從提示詞物件遷移

OpenAI 正在棄用 API 中可重複使用的提示詞物件。自 2026 年 6 月 3 日起，將不再強調提示詞建立功能，而 v1/prompts 預定於 2026 年 11 月 30 日關閉。請參閱已棄用項目頁面，瞭解最新時程。

若要停止使用 OpenAI API 平台中的 提示詞 功能，請將提示詞內容從受管理的 prompt 物件移至應用程式碼中。這樣能讓你更充分地掌控審查、測試、部署和版本控管。

遷移前：使用提示詞物件

使用提示詞物件

JavaScript

1
2
3
4
5
6
7
8
9
10
11
12
13
14import OpenAI from "openai";

const client = new OpenAI();

const response = await client.responses.create({
  prompt: {
    id: "pmpt_123",
    version: "1",
    variables: {
      customer_name: "Acme",
      issue: "billing question",
    },
  },
});

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17# Replace the illustrative IDs and URLs below with your own resource values.

from openai import OpenAI

client = OpenAI()
prompt_id = "pmpt_123"

response = client.responses.create(
    prompt={
        "prompt_id": prompt_id,
        "version": "1",
        "variables": {
            "customer_name": "Acme",
            "issue": "billing question",
        },
    }
)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27package main

import (
	"context"
	"fmt"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	response, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Prompt: responses.ResponsePromptParam{
			ID:      "pmpt_123",
			Version: openai.String("1"),
			Variables: map[string]responses.ResponsePromptVariableUnionParam{
				"customer_name": {OfString: openai.String("Acme")},
				"issue":         {OfString: openai.String("billing question")},
			},
		},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.OutputText())
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.JsonValue;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.ResponsePrompt;

String promptId = "pmpt_123";

ResponseCreateParams params =
    ResponseCreateParams.builder()
        .prompt(
            ResponsePrompt.builder()
                .id(promptId)
                .version("1")
                .variables(
                    ResponsePrompt.Variables.builder()
                        .putAdditionalProperty("customer_name", JsonValue.from("Acme"))
                        .putAdditionalProperty("issue", JsonValue.from("billing question"))
                        .build())
                .build())
        .build();

client.responses().create(params).output().stream()
    .flatMap(item -> item.message().stream())
    .flatMap(message -> message.content().stream())
    .flatMap(content -> content.outputText().stream())
    .forEach(text -> System.out.println(text.text()));

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16require "openai"

client = OpenAI::Client.new

response = client.responses.create(
  prompt: {
    id: "pmpt_123",
    version: "1",
    variables: {
      customer_name: "Acme",
      issue: "billing question"
    }
  }
)

puts(response.output_text)

1
2
3
4
5
6
7
8
9
10
11
12
13curl https://api.openai.com/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "prompt": {
      "prompt_id": "pmpt_123",
      "version": "1",
      "variables": {
        "customer_name": "Acme",
        "issue": "billing question"
      }
    }
  }'

遷移後：將提示詞直接寫入程式碼

將提示詞直接寫入程式碼

JavaScript

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21import OpenAI from "openai";

const client = new OpenAI();

const response = await client.responses.create({
  model: "gpt-6-astra",
  input: [
    {
      role: "system",
      content:
        "You are a helpful support assistant. Be concise, accurate, and friendly.",
    },
    {
      role: "user",
      content:
        "Customer name: Acme. Issue: billing question. Write a response to the customer.",
    },
  ],
});

console.log(response.output_text);

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19from openai import OpenAI

client = OpenAI()

response = client.responses.create(
    model="gpt-6-astra",
    input=[
        {
            "role": "system",
            "content": "You are a helpful support assistant. Be concise, accurate, and friendly.",
        },
        {
            "role": "user",
            "content": "Customer name: Acme. Issue: billing question. Write a response to the customer.",
        },
    ],
)

print(response.output_text)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24package main

import (
	"context"
	"fmt"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	response, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{OfInputItemList: responses.ResponseInputParam{
			responses.ResponseInputItemParamOfMessage("You are a helpful support assistant. Be concise, accurate, and friendly.", responses.EasyInputMessageRoleSystem),
			responses.ResponseInputItemParamOfMessage("Customer name: Acme. Issue: billing question. Write a response to the customer.", responses.EasyInputMessageRoleUser),
		}},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.OutputText())
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.responses.EasyInputMessage;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.ResponseInputItem;
import java.util.List;

