文档

三种协议、一套密钥。把 base_url 指向 FanQieAPI,原有代码无需重写。

概览

FanQieAPI 提供与主流大模型厂商一致的 HTTP 接口。你只需要把 base_url 指向本站,并换成 FanQieAPI 的密钥,即可调用 Kimi、DeepSeek、MiniMax 等模型。

协议端点说明
Chat Completions POST /v1/chat/completions 最通用,兼容 OpenAI SDK
Responses POST /v1/responses 面向多轮与工具调用
Anthropic Messages POST /v1/messages 兼容 Anthropic SDK
模型列表 GET /v1/models 返回当前可用模型

根地址为 https://fanqieapi.com/v1, 请勿在请求中指定其他域名。接口已开放跨域访问,可在浏览器中直接调用; 但密钥等同于余额,正式环境请仍从服务端发起请求。

鉴权

所有接口都需要携带 API Key。密钥由 FanQieAPI 控制台创建,格式形如 sk-fq-xxxxxxxx,仅在创建时显示一次。

支持两种传递方式,任选其一:

HTTP Header
# 方式一:OpenAI 风格
Authorization: Bearer sk-fq-xxxxxxxxxxxxxxxx

# 方式二:Anthropic 风格
x-api-key: sk-fq-xxxxxxxxxxxxxxxx

密钥等同于账户余额,请只在服务端保存,不要写进前端代码或公开仓库。

模型与价格

调用时 model 字段必须使用下表列出的名称, 大小写与连字符需完全一致。完整价格见 模型广场。

模型名称 上下文 输入 输出 缓存输入
Kimi-K3.1-Pro 1M $0.50 $5.00 $0.30
Kimi-K3.1-Pro-Abliterated 1M $5.00 $10.00 $0.50
DeepSeek-V4.1-Flash-1017 1M $0.02 $0.16 $0.004
DeepSeek-V4.1-Flash-1017-Abliterated 1M $0.24 $0.90 $0.05
MiniMax-M3.1-Flash 1M $0.04 $0.16 $0.004
GLM-5.3-Abliterated 1M $2.00 $4.00 $0.30
GLM-5.3-Flash-Abliterated 1M $0.24 $1.00 $0.05

价格单位:美元 / 每 100 万 tokens。

也可以直接查询模型列表:

cURL
curl https://fanqieapi.com/v1/models \
  -H "Authorization: Bearer $FANQIE_API_KEY"

第一个请求

用一条 curl 命令验证密钥是否可用:

cURL
curl https://fanqieapi.com/v1/chat/completions \
  -H "Authorization: Bearer $FANQIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4.1-Flash-1017",
    "messages": [
      { "role": "user", "content": "你好,介绍一下你自己" }
    ]
  }'

成功时返回:

JSON
{
  "id": "chatcmpl-8b34189b5e08aba0",
  "object": "chat.completion",
  "created": 1794300568,
  "model": "DeepSeek-V4.1-Flash-1017",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "你好!我是 AI 助手。" },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 187, "completion_tokens": 11, "total_tokens": 198 }
}

Chat Completions

POST /v1/chat/completions,与 OpenAI 的 Chat Completions 接口一致。

请求参数

字段类型必填说明
model string是模型名称,见上表
messages array是对话消息,支持 system / user / assistant
stream boolean否为 true 时以 SSE 流式返回
max_tokens integer否最大输出 token 数,上限 32768,超出部分按 32768 处理
temperature number否采样温度,建议 0–2
stream_options object否设为 {"include_usage": true} 时,流式末尾返回 usage

多轮对话

cURL
curl https://fanqieapi.com/v1/chat/completions \
  -H "Authorization: Bearer $FANQIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Kimi-K3.1-Pro",
    "messages": [
      { "role": "system", "content": "你是一位严谨的技术顾问。" },
      { "role": "user", "content": "什么是 KV Cache?" }
    ],
    "temperature": 0.7,
    "max_tokens": 1024
  }'

