模型服务 API 文档

1. 接入地址

在管理后台「模型网关 → 密钥管理」生成 Key，并一键复制调用示例。

2. 重要：默认流式输出

3. 请求示例

cURL（流式，默认）

-N 关闭 curl 自身缓冲，才能看到逐字输出。

curl -N https://www.muyaotech.com/api/gateway/chat/completions \
  -H "Authorization: Bearer sk-xxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4-Flash",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "你好，介绍一下你自己"}
    ]
  }'

cURL（非流式，需显式 stream:false）

curl https://www.muyaotech.com/api/gateway/chat/completions \
  -H "Authorization: Bearer sk-xxxx" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "DeepSeek-V4-Flash",
    "messages": [{"role": "user", "content": "你好"}],
    "stream": false
  }'

Python（OpenAI SDK 兼容，流式）

from openai import OpenAI

client = OpenAI(base_url="https://www.muyaotech.com/api/gateway", api_key="sk-xxxx")

stream = client.chat.completions.create(
    model="DeepSeek-V4-Flash",
    messages=[{"role": "user", "content": "你好"}],
    stream=True,                       # 与服务端默认一致
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)

# 如需一次性返回，改为 stream=False
# resp = client.chat.completions.create(..., stream=False)
# print(resp.choices[0].message.content)

Node.js（原生 fetch 读 SSE）

const res = await fetch("https://www.muyaotech.com/api/gateway/chat/completions", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: "Bearer sk-xxxx",
  },
  body: JSON.stringify({
    model: "DeepSeek-V4-Flash",
    messages: [{ role: "user", content: "你好" }],
    // 不传 stream 即为流式
  }),
});

const reader = res.body.getReader();
const decoder = new TextDecoder();
let buf = "";
while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  buf += decoder.decode(value, { stream: true });
  const lines = buf.split("\n");
  buf = lines.pop() ?? "";
  for (const line of lines) {
    const s = line.trim();
    if (!s.startsWith("data:")) continue;
    const data = s.slice(5).trim();
    if (data === "[DONE]") continue;            // 流结束标志
    const obj = JSON.parse(data);
    if (obj.usage) continue;                    // 末帧为用量统计
    const delta = obj.choices?.[0]?.delta?.content;
    if (delta) process.stdout.write(delta);
  }
}

4. 请求参数

5. 可调用模型清单

以下模型经本服务实测可用，可直接在 model 字段指定。未列出的模型（图像/语音/视频/Embedding/Rerank 等专用端点，或无权限模型）本接口不对外开放。

6. 函数调用（Function Calling / tool_calls）

本服务支持 OpenAI 原生 tool_calls：在请求中传入 tools（可选 tool_choice）， 模型可在需要时返回要调用的函数名与参数；客户端执行后，将结果以 role: "tool" 消息回传，即可完成多轮函数调用。所有白名单模型（含 DeepSeek / Qwen 系列）均经实测支持。

Python（非流式，单轮函数调用）

from openai import OpenAI

client = OpenAI(base_url="https://www.muyaotech.com/api/gateway", api_key="sk-xxxx")

tools = [{
    "type": "function",
    "function": {
        "name": "get_weather",
        "description": "获取指定城市的当前天气",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string", "description": "城市名"}},
            "required": ["city"],
        },
    },
}]

# 非流式需显式 stream=False
resp = client.chat.completions.create(
    model="DeepSeek-V4-Flash",
    messages=[{"role": "user", "content": "北京今天天气怎么样？"}],
    tools=tools,
    stream=False,
)
msg = resp.choices[0].message
if msg.tool_calls:
    call = msg.tool_calls[0]
    print("调用函数:", call.function.name, call.function.arguments)
    # => 调用函数: get_weather {"city": "北京"}
    # 客户端在此执行 get_weather("北京")，把结果回传后再次调用即可拿到自然语言回答：
    # messages = [{"role":"user","content":"北京今天天气怎么样？"}, msg,
    #             {"role":"tool","tool_call_id":call.id,"content":"晴 26℃"}]
    # answer = client.chat.completions.create(model="DeepSeek-V4-Flash", messages=messages, tools=tools, stream=False)

Node.js（流式增量拼装 tool_calls）

// 流式默认开启；按 index 累积 tool_calls 片段
const toolAcc = {};  // index -> { name, arguments }
const reader = res.body.getReader();
const decoder = new TextDecoder();
let buf = "";
while (true) {
  const { done, value } = await reader.read();
  if (done) break;
  buf += decoder.decode(value, { stream: true });
  const lines = buf.split("\n");
  buf = lines.pop() ?? "";
  for (const line of lines) {
    const s = line.trim();
    if (!s.startsWith("data:")) continue;
    const data = s.slice(5).trim();
    if (data === "[DONE]") continue;
    const obj = JSON.parse(data);
    if (obj.usage) continue;
    for (const part of obj.choices?.[0]?.delta?.tool_calls ?? []) {
      const i = part.index ?? 0;
      toolAcc[i] = toolAcc[i] || { name: "", arguments: "" };
      if (part.function?.name) toolAcc[i].name += part.function.name;
      if (part.function?.arguments) toolAcc[i].arguments += part.function.arguments;
    }
  }
}
// 全部读完后，toolAcc 中即为完整函数名 + JSON 参数字符串
console.log(toolAcc);

7. 响应格式

流式（默认）

响应头 Content-Type: text/event-stream。每帧一行 data:，增量文本在choices[0].delta.content：

data:{"choices":[{"delta":{"content":"你"},"index":0,"finish_reason":null}],"object":"chat.completion.chunk","model":"deepseek-v4-flash","id":"chatcmpl-xxx"}

data:{"choices":[{"delta":{"content":"好"},"index":0,"finish_reason":null}],"object":"chat.completion.chunk","model":"deepseek-v4-flash","id":"chatcmpl-xxx"}

data:{"choices":[{"delta":{"content":""},"index":0,"finish_reason":"stop"}],"object":"chat.completion.chunk","model":"deepseek-v4-flash","id":"chatcmpl-xxx"}

data:{"choices":[],"usage":{"prompt_tokens":7,"completion_tokens":42,"total_tokens":49},"object":"chat.completion.chunk","model":"deepseek-v4-flash","id":"chatcmpl-xxx"}

data: [DONE]

• 用量统计在倒数第二帧：该帧 choices 为空数组，携带 usage。
• 流结束标志为 data: [DONE]，收到后即可停止读取。
• 部分模型（如 R1 系列）会额外返回 delta.reasoning_content 思考过程，可按需忽略。

非流式（stream: false）

{
  "id": "chatcmpl-xxx",
  "object": "chat.completion",
  "model": "deepseek-v4-flash",
  "choices": [
    { "index": 0, "message": { "role": "assistant", "content": "你好…" }, "finish_reason": "stop" }
  ],
  "usage": { "prompt_tokens": 7, "completion_tokens": 42, "total_tokens": 49 }
}

8. 错误码

错误统一返回 JSON（流式请求若在建流前失败，同样返回 JSON 而非 SSE）：

{ "error": { "message": "Invalid API key", "code": "invalid_api_key" } }

9. 说明

• 多轮对话：在 messages 中保留历史消息即可，服务端不存储上下文。
• 计量：每次调用按 Key 归属记录调用次数与 token（流式在末帧结算），可在管理后台查看每日/模型/Key 维度消耗。
• 默认模型为 DeepSeek-V4-Flash（高速响应）；不指定 model 时使用它。
• 流式响应已关闭中间层缓冲，可实现逐字输出；客户端请勿开启响应缓冲（如 curl 需加 -N）。