ResponseCreateParams params =
    ResponseCreateParams.builder()
        .model("gpt-6-astra")
        .inputOfResponse(
            List.of(
                ResponseInputItem.ofEasyInputMessage(
                    EasyInputMessage.builder()
                        .role(EasyInputMessage.Role.SYSTEM)
                        .content(
                            "You are a helpful support assistant. Be concise, accurate, and friendly.")
                        .build()),
                ResponseInputItem.ofEasyInputMessage(
                    EasyInputMessage.builder()
                        .role(EasyInputMessage.Role.USER)
                        .content(
                            "Customer name: Acme. Issue: billing question. Write a response to the customer.")
                        .build())))
        .build();

client.responses().create(params).output().stream()
    .flatMap(item -> item.message().stream())
    .flatMap(message -> message.content().stream())
    .flatMap(content -> content.outputText().stream())
    .forEach(text -> System.out.println(text.text()));

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19using OpenAI.Responses;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ResponsesClient client = new(key);

ResponseResult response = await client.CreateResponseAsync(
    "gpt-6-astra",
    [
        ResponseItem.CreateSystemMessageItem(
            "You are a helpful support assistant. Be concise, accurate, and friendly."
        ),
        ResponseItem.CreateUserMessageItem(
            "Customer name: Acme. Issue: billing question. Write a response to the customer."
        ),
    ]
);

Console.WriteLine(response.GetOutputText());

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19require "openai"

client = OpenAI::Client.new

response = client.responses.create(
  model: "gpt-6-astra",
  input: [
    {
      role: :system,
      content: "You are a helpful support assistant. Be concise, accurate, and friendly."
    },
    {
      role: :user,
      content: "Customer name: Acme. Issue: billing question. Write a response to the customer."
    }
  ]
)

puts(response.output_text)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16curl https://api.openai.com/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -d '{
    "model": "gpt-6-astra",
    "input": [
      {
        "role": "system",
        "content": "You are a helpful support assistant. Be concise, accurate, and friendly."
      },
      {
        "role": "user",
        "content": "Customer name: Acme. Issue: billing question. Write a response to the customer."
      }
    ]
  }'

使用 Codex 進行遷移

使用 OpenAI Developers 外掛程式和 OpenAI Docs 技能，自動完成遷移，並加快使用 OpenAI API 進行開發的速度。

$openai-docs update this project to store prompts in code instead of using a prompts object

有哪些變更

不再從 API 請求中參照已儲存的提示詞物件，而是將提示詞文字儲存在程式碼庫中，並在呼叫 Responses API 時，直接將產生的訊息作為 input 傳入。

將提示詞內容移至原始碼中 ，讓提示詞的變更與產品邏輯遵循相同的審查和發布流程。
以函式引數取代提示詞變數 ，讓應用程式中的動態值明確且具有型別。
呼叫 Responses API 時，透過 input 傳入訊息 ，而不使用 prompt 物件。
將版本控管移至程式碼庫 ，使用 git 提交、PR 審查，以及測試或評估來管理。
將靜態內容放在前面，動態內容放在後面 ，以保留提示詞快取的效益，因為快取命中取決於前綴是否完全相符。

範例

使用輔助函式建立提示詞

JavaScript

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25import OpenAI from "openai";

const client = new OpenAI();

function buildSupportPrompt({ customerName, issue }) {
  return [
    {
      role: "system",
      content:
        "You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details.",
    },
    {
      role: "user",
      content: `Customer name: ${customerName}. Issue: ${issue}. Write a response to the customer.`,
    },
  ];
}

const response = await client.responses.create({
  model: "gpt-6-astra",
  input: buildSupportPrompt({
    customerName: "Acme",
    issue: "billing question",
  }),
});

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25from openai import OpenAI

client = OpenAI()


def build_support_prompt(customer_name, issue):
    return [
        {
            "role": "system",
            "content": "You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details.",
        },
        {
            "role": "user",
            "content": f"Customer name: {customer_name}. Issue: {issue}. Write a response to the customer.",
        },
    ]


response = client.responses.create(
    model="gpt-6-astra",
    input=build_support_prompt(
        customer_name="Acme",
        issue="billing question",
    ),
)

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28package main

import (
	"context"
	"fmt"

	"github.com/openai/openai-go/v3"
	"github.com/openai/openai-go/v3/responses"
)

func main() {
	client := openai.NewClient()
	response, err := client.Responses.New(context.Background(), responses.ResponseNewParams{
		Model: "gpt-6-astra",
		Input: responses.ResponseNewParamsInputUnion{OfInputItemList: buildSupportPrompt("Acme", "billing question")},
	})
	if err != nil {
		panic(err)
	}
	fmt.Println(response.OutputText())
}

func buildSupportPrompt(customerName string, issue string) responses.ResponseInputParam {
	return responses.ResponseInputParam{
		responses.ResponseInputItemParamOfMessage("You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details.", responses.EasyInputMessageRoleSystem),
		responses.ResponseInputItemParamOfMessage(fmt.Sprintf("Customer name: %s. Issue: %s. Write a response to the customer.", customerName, issue), responses.EasyInputMessageRoleUser),
	}
}