Responses

POST /v1/responses,输入用 input,输出为结构化的 output 数组,推理过程会以 reasoning 类型单独给出。

cURL
curl https://fanqieapi.com/v1/responses \
  -H "Authorization: Bearer $FANQIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4.1-Flash-1017",
    "input": "用一句话解释什么是向量数据库",
    "max_output_tokens": 512
  }'
JSON
{
  "id": "resp-fa1b2fc4e5499163",
  "object": "response",
  "created_at": 1794299910,
  "model": "DeepSeek-V4.1-Flash-1017",
  "status": "completed",
  "output": [
    {
      "type": "reasoning",
      "id": "rs_36eeff13159f87f3",
      "summary": [
        { "type": "summary_text", "text": "用户想用一句话了解向量数据库。" }
      ]
    },
    {
      "type": "message",
      "id": "msg_d3c0ff98b3430950",
      "status": "completed",
      "role": "assistant",
      "content": [
        {
          "type": "output_text",
          "text": "向量数据库是一种专门用于存储和检索高维向量的数据库。",
          "annotations": []
        }
      ]
    }
  ],
  "usage": { "input_tokens": 23, "output_tokens": 64, "total_tokens": 87 }
}

Anthropic Messages

POST /v1/messages,与 Anthropic Messages 接口一致。按 Anthropic 规范,max_tokens 为必填参数。为兼容部分客户端,网关在缺省时不做拦截, 但不保证各模型行为一致,建议始终显式传入。

cURL
curl https://fanqieapi.com/v1/messages \
  -H "x-api-key: $FANQIE_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "Kimi-K3.1-Pro",
    "max_tokens": 1024,
    "system": "你是一位中文写作助手。",
    "messages": [
      { "role": "user", "content": "帮我写一段产品介绍" }
    ]
  }'

返回 content 数组,其中 thinking 为推理内容、 text 为正式回答。

流式输出

三个接口都支持流式。在请求体中加入 "stream": true,响应类型变为 text/event-stream, 按空行分隔的若干个事件依次推送。

三种协议的流式事件格式并不相同,结束标记也不同, 解析时请按各自协议处理:

协议事件行结束标记
Chat Completions 仅 data: data: [DONE]
Responses 仅 data: response.completed 事件
Anthropic Messages event: + data: message_stop 事件

Chat Completions

cURL
curl -N https://fanqieapi.com/v1/chat/completions \
  -H "Authorization: Bearer $FANQIE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "MiniMax-M3.1-Flash",
    "messages": [ { "role": "user", "content": "数到 3" } ],
    "stream": true,
    "stream_options": { "include_usage": true }
  }'

返回:

SSE
data: {"id":"chatcmpl-25c769954826e3b2","object":"chat.completion.chunk","created":1794301860,"model":"DeepSeek-V4.1-Flash-1017","choices":[{"index":0,"delta":{"role":"assistant"},"logprobs":null,"finish_reason":null}]}

data: {"id":"chatcmpl-25c769954826e3b2","object":"chat.completion.chunk","created":1794301860,"model":"DeepSeek-V4.1-Flash-1017","choices":[{"index":0,"delta":{"reasoning_content":"需要简单:"},"logprobs":null,"finish_reason":null}]}

data: {"id":"chatcmpl-25c769954826e3b2","object":"chat.completion.chunk","created":1794301860,"model":"DeepSeek-V4.1-Flash-1017","choices":[{"index":0,"delta":{"content":"1、2、3"},"logprobs":null,"finish_reason":null}]}

data: {"id":"chatcmpl-25c769954826e3b2","object":"chat.completion.chunk","created":1794301860,"model":"DeepSeek-V4.1-Flash-1017","choices":[{"index":0,"delta":{"role":"assistant"},"logprobs":null,"finish_reason":"stop"}],"usage":{"prompt_tokens":24,"completion_tokens":43,"total_tokens":67}}

data: [DONE]

只有显式设置 include_usage: true 时,Chat 流式响应才会附带 usage;否则不返回,用量仍会正常计费。

三个模型都会输出推理过程,在 Chat 协议中表现为 delta.reasoning_content, 正式回答在 delta.content。 两者可能交替出现,渲染给终端用户时应只取 content。 不同模型的结束分片形态略有差异,delta 可能为空对象,解析时请容错。

Responses

事件类型由 type 字段区分,文本增量在 response.output_text.delta, 全部结束后推送 response.completed, 没有 [DONE] 标记。

