OpenAI 最近提供了一个创建模型响应的接口。提供文本或图像输入以生成文本或图像输出。让模型调用您自己的自定义代码或使用内置工具，如 web 搜索或文件搜索，以使用您自己的数据作为模型响应的输入。

本文档主要介绍 OpenAI Responses API 操作的使用流程，利用它我们可以轻松使用官方 OpenAI 的创建模型响应功能。

申请流程

要使用 OpenAI Responses API，首先可以到 OpenAI Responses API 页面点击「Acquire」按钮，获取请求所需要的凭证：

如果你尚未登录或注册，会自动跳转到登录页面邀请您来注册和登录，登录注册之后会自动返回当前页面。

在首次申请时会有免费额度赠送，可以免费使用该 API。

基本使用

接下来就可以在界面上填写对应的内容，如图所示：

在第一次使用该接口时，我们至少需要填写三个内容，一个是 authorization，直接在下拉列表里面选择即可。另一个参数是 model， model 就是我们选择使用 OpenAI ChatGPT 官网模型类别，这里我们主要有 20 种模型，详情可以看我们提供的模型。最后一个参数是input，input是我们输入的提问词数组，它是一个数组，表示可以同时上传多个提问词，每个提问词包含了 role 和 content，其中 role 表示提问者的角色，我们提供了三种身份，分别为 user 、assistant、system 。另一个 content 就是我们提问的具体内容。

同时您可以注意到右侧有对应的调用代码生成，您可以复制代码直接运行，也可以直接点击「Try」按钮进行测试。

调用之后，我们发现返回结果如下：

json { "id": "resp_68a98322e3c88191a027de2711a02a490554cad0b36c0400", "object": "response", "created_at": 1755939618, "status": "completed", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1", "output": [ { "id": "msg_68a98323422c8191a7f383eea48ba5160554cad0b36c0400", "type": "message", "status": "completed", "content": [ { "type": "output_text", "annotations": [], "text": "Hello! How can I assist you today?" } ], "role": "assistant" } ], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": { "effort": null, "summary": null }, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1, "text": { "format": { "type": "text" } }, "tool_choice": "auto", "tools": [], "top_p": 1, "truncation": "disabled", "usage": { "input_tokens": 8, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 10, "output_tokens_details": { "reasoning_tokens": 0 }, "total_tokens": 18 }, "user": null, "metadata": {} }

返回结果一共有多个字段，介绍如下：

• id，生成此次对话任务的 ID，用于唯一标识此次对话任务。
• model，选择的 OpenAI ChatGPT 官网模型。
• output，ChatGPT 针对提问词给于的回答信息。
• usage：针对本次问答对 token 的统计信息。

其中 output 是包含了 ChatGPT 的回答信息，它里面的 output 是 ChatGPT，可以发现如图所示。

可以看到，output 里面的 content 字段包含了 ChatGPT 回复的具体内容。

流式响应

该接口也支持流式响应，这对网页对接十分有用，可以让网页实现逐字显示效果。

如果想流式返回响应，可以更改请求头里面的 stream 参数，修改为 true。

修改如图所示，不过调用代码需要有对应的更改才能支持流式响应。

将 stream 修改为 true 之后，API 将逐行返回对应的 JSON 数据，在代码层面我们需要做相应的修改来获得逐行的结果。

Python 样例调用代码：

```python import requests

url = "https://api.acedata.cloud/openai/responses"

headers = { "accept": "application/json", "authorization": "Bearer {token}", "content-type": "application/json" }

payload = { "model": "gpt-4.1", "input": [{"role":"user","content":"Hello"}], "stream": True }

response = requests.post(url, json=payload, headers=headers) print(response.text) ```

输出效果如下：