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.responses.EasyInputMessage;
import com.openai.models.responses.ResponseCreateParams;
import com.openai.models.responses.ResponseInputItem;
import java.util.List;

private static List<ResponseInputItem> buildSupportPrompt(String customerName, String issue) {
  return List.of(
      ResponseInputItem.ofEasyInputMessage(
          EasyInputMessage.builder()
              .role(EasyInputMessage.Role.SYSTEM)
              .content(
                  "You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details.")
              .build()),
      ResponseInputItem.ofEasyInputMessage(
          EasyInputMessage.builder()
              .role(EasyInputMessage.Role.USER)
              .content(
                  "Customer name: "
                      + customerName
                      + ". Issue: "
                      + issue
                      + ". Write a response to the customer.")
              .build()));
}

ResponseCreateParams params =
    ResponseCreateParams.builder()
        .model("gpt-6-astra")
        .inputOfResponse(buildSupportPrompt("Acme", "billing question"))
        .build();

client.responses().create(params).output().stream()
    .flatMap(item -> item.message().stream())
    .flatMap(message -> message.content().stream())
    .flatMap(content -> content.outputText().stream())
    .forEach(text -> System.out.println(text.text()));

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21using OpenAI.Responses;
#pragma warning disable OPENAI001

string key = Environment.GetEnvironmentVariable("OPENAI_API_KEY")!;
ResponsesClient client = new(key);

static ResponseItem[] BuildSupportPrompt(string customerName, string issue) =>
[
    ResponseItem.CreateSystemMessageItem(
        "You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details."
    ),
    ResponseItem.CreateUserMessageItem(
        $"Customer name: {customerName}. Issue: {issue}. Write a response to the customer."
    ),
];

ResponseResult response = await client.CreateResponseAsync(
    "gpt-6-astra",
    BuildSupportPrompt("Acme", "billing question")
);
Console.WriteLine(response.GetOutputText());

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23require "openai"

def build_support_prompt(customer_name, issue)
  [
    {
      role: :system,
      content: "You are a helpful support assistant. Be concise, accurate, and friendly. Do not invent policy details."
    },
    {
      role: :user,
      content: "Customer name: #{customer_name}. Issue: #{issue}. Write a response to the customer."
    }
  ]
end

client = OpenAI::Client.new

response = client.responses.create(
  model: "gpt-6-astra",
  input: build_support_prompt("Acme", "billing question")
)

puts(response.output_text)

遷移的好處

你能更嚴密地掌控工程流程：提示詞與產品程式碼放在一起，變更透過 PR 進行，測試和評估可在 CI 中執行，也能透過自己的設定或功能旗標來管理推出流程或實驗。

不要將提示詞直接寫在程式碼庫的各個角落。請建立一個小型的 prompts/ 模組，以具名的建構函式定義每個提示詞，並加入輕量的評估測試資料，讓提示詞變更能像產品邏輯一樣接受審查。

建議

開始使用

核心概念

SDK 與 CLI

資源

舊版 API

選擇模型

文字與程式碼

提示詞

推理

圖像與影片

即時互動與音訊

專用模型

智慧體 API

智慧體 SDK

ChatKit

搜尋與檢索

連接工具與資料

建立工具工作流程

電腦與程式碼

媒體

GPT-Live

Realtime API

開發語音應用

連線

音訊處理

正式上線

效能與品質

成本與吞吐量

安全與治理

基礎架構與存取

營運

核心概念

Plan

Build

Test and publish

Conversion specs

指南

資源

開始使用

指南

File Upload

API

Measurement

Advertiser API

API 參考文件

開始使用

基礎

探索

適用平台

版本資訊

工作流程

能力

參考資料

自訂

設定檔

智慧體組態

擴充 ChatGPT 與 Codex 的功能

Linux

Windows

開發工作流程

擴充與自動化

環境

使用 Codex 進行建置

第三方整合

參考資料

權限

Codex Security

網路安全

開始使用

ChatGPT Work

身分識別與身分驗證

工作區存取權、政策與模型

外掛程式與連接器控制措施

使用情況、治理與合規

部署與模型供應商

社群

部落格

社群

部落格