AI 应用人才和算法工程师差在哪？写一个提示词评估器就清楚了

"""提示词评估器：用固定测试集比较提示词模板的输出质量。"""
import json

import requests

API_URL = "https://你的模型服务地址/v1/chat/completions"
API_KEY = "你的API密钥"
MODEL = "你的模型名"


def chat(prompt: str) -> str:
    resp = requests.post(
        API_URL,
        headers={"Authorization": f"Bearer {API_KEY}"},
        json={"model": MODEL, "messages": [{"role": "user", "content": prompt}], "temperature": 0},
        timeout=60,
    )
    resp.raise_for_status()
    return resp.json()["choices"][0]["message"]["content"]


# 测试集：换成你自己的业务材料与验收要点
TEST_CASES = [
    {"input": "（粘贴一段真实业务材料，如产品说明原文）",
     "must_include": ["金额", "日期"],
     "max_len": 300},
]


def score(output: str, case: dict) -> dict:
    return {
        "要点齐全": all(k in output for k in case["must_include"]),
        "长度合格": len(output) <= case["max_len"],
    }


def evaluate(template: str) -> dict:
    results = [score(chat(template.format(material=c["input"])), c) for c in TEST_CASES]
    passed = sum(1 for r in results if all(r.values()))
    return {"用例数": len(results), "全部通过": passed,
            "一次通过率": f"{passed / len(results):.1%}"}


if __name__ == "__main__":
    tpl = "请从以下材料中提取金额与日期，200 字以内输出：\n{material}"
    print(json.dumps(evaluate(tpl), ensure_ascii=False, indent=2))