SDK 快速开始

1. 安装

先确认本地运行时版本：TypeScript SDK 需要 Node.js 18.17+；Python SDK 需要 Python 3.10+，安装后用 seacloud_sdk 导入。

TypeScript

pnpm add @seacloudai/sdk

Python

pip install seacloudai-sdk

2. 初始化客户端

API Key 必须由调用方显式传入。你可以从环境变量读取后再传给 SDK，但 SDK 本身不会自动读取环境变量。

TypeScript

import { SeaCloud, getSeaCloudDocs } from "@seacloudai/sdk";

const docs = getSeaCloudDocs({ locale: "zh-CN" });

const client = new SeaCloud({
  apiKey: process.env.SEACLOUD_API_KEY!,
  timeout: 600_000,
});

console.table(docs.methods);

Python

import os

from seacloud_sdk import SeaCloud, getSeaCloudDocs

docs = getSeaCloudDocs({"locale": "zh-CN"})

client = SeaCloud(
    api_key=os.environ["SEACLOUD_API_KEY"],
    timeout=600_000,
)

print(docs["methods"])

3. 发送对话请求

TypeScript

const text = await client.chat.send(
  "gpt-4.1",
  [{ role: "user", content: "用一句话介绍 SeaCloud SDK" }],
  { temperature: 0.2, maxTokens: 128 },
);

console.log(text);

Python

text = await client.chat.send(
    "gpt-4.1",
    [{"role": "user", "content": "用一句话介绍 SeaCloud SDK"}],
    {"temperature": 0.2, "maxTokens": 128},
)

print(text)

4. 创建生成任务并等待结果

run 只创建任务并立即返回任务句柄；runSync / run_sync 会创建任务、轮询状态并读取最终 response。

TypeScript

const result = await client.runSync("wan25_t2i_preview", {
  prompt: "a cinematic photo of a cat astronaut",
  n: 1,
  prompt_extend: true,
  size: "1024*1024",
  watermark: false,
});

console.log(result.output?.urls);

Python

result = await client.run_sync("wan25_t2i_preview", {
    "prompt": "a cinematic photo of a cat astronaut",
    "n": 1,
    "prompt_extend": True,
    "size": "1024*1024",
    "watermark": False,
})

print(result.get("output", {}).get("urls"))

5. 下一步

• 需要理解 dryRun、任务轮询、模型规格和错误处理：继续看操作手册。