SSE
data: {"type":"response.created","response":{"id":"resp-8d48cda9afc5fa71","status":"in_progress"}}

data: {"type":"response.in_progress","response":{"id":"resp-8d48cda9afc5fa71","status":"in_progress"}}

data: {"type":"response.output_item.added","output_index":0,"item":{"type":"reasoning","id":"rs_4d1d6ebd22e4118c","summary":[]}}

data: {"type":"response.reasoning_summary_text.delta","item_id":"rs_4d1d6ebd22e4118c","output_index":0,"summary_index":0,"delta":"用户想打个招呼。"}

data: {"type":"response.output_item.added","output_index":1,"item":{"type":"message","id":"msg_b12e48994b391682","status":"in_progress","role":"assistant","content":[]}}

data: {"type":"response.output_text.delta","item_id":"msg_b12e48994b391682","output_index":1,"content_index":0,"delta":"你好"}

data: {"type":"response.output_text.done","item_id":"msg_b12e48994b391682","output_index":1,"content_index":0,"text":"你好"}

data: {"type":"response.output_item.done","output_index":1,"item":{"type":"message","id":"msg_b12e48994b391682","status":"completed","role":"assistant","content":[{"type":"output_text","text":"你好","annotations":[]}]}}

data: {"type":"response.completed","response":{"id":"resp-8d48cda9afc5fa71","status":"completed","usage":{"input_tokens":22,"output_tokens":36,"total_tokens":58}}}

Anthropic Messages

每个事件都有 event: 行与 data: 行,以 message_stop 结束, 同样没有 [DONE]。

SSE
event: message_start
data: {"type":"message_start","message":{"id":"msg_0b0a35a9bc89f12e","role":"assistant","model":"Kimi-K3.1-Pro","content":[],"usage":{"input_tokens":9,"output_tokens":1}}}

event: content_block_delta
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"你好"}}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":36}}

event: message_stop
data: {"type":"message_stop"}

SDK 接入

官方 SDK 均可直接使用,只需替换 base_url 与密钥。

Python(OpenAI)

Python
from openai import OpenAI

client = OpenAI(
    base_url="https://fanqieapi.com/v1",
    api_key="sk-fq-xxxxxxxxxxxxxxxx",
)

resp = client.chat.completions.create(
    model="DeepSeek-V4.1-Flash-1017",
    messages=[{"role": "user", "content": "你好"}],
)
print(resp.choices[0].message.content)

Python(Anthropic)

Python
import anthropic

client = anthropic.Anthropic(
    base_url="https://fanqieapi.com",
    api_key="sk-fq-xxxxxxxxxxxxxxxx",
)

msg = client.messages.create(
    model="Kimi-K3.1-Pro",
    max_tokens=1024,
    messages=[{"role": "user", "content": "你好"}],
)
print(msg.content[0].text)

Node.js

JavaScript
import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://fanqieapi.com/v1",
  apiKey: process.env.FANQIE_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "MiniMax-M3.1-Flash",
  messages: [{ role: "user", content: "你好" }],
});

console.log(resp.choices[0].message.content);

Anthropic SDK 会自动拼接 /v1/messages, 因此 base_url 填 https://fanqieapi.com 即可。

计费与额度

账户采用预充值模式,按实际消耗的 token 扣减余额。每次请求结束后立即结算, 可在响应体的 usage 字段看到本次用量。

  • 输入与输出分别计价,缓存命中的输入按缓存价计费,价格更低。
  • 余额不足时请求直接返回 402,不会调用模型。
  • 流式请求中途断开,已产生的用量仍会结算。
  • 扣费金额 = 输入 tokens × 输入价 + 缓存输入 tokens × 缓存价 + 输出 tokens × 输出价。

max_tokens 限制的是下发给你的可见输出长度, 网关会对超出部分做截断;计费则以模型实际生成的 token 为准。 推理模型的思考过程同样计入输出,因此当 max_tokens 设得很小时,可能出现 content 为空、 finish_reason 为 length, 但仍按较多 token 计费的情况。若需严格控制成本,建议同时约束输入长度, 或选择输出价更低的模型。

错误码

错误响应遵循所调用协议的标准格式,error.message 为可直接展示给终端用户的中文说明。

状态码含义处理建议
400请求参数不合法检查 JSON 格式、必填字段与模型名
401未提供或密钥无效确认请求头中的密钥是否正确
402余额不足充值后重试
403密钥或账户已被禁用联系商务恢复
404模型不存在或接口路径错误使用模型广场中的准确名称,并检查请求路径
413请求体过大单次请求不超过 8 MB
429请求过于频繁或并发超限降低频率或减少并发后重试
502上游服务异常稍后重试
504请求超时缩短输入或稍后重试
JSON
{
  "error": {
    "message": "账户余额不足,请充值后重试。",
    "type": "insufficient_quota",
    "param": null,
    "code": "insufficient_quota"
  }
}

速率限制

项目限制
单密钥请求频率120 次 / 分钟
单密钥并发请求8 个
单次请求体大小8 MB
单次请求最长处理时间600 秒

如需更高配额,请联系 商务咨询。