模型api接口规范-j9九游会登录

适用场景

本章节适用于以下场景：

• 需要将第三方模型服务（如deepseek、moonshot等）接入agentarts，用于确认模型api是否满足平台接口协议规范。
• 自行部署了模型推理服务，需要按平台支持的协议实现/适配api接口后再接入agentarts。
• 模型接入或调用过程中出现异常（例如请求失败、响应字段缺失、格式不兼容等），需要对照接口规范排查请求与响应格式问题。

费用说明

调用模型api会产生相应费用，计费维度为token数量（包括输入token和输出token）。具体计费规则请参考接入的模型的供应商j9九游会登录官网。

建议在测试阶段设置合理的max_tokens参数以控制成本。

文本对话（chat）api规范

接口格式

类型：post

协议：http/https

请求体参数

• 工具调用请求示例
  • 流式请求示例

    {
        "model": "my-chat-model",
        "messages": [
            {
                "role": "user",
                "content": "请帮我查询南京的天气"
            }
        ],
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "description": "获取给定地点的天气",
                    "parameters": {
                        "type": "object",
                        "properties": {
                            "location": {
                                "type": "string",
                                "description": "地点，例如南京。"
                            }
                        },
                        "required": ["location"]
                    }
                }
            }
        ],
        "max_tokens": 200,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": true
    }

  • 非流式请求示例

    {
        "model": "my-chat-model",
        "messages": [
            {
                "role": "user",
                "content": "请帮我查询南京的天气"
            }
        ],
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "description": "获取给定地点的天气",
                    "parameters": {
                        "type": "object",
                        "properties": {
                            "location": {
                                "type": "string",
                                "description": "地点，例如南京。"
                            }
                        },
                        "required": ["location"]
                    }
                }
            }
        ],
        "max_tokens": 200,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": false
    }

• 非工具调用请求示例
  • 流式请求示例

    {
        "model": "my-chat-model",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": "你好！"
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": true
    }

  • 非流式请求示例

    {
        "model": "my-chat-model",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": "你好！"
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": false
    }

文本向量化（embeddings）api规范

接口格式

类型：post

协议：http/https

请求体参数

请求示例：

{
    "model": "my-embedding-model",
    "input": "你好"
}

响应体参数

响应示例：

{
    "data": [
        {
            "index": 0,
            "embedding": [
                0.02513289265334606,
                -0.017512470483779907,
                -0.029955564066767693,
                ...
            ],
            "object": "embedding"
        }
    ],
    "usage": {
        "prompt_tokens": 5,
        "total_tokens": 5
    },
    "model": "my-embedding-model",
    "object": "list"
}

文本排序（rerank）api规范

接口格式

类型：post

协议：http/https

请求体参数

请求示例：

{
    "model": "my-rerank-model",
    "query": "请问ai原生应用引擎提供了什么能力？",
    "docs": ["ai原生应用引擎提供了应用开发、模型网关等能力。", "ai原生应用引擎正在逐步完善、提高竞争力。"],
    "top_n": 3
}

响应体参数

响应示例：

{
    "model": "my-rerank-model",
    "usage": {
        "prompt_tokens": 5,
        "total_tokens": 5
    },
"results": [
      {
           "index": 0,
           "document": {"text": "ai原生应用引擎提供了应用开发、模型网关等能力。"},
           "relevance_score": 0.9
      },
     {
          "index": 1,
          "document": {"text": "ai原生应用引擎正在逐步完善、提高竞争力。"},
          "relevance_score": 0.5
     }
   ]
}

图像理解api规范

接口格式

类型：post

协议：http/https

请求体参数

[
{
"type": "image_url",
"image_url": {
"url": "data:image/png;base64"
}
},
{
"type": "text",
"text": "图里有什么"
}
]

• 工具调用请求示例
  • 流式请求示例

    {
        "model": "model-img2text",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "图里面有什么"
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": "一个图片链接"
                        }
                    }
                ],
            }
        ],
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "description": "获取给定地点的天气",
                    "parameters": {
                        "type": "object",
                        "properties": {
                            "location": {
                                "type": "string",
                                "description": "地点，例如南京。"
                            }
                        },
                        "required": [
                            "location"
                        ]
                    }
                }
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": true
    }