```json data: {"type": "response.created", "sequence_number": 0, "response": {"id": "resp_68a9837bb9bc8190b403947311db6faa0721186e8fbb89d0", "object": "response", "created_at": 1755939707, "status": "in_progress", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1-data", "output": [], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": {"effort": null, "summary": null}, "safety_identifier": null, "service_tier": "auto", "store": true, "temperature": 1.0, "text": {"format": {"type": "text"}}, "tool_choice": "auto", "tools": [], "top_p": 1.0, "truncation": "disabled", "usage": null, "user": null, "metadata": {}}, "model": "gpt-4.1"}

data: {"type": "response.in_progress", "sequence_number": 1, "response": {"id": "resp_68a9837bb9bc8190b403947311db6faa0721186e8fbb89d0", "object": "response", "created_at": 1755939707, "status": "in_progress", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1-data", "output": [], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": {"effort": null, "summary": null}, "safety_identifier": null, "service_tier": "auto", "store": true, "temperature": 1.0, "text": {"format": {"type": "text"}}, "tool_choice": "auto", "tools": [], "top_p": 1.0, "truncation": "disabled", "usage": null, "user": null, "metadata": {}}, "model": "gpt-4.1"}

data: {"type": "response.output_item.added", "sequence_number": 2, "output_index": 0, "item": {"id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "type": "message", "status": "in_progress", "content": [], "role": "assistant"}, "model": "gpt-4.1"}

data: {"type": "response.content_part.added", "sequence_number": 3, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "part": {"type": "output_text", "annotations": [], "text": ""}, "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 4, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": "Hello", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 5, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": "!", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 6, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " How", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 7, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " can", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 8, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " I", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 9, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " help", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 10, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " you", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 11, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " today", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 12, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": "?", "model": "gpt-4.1"}

data: {"type": "response.output_text.delta", "sequence_number": 13, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "delta": " 😊", "model": "gpt-4.1"}

data: {"type": "response.output_text.done", "sequence_number": 14, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "text": "Hello! How can I help you today? 😊", "model": "gpt-4.1"}

data: {"type": "response.content_part.done", "sequence_number": 15, "item_id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "output_index": 0, "content_index": 0, "part": {"type": "output_text", "annotations": [], "text": "Hello! How can I help you today? 😊"}, "model": "gpt-4.1"}

data: {"type": "response.output_item.done", "sequence_number": 16, "output_index": 0, "item": {"id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "type": "message", "status": "completed", "content": [{"type": "output_text", "annotations": [], "text": "Hello! How can I help you today? 😊"}], "role": "assistant"}, "model": "gpt-4.1"}

data: {"type": "response.completed", "sequence_number": 17, "response": {"id": "resp_68a9837bb9bc8190b403947311db6faa0721186e8fbb89d0", "object": "response", "created_at": 1755939707, "status": "completed", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1-data", "output": [{"id": "msg_68a9837c49f081908f568bf9c6065c620721186e8fbb89d0", "type": "message", "status": "completed", "content": [{"type": "output_text", "annotations": [], "text": "Hello! How can I help you today? 😊"}], "role": "assistant"}], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": {"effort": null, "summary": null}, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1.0, "text": {"format": {"type": "text"}}, "tool_choice": "auto", "tools": [], "top_p": 1.0, "truncation": "disabled", "usage": {"input_tokens": 8, "input_tokens_details": {"cached_tokens": 0}, "output_tokens": 11, "output_tokens_details": {"reasoning_tokens": 0}, "total_tokens": 19}, "user": null, "metadata": {}}, "model": "gpt-4.1"}

```

可以看到，响应里面有许多 data ，data 里面的 delta 即为最新的回答内容，与上文介绍的内容一致。delta 是新增的回答内容，您可以根据结果来对接到您的系统中。同时流式响应的结束是根据 data 的内容来判断的，如果 type 的内容为 response.completed，则表示流式响应回答已经全部结束。返回的 data 结果一共有多个字段，介绍如下：

• item_id，生成此次对话任务的 ID，用于唯一标识此次对话任务。
• type，生成此次对话 Responses 任务的类型。
• model，选择的 OpenAI ChatGPT 官网模型。
• delta，ChatGPT 针对提问词给于的回答信息。

JavaScript 也是支持的，比如 Node.js 的流式调用代码如下：

