否	生成的最大令牌数
presence_penalty	number	否	存在惩罚，范围 -2.0 到 2.0，默认为 0
frequency_penalty	number	否	频率惩罚，范围 -2.0 到 2.0，默认为 0
logit_bias	object	否	修改指定令牌出现在完成中的可能性
response_format	object	否	指定模型必须输出的格式
seed	integer	否	用于确定性采样的种子值
parallel_tool_calls	boolean	否	是否允许并行工具调用
user	string	否	表示最终用户的唯一标识符

Message对象类型

Developer 消息（开发者消息）

{
  "role": "developer",
  "content": "你是一个有帮助的助手，专注于回答用户的问题。"
}

开发者提供的指令，无论用户发送什么消息，模型都应遵循。在较新的OpenAI模型（如o1系列）中，开发者消息替代了之前的系统消息。

System 消息（系统消息）

{
  "role": "system",
  "content": "你是一个有帮助的助手，专注于回答用户的问题。"
}

开发者提供的指令，无论用户发送什么消息，模型都应遵循。在较新的OpenAI模型（如o1系列）中，建议使用开发者消息代替系统消息。

User 消息（用户消息）

{
  "role": "user",
  "content": "你好，请介绍一下自己。"
}

终端用户发送的消息，包含提示或额外的上下文信息。

Assistant 消息（助手消息）

{
  "role": "assistant",
  "content": "我是一个AI助手，可以回答问题和提供信息。"
}

模型响应用户消息发送的消息。

Tool 消息（工具消息）

{
  "role": "tool",
  "content": "{\"temperature\":32,\"unit\":\"celsius\",\"description\":\"晴朗\",\"humidity\":45}",
  "tool_call_id": "call_abc123"
}

工具消息包含工具调用的结果，必须包含以下字段：

content: 工具消息的内容（字符串）
role: 消息作者的角色，在这种情况下为"tool"
tool_call_id: 此消息响应的工具调用ID

工具和函数定义

Tool 对象

{
  "type": "function",
  "function": {
    "name": "get_weather",
    "description": "获取指定城市的天气信息",
    "parameters": {
      "type": "object",
      "properties": {
        "location": {
          "type": "string",
          "description": "城市名称，如北京、上海"
        },
        "unit": {
          "type": "string",
          "enum": ["celsius", "fahrenheit"],
          "description": "温度单位"
        }
      },
      "required": ["location"]
    },
    "strict": true
  }
}

工具对象包含以下字段：

type: 工具类型，目前仅支持"function"
function: 函数定义，包含以下字段：
- name: 函数名称（必填）。必须是a-z、A-Z、0-9或包含下划线和破折号，最大长度为64
- description: 函数功能描述（可选）。模型用它来决定何时以及如何调用函数
- parameters: 函数接受的参数，描述为JSON Schema对象（可选）
- strict: 是否在生成函数调用时启用严格的模式遵循（可选，默认为false）

Tool Choice 对象

用于指定特定工具：

{
  "type": "function",
  "function": {
    "name": "get_weather"
  }
}

消息内容格式

文本内容

{
  "role": "user",
  "content": "你好，请介绍一下自己。"
}

图片内容

{
  "role": "user",
  "content": [
    {
      "type": "text",
      "text": "这张图片是什么内容？"
    },
    {
      "type": "image_url",
      "image_url": {
        "url": "https://example.com/image.jpg",
        "detail": "high"
      }
    }
  ]
}

工具调用内容

{
  "role": "assistant",
  "content": null,
  "tool_calls": [
    {
      "id": "call_abc123",
      "type": "function",
      "function": {
        "name": "get_weather",
        "arguments": "{\"location\":\"北京\",\"unit\":\"celsius\"}"
      }
    }
  ]
}

响应

非流式响应

{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1677652288,
  "model": "gpt-4o",
  "system_fingerprint": "fp_44709d6fcb",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "你好！我能帮你什么忙吗？",
      "reasoning_content": "用户用中文问候，我应该用中文回复。"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 9,
    "completion_tokens": 12,
    "total_tokens": 21
  }
}

