Responses

POST https://api.tokentosea.com/v1/responses

按 OpenAI Responses 协议发送请求，原样返回厂商的 Response 对象。TokenToSea 只开放无状态用法：每次请求都携带完整上下文，不在平台或厂商侧保存会话。

前提：全能密钥或模型密钥，账户可用余额足以覆盖本次预留。Python 示例需要 pip install openai，Node.js 示例需要 Node.js 18+ 与 npm install openai。

支持的型号#

截至 2026-10-21，62 个型号开放本接口：千问 39、豆包 11、MiniMax 6、DeepSeek 2、阶跃星辰 2、Kimi 1（kimi-k3）、智谱 1（glm-5.3）。以 GET /v1/models 中 verified_endpoints 包含 /v1/responses 的型号为准。所有开放的型号都支持 JSON 与 SSE 两种响应方式。

请求头#

请求头	必填	说明
`Authorization`	必填	`Bearer sk-relay_…`，也可用 `x-api-key`
`Content-Type`	必填	`application/json`
`Idempotency-Key`	建议	见幂等

请求体#

参数	类型	必填	说明
`model`	string	必填	型号 ID
`input`	string \| array	必填	文本，或消息 / 内容项数组。图片使用 `{ "type": "input_image", "image_url": "https://…" }`，仅对已验收图片输入的型号开放
`instructions`	string	可选	系统指令，原样转发
`max_output_tokens`	integer	可选	输出上限。不传时按型号默认值补上（多数为 4096），不能超过 `max_output_tokens`
`stream`	boolean	可选	`true` 时以 SSE 事件返回
`store`	boolean	可选	只能为 `false` 或不传。网关总是以 `store: false` 发给厂商（`doubao-seed-translation-250915` 不发送该字段）
`tools`	array	可选	只接受 `type` 为 `function` 或 `custom` 的工具；托管工具（联网搜索、文件检索、代码执行等）返回 `422 hosted_tools_not_priced`
`tool_choice`	string \| object	可选	原样转发
`reasoning`	object	可选	思考配置，原样转发。DeepSeek 示例使用 `{ "effort": "none" }`
其他字段	—	可选	`temperature`、`top_p`、`text` 等原样转发，是否支持以厂商为准

不支持的用法#

请求中出现	返回
`store: true`、`previous_response_id`、`conversation`、`background: true`	`422 stateful_responses_not_supported`
`cache_control`（本接口只有 `kimi-k3` 开放显式缓存）	`422 explicit_cache_not_priced`
`prompt_cache_options`（`kimi-k3` 以外）	`422 cache_options_not_priced`
托管工具、`mcp_servers`、`web_search_options` 等附加产品	`422 hosted_tools_not_priced` / `auxiliary_product_not_priced`

需要多轮对话时，把之前的输出作为 input 的一部分重新发送。

响应#

响应示例

{
  "id": "resp_…",
  "object": "response",
  "created_at": 1792483200,
  "status": "completed",
  "model": "glm-5.3",
  "output": [
    {
      "type": "message",
      "id": "msg_…",
      "status": "completed",
      "role": "assistant",
      "content": [{ "type": "output_text", "text": "向量检索把内容转成向量，再按相似度查找。", "annotations": [] }]
    }
  ],
  "usage": {
    "input_tokens": 16,
    "output_tokens": 19,
    "total_tokens": 35,
    "input_tokens_details": { "cached_tokens": 0 }
  }
}

字段以厂商返回为准。结算未完成时返回 202 和结果信封。

流式事件#

"stream": true 时返回 text/event-stream，事件形如 event: response.output_text.delta + data: {…}。

中间事件实时转发。response.completed 在主站完成结算后才发出，usage 位于该事件的 response.usage。
结算失败时以 event: error、data: {"error":{"code":"execution_result_pending"}} 结束，用同一个键查询 /v1/results/{key}。
客户端断开不影响结算，完整 SSE 文本保存在结果的 output.stream。

示例#

文本请求#

curl https://api.tokentosea.com/v1/responses \
  -H "Authorization: Bearer $TOKENTOSEA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "model": "glm-5.3",
    "instructions": "用中文简洁回答。",
    "input": "什么是向量检索？",
    "max_output_tokens": 1024,
    "store": false
  }'

import os
import uuid

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TOKENTOSEA_API_KEY"],
    base_url="https://api.tokentosea.com/v1",
    max_retries=0,
)
response = client.responses.create(
    model="glm-5.3",
    instructions="用中文简洁回答。",
    input="什么是向量检索？",
    max_output_tokens=1024,
    store=False,
    extra_headers={"Idempotency-Key": str(uuid.uuid4())},
)
print(response.output_text)
print(response.usage)

import OpenAI from 'openai';
import { randomUUID } from 'node:crypto';

const client = new OpenAI({
  apiKey: process.env.TOKENTOSEA_API_KEY,
  baseURL: 'https://api.tokentosea.com/v1',
  maxRetries: 0,
});
const response = await client.responses.create(
  {
    model: 'glm-5.3',
    instructions: '用中文简洁回答。',
    input: '什么是向量检索？',
    max_output_tokens: 1024,
    store: false,
  },
  { headers: { 'Idempotency-Key': randomUUID() } },
);
console.log(response.output_text);
console.log(response.usage);