```javascript const options = { method: "post", headers: { accept: "application/json", authorization: "Bearer b82d32f570bc434d9ba9923aa0e7dce0", "content-type": "application/json", }, body: JSON.stringify({ model: "gpt-4.1", input: [{ role: "user", content: "Hello" }], stream: true, }), };

fetch("https://api.acedata.cloud/openai/responses", options) .then((response) => response.json()) .then((response) => console.log(response)) .catch((err) => console.error(err)); ```

Java 样例代码：

```java JSONObject jsonObject = new JSONObject(); jsonObject.put("model", "gpt-4.1"); jsonObject.put("input", [{"role":"user","content":"Hello"}]); jsonObject.put("stream", true); MediaType mediaType = "application/json; charset=utf-8".toMediaType(); RequestBody body = jsonObject.toString().toRequestBody(mediaType); Request request = new Request.Builder() .url("https://api.acedata.cloud/openai/responses") .post(body) .addHeader("accept", "application/json") .addHeader("authorization", "Bearer b82d32f570bc434d9ba9923aa0e7dce0") .addHeader("content-type", "application/json") .build();

OkHttpClient client = new OkHttpClient(); Response response = client.newCall(request).execute(); System.out.print(response.body!!.string()) ```

其他语言可以另外自行改写，原理都是一样的。

多轮对话

如果您想要对接多轮对话功能，需要对 input 字段上传多个提问词，多个提问词的具体示例如下图所示：

Python 样例调用代码：

```python import requests

url = "https://api.acedata.cloud/openai/responses"

headers = { "accept": "application/json", "authorization": "Bearer {token}", "content-type": "application/json" }

payload = { "model": "gpt-4.1", "input": [{"role":"user","content":"Hello"},{"role":"assistant","content":"Hello! How can I help you today? 😊"},{"role":"user","content":"What did I just say?"}] }

response = requests.post(url, json=payload, headers=headers) print(response.text) ```

通过上传多个提问词，就可以轻松实现多轮对话，可以得到如下回答：

json { "id": "resp_68a989c03c508191a1dd82ce2e37e88a0932a4328c0a5d5b", "object": "response", "created_at": 1755941312, "status": "completed", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1", "output": [ { "id": "msg_68a989c092e4819189821a9eb8247e1e0932a4328c0a5d5b", "type": "message", "status": "completed", "content": [ { "type": "output_text", "annotations": [], "text": "You just said \"Hello.\" \n\nWould you like to continue the conversation or ask a question?" } ], "role": "assistant" } ], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": { "effort": null, "summary": null }, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1, "text": { "format": { "type": "text" } }, "tool_choice": "auto", "tools": [], "top_p": 1, "truncation": "disabled", "usage": { "input_tokens": 32, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 20, "output_tokens_details": { "reasoning_tokens": 0 }, "total_tokens": 52 }, "user": null, "metadata": {} }

可以看到，output 包含的信息与基本使用的内容是一致的，这个包含了 ChatGPT 针对多个对话进行回复的具体内容，这样就可以根据多个对话内容来回答对应的问题了。

视觉模型

gpt-4o 是 OpenAI 开发的多模态大型语言模型,它在 GPT-4 的基础上增加了视觉理解能力。这个模型可以同时处理文本和图像输入,实现了跨模态的理解和生成。

使用 gpt-4o 模型的文本处理是与上文的基本使用内容一致的，下面将简要介绍一下如果使用模型的图像处理能力。

使用 gpt-4o 模型的图像处理能力主要是通过在原有的 content 内容基础上添加一个 type 字段，通过该字段可以知道上传的是文本还是图片，从而使用 gpt-4o 模型的图像处理能力，下面主要讲述采用 Curl 和 Python 俩种方式来调用该功能。