流式响应

当 stream=true 时，服务器将发送一系列 Server-Sent Events (SSE)，每个事件包含部分响应。每个事件的格式如下：

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-3.5-turbo","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"reasoning_content":"用户用中文问候，"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-3.5-turbo","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"reasoning_content":"我应该用中文回复。"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-3.5-turbo","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"content":"你"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-3.5-turbo","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"content":"好"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-3.5-turbo","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]

reasoning_content 字段

Bella OpenAPI 在标准 OpenAI 接口基础上扩展了 reasoning_content 字段，用于提供模型的推理过程：

只有输出推理过程的模型才会返回 reasoning_content 字段
在非流式响应中，reasoning_content 作为 Message 对象的一个属性返回
在流式响应中，reasoning_content 通过 delta 对象分块返回
推理内容可以帮助开发者理解模型的思考过程和决策依据

响应参数

参数	类型	描述
id	string	响应的唯一标识符
object	string	对象类型，通常为 "chat.completion" 或 "chat.completion.chunk"
created	integer	响应创建的时间戳
model	string	使用的模型
system_fingerprint	string	模型运行的后端配置指纹
choices	array	完成选项数组
usage	object	令牌使用统计

Choice 对象

参数	类型	描述
index	integer	选项的索引
message	object	在非流式响应中，包含完整的消息
delta	object	在流式响应中，包含消息的增量部分
finish_reason	string	完成原因，可能的值有 "stop"、"length"、"tool_calls"、"content_filter" 或 null

Message/Delta 对象

参数	类型	描述
role	string	消息的角色，通常为 "assistant"
content	string	消息的内容
reasoning_content	string	包含模型的推理过程
tool_calls	array	工具调用列表

Usage 对象

参数	类型	描述
prompt_tokens	integer	提示使用的令牌数
completion_tokens	integer	完成使用的令牌数
total_tokens	integer	总共使用的令牌数

错误码

错误码	描述
400	请求参数错误
401	认证失败，无效的 API 密钥
403	权限不足，API 密钥没有权限访问请求的资源
404	请求的资源不存在
429	请求过多，超出速率限制
500	服务器内部错误
503	服务暂时不可用

示例

基本请求示例

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "system",
      "content": "你是一个有帮助的助手。"
    },
    {
      "role": "user",
      "content": "你好，请介绍一下自己。"
    }
  ],
  "temperature": 0.7,
  "stream": false
}

工具调用请求示例

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": "北京今天的天气怎么样？"
    }
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_weather",
        "description": "获取指定城市的天气信息",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "城市名称，如北京、上海"
            },
            "unit": {
              "type": "string",
              "enum": ["celsius", "fahrenheit"],
              "description": "温度单位"
            }
          },
          "required": ["location"]
        },
        "strict": true
      }
    }
  ],
  "tool_choice": "auto"
}

工具调用响应示例

当模型决定调用工具时，响应将包含 tool_calls 字段：

{
  "id": "chatcmpl-123",
  "object": "chat.completion",
  "created": 1677652288,
  "model": "gpt-4o",
  "system_fingerprint": "fp_44709d6fcb",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": null,
      "reasoning_content": "用户询问北京的天气，我需要调用",
      "tool_calls": [
        {
          "id": "call_abc123",
          "type": "function",
          "function": {
            "name": "get_weather",
            "arguments": "{\"location\":\"北京\",\"unit\":\"celsius\"}"
          }
        }
      ]
    },
    "finish_reason": "tool_calls"
  }],
  "usage": {
    "prompt_tokens": 82,
    "completion_tokens": 25,
    "total_tokens": 107
  }
}

工具结果提交示例

在获取工具执行结果后，需要将结果提交回对话：

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": "北京今天的天气怎么样？"
    },
    {
      "role": "assistant",
      "content": null,
      "tool_calls": [
        {
          "id": "call_abc123",
          "type": "function",
          "function": {
            "name": "get_weather",
            "arguments": "{\"location\":\"北京\",\"unit\":\"celsius\"}"
          }
        }
      ]
    },
    {
      "role": "tool",
      "tool_call_id": "call_abc123",
      "name": "get_weather",
      "content": "{\"temperature\":32,\"unit\":\"celsius\",\"description\":\"晴朗\",\"humidity\":45}"
    }
  ]
}

