Kimi K3 企业接入全解：官方 API 与多模型聚合路径怎么选？

pip install openai

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MOONSHOT_API_KEY"],
    base_url="https://api.moonshot.cn/v1",
)

completion = client.chat.completions.create(
    model="kimi-k3",
    messages=[
        {"role": "system", "content": "你是一位专业的技术文档工程师。"},
        {"role": "user", "content": "用三句话解释什么是 Transformer 注意力机制。"},
    ],
)
print(completion.choices[0].message.content)

npm install openai

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.MOONSHOT_API_KEY,
  baseURL: "https://api.moonshot.cn/v1",
});

const completion = await client.chat.completions.create({
  model: "kimi-k3",
  messages: [
    { role: "system", content: "你是一位专业的技术文档工程师。" },
    { role: "user", content: "用三句话解释什么是 Transformer 注意力机制。" },
  ],
});
console.log(completion.choices[0].message.content);

curl https://api.moonshot.cn/v1/chat/completions \
  -H "Authorization: Bearer $MOONSHOT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3",
    "messages": [
      {"role": "system", "content": "你是一位专业的技术文档工程师。"},
      {"role": "user", "content": "用三句话解释什么是 Transformer 注意力机制。"}
    ]
  }'

completion = client.chat.completions.create(
    model="kimi-k3",
    reasoning_effort="low",   # low / high / max
    messages=[{"role": "user", "content": "2 + 2 等于几？"}],
)

from pathlib import Path

# 长文档作为固定前缀
knowledge = Path("product-manual.md").read_text(encoding="utf-8")

questions = [
    "第三章的核心结论是什么？",
    "列出所有提到的技术限制。",
    "对比第二章和第四章的方案差异。",
]

for question in questions:
    completion = client.chat.completions.create(
        model="kimi-k3",
        messages=[
            {"role": "system", "content": knowledge},  # 固定前缀，命中缓存
            {"role": "user", "content": question},     # 每次变化的部分
        ],
    )
    print(f"Q: {question}")
    print(f"A: {completion.choices[0].message.content}\n")

总成本 = 输出量 × ¥100/M + 缓存命中输入 × ¥2/M + 未命中输入 × ¥20/M

import json

tools = [
    {
        "type": "function",
        "function": {
            "name": "query_database",
            "description": "查询企业内部数据库，返回指定表的记录",
            "parameters": {
                "type": "object",
                "properties": {
                    "table": {"type": "string", "description": "表名"},
                    "filter": {"type": "string", "description": "查询条件（SQL WHERE 子句格式）"},
                },
                "required": ["table"],
            },
        },
    }
]

messages = [{"role": "user", "content": "查询销售表中 2026 年 Q2 的总收入"}]

# 第一轮：模型决定调用哪个工具
response = client.chat.completions.create(
    model="kimi-k3",
    messages=messages,
    tools=tools,
)

assistant_msg = response.choices[0].message
messages.append(assistant_msg)

# 执行工具并将结果回传
for tool_call in assistant_msg.tool_calls or []:
    args = json.loads(tool_call.function.arguments)
    # 实际业务中替换为真实查询逻辑
    result = {"total_revenue": "¥12,430,000", "period": "2026-Q2"}
    messages.append({
        "role": "tool",
        "tool_call_id": tool_call.id,
        "content": json.dumps(result, ensure_ascii=False),
    })

# 第二轮：获取最终回答
final = client.chat.completions.create(
    model="kimi-k3",
    messages=messages,
    tools=tools,
)
print(final.choices[0].message.content)

import base64
from pathlib import Path

def encode_image(path: str) -> str:
    return base64.b64encode(Path(path).read_bytes()).decode()

completion = client.chat.completions.create(
    model="kimi-k3",
    messages=[
        {
            "role": "user",
            "content": [
                {
                    "type": "image_url",
                    "image_url": {
                        "url": f"data:image/png;base64,{encode_image('diagram.png')}"
                    },
                },
                {"type": "text", "text": "分析这张架构图，指出潜在的单点故障。"},
            ],
        }
    ],
)

stream = client.chat.completions.create(
    model="kimi-k3",
    messages=[{"role": "user", "content": "写一份 2000 字的技术调研报告大纲"}],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta
    if delta.content:
        print(delta.content, end="", flush=True)