本文介绍 Responses API 创建模型请求时的输入输出参数,供您使用接口时查阅字段含义。
请求体参数说明如下。
参数 | 类型 | 必填 | 默认值 | 说明 |
|---|---|---|---|---|
model | string | 是 | 您需要调用的模型的 ID (Model ID)。 | |
input | string / array | 是 | 输入的内容,模型需要处理的输入信息。支持以下类型:
| |
instructions | string | 否 | null | 在模型上下文中插入系统消息或者开发者作为第一条指令。当与 previous_response_id 一起使用时,前一个回复中的指令不会被继承到下一个回复中。这样可以方便地在新的回复中替换系统(或开发者)消息。 不可与缓存能力一起使用。 |
previous_response_id | string | 否 | null | 上一个模型回复的唯一标识符。使用该标识符可以实现多轮对话。工作原理可参见多轮对话场景。 |
expire_at | integer | 否 | 设置存储的过期时刻,需传入 UTC Unix 时间戳。默认保留 3 天,最多 7 天。 | |
max_output_tokens | integer | 否 | null | 模型输出最大 token 数,包含模型回答和思维链内容。 |
thinking | object | 否 | 控制模型是否开启深度思考模式。 | |
reasoning | object | 否 | {"effort": "medium"} | 限制深度思考的工作量。 |
caching | object | 否 | {"type": "disabled"} | 是否开启缓存。 |
store | boolean | 否 | true | 是否储存生成的模型响应,以便后续通过 API 检索。 |
stream | boolean | 否 | false | 响应内容是否流式返回。 |
temperature | float | 否 | 1 | 采样温度,取值范围 [0, 2]。 |
top_p | float | 否 | 0.7 | 核采样概率阈值,取值范围 [0, 1]。 |
text | object | 否 | 模型文本输出的格式定义,可以是自然语言,也可以是结构化的 JSON 数据。详情请看结构化输出。 | |
tools | array | 否 | 模型可以调用的工具列表。 | |
tool_choice | string / object | 否 | auto | 控制模型返回是否包含待调用的工具。 |
max_tool_calls | integer | 否 | 最大工具调用轮次,取值范围 [1, 10]。 | |
context_management | object | 否 | 上下文管理策略。 |
详情请参见The response object。
详情请参见流式响应。
request
curl --location "https://operator.las.cn-beijing.volces.com/api/v1/responses" \ --header "Authorization: Bearer $LAS_API_KEY" \ --header "Content-Type: application/json" \ --data '{ "model": "doubao-seed-1-6-250615", "input": "你好呀。" }'
response
{ "id": "resp_0217****", "created_at": 1756280722.0, "error": null, "incomplete_details": null, "instructions": null, "model": "doubao-seed-1-6-250615", "object": "response", "output": [ { "id": "rs_0217****404a", "summary": [ { "text": "\n用户说“你好呀。”,这是一个很友好的问候,我需要用同样友好的方式回应。首先,应该回应用户的问候,然后可以表达一下愿意帮助的态度,让对话能够继续下去。用户可能接下来会有其他问题或者想聊聊天,所以回应要自然、亲切。\n\n首先,直接回“你好呀!”来呼应对方的问候,保持一致的语气。然后加上一句“很高兴见到你~ 有什么我可以帮你的吗?”这样既表达了友好,又主动提供帮助,让用户知道我在这里可以协助他们。这样的回应比较温暖,也符合日常交流的习惯,不会显得太生硬。\n\n需要注意的是,用户可能只是想打个招呼,所以回应不要太复杂,保持简洁和亲切就好。避免使用太正式的语言,保持口语化,让用户感觉舒适。总结一下,回应应该包含问候、表达高兴见到对方,以及提供帮助的意愿。", "type": "summary_text" } ], "type": "reasoning", "status": "completed" }, { "id": "msg_0217****a93c", "content": [ { "text": "你好呀!很高兴见到你~ 有什么我可以帮你的吗? 😊", "type": "output_text", "annotations": null } ], "role": "assistant", "status": "completed", "type": "message" } ], "parallel_tool_calls": null, "temperature": null, "tool_choice": null, "tools": null, "top_p": null, "max_output_tokens": 32768, "previous_response_id": null, "thinking": null, "service_tier": "default", "status": "completed", "text": null, "usage": { "input_tokens": 88, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 230, "output_tokens_details": { "reasoning_tokens": 211 }, "total_tokens": 318 }, "caching": { "type": "disabled" }, "store": true, "expire_at": 1756539922 }
request
curl --location 'https://operator.las.cn-beijing.volces.com/api/v1/responses' \ --header "Authorization: Bearer $LAS_API_KEY" \ --header 'Content-Type: application/json' \ --data '{ "model": "doubao-seed-1-6-250615", "input": "查询北京今天的天气", "tools":[ { "type": "function", "name": "获取天气信息", "description": "根据给出的位置获取天气信息", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "城市名称,例如:北京" } }, "required": [ "location" ] } } ] }'
response
{ "id": "resp_0217****b63e", "created_at": 1756477803.0, "error": null, "incomplete_details": null, "instructions": null, "model": "doubao-seed-1-6-250615", "object": "response", "output": [ { "id": "rs_0217****6b15", "summary": [ { "text": "让我思考一下,我现在需要回答用户的问题:“北京的天气怎么样?”根据系统提示,我需要调用工具来获取天气信息。首先,我要确认用户的问题是否需要调用工具。用户询问的是当前北京的天气,这属于实时信息,无法通过常识直接回答,所以必须使用提供的工具。\n\n接下来,查看可用的工具列表,只有一个工具叫做“获取天气信息”,它的参数需要一个“location”,类型是字符串,必填项。用户的问题中明确提到了“北京”,所以location参数应该是“北京”。\n\n然后,我需要按照格式要求生成工具调用。系统提示说调用工具需要用包裹,里面是一个JSON数组。这里只需要调用一个工具,所以是单工具调用。参数部分应该是{\"location\": \"北京\"}。\n\n检查是否有其他需要注意的地方,比如参数是否正确,格式是否符合要求。确认无误后,生成调用语句。不需要并行调用,因为只有一个工具可用,且只需要一次调用就能获取所需信息。因此,最终的调用应该是:\n\n", "type": "summary_text" } ], "type": "reasoning", "status": "completed" }, { "arguments": "{\"location\":\"北京\"}", "call_id": "call_****a6al", "name": "获取天气信息", "type": "function_call", "id": "fc_0217****c976b", "status": "completed" } ], "parallel_tool_calls": null, "temperature": null, "tool_choice": null, "tools": [ { "name": "获取天气信息", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "城市名称,例如:北京" } }, "required": ["location"] }, "type": "function", "description": "根据给出的位置获取天气信息" } ], "top_p": null, "max_output_tokens": 32768, "previous_response_id": null, "thinking": null, "service_tier": "default", "status": "completed", "text": null, "usage": { "input_tokens": 454, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 306, "output_tokens_details": { "reasoning_tokens": 279 }, "total_tokens": 760 }, "caching": { "type": "disabled" }, "store": true, "expire_at": 1756737003 }