Prompt Engineering Instructor 使用示例

# 真实世界示例 使用 Instructor 进行结构化数据提取的实用示例。 ## 数据提取 python class CompanyInfo(BaseModel): name: str founded: int industry: str employees: int text = "Apple 成立于 1976 年，属于科技行业，拥有 164,000 名员工。" company = client.messages.create( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": f"提取：{text}"}], response_model=CompanyInfo ) ## 分类 python class Sentiment(str, Enum): POSITIVE = "positive" NEGATIVE = "negative" NEUTRAL = "neutral" class Review(BaseModel): sentiment: Sentiment confidence: float = Field(ge=0.0, le=1.0) review = client.messages.create( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": "这个产品太棒了！"}], response_model=Review ) ## 多实体提取 python class Person(BaseModel): name: str role: str class Entities(BaseModel): people: list[Person] organizations: list[str] locations: list[str] entities = client.messages.create( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": "苹果公司首席执行官 Tim Cook 在库比蒂诺发表了讲话..."}], response_model=Entities ) ## 结构化分析 python class Analysis(BaseModel): summary: str key_points: list[str] sentiment: Sentiment actionable_items: list[str] analysis = client.messages.create( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": "分析：[长文本]"}], response_model=Analysis ) ## 批处理 python texts = ["text1", "text2", "text3"] results = [ client.messages.create( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": text}], response_model=YourModel ) for text in texts ] ## 流式处理 python for partial in client.messages.create_partial( model="claude-sonnet-4-5-20250929", max_tokens=1024, messages=[{"role": "user", "content": "生成报告..."}], response_model=Report ): print(f"进度：{partial.title}") # 实时更新 UI