Ollama 启动后，默认在 http://localhost:11434 提供 API 服务。

1、对话接口

curl http://localhost:11434/api/chat -d '{
    "model":"qwen3:4b",
    "messages":[
        {"role":"user","content":"你好"}
    ]
}'

2、流式输出

curl http://localhost:11434/api/chat -d '{
    "model":"qwen3:4b",
    "messages":[
        {"role":"user","content":"你好"}
    ],
    "stream":true
}'

3、Python调用

import requests

response = requests.post(
    "http://localhost:11434/api/chat",
    json = {
        "model":"qwen3:4b",
        "messages":[
            {"role":"user", "content":"你好"}
        ],
        "stream":False
    }
)
print(response.json()["message"]["content"])

4、openai接口

from openai import OpenAI

client=OpenAI(
    base_url="http://localhost:11434/v1",
    api_key="ollama"
)
response = client.chat.completions.create(
    model="qwen2.5:0.5b",
    messages=[
        {"role":"system","content":"你是一个Python专家"},
        {"role":"user","content":"解释一下装饰器"}
    ]
)

print(response.choices[0].message.content)

或流式输出

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:11434/v1",
    api_key="ollama"
)

response = client.chat.completions.create(
    model = "qwen2.5:0.5b",
    messages = [
        {"role":"user","content":"你好"}
    ],
    stream = True
)

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