工具结果响应示例

模型处理工具结果后的响应：

{
  "id": "chatcmpl-456",
  "object": "chat.completion",
  "created": 1677652290,
  "model": "gpt-4o",
  "system_fingerprint": "fp_44709d6fcb",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "北京今天的天气晴朗，气温32°C，湿度45%。天气较热，建议做好防晒措施，多补充水分。",
      "reasoning_content": "根据天气API返回的数据，北京今天天气晴朗，温度32摄氏度，湿度45%。这属于较热的天气，应该提醒用户注意防晒和补水。"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 110,
    "completion_tokens": 42,
    "total_tokens": 152
  }
}

图片输入示例

支持在用户消息中包含图片：

{
  "model": "gpt-4o",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "这张图片是什么内容？"
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "https://example.com/image.jpg",
            "detail": "high"
          }
        }
      ]
    }
  ]
}

图片URL格式

图片URL可以是以下格式之一：

互联网可访问的URL：

{
  "type": "image_url",
  "image_url": {
    "url": "https://example.com/image.jpg"
  }
}

Base64编码的图片数据：

{
  "type": "image_url",
  "image_url": {
    "url": "data:image/jpeg;base64,/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0aHBwgJC4nICIsIxwcKDcpLDAxNDQ0Hyc5PTgyPC4zNDL..."
  }
}

图片细节级别：可以通过 detail 参数指定图片分析的细节级别，可选值为：
- "high"：高细节分析，适用于需要识别细小文本或细节的场景
- "low"：低细节分析，处理速度更快，消耗的令牌更少
- 不指定时默认为 "auto"

图片输入响应示例

{
  "id": "chatcmpl-789",
  "object": "chat.completion",
  "created": 1677652295,
  "model": "gpt-4o",
  "system_fingerprint": "fp_44709d6fcb",
  "choices": [{
    "index": 0,
    "message": {
      "role": "assistant",
      "content": "这张图片展示了一只橙色的猫咪坐在窗台上，望向窗外。窗外可以看到一些绿色的树木和蓝色的天空。猫咪看起来很放松，尾巴卷曲在身体旁边。",
      "reasoning_content": "图片中有一只橙色的猫，它坐在窗台上看向窗外。我可以看到窗外有树木和天空。猫咪的姿势显示它很放松。我应该详细描述我看到的内容，包括猫的颜色、位置、周围环境和姿态。"
    },
    "finish_reason": "stop"
  }],
  "usage": {
    "prompt_tokens": 1042,
    "completion_tokens": 65,
    "total_tokens": 1107
  }
}

流式工具调用示例

当在流式模式下使用工具调用时，响应会分多个事件返回：

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"role":"assistant"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"reasoning_content":"用户询问北京的天气，我需要调用"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"reasoning_content":"天气查询函数来获取这一信息。"},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"id":"call_abc123","type":"function","function":{"name":"get_weather"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"{\""}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"location"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"北京"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\",\""}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"unit"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\":\""}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"celsius"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","choices":[{"index":0,"delta":{"tool_calls":[{"index":0,"function":{"arguments":"\"}"}}]},"finish_reason":null}]}

data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1677652288,"model":"gpt-4o","system_fingerprint":"fp_44709d6fcb","usage":{"prompt_tokens": 1042,"completion_tokens": 65,"total_tokens": 1107},"choices":[{"index":0,"delta":{},"finish_reason":"tool_calls"}]}

data: [DONE]

目录​

接口描述​

请求​

HTTP 请求​

请求体​

Message对象类型​

Developer 消息（开发者消息）​

System 消息（系统消息）​

User 消息（用户消息）​

Assistant 消息（助手消息）​

Tool 消息（工具消息）​

工具和函数定义​

Tool 对象​

Tool Choice 对象​

消息内容格式​

文本内容​

图片内容​

工具调用内容​

响应​

非流式响应​

流式响应​

reasoning_content 字段​

响应参数​

Choice 对象​

Message/Delta 对象​

Usage 对象​

错误码​

示例​

基本请求示例​

工具调用请求示例​

工具调用响应示例​

工具结果提交示例​

工具结果响应示例​

图片输入示例​

图片URL格式​

图片输入响应示例​

流式工具调用示例​

目录