Chat Completions
对话生成接口,兼容 OpenAI Chat Completions API。
端点
POST /v1/chat/completions请求参数
| 参数 | 类型 | 必填 | 说明 |
|---|---|---|---|
model | string | ✅ | 模型名称 |
messages | array | ✅ | 消息列表 |
stream | boolean | ❌ | 是否启用流式响应(推荐开启) |
messages 格式
json
{
"messages": [
{
"role": "user",
"content": "生成一只可爱的猫"
}
]
}多模态请求(图生图)
json
{
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": "让这张图片更加鲜艳"
},
{
"type": "image_url",
"image_url": {
"url": "data:image/jpeg;base64,/9j/4AAQ..."
}
}
]
}
]
}图片限制
| 限制项 | 说明 |
|---|---|
| 支持格式 | PNG, JPEG, GIF, WebP |
| 数量限制 | 默认 5 张,最大 10 张 |
| 数据格式 | Base64 Data URL (data:image/jpeg;base64,...) |
| 自动转换 | 服务器会自动转换为 JPG 格式 |
非流式响应
请求示例
bash
curl -X POST http://localhost:3000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-your-key" \
-d '{
"model": "gemini-3-pro-image-preview",
"messages": [
{
"role": "user",
"content": "generate a cat"
}
]
}'响应示例
json
{
"id": "chatcmpl-1732374740123",
"object": "chat.completion",
"created": 1732374740,
"model": "gemini-3-pro-image-preview",
"choices": [
{
"index": 0,
"message": {
"role": "assistant",
"content": ""
},
"finish_reason": "stop"
}
]
}流式响应
推荐使用
流式模式包含心跳保活机制,可以避免长时间生成导致的连接超时。
请求示例
bash
curl -X POST http://localhost:3000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer sk-your-key" \
-d '{
"model": "gemini-3-pro-image-preview",
"stream": true,
"messages": [
{
"role": "user",
"content": "generate a cat"
}
]
}'响应示例
data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1732374740,"model":"gemini-3-pro-image-preview","choices":[{"index":0,"delta":{"role":"assistant","content":""},"finish_reason":null}]}
: keep-alive
: keep-alive
data: {"id":"chatcmpl-123","object":"chat.completion.chunk","created":1732374740,"model":"gemini-3-pro-image-preview","choices":[{"index":0,"delta":{"content":""},"finish_reason":"stop"}]}
data: [DONE]错误处理
队列已满 (429)
json
{
"error": {
"message": "队列已满",
"type": "rate_limit_exceeded",
"code": "QUEUE_FULL"
}
}解决方案
启用流式模式 (stream: true) 可以无限排队,避免 429 错误。
模型不支持 (400)
json
{
"error": {
"message": "没有 Worker 支持模型: invalid-model",
"type": "invalid_request_error",
"code": "MODEL_NOT_FOUND"
}
}