429 时，使用指数退避和抖动进行重试；如果重复重试仍然发生，则降低突发流量。
限制并发
下面的 Python 示例使用异步信号量来限制并发聊天请求：import asyncio
import os
from openai import AsyncOpenAI
client = AsyncOpenAI(
api_key=os.environ["COMETAPI_KEY"],
base_url="https://api.cometapi.com/v1",
)
semaphore = asyncio.Semaphore(5)
async def ask(prompt):
async with semaphore:
completion = await client.chat.completions.create(
model="your-model-id",
messages=[{"role": "user", "content": prompt}],
)
return completion.choices[0].message.content
async def main():
prompts = ["Say hello.", "Write a title.", "Return one JSON key."]
results = await asyncio.gather(*(ask(prompt) for prompt in prompts))
print(results)
asyncio.run(main())
[
"Hello.",
"A concise title",
"{\"key\":\"value\"}"
]
重试速率限制
下面的 JavaScript 示例会对429 响应进行带抖动的重试：
import OpenAI from "openai";
const client = new OpenAI({
apiKey: process.env.COMETAPI_KEY,
baseURL: "https://api.cometapi.com/v1",
});
async function sleep(milliseconds) {
return new Promise((resolve) => setTimeout(resolve, milliseconds));
}
async function createCompletion() {
for (let attempt = 0; attempt < 5; attempt += 1) {
try {
return await client.chat.completions.create({
model: "your-model-id",
messages: [{ role: "user", content: "Say hello." }],
});
} catch (error) {
if (error.status !== 429 || attempt === 4) {
throw error;
}
const delay = Math.min(30000, 1000 * 2 ** attempt);
await sleep(delay + Math.random() * 1000);
}
}
}
const completion = await createCompletion();
console.log(completion.choices[0].message.content);
{
"choices": [
{
"message": {
"content": "Hello."
}
}
]
}
常见错误
| 错误 | 解决方法 |
|---|---|
| 无限并行请求 | 添加信号量、队列或工作池。 |
| 重试所有失败 | 仅重试 429 和临时服务器故障。 |
| 没有按模型统计指标 | 为每个请求记录 route、model ID、状态和延迟。 |
| 重试风暴 | 添加抖动并限制最大重试延迟。 |