Compare commits
2 Commits
| Author | SHA1 | Date | |
|---|---|---|---|
 zk
|
e0cb4493e3 | ||
|
zk
|
b9d8c2724a |
+2
-2
@@ -1,5 +1,5 @@
|
||||
*/target
|
||||
**/.idea
|
||||
.idea/
|
||||
./log
|
||||
./logs
|
||||
logs/
|
||||
log/
|
||||
@@ -1,202 +0,0 @@
|
||||
---
|
||||
inclusion: manual
|
||||
---
|
||||
|
||||
# 数据清洗方案
|
||||
|
||||
## 总体架构
|
||||
|
||||
```
|
||||
爬虫(公司网络) → app_job_data(原始数据)
|
||||
↓
|
||||
Java定时任务读取（多线程）
|
||||
↓
|
||||
三次AI调用：结构化 → 专业匹配 → 技能提取
|
||||
↓
|
||||
写入业务表(bg_company + bg_job + 关联表 + bg_skill_tag)
|
||||
↓
|
||||
公司信息不完整的 → 调AI补充
|
||||
```
|
||||
|
||||
所有清洗逻辑放在 `manager` 模块，通过 `@Scheduled` 定时任务触发。
|
||||
|
||||
---
|
||||
|
||||
## 一、岗位清洗触发逻辑
|
||||
|
||||
### 1.1 状态管理
|
||||
|
||||
`app_job_data.clean_status`：
|
||||
- 0=待清洗：新爬到的数据，默认值
|
||||
- 1=清洗中：定时任务已锁定，正在处理
|
||||
- 2=已入库：清洗成功，已写入 bg_job
|
||||
- 3=已丢弃：AI判定为无效数据
|
||||
|
||||
### 1.2 两个定时任务
|
||||
|
||||
#### 任务A：岗位清洗（每5分钟）
|
||||
1. 事务内 `SELECT FOR UPDATE` + `UPDATE clean_status=1` 批量锁定
|
||||
2. 丢入线程池，多线程并发处理
|
||||
3. 每条独立更新状态
|
||||
|
||||
#### 任务B：僵尸恢复（每30分钟）
|
||||
将超时10分钟仍在清洗中的数据重置为待清洗。
|
||||
|
||||
---
|
||||
|
||||
## 二、岗位清洗逻辑（三次AI调用）
|
||||
|
||||
### 2.1 前置校验
|
||||
- `description` 为空或长度 < 20 → 标记丢弃，跳过
|
||||
|
||||
### 2.2 参考数据缓存（DictCacheService）
|
||||
|
||||
启动时加载：
|
||||
- `bg_job_category` 叶子节点（level=3），格式 `id:name(一级/二级)`
|
||||
- `bg_industry` 叶子节点（level=2），格式 `id:name(一级)`
|
||||
- `bg_major_category` 叶子节点（level=3），格式 `id:name(一级/二级)`
|
||||
- `bg_china_regions_code` 省市级，供城市名匹配
|
||||
|
||||
### 2.3 第一次AI：提取岗位结构化信息
|
||||
|
||||
#### 输入
|
||||
- 原始字段：job_title、salary、location、company、experience、education、description
|
||||
- 参考列表：岗位分类、行业
|
||||
|
||||
#### 返回JSON
|
||||
```json
|
||||
{
|
||||
"valid": true,
|
||||
"title": "Java高级开发工程师",
|
||||
"salary": "15-25K",
|
||||
"education": 2,
|
||||
"minExperience": 3,
|
||||
"employmentType": 0,
|
||||
"categoryId": 12,
|
||||
"requiredIndustryId": 5,
|
||||
"description": "岗位职责...",
|
||||
"requirement": "任职要求...",
|
||||
"bonus": "加分项...",
|
||||
"tags": ["数据分析", "产品策略"],
|
||||
"skillTags": ["Java", "Spring Boot"],
|
||||
"companyShortName": "字节跳动",
|
||||
"cities": ["北京", "上海"]
|
||||
}
|
||||
```
|
||||
|
||||
#### Java处理流程
|
||||
1. valid=false → 丢弃
|
||||
2. source_id 去重检查
|
||||
3. 公司处理（查或创建 bg_company）
|
||||
4. 地区匹配（cities → region_code）
|
||||
5. 写入 bg_job + bg_job_region_relation + 更新 clean_status=2
|
||||
|
||||
### 2.4 第二次AI：专业匹配
|
||||
|
||||
岗位入库后，单独调用AI匹配专业要求。
|
||||
|
||||
#### 输入
|
||||
- 岗位标题、职责、要求
|
||||
- 三级专业分类列表（845条，格式 `id:name(一级/二级)`）
|
||||
|
||||
#### 返回JSON
|
||||
```json
|
||||
{
|
||||
"requiredMajorIds": [101, 203],
|
||||
"majorSensitivity": 1
|
||||
}
|
||||
```
|
||||
|
||||
#### 规则
|
||||
- requiredMajorIds：从专业列表中选最相关的，最多3个，无明确要求则空数组
|
||||
- majorSensitivity：0=不限 1=优先 2=强制
|
||||
- majorSensitivity=0 时，requiredMajorIds 应为空数组
|
||||
- 结果更新到 bg_job 的 required_major_ids 和 major_sensitivity 字段
|
||||
|
||||
### 2.5 第三次AI：技能提取
|
||||
|
||||
自由提取岗位核心技能，不再依赖预定义标签库。
|
||||
|
||||
#### 输入
|
||||
- 岗位标题、职责、要求
|
||||
|
||||
#### 返回JSON
|
||||
```json
|
||||
["java", "spring boot", "mysql", "redis"]
|
||||
