开发者实践：面向工程环境的提示词设计

def create_analysis_prompt(context: str, question: str) -> str:
    return f"""
    基于以下上下文信息回答问题：

    上下文：{context}

    问题：{question}

    请提供详细、准确的回答。如果信息不足，请明确说明。
    """

# prompt_config.yaml
code_review:
  role: "资深软件工程师"
  constraints:
    - "重点关注安全性"
    - "检查代码规范"
    - "评估性能影响"
  output_format: "markdown"

class PromptTestSuite:
    def test_basic_functionality(self):
        prompt = generate_prompt("简单问题")
        response = llm.generate(prompt)
        assert len(response) > 0
        assert "答案" in response

    def test_edge_cases(self):
        prompt = generate_prompt("边界情况输入")
        response = llm.generate(prompt)
        assert "无法回答" in response or "需要更多信息" in response

def ab_test_prompts(prompt_a: str, prompt_b: str, test_cases: List[str]):
    results = []
    for test_case in test_cases:
        response_a = evaluate_prompt(prompt_a, test_case)
        response_b = evaluate_prompt(prompt_b, test_case)
        results.append(compare_responses(response_a, response_b))
    return analyze_results(results)

{
  "timestamp": "2025-01-18T10:00:00Z",
  "prompt_version": "v2.1.0",
  "input_hash": "abc123",
  "output_metrics": {
    "tokens_used": 150,
    "response_time_ms": 1200,
    "quality_score": 0.85
  },
  "user_feedback": "满意"
}