• Curl 脚本方式

curl -X POST 'https://api.acedata.cloud/openai/responses' \ -H 'accept: application/json' \ -H 'authorization: Bearer {token}' \ -H 'content-type: application/json' \ -d '{ "model": "gpt-4.1", "input": [ { "role": "user", "content": [ {"type": "input_text", "text": "what is in this image?"}, { "type": "input_image", "image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" } ] } ] }'

• Python 脚本方式

```python import requests

url = "https://api.acedata.cloud/openai/chat/completions"

headers = { "accept": "application/json", "authorization": "Bearer {token}", "content-type": "application/json" }

payload = { "model": "gpt-4.1", "input": [ { "role": "user", "content": [ {"type": "input_text", "text": "what is in this image?"}, { "type": "input_image", "image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg" } ] } ] }

response = requests.post(url, json=payload, headers=headers) print(response.text) ```

然后可以得到下面的结果，结果里面的字段信息是与上文一致的，具体的如下：

json { "id": "resp_68a98c1bb784819e9b9f622007a2d37602483949012d2193", "object": "response", "created_at": 1755941915, "status": "completed", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1", "output": [ { "id": "msg_68a98c1dd030819e97fb71e6ee33f5a902483949012d2193", "type": "message", "status": "completed", "content": [ { "type": "output_text", "annotations": [], "text": "This image shows a scenic path, possibly a boardwalk, running through a lush green field or meadow. The sky above is bright blue with some white clouds, and there are green trees and bushes in the background. It looks like a peaceful nature scene, possibly in a park, wetland, or prairie area. The image conveys a sense of tranquility and natural beauty." } ], "role": "assistant" } ], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": { "effort": null, "summary": null }, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1, "text": { "format": { "type": "text" } }, "tool_choice": "auto", "tools": [], "top_p": 1, "truncation": "disabled", "usage": { "input_tokens": 1118, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 75, "output_tokens_details": { "reasoning_tokens": 0 }, "total_tokens": 1193 }, "user": null, "metadata": {} }

可以看到回答的内容是基于图片进行回答的，因此通过上述俩种方式可以轻松使用 gpt-4.1 模型的文本和图像处理能力。

除了，gpt-4.1，还有一个更低成本的模型，叫做 gpt-4o-mini。gpt-4o-mini 是 OpenAI 开发的最新一代大型语言模型，它不仅响应速度快，同时价格也更便宜，也支持多模态。vision 功能的使用可参考上文 gpt-4.1 模型的使用的内容。

文件处理模型的创建

请求样例：

json { "model": "gpt-4.1", "input": [ { "role": "user", "content": [ { "type": "input_text", "text": "what is in this file?" }, { "type": "input_file", "file_url": "https://www.berkshirehathaway.com/letters/2024ltr.pdf" } ] } ] }

样例结果：

json { "id": "resp_68a98d7bb57c819ba25424f5f50a29a300a1af2af822e88a", "object": "response", "created_at": 1755942267, "status": "completed", "background": false, "content_filters": null, "error": null, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-4.1", "output": [ { "id": "msg_68a98d7d9b80819b9b0f09b7bcd00bf900a1af2af822e88a", "type": "message", "status": "completed", "content": [ { "type": "output_text", "annotations": [], "text": "The file you posted contains the **2024 annual letter to shareholders from Berkshire Hathaway Inc.**, written by Warren E. Buffett, Chairman of the Board. This document is a comprehensive communication that is typically included in Berkshire's annual report to shareholders.\n\n### What's Inside the File:\n\n#### 1. **Chairman's Letter to Shareholders**\n - **Introduction & Philosophy:** Warren Buffett discusses the purpose of the annual report, Berkshire Hathaway’s communication style, and his philosophy for transparency and candid discussion of both successes and failures.\n - **Discussion of Mistakes:** He talks openly about the mistakes made in capital allocation and personnel decisions, emphasizing the importance of admitting errors and acting promptly to correct them.\n - **Succession Comments:** Buffett references his eventual retirement, and that Greg Abel will succeed him as CEO and writer of these letters.