```
|
||||
|
||||
#### prompt 规则
|
||||
- 统一小写字母
|
||||
- 尽量简短，使用业界通用缩写
|
||||
- 提取范围：技术栈、专业领域知识、行业工具、专业资质能力
|
||||
- 不提取纯软技能（沟通、协作、学习能力等）
|
||||
- 无专业能力要求的岗位（销售、行政等）返回空数组
|
||||
- 最多15个
|
||||
|
||||
#### 技能入库流程
|
||||
1. 遍历AI返回的技能名，统一转小写
|
||||
2. `INSERT IGNORE INTO bg_skill_tag`（依靠 name 唯一索引去重，ID 由 IdWorker 生成）
|
||||
3. `SELECT id FROM bg_skill_tag WHERE name=?` 拿到ID
|
||||
4. 写入 bg_job_skill_tag_relation
|
||||
|
||||
#### 并发安全
|
||||
依靠数据库唯一索引保证，不加应用层锁。多线程同时插入相同技能名时，INSERT IGNORE 自动忽略重复。
|
||||
|
||||
### 2.6 容错设计
|
||||
- 第二次、第三次AI调用失败不影响岗位入库
|
||||
- 每次调用独立 try-catch，仅日志记录
|
||||
|
||||
---
|
||||
|
||||
## 三、公司数据触发逻辑
|
||||
|
||||
### 3.1 状态
|
||||
`bg_company.status`：0=待完善 1=已完善 2=禁用 3=补充中 4=补充失败
|
||||
|
||||
### 3.2 定时任务
|
||||
- 任务C：公司补充（每小时），SELECT FOR UPDATE 锁定 status=0 的数据
|
||||
- 任务D：僵尸恢复（每小时，与C错开），重置超时10分钟的 status=3 数据
|
||||
|
||||
---
|
||||
|
||||
## 四、公司数据补充逻辑（AI补充）
|
||||
|
||||
### 4.1 流程
|
||||
1. 拿 short_name 调AI
|
||||
2. prompt 附带行业列表，AI直接返回 industryId
|
||||
3. valid=false → status=4（补充失败）
|
||||
4. regionCode：AI返回城市名，Java侧匹配
|
||||
5. 回填 bg_company 各字段，status=1
|
||||
|
||||
### 4.2 AI返回JSON
|
||||
```json
|
||||
{
|
||||
"valid": true,
|
||||
"name": "北京字节跳动科技有限公司",
|
||||
"city": "北京",
|
||||
"companyType": "独角兽",
|
||||
"industryId": 5,
|
||||
"tags": ["短视频", "人工智能"],
|
||||
"summary": "全球领先的内容平台和科技公司",
|
||||
"description": "字节跳动成立于2012年...",
|
||||
"foundedYear": "2012",
|
||||
"address": "北京市海淀区...",
|
||||
"scale": "10000人以上",
|
||||
"website": "https://www.bytedance.com",
|
||||
"financingStage": "已上市",
|
||||
"latestValuation": "2200亿美元",
|
||||
"news": ["新闻1", "新闻2", "新闻3"]
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## 五、关键设计决策
|
||||
|
||||
| 决策点 | 结论 | 原因 |
|
||||
|--------|------|------|
|
||||
| 技能标签来源 | AI自由提取，自动入库 bg_skill_tag | 去掉预定义标签限制，覆盖面更广 |
|
||||
| 技能并发去重 | INSERT IGNORE + 唯一索引 | 不加应用层锁，性能好 |
|
||||
| 专业匹配 | 单独一次AI调用 | 专业列表845条，和第一次prompt合并会超token |
|
||||
| AI调用次数 | 三次（结构化 + 专业 + 技能） | 各维度独立，容错互不影响 |
|
||||
| 公司数据来源 | AI补充，不用工商API | 公司简称查API不精确，AI覆盖率高 |
|
||||
@@ -97,22 +97,7 @@ app:
|
||||
# base-url 配到版本路径，如 DeepSeek: https://api.deepseek.com/v1，豆包: https://ark.cn-beijing.volces.com/api/v3
|
||||
ai:
|
||||
providers:
|
||||
job-clean:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-seed-2-0-lite-260215}
|
||||
job-recommend:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
|
||||
|
||||
# 岗位清洗配置
|
||||
job-clean:
|
||||
batch-size: 60
|
||||
thread-pool-size: 60
|
||||
|
||||
# 公司数据补充配置
|
||||
company-clean:
|
||||
batch-size: 10
|
||||
thread-pool-size: 3
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
@@ -97,22 +97,7 @@ app:
|
||||
# base-url 配到版本路径，如 DeepSeek: https://api.deepseek.com/v1，豆包: https://ark.cn-beijing.volces.com/api/v3
|
||||
ai:
|
||||
providers:
|
||||
job-clean:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-seed-2-0-lite-260215}
|
||||
job-recommend:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
|
||||
|
||||
# 岗位清洗配置
|
||||
job-clean:
|
||||
batch-size: 60
|
||||
thread-pool-size: 60
|
||||
|
||||
# 公司数据补充配置
|
||||
company-clean:
|
||||
batch-size: 10
|
||||
thread-pool-size: 3