流式事件#

curl -N https://api.tokentosea.com/v1/responses \
  -H "Authorization: Bearer $TOKENTOSEA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "model": "glm-5.3",
    "input": "列出三个使用向量检索的场景。",
    "max_output_tokens": 1024,
    "store": false,
    "stream": true
  }'

import os
import uuid

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TOKENTOSEA_API_KEY"],
    base_url="https://api.tokentosea.com/v1",
    max_retries=0,
)
stream = client.responses.create(
    model="glm-5.3",
    input="列出三个使用向量检索的场景。",
    max_output_tokens=1024,
    store=False,
    stream=True,
    extra_headers={"Idempotency-Key": str(uuid.uuid4())},
)
for event in stream:
    if event.type == "response.output_text.delta":
        print(event.delta, end="", flush=True)
    elif event.type == "response.completed":
        print("\n", event.response.usage)

import OpenAI from 'openai';
import { randomUUID } from 'node:crypto';

const client = new OpenAI({
  apiKey: process.env.TOKENTOSEA_API_KEY,
  baseURL: 'https://api.tokentosea.com/v1',
  maxRetries: 0,
});
const stream = await client.responses.create(
  { model: 'glm-5.3', input: '列出三个使用向量检索的场景。', max_output_tokens: 1024, store: false, stream: true },
  { headers: { 'Idempotency-Key': randomUUID() } },
);
for await (const event of stream) {
  if (event.type === 'response.output_text.delta') process.stdout.write(event.delta);
  if (event.type === 'response.completed') console.log('\n', event.response.usage);
}

图片输入#

curl https://api.tokentosea.com/v1/responses \
  -H "Authorization: Bearer $TOKENTOSEA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $(uuidgen)" \
  -d '{
    "model": "qwen3.8-max",
    "input": [{
      "role": "user",
      "content": [
        {"type": "input_text", "text": "描述这张图片。"},
        {"type": "input_image", "image_url": "https://example.com/photo.jpg"}
      ]
    }],
    "max_output_tokens": 1024,
    "store": false
  }'

import os
import uuid

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["TOKENTOSEA_API_KEY"],
    base_url="https://api.tokentosea.com/v1",
    max_retries=0,
)
IMAGE_URL = "https://example.com/photo.jpg"

response = client.responses.create(
    model="qwen3.8-max",
    input=[{
        "role": "user",
        "content": [
            {"type": "input_text", "text": "描述这张图片。"},
            {"type": "input_image", "image_url": IMAGE_URL},
        ],
    }],
    max_output_tokens=1024,
    store=False,
    extra_headers={"Idempotency-Key": str(uuid.uuid4())},
)
print(response.output_text)

import OpenAI from 'openai';
import { randomUUID } from 'node:crypto';

const client = new OpenAI({
  apiKey: process.env.TOKENTOSEA_API_KEY,
  baseURL: 'https://api.tokentosea.com/v1',
  maxRetries: 0,
});
const IMAGE_URL = 'https://example.com/photo.jpg';

const response = await client.responses.create(
  {
    model: 'qwen3.8-max',
    input: [{
      role: 'user',
      content: [
        { type: 'input_text', text: '描述这张图片。' },
        { type: 'input_image', image_url: IMAGE_URL },
      ],
    }],
    max_output_tokens: 1024,
    store: false,
  },
  { headers: { 'Idempotency-Key': randomUUID() } },
);
console.log(response.output_text);

错误#

状态码	`code`	原因与处理
`422`	`stateful_responses_not_supported`	使用了 `store: true`、`previous_response_id`、`conversation` 或 `background`
`422`	`endpoint_not_verified`	该型号未开放 Responses，改用 Chat Completions
`422`	`input_modality_not_verified`	型号未开放图片输入
`422`	`hosted_tools_not_priced`、`explicit_cache_not_priced`、`cache_options_not_priced`	见不支持的用法
`400` / `409`	`main_wallet_authorization_failed`	见 API 概览

下一步#

Chat Completions：覆盖型号最多的对话接口
Messages：Anthropic 协议
流式输出指南