  • 非流式请求示例

    {
        "model": "model-img2text",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "图里面有什么"
                    },
                    {
                        "type": "image_url",
                        "image_url": {
                            "url": "一个图片链接"
                        }
                    }
                ],
            }
        ],
        "tools": [
            {
                "type": "function",
                "function": {
                    "name": "get_weather",
                    "description": "获取给定地点的天气",
                    "parameters": {
                        "type": "object",
                        "properties": {
                            "location": {
                                "type": "string",
                                "description": "地点，例如南京。"
                            }
                        },
                        "required": [
                            "location"
                        ]
                    }
                }
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": false
    }

• 非工具调用请求示例
  • 流式请求示例

    {
        "model": "model-img2text",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": [ { "type": "text", "text": "图里面有什么" }, { "type": "image_url", "image_url": { "url": "一个图片链接" } } ],
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": true
    }

  • 非流式请求示例

    {
        "model": "model-img2text",
        "messages": [
            {
                "role": "system",
                "content": " you are a helpful assistant. "
            },
            {
                "role": "user",
                "content": [ { "type": "text", "text": "图里面有什么" }, { "type": "image_url", "image_url": { "url": "一个图片链接" } } ],
            }
        ],
        "max_tokens": 20,
        "presence_penalty": 1.2,
        "frequency_penalty": 1.0,
        "temperature": 0.5,
        "top_p": 0.95,
        "stream": false
    }

响应体参数

• 工具调用响应示例
  • 流式响应示例

    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"role":"assistant","content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[{"id":"call_abc123","type":"function","function":{"name":"search_web","arguments":{}}}],"content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[{"id":"call_abc123","type":"function","function":{"name":"search_web","arguments":{"query":"最新款iphone发布信息","limit":5}}}],"content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[{"id":"call_abc123","type":"function","function":{"name":"search_web","arguments":{"query":"最新款iphone发布信息","limit":5}},"index":0}],"content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"正在为您搜索最新款iphone的发布信息……"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"finish_reason":"tool_calls","delta":{"tool_calls":[],"content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"usage":{"completion_tokens":67,"prompt_tokens":35,"total_tokens":102},"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:[done]

  • 非流式响应示例

    {
      "choices": [
        {
          "message": {
            "role": "assistant",
            "tool_calls": [
              {
                "id": "call_12345",
                "type": "function",
                "function": {
                  "name": "image_analysis",
                  "arguments": "{\"detail_level\": \"high\", \"output_format\": \"json\"}"
                }
              }
            ]
          },
          "finish_reason": "tool_calls",
          "index": 0
        }
      ],
      "created": 1753965925754,
      "id": "model-img2text",
      "object": "chat.completions",
      "usage": {
        "prompt_tokens": 142,
        "completion_tokens": 28,
        "total_tokens": 170
      },
      "request_id": "xxx"
    }

• 非工具调用响应示例
  • 流式响应示例

    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"role":"assistant","content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"这是"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"google"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":" chrome"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"浏览器"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"的"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"图标"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"delta":{"tool_calls":[],"content":"。"},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[{"finish_reason":"stop","delta":{"tool_calls":[],"content":""},"index":0}],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:{"created":1767494144,"usage":{"completion_tokens":43,"prompt_tokens":35,"total_tokens":78},"model":"qwen2.5-vl-72b","id":"chatcmpl-417592f9469edb89448e18ce650333ae","choices":[],"request_id":"74fb6469180f3e516732b36e3506e5f7","object":"chat.completion.chunk"}
    data:[done]

  • 非流式响应示例

    {
        "choices": [
            {
                "delta": {
                    "role": "assistant",
                    "content": "图像整体呈现出简洁、抽象的风格，主要内容是一个灰色的圆形头像轮廓。"
                },
                "finish_reason": "stop",
                "index": 0
            }
        ],
        "created": 1753965925754,
        "id": "model-img2text",
        "object": "chat.completions",
        "usage": {
            "prompt_tokens": 142,
            "completion_tokens": 118,
            "total_tokens": 260
        },
        "request_id": "xxx"
    }