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
@@ -97,22 +97,7 @@ app:
|
||||
# base-url 配到版本路径，如 DeepSeek: https://api.deepseek.com/v1，豆包: https://ark.cn-beijing.volces.com/api/v3
|
||||
ai:
|
||||
providers:
|
||||
job-clean:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-seed-2-0-lite-260215}
|
||||
job-recommend:
|
||||
base-url: ${AI_BASE_URL:https://ark.cn-beijing.volces.com/api/v3}
|
||||
api-key: ${AI_API_KEY:fd065993-bee2-4f31-8bf2-56d5d3012c02}
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
|
||||
|
||||
# 岗位清洗配置
|
||||
job-clean:
|
||||
batch-size: 60
|
||||
thread-pool-size: 60
|
||||
|
||||
# 公司数据补充配置
|
||||
company-clean:
|
||||
batch-size: 10
|
||||
thread-pool-size: 3
|
||||
model: ${AI_MODEL:doubao-1-5-lite-32k-250115}
|
||||
@@ -1,24 +0,0 @@
|
||||
package org.jiayunet.mapper;
|
||||
|
||||
import org.apache.ibatis.annotations.Mapper;
|
||||
import org.apache.ibatis.annotations.Param;
|
||||
import org.apache.ibatis.annotations.Select;
|
||||
import org.jiayunet.pojo.po.AppJobData;
|
||||
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* 爬虫岗位原始数据Mapper
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Mapper
|
||||
public interface AppJobDataMapper extends CommonMapper<AppJobData> {
|
||||
|
||||
/**
|
||||
* 查询待清洗数据并加行锁（SELECT ... FOR UPDATE）
|
||||
* <p>必须在事务内调用，配合状态更新实现原子锁定</p>
|
||||
*/
|
||||
@Select("SELECT * FROM app_job_data WHERE clean_status = 0 AND is_valid = 1 LIMIT #{limit} FOR UPDATE")
|
||||
List<AppJobData> selectForUpdate(@Param("limit") int limit);
|
||||
}
|
||||
@@ -1,79 +0,0 @@
|
||||
package org.jiayunet.pojo.po;
|
||||
|
||||
import com.baomidou.mybatisplus.annotation.IdType;
|
||||
import com.baomidou.mybatisplus.annotation.TableId;
|
||||
import com.baomidou.mybatisplus.annotation.TableName;
|
||||
import lombok.Data;
|
||||
|
||||
import java.time.Instant;
|
||||
|
||||
/**
|
||||
* 爬虫岗位原始数据表（app_job_data）
|
||||
* <p>存储爬虫抓取的原始岗位数据，供清洗服务读取并写入业务表</p>
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Data
|
||||
@TableName(value = "app_job_data")
|
||||
public class AppJobData {
|
||||
|
||||
@TableId(type = IdType.AUTO)
|
||||
private Long id;
|
||||
|
||||
/** 关联爬取任务ID */
|
||||
private Long taskCrawlId;
|
||||
|
||||
/** 职位名称 */
|
||||
private String jobTitle;
|
||||
|
||||
/** 薪资 */
|
||||
private String salary;
|
||||
|
||||
/** 工作地点 */
|
||||
private String location;
|
||||
|
||||
/** 公司名称 */
|
||||
private String company;
|
||||
|
||||
/** 经验要求 */
|
||||
private String experience;
|
||||
|
||||
/** 学历要求 */
|
||||
private String education;
|
||||
|
||||
/** 岗位详情(职责+要求+介绍) */
|
||||
private String description;
|
||||
|
||||
/** 详情页URL */
|
||||
private String detailUrl;
|
||||
|
||||
/** 内容哈希值，用于查重 */
|
||||
private String contentHash;
|
||||
|
||||
/** 数据来源 0=官网 1=平台 */
|
||||
private Integer sources;
|
||||
|
||||
/** 是否独立URL 0=页内展示 1=独立页面 */
|
||||
private Integer isIndependentUrl;
|
||||
|
||||
/** 是否有效 0=无效 1=有效 */
|
||||
private Integer isValid;
|
||||
|
||||
/** 有效期 */
|
||||
private Instant expireAt;
|
||||
|
||||
/** 验证状态 pending=待验证 checking=验证中 checked=已验证 */
|
||||
private String checkStatus;
|
||||
|
||||