\n - **Anecdotal Story:** The story of Pete Liegl, founder of Forest River (an RV manufacturer acquired by Berkshire), is told to illustrate management philosophy and business decision-making.\n\n#### 2. **2024 Business and Financial Performance**\n - **Key Results:** Summary of how Berkshire performed financially in 2024 vs. 2023, including operating earnings breakdown by business segments such as insurance, BNSF railroad, and energy.\n - **Insurance Business:** GEICO and the property-casualty insurance division had a standout year, with commentary on the industry and how Berkshire approaches insurance risk, pricing, and investment of insurance \"float.\"\n - **Investments:** Discussion on Berkshire’s strategy of owning both full businesses and partial stakes (marketable securities) in large companies (e.g., Apple, American Express, Coca-Cola), and its deployment of cash.\n - **Taxes:** Reference to Berkshire breaking records in corporate tax payments ($26.8 billion to the IRS in 2024).\n\n#### 3. **Long-term Philosophy & Capitalism Commentary**\n - **On Equities:** Buffett explains why Berkshire prioritizes ownership of businesses (equities) over cash or bonds, and why the company favors long-term investments.\n - **On Capitalism:** There’s a reflection on America’s growth, the role of capitalism, savings, and capital allocation in the nation’s success, and a nod to the importance of maintaining a stable currency.\n\n#### 4. **Japanese Investments**\n - **Update on Japanese Holdings:** Berkshire’s growing investments in five Japanese trading companies, and the positive view of their management and governance.\n\n#### 5. **Berkshire Hathaway Annual Meeting**\n - **Annual Gathering Info:** Details about the annual meeting in Omaha, including social events, book sales, and charitable initiatives related to the meeting.\n - **Personal Stories:** Personal anecdotes involving Buffett’s family, (including his sister Bertie), to add a human touch to the letter.\n\n#### 6. **Performance Tables**\n - **Berkshire vs S&P 500 (1965-2024):** Two detailed tables showing annual percentage change in Berkshire’s share price vs. total return for the S&P 500, as well as long-term compounded and overall gains.\n\n---\n\n### In Summary\n\nThis file is the **2024 Berkshire Hathaway annual letter to shareholders**, primarily written by Warren Buffett. It covers business performance, management philosophy, investment strategy, earnings and taxes, insurance operations, significant holdings, capital allocation, succession updates, and more. Tables show a remarkable outperformance of Berkshire Hathaway vs. the S&P 500 over nearly six decades – a central point of pride in the letter.\n\nIf you want specifics from any particular section, let me know!" } ], "role": "assistant" } ], "parallel_tool_calls": true, "previous_response_id": null, "prompt_cache_key": null, "reasoning": { "effort": null, "summary": null }, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1, "text": { "format": { "type": "text" } }, "tool_choice": "auto", "tools": [], "top_p": 1, "truncation": "disabled", "usage": { "input_tokens": 8438, "input_tokens_details": { "cached_tokens": 0 }, "output_tokens": 731, "output_tokens_details": { "reasoning_tokens": 0 }, "total_tokens": 9169 }, "user": null, "metadata": {} }

可以看到，我们对输入的文件也进行了处理文件，结果与上文类似。

错误处理

在调用 API 时，如果遇到错误，API 会返回相应的错误代码和信息。例如：

• 400 token_mismatched：Bad request, possibly due to missing or invalid parameters.
• 400 api_not_implemented：Bad request, possibly due to missing or invalid parameters.
• 401 invalid_token：Unauthorized, invalid or missing authorization token.
• 429 too_many_requests：Too many requests, you have exceeded the rate limit.
• 500 api_error：Internal server error, something went wrong on the server.

错误响应示例

{ "success": false, "error": { "code": "api_error", "message": "fetch failed" }, "trace_id": "2cf86e86-22a4-46e1-ac2f-032c0f2a4e89" }

结论

通过本文档，您已经了解了如何使用 OpenAI Responses API 轻松实现官方 OpenAI 的创建 Responses 功能。希望本文档能帮助您更好地对接和使用该 API。如有任何问题，请随时联系我们的技术支持团队。