File size: 3,534 Bytes
61fd574
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
import os

from app.agents.base_agent import BaseAgent

MOCK_RESULT = {
    "must_have": [
        "Java", "SpringBoot", "MyBatis", "MySQL", "Redis",
        "分布式系统", "微服务架构", "本科及以上学历", "计算机相关专业",
    ],
    "nice_to_have": [
        "Kubernetes", "Docker", "消息队列", "Elasticsearch",
        "大流量高并发经验", "开源项目贡献",
    ],
    "buzzwords": [
        "具有较强的学习能力和团队协作精神",
        "工作积极主动,责任心强",
        "具备良好的沟通表达能力",
        "有较强的抗压能力",
    ],
    "expectation_signals": [
        {"tag": "加班文化", "evidence": "能承受较大工作压力", "confidence": "medium"},
        {"tag": "技术氛围", "evidence": "钻研前沿技术,追求卓越", "confidence": "high"},
        {"tag": "团队协作", "evidence": "具备良好的团队协作精神", "confidence": "medium"},
        {"tag": "结果导向", "evidence": "以结果为导向,注重交付质量", "confidence": "high"},
    ],
    "position_summary": "Java后端开发工程师,负责核心业务系统的设计与开发,要求具备扎实的Java基础和微服务实践经验",
}


class JdParser(BaseAgent):
    name = "jd_parser"

    def build_prompt(self, jd_text: str) -> tuple[str, dict]:
        system = (
            "你是一个JD解析专家。请对提供的JD文本执行以下四项任务:\n\n"
            "1. 去噪(Denoise):识别并过滤掉模板话术/通用套话(buzzwords),"
            "这些描述通常在任何岗位中都能出现,缺乏具体信息量。\n\n"
            "2. 分层(Layer):将JD中要求的能力/条件分为两类:\n"
            "   - must_have:明确要求的必备技能或条件\n"
            "   - nice_to_have:加分项、优先考虑项、期望但不强制\n\n"
            "3. 翻译(Translate):将模糊描述转换为具体的技术栈或技能标签。\n"
            "   例如:'熟悉主流框架' → SpringBoot, MyBatis\n"
            "   例如:'熟悉数据库技术' → MySQL, Redis\n"
            "   例如:'熟练掌握Linux操作' → Linux\n\n"
            "4. 期望信号提取(Expectation Signals):提取JD中隐含的关于加班文化、"
            "团队氛围、薪酬待遇、技术风格等信号。\n"
            "   每条信号必须包含:\n"
            "   - tag:信号类别(如'加班文化''技术氛围''结果导向''团队协作'等)\n"
            "   - evidence:原文引用(精确的JD原文段落)\n"
            "   - confidence:置信度(high/medium/low)\n\n"
            "请严格按以下JSON格式输出:\n"
            "{\n"
            '  "must_have": ["技能1", "技能2", ...],\n'
            '  "nice_to_have": ["技能1", "技能2", ...],\n'
            '  "buzzwords": ["套话1", "套话2", ...],\n'
            '  "expectation_signals": [\n'
            '    {"tag": "信号类别", "evidence": "原文引用", "confidence": "high/medium/low"}\n'
            "  ],\n"
            '  "position_summary": "一句话总结该岗位"\n'
            "}"
        )
        user = {"jd_text": jd_text}
        return system, user

    async def parse_jd(self, jd_text: str) -> dict:
        if not os.getenv("DEEPSEEK_API_KEY"):
            return dict(MOCK_RESULT)
        system, user_input = self.build_prompt(jd_text)
        result = await self.call_llm(system, user_input)
        return result