/** 清洗状态 0=待清洗 1=清洗中 2=已入库 3=已丢弃 */
|
||||
private Integer cleanStatus;
|
||||
|
||||
/** 上次验证时间 */
|
||||
private Instant lastCheckAt;
|
||||
|
||||
/** 创建时间 */
|
||||
private Instant createdAt;
|
||||
|
||||
/** 更新时间 */
|
||||
private Instant updatedAt;
|
||||
}
|
||||
@@ -77,6 +77,12 @@ public class Job {
|
||||
/** 爬虫原始数据ID，用于去重 */
|
||||
private String sourceId;
|
||||
|
||||
/** 招聘分类 0=校招 1=实习 2=社招 3=其他 */
|
||||
private Integer recruitCategory;
|
||||
|
||||
/** 发布日期 */
|
||||
private Instant expireAt;
|
||||
|
||||
/** 状态 0=上架 1=下架 2=已失效 */
|
||||
private Integer status;
|
||||
|
||||
|
||||
@@ -1,184 +0,0 @@
|
||||
package org.jiayunet.service;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import org.jiayunet.ai.AiChatAbility;
|
||||
import org.jiayunet.mapper.CompanyMapper;
|
||||
import org.jiayunet.pojo.po.Company;
|
||||
import org.jiayunet.tool.HttpTool;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.beans.factory.annotation.Value;
|
||||
import org.springframework.scheduling.annotation.Scheduled;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import java.time.Instant;
|
||||
import java.util.List;
|
||||
import java.util.concurrent.ExecutorService;
|
||||
import java.util.concurrent.Executors;
|
||||
|
||||
/**
|
||||
* 公司数据补充服务
|
||||
* <p>定时从 bg_company 捞取待完善数据，调用 AI 补充公司信息</p>
|
||||
* <p>依赖：AiChatAbility（AI调用）、DictCacheService（行业列表/地区匹配）、CompanyCleanTransactionService（事务操作）</p>
|
||||
* <p>使用表：bg_company（读取/更新）</p>
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Slf4j
|
||||
@Service
|
||||
public class CompanyCleanService {
|
||||
|
||||
@Autowired
|
||||
private AiChatAbility aiChatAbility;
|
||||
|
||||
@Autowired
|
||||
private DictCacheService dictCacheService;
|
||||
|
||||
@Autowired
|
||||
private CompanyCleanTransactionService companyCleanTransactionService;
|
||||
|
||||
@Autowired
|
||||
private CompanyMapper companyMapper;
|
||||
|
||||
@Value("${app.company-clean.batch-size:10}")
|
||||
private int batchSize;
|
||||
|
||||
@Value("${app.company-clean.thread-pool-size:3}")
|
||||
private int threadPoolSize;
|
||||
|
||||
private ExecutorService executorService;
|
||||
|
||||
/** 初始化线程池 */
|
||||
@javax.annotation.PostConstruct
|
||||
public void init() {
|
||||
executorService = Executors.newFixedThreadPool(threadPoolSize);
|
||||
}
|
||||
|
||||
/**
|
||||
* 任务C：公司数据补充（每小时）
|
||||
* <p>1. 批量锁定待完善公司 2. 多线程并发调用AI补充 3. 回填数据</p>
|
||||
*/
|
||||
@Scheduled(cron = "0 */1 * * * ?")
|
||||
public void cleanCompany() {
|
||||
List<Company> companyList = companyCleanTransactionService.lockBatch(batchSize);
|
||||
if (companyList.isEmpty()) {
|
||||
return;
|
||||
}
|
||||
log.info("公司补充：锁定{}条数据", companyList.size());
|
||||
|
||||
// 多线程并发处理
|
||||
for (Company company : companyList) {
|
||||
executorService.submit(() -> {
|
||||
try {
|
||||
cleanOne(company);
|
||||
} catch (Exception e) {
|
||||
log.error("公司补充异常, id={}, shortName={}", company.getId(), company.getShortName(), e);
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 任务D：公司僵尸恢复（每小时，与任务C错开30分钟）
|
||||
* <p>将超时10分钟仍在补充中的数据重置为待完善</p>
|
||||
*/
|
||||
@Scheduled(cron = "0 30 */1 * * ?")
