Skip to content

generate.md

Latest commit

 

History

History
87 lines (66 loc) · 1.87 KB

generate.md

File metadata and controls

87 lines (66 loc) · 1.87 KB

POST /api/generate

文本生成接口，支持单次提示词生成和流式输出。

请求

POST /api/generate
Content-Type: application/json

请求体

{
  "model": "Qwen/Qwen3-0.6B-GGUF",
  "prompt": "Write a haiku about programming",
  "stream": true,
  "options": {
    "temperature": 0.7,
    "max_tokens": 256
  }
}
字段 类型 必填 说明
model string 是 模型名称
prompt string 是 提示词
stream bool 否 是否流式输出（默认 true）
options object 否 生成参数

响应

流式响应（默认）

data: {"model":"Qwen/Qwen3-0.6B-GGUF","response":"Code","done":false,"created_at":"2026-03-11T00:43:32.205Z"}

data: {"model":"Qwen/Qwen3-0.6B-GGUF","response":" is","done":false,"created_at":"2026-03-11T00:43:32.212Z"}

data: {"model":"Qwen/Qwen3-0.6B-GGUF","response":"","done":true,"created_at":"2026-03-11T00:43:32.343Z"}

非流式响应

{
  "model": "Qwen/Qwen3-0.6B-GGUF",
  "response": "Code is written,\nCollaborating on the code.\nAnd solved the challenge.",
  "done": true,
  "created_at": "2026-03-11T00:43:32.343Z"
}

与 /api/chat 的区别

/api/generate /api/chat
输入 单个 prompt 字符串 messages 数组（多轮对话）
响应字段 response message
适用场景 单次生成、补全 对话、多轮交互

示例

curl

curl http://localhost:11435/api/generate -d '{
  "model": "Qwen/Qwen3-0.6B-GGUF",
  "prompt": "Explain quantum computing in one sentence",
  "stream": false
}'

Python

import requests

resp = requests.post("http://localhost:11435/api/generate", json={
    "model": "Qwen/Qwen3-0.6B-GGUF",
    "prompt": "Write a Python hello world",
    "stream": False
})
print(resp.json()["response"])