用 Python 写 AI 命令行工具：3个实用案例手把手教学

pip install click openai

import click
from openai import OpenAI

client = OpenAI()

@click.group()
def cli():
    """AI 命令行工具箱"""
    pass

@cli.command()
@click.argument("text")
@click.option("--model", default="gpt-4o", help="使用的模型")
def summarize(text, model):
    """快速总结文本"""
    response = client.chat.completions.create(
        model=model,
        messages=[
            {"role": "system", "content": "你是一个简洁的文本总结助手。请用不超过100字总结以下内容。"},
            {"role": "user", "content": text}
        ]
    )
    click.echo(response.choices[0].message.content)

if __name__ == "__main__":
    cli()

$ python ai-cli.py summarize "人工智能正在改变软件开发的方式..."

@cli.command()
@click.argument("file_path", type=click.Path(exists=True))
@click.option("--model", default="claude-sonnet-4-20250514")
@click.option("--strict/--loose", default=False, help="严格模式：只报错误")
def review(file_path, model, strict):
    """审查 Python 文件的代码质量"""
    with open(file_path, "r") as f:
        code = f.read()

    level = "仅报告严重问题" if strict else "报告所有问题和建议"

    response = client.chat.completions.create(
        model=model,
        messages=[
            {
                "role": "system",
                "content": f"你是一个资深Python工程师。请审查以下代码，{level}。用中文回答，每个问题标注行号。"
            },
            {"role": "user", "content": code}
        ],
        temperature=0.3
    )

    click.secho(f"\n📋 代码审查报告 ({model})\n{'─'*40}", fg="cyan")
    click.echo(response.choices[0].message.content)

# 宽松模式：所有建议
python ai-cli.py review main.py

# 严格模式：只看错误
python ai-cli.py review --strict main.py

@cli.command()
@click.argument("log_file", type=click.Path(exists=True))
@click.option("--error-only", is_flag=True, help="只显示错误相关分析")
def analyze(log_file, error_only):
    """分析日志文件，提取关键信息"""
    # 只读最后 5000 行，避免内存爆炸
    with open(log_file, "r") as f:
        lines = f.readlines()[-5000:]
    content = "".join(lines)

    prompt = "请分析以下日志文件：\n1. 统计错误数量\n2. 找出最常见的错误类型\n3. 推测可能的原因\n4. 给出修复建议"
    if error_only:
        prompt = "只分析日志中的错误条目，忽略正常日志。"

    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[
            {"role": "system", "content": "你是一个运维专家。请分析日志并提供可操作的修复建议。"},
            {"role": "user", "content": f"{prompt}\n\n日志内容：\n{content}"}
        ]
    )

    click.echo(response.choices[0].message.content)

# 完整分析
python ai-cli.py analyze /var/log/app.log

# 只看错误
python ai-cli.py analyze --error-only /var/log/app.log

@cli.command()
@click.argument("query", nargs=-1, required=True)
@click.option("--model", default="gpt-4o")
@click.option("--max-length", default=300, help="回答最大长度（字符）")
def search(*query, model, max_length):
    """用 AI 搜索知识"""
    question = " ".join(query)

    response = client.chat.completions.create(
        model=model,
        messages=[
            {
                "role": "system",
                "content": f"你是一个知识问答助手。回答要简洁实用，控制在{max_length}字以内。不要编造你不知道的信息。"
            },
            {"role": "user", "content": question}
        ],
        max_tokens=max_length
    )

    result = response.choices[0].message.content
    click.secho(f"🔍 搜索: {question}\n", fg="green")
    click.echo(result)

python ai-cli.py search "Python 中 list 和 tuple 的区别"
python ai-cli.py search "Docker volume 和 bind mount 哪个更好"

click.secho("错误:", fg="red", bold=True)
click.secho("警告:", fg="yellow")
click.secho("成功:", fg="green")

import hashlib
import json
from pathlib import Path

CACHE_DIR = Path.home() / ".ai-cli-cache"

def cached_call(prompt, model):
    cache_key = hashlib.md5(f"{prompt}:{model}".encode()).hexdigest()
    cache_file = CACHE_DIR / f"{cache_key}.json"

    if cache_file.exists():
        with open(cache_file) as f:
            return json.load(f)["result"]

    response = client.chat.completions.create(
        model=model,
        messages=[{"role": "user", "content": prompt}]
    )

    result = response.choices[0].message.content
    CACHE_DIR.mkdir(exist_ok=True)
    with open(cache_file, "w") as f:
        json.dump({"prompt": prompt, "result": result}, f)

    return result

pip install -e .

setup(
    name="ai-cli",
    entry_points={
        "console_scripts": [
            "ai=ai_cli:cli",
        ],
    },
)

from litellm import completion  # 用 LiteLLM 统一接口

response = completion(
    model="anthropic/claude-sonnet-4-20250514",
    messages=[{"role": "user", "content": text}]
)