|
||||
public void recoverZombie() {
|
||||
int recovered = companyMapper.update(null,
|
||||
new LambdaUpdateWrapper<Company>()
|
||||
.set(Company::getStatus, 0)
|
||||
.eq(Company::getStatus, 3)
|
||||
.lt(Company::getUpdateTime, Instant.now().minusSeconds(600)));
|
||||
|
||||
if (recovered > 0) {
|
||||
log.info("公司僵尸恢复：重置{}条数据", recovered);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 补充单条公司数据
|
||||
* <p>1. 拼prompt调AI 2. 解析结果 3. 回填数据</p>
|
||||
*/
|
||||
public void cleanOne(Company company) {
|
||||
log.info("公司补充开始, id={}, shortName={}", company.getId(), company.getShortName());
|
||||
|
||||
String systemPrompt = buildSystemPrompt();
|
||||
String userMessage = buildUserMessage(company.getShortName());
|
||||
|
||||
String aiResponse = aiChatAbility.chat(systemPrompt, userMessage);
|
||||
|
||||
try {
|
||||
// 去掉可能的 markdown 代码块标记
|
||||
String json = aiResponse.trim();
|
||||
if (json.startsWith("```")) {
|
||||
json = json.replaceAll("^```\\w*\\n?", "").replaceAll("\\n?```$", "").trim();
|
||||
}
|
||||
// 清除控制字符（Tab等），保留换行符
|
||||
json = json.replaceAll("[\\x00-\\x08\\x0B\\x0C\\x0E-\\x1F]", "");
|
||||
|
||||
JsonNode root = HttpTool.objectMapper.readTree(json);
|
||||
|
||||
// valid 校验：AI不认识该公司
|
||||
if (!root.path("valid").asBoolean(false)) {
|
||||
log.info("公司补充：AI不认识该公司, id={}, shortName={}", company.getId(), company.getShortName());
|
||||
companyCleanTransactionService.updateCompanyStatus(company.getId(), 4);
|
||||
return;
|
||||
}
|
||||
|
||||
// 回填数据
|
||||
companyCleanTransactionService.saveCompanyData(root, company);
|
||||
log.info("公司补充完成, id={}, shortName={}", company.getId(), company.getShortName());
|
||||
|
||||
} catch (Exception e) {
|
||||
log.error("公司AI返回解析失败, id={}, shortName={}, response={}",
|
||||
company.getId(), company.getShortName(), aiResponse, e);
|
||||
// 保持 status=3，由僵尸恢复任务重置
|
||||
}
|
||||
}
|
||||
|
||||
/** 构建系统提示词 */
|
||||
private String buildSystemPrompt() {
|
||||
return """
|
||||
你是一个企业信息补充助手。根据提供的公司简称，补充该公司的详细信息。
|
||||
|
||||
返回JSON格式要求：
|
||||
{
|
||||
"valid": true/false,
|
||||
"name": "公司全称",
|
||||
"city": "总部所在城市，精确到市",
|
||||
"companyType": "企业类型",
|
||||
"industryId": 行业ID,
|
||||
"tags": ["公司标签，最多5个"],
|
||||
"summary": "一句话简介，100字以内",
|
||||
"description": "公司详细描述，500字以内",
|
||||
"foundedYear": "成立年份",
|
||||
"address": "总部/注册地址",
|
||||
"scale": "企业规模",
|
||||
"website": "官网地址",
|
||||
"financingStage": "融资状态",
|
||||
"latestValuation": "最新估值",
|
||||
"news": ["相关新闻，最多3条，每条50字以内"]
|
||||
}
|
||||
|
||||
规则：
|
||||
1. 如果不认识该公司，返回 {"valid": false}
|
||||
2. name 根据公司简称推断完整的企业注册名称
|
||||
3. companyType 取值：上市企业、独角兽、国企、央企、民营企业、外资企业、合资企业、事业单位、其他
|
||||
4. industryId 必须从给定行业列表中选择，不确定则null
|
||||
5. scale 取值：少于50人、50-150人、150-500人、500-1000人、1000-5000人、5000-10000人、10000人以上
|
||||
6. tags 体现公司核心业务特征，最多5个
|
||||
7. news 基于你的知识提供该公司最新的3条相关新闻，每条50字以内
|
||||
8. latestValuation 知道就提供，不知道则null
|
||||
9. 不确定的字段返回null，不要编造
|
||||
10. 字符串值中不允许出现Tab、换行等控制字符
|
||||
11. 只返回JSON，不要其他内容
|
||||
""";
|
||||
}
|
||||
|
||||
/** 构建用户消息 */
|
||||
private String buildUserMessage(String shortName) {
|
||||
return "【公司简称】\n" + shortName +
|
||||
"\n\n【行业列表】\n" + dictCacheService.getIndustryText();
|
||||
}
|
||||
}
|
||||
@@ -1,165 +0,0 @@
|
||||
package org.jiayunet.service;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import org.jiayunet.mapper.CompanyMapper;
|
||||
import org.jiayunet.pojo.po.Company;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.springframework.transaction.annotation.Transactional;
|
||||
|
||||
import java.time.Instant;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* 公司数据补充事务服务
|
||||
* <p>独立出来解决 @Transactional 同类自调用失效问题</p>
|
||||
* <p>依赖：CompanyMapper、DictCacheService（地区匹配、行业校验）</p>
|
||||
* <p>使用表：bg_company（更新）</p>
