( API ) — YuazLLM 3B MoE

API 三种格式

同一个模型，三种调用方式：原子格式（网页服务在用的原生格式）、 OpenAI 兼容、Claude 兼容。三种都支持流式与非流式。

鉴权：公开免鉴权 限流：150 次/分钟 max_tokens：≤ 500 思考链：≤ 500 字（与回复分开） 上下文：1024 字 · 最多 8 轮 模型名：yuaz-llm-3b-moe
POSThttps://ai.yuaz.club/api/ai/chat

网页对话页正在使用的原生格式：只传本轮 message 与历史 history，超长由服务端自动剔除最旧轮次（最多保留最近 8 轮）。

请求体
{
  "message": "你好",
  "history": [
    { "role": "user", "content": "上一句" },
    { "role": "assistant", "content": "上一句的回复" }
  ],
  "stream": false,
  "max_new_tokens": 500,
  "temperature": 0.7,
  "top_k": 20,
  "top_p": 0.9,
  "repetition_penalty": 1.15,
  "force_think": false
}
参数说明
max_new_tokens   回复字数上限，1–500（默认 500）
force_think      强制思考模式：true 时服务端在助手回合前预填思考标签，
                 让模型先输出思考过程；思考最多 500 字，与回复字数分开计算。
                 思考超过 500 字仍未作答时，服务端会自动截断思考链并继续生成答案。
history          最多保留最近 8 轮，超出自动丢弃最早的整轮。
响应（非流式）
{
  "success": true,
  "reply": "我是元晦（Yuaz LLM YuAn Hui），原子漫画社……",
  "time": "8.2s",
  "device": "Vulkan GPU (元晦 3B MoE SFT-v3)",
  "thinking": "用户问我是谁，先简单介绍自己。",
  "sources": [],
  "prompt_tokens": 12,
  "completion_tokens": 42,
  "tools_used": 0,
  "rounds": 1,
  "context_truncated": false,
  "dropped_rounds": 0
}
流式（stream: true）
event: think                      ← 思维链增量（前端可折叠展示）
data: {"delta":"我先想想…"}

event: sources                    ← 联网搜索命中时返回来源
data: {"items":[{"title":"原神_百度百科","url":"https://baike.baidu.com/…"}]}

event: meta                       ← 进度 / 上下文状态
data: {"status":"thinking","context_truncated":false,"dropped_rounds":0}

data: {"token":"我"}
data: {"token":"是"}

event: usage                      ← 本次 token 用量
data: {"prompt_tokens":12,"completion_tokens":8}

event: meta
data: {"tools_used":0,"rounds":1,"context_truncated":false,"dropped_rounds":0}

data: [DONE]
curl
curl -N -X POST https://ai.yuaz.club/api/ai/chat \
  -H 'Content-Type: application/json' \
  -d '{"message":"你好","stream":true}'
POSThttps://ai.yuaz.club/v1/chat/completions
GEThttps://ai.yuaz.club/v1/models

兼容 OpenAI Chat Completions。把 SDK 的 base_url 指到 https://ai.yuaz.club/v1，api_key 随便填即可。

请求体
{
  "model": "yuaz-llm-3b-moe",
  "messages": [
    { "role": "system", "content": "你是一个友好的助手" },
    { "role": "user", "content": "你好" }
  ],
  "max_tokens": 500,
  "temperature": 0.7,
  "stream": false
}
响应（非流式）
{
  "id": "chatcmpl-xxxxxxxx",
  "object": "chat.completion",
  "created": 1789187000,
  "model": "yuaz-llm-3b-moe",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "我是元晦……" },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 0, "completion_tokens": 0, "total_tokens": 0 }
}
流式（标准 SSE chunk）
data: {"id":"chatcmpl-xxx","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"我"},"finish_reason":null}]}

data: {"...":"...","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]
Python（openai SDK）
from openai import OpenAI

client = OpenAI(base_url="https://ai.yuaz.club/v1", api_key="not-needed")
resp = client.chat.completions.create(
    model="yuaz-llm-3b-moe",
    messages=[{"role": "user", "content": "你好"}],
)
print(resp.choices[0].message.content)
curl
curl -N https://ai.yuaz.club/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer not-needed' \
  -d '{"model":"yuaz-llm-3b-moe","messages":[{"role":"user","content":"你好"}],"stream":true}'
POSThttps://ai.yuaz.club/v1/messages

兼容 Anthropic Messages API（含 system 字段与内容块数组，以及助手预填续写）。

请求体
{
  "model": "yuaz-llm-3b-moe",
  "max_tokens": 500,
  "system": "你是一个友好的助手",
  "messages": [
    { "role": "user", "content": "你好" }
  ],
  "stream": false
}
响应（非流式）
{
  "id": "msg_xxxxxxxx",
  "type": "message",
  "role": "assistant",
  "model": "yuaz-llm-3b-moe",
  "content": [ { "type": "text", "text": "我是元晦……" } ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": { "input_tokens": 0, "output_tokens": 0 }
}
流式（Anthropic 事件序列）
event: message_start
data: {"type":"message_start","message":{"id":"msg_xxx","role":"assistant","content":[]}}

event: content_block_start
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}}

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

event: content_block_stop
data: {"type":"content_block_stop","index":0}

event: message_delta
data: {"type":"message_delta","delta":{"stop_reason":"end_turn"},"usage":{"output_tokens":1}}

event: message_stop
data: {"type":"message_stop"}
Python（anthropic SDK）
from anthropic import Anthropic

client = Anthropic(base_url="https://ai.yuaz.club", api_key="not-needed")
msg = client.messages.create(
    model="yuaz-llm-3b-moe",
    max_tokens=500,
    messages=[{"role": "user", "content": "你好"}],
)
print(msg.content[0].text)
curl
curl -N https://ai.yuaz.club/v1/messages \
  -H 'Content-Type: application/json' \
  -H 'anthropic-version: 2023-06-01' \
  -H 'x-api-key: not-needed' \
  -d '{"model":"yuaz-llm-3b-moe","max_tokens":500,"messages":[{"role":"user","content":"你好"}]}'

试一试

结果会显示在这里（非流式）。