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Slf4j
|
||||
@Service
|
||||
public class CompanyCleanTransactionService {
|
||||
|
||||
@Autowired
|
||||
private CompanyMapper companyMapper;
|
||||
|
||||
@Autowired
|
||||
private DictCacheService dictCacheService;
|
||||
|
||||
/**
|
||||
* 回填公司数据（事务）
|
||||
* <p>1. 解析AI返回的各字段 2. 地区匹配 3. 行业校验 4. 更新bg_company</p>
|
||||
*/
|
||||
@Transactional(rollbackFor = Exception.class)
|
||||
public void saveCompanyData(JsonNode root, Company company) {
|
||||
// name：AI根据shortName推断的全称
|
||||
String name = root.path("name").asText(null);
|
||||
if (name != null && !name.isBlank()) {
|
||||
company.setName(name);
|
||||
}
|
||||
|
||||
// regionCode：AI返回城市名，Java侧匹配
|
||||
String city = root.path("city").asText(null);
|
||||
if (city != null && !city.isBlank()) {
|
||||
String regionCode = dictCacheService.matchRegionCode(city);
|
||||
company.setRegionCode(regionCode);
|
||||
}
|
||||
|
||||
// companyType
|
||||
String companyType = root.path("companyType").asText(null);
|
||||
if (companyType != null && !"null".equals(companyType)) {
|
||||
company.setCompanyType(companyType);
|
||||
}
|
||||
|
||||
// industryId：校验是否存在于行业列表中
|
||||
long industryId = root.path("industryId").asLong(0);
|
||||
company.setIndustryId(industryId > 0 ? industryId : null);
|
||||
|
||||
// tags：JSON数组
|
||||
JsonNode tagsNode = root.path("tags");
|
||||
if (tagsNode.isArray() && !tagsNode.isEmpty()) {
|
||||
List<String> tags = new ArrayList<>();
|
||||
tagsNode.forEach(node -> tags.add(node.asText()));
|
||||
company.setTags(tags);
|
||||
}
|
||||
|
||||
// summary
|
||||
String summary = root.path("summary").asText(null);
|
||||
if (summary != null && !"null".equals(summary)) {
|
||||
company.setSummary(summary);
|
||||
}
|
||||
|
||||
// description
|
||||
String description = root.path("description").asText(null);
|
||||
if (description != null && !"null".equals(description)) {
|
||||
company.setDescription(description);
|
||||
}
|
||||
|
||||
// foundedYear
|
||||
String foundedYear = root.path("foundedYear").asText(null);
|
||||
if (foundedYear != null && !"null".equals(foundedYear)) {
|
||||
company.setFoundedYear(foundedYear);
|
||||
}
|
||||
|
||||
// address
|
||||
String address = root.path("address").asText(null);
|
||||
if (address != null && !"null".equals(address)) {
|
||||
company.setAddress(address);
|
||||
}
|
||||
|
||||
// scale
|
||||
String scale = root.path("scale").asText(null);
|
||||
if (scale != null && !"null".equals(scale)) {
|
||||
company.setScale(scale);
|
||||
}
|
||||
|
||||
// website
|
||||
String website = root.path("website").asText(null);
|
||||
if (website != null && !"null".equals(website)) {
|
||||
company.setWebsite(website);
|
||||
}
|
||||
|
||||
// financingStage
|
||||
String financingStage = root.path("financingStage").asText(null);
|
||||
if (financingStage != null && !"null".equals(financingStage)) {
|
||||
company.setFinancingStage(financingStage);
|
||||
}
|
||||
|
||||
// latestValuation
|
||||
String latestValuation = root.path("latestValuation").asText(null);
|
||||
if (latestValuation != null && !"null".equals(latestValuation)) {
|
||||
company.setLatestValuation(latestValuation);
|
||||
}
|
||||
|
||||
// news：JSON数组
|
||||
JsonNode newsNode = root.path("news");
|
||||
if (newsNode.isArray() && !newsNode.isEmpty()) {
|
||||
List<String> news = new ArrayList<>();
|
||||
newsNode.forEach(node -> news.add(node.asText()));
|
||||
company.setNews(news);
|
||||
}
|
||||
|
||||
// 更新状态和时间
|
||||
company.setStatus(1);
|
||||
company.setUpdateTime(Instant.now());
|
||||
|
||||
companyMapper.updateById(company);
|
||||
}
|
||||
|
||||
/** 更新公司状态 */
|
||||
public void updateCompanyStatus(Long id, int status) {
|
||||
companyMapper.update(null,
|
||||
new LambdaUpdateWrapper<Company>()
|
||||
.set(Company::getStatus, status)
|
||||
.set(Company::getUpdateTime, Instant.now())
|
||||
.eq(Company::getId, id));
|
||||
}
|
||||
|
||||
/**
|
||||
* 原子锁定一批待完善公司（事务内 SELECT FOR UPDATE + UPDATE 状态）
|
||||
* <p>行锁保证并发安全，其他线程会阻塞直到事务提交</p>
|
||||
*
|
||||
* @return 锁定成功的公司列表，可能为空
|
||||
*/
|
||||
@Transactional(rollbackFor = Exception.class)
|
||||
public List<Company> lockBatch(int batchSize) {
|
||||
List<Company> companyList = companyMapper.selectForUpdate(batchSize);
|
||||
if (companyList.isEmpty()) {
|
||||
return companyList;
|
||||
}
|
||||
|
||||
List<Long> ids = companyList.stream().map(Company::getId).toList();
|
||||
companyMapper.update(null,
|
||||
new LambdaUpdateWrapper<Company>()
|
||||
.set(Company::getStatus, 3)
|
||||
.set(Company::getUpdateTime, Instant.now())
|
||||
.in(Company::getId, ids));
|
||||
|
||||
return companyList;
|
||||
}
|
||||
}
|
||||
@@ -116,15 +116,6 @@ public class DictCacheService {
|
||||
regionList.size(), majorCategoryList.size(), majorLeafCount);
|
||||
}
|
||||
|
||||
/** 获取岗位分类文本（叶子节点，带父级路径，逗号分隔） */
|
||||
public String getJobCategoryText() {
|
||||
return jobCategoryText;
|
||||
}
|
||||
|
||||
/** 获取行业文本（叶子节点，带父级路径，逗号分隔） */
|
||||
public String getIndustryText() {
|
||||
return industryText;
|
||||
}
|
||||
|
||||
/** 获取专业分类文本（三级叶子节点，带父级路径，逗号分隔） */
|
||||
public String getMajorCategoryText() {
|
||||
|
||||
@@ -1,384 +0,0 @@
|
||||
package org.jiayunet.service;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import org.jiayunet.ai.AiChatAbility;
|
||||
import org.jiayunet.ai.AiResponseCleanTool;
|
||||
import com.baomidou.mybatisplus.core.toolkit.IdWorker;
|
||||
import org.jiayunet.mapper.AppJobDataMapper;
|
||||
import org.jiayunet.mapper.JobMapper;
|
||||
import org.jiayunet.mapper.SkillTagMapper;
|
||||
import org.jiayunet.pojo.po.AppJobData;
|
||||
import org.jiayunet.pojo.po.Job;
|
||||
import org.jiayunet.pojo.po.SkillTag;
|
||||
import org.jiayunet.tool.HttpTool;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.beans.factory.annotation.Value;
|
||||
import org.springframework.scheduling.annotation.Scheduled;
|
||||
import org.springframework.stereotype.Service;
|
||||
|
||||
import java.time.Instant;
|
||||
import java.util.ArrayList;
|
||||
import java.util.List;
|
||||
import java.util.concurrent.ExecutorService;
|
||||
import java.util.concurrent.Executors;
|
||||
|
||||
/**
|
||||
* 岗位清洗服务
|
||||
* <p>定时从 app_job_data 捞取待清洗数据，调用 AI 清洗后写入业务表</p>
|
||||
* <p>依赖：AiChatAbility（AI调用）、DictCacheService（字典缓存）、JobCleanTransactionService（事务操作）</p>
|
||||
* <p>使用表：app_job_data（读取/更新状态）、bg_job（去重查询/更新专业）、bg_skill_tag（技能入库）</p>
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Slf4j
|
||||
@Service
|
||||
public class JobCleanService {
|
||||
|
||||
@Autowired
|
||||
private AiChatAbility aiChatAbility;
|
||||
|
||||
@Autowired
|
||||
private DictCacheService dictCacheService;
|
||||
|
||||
@Autowired
|
||||
private JobCleanTransactionService jobCleanTransactionService;
|
||||
|
||||
@Autowired
|
||||
private AppJobDataMapper appJobDataMapper;
|
||||
|
||||
@Autowired
|
||||
private JobMapper jobMapper;
|
||||
|
||||
@Autowired
|
||||
private SkillTagMapper skillTagMapper;
|
||||
|
||||
@Value("${app.job-clean.batch-size:20}")
|
||||
private int batchSize;
|
||||
|
||||
@Value("${app.job-clean.thread-pool-size:5}")
|
||||
private int threadPoolSize;
|
||||
|
||||
private ExecutorService executorService;
|
||||
|
||||
/** 初始化线程池 */
|
||||
@javax.annotation.PostConstruct
|
||||
public void init() {
|
||||
executorService = Executors.newFixedThreadPool(threadPoolSize);
|
||||
}
|
||||
|
||||
/**
|
||||
* 定时任务A：岗位清洗（每3分钟）
|
||||
* <p>1. 批量锁定待清洗数据 2. 多线程并发调用AI清洗 3. 写入业务表</p>
|
||||
*/
|
||||
@Scheduled(cron = "0 */3 * * * ?")
|
||||
public void cleanJob() {
|
||||
List<AppJobData> dataList = jobCleanTransactionService.lockBatch(batchSize);
|
||||
if (dataList.isEmpty()) {
|
||||
return;
|
||||
}
|
||||
log.info("岗位清洗：锁定{}条数据", dataList.size());
|
||||
|
||||
// 多线程并发处理
|
||||
for (AppJobData data : dataList) {
|
||||
executorService.submit(() -> {
|
||||
try {
|
||||
cleanOne(data);
|
||||
} catch (Exception e) {
|
||||
log.error("岗位清洗异常, id={}", data.getId(), e);
|
||||
// 异常时保持 clean_status=1，由僵尸恢复任务重置
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 定时任务B：僵尸恢复（每30分钟）
|
||||
* <p>将超时10分钟仍在清洗中的数据重置为待清洗</p>
|
||||
*/
|
||||
@Scheduled(cron = "0 */30 * * * ?")
|
||||
public void recoverZombie() {
|
||||
int recovered = appJobDataMapper.update(null,
|
||||
new LambdaUpdateWrapper<AppJobData>()
|
||||
.set(AppJobData::getCleanStatus, 0)
|
||||
.eq(AppJobData::getCleanStatus, 1)
|
||||
.lt(AppJobData::getUpdatedAt, Instant.now().minusSeconds(600)));
|
||||
|
||||
if (recovered > 0) {
|
||||
log.info("僵尸恢复：重置{}条数据", recovered);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 清洗单条岗位数据
|
||||
* <p>1. 前置校验 2. 第一次AI提取结构化信息 3. 写入业务表 4. 第二次AI匹配专业 5. 第三次AI提取技能</p>
|
||||
*/
|
||||
public void cleanOne(AppJobData data) {
|
||||
// 1. 前置校验
|
||||
if (data.getDescription() == null || data.getDescription().length() < 20) {
|
||||
jobCleanTransactionService.updateCleanStatus(data.getId(), 3);
|
||||
return;
|
||||
}
|
||||
|
||||
// 2. 第一次AI：提取岗位结构化信息
|
||||
String systemPrompt = buildSystemPrompt();
|
||||
String userMessage = buildUserMessage(data);
|
||||
String aiResponse = aiChatAbility.chat(systemPrompt, userMessage);
|
||||
|
||||
// 3. 解析JSON
|
||||
try {
|
||||
String json = AiResponseCleanTool.clean(aiResponse);
|
||||
JsonNode root = HttpTool.objectMapper.readTree(json);
|
||||
|
||||
// valid 校验
|
||||
if (!root.path("valid").asBoolean(false)) {
|
||||
jobCleanTransactionService.updateCleanStatus(data.getId(), 3);
|
||||
return;
|
||||
}
|
||||
|
||||
// 4. 去重检查
|
||||
String sourceId = String.valueOf(data.getId());
|
||||
Long existJob = jobMapper.selectCount(new LambdaQueryWrapper<Job>().eq(Job::getSourceId, sourceId));
|
||||
if (existJob > 0) {
|
||||
jobCleanTransactionService.updateCleanStatus(data.getId(), 2);
|
||||
return;
|
||||
}
|
||||
|
||||
// 5. 公司处理（加锁防并发重复）
|
||||
String companyShortName = root.path("companyShortName").asText("");
|
||||
if (companyShortName.isBlank()) {
|
||||
companyShortName = data.getCompany();
|
||||
}
|
||||
Long companyId = jobCleanTransactionService.findOrCreateCompany(companyShortName);
|
||||
|
||||
// 6. 地区处理
|
||||
List<String> regionCodes = new ArrayList<>();
|
||||
JsonNode citiesNode = root.path("cities");
|
||||
if (citiesNode.isArray()) {
|
||||
for (JsonNode city : citiesNode) {
|
||||
String code = dictCacheService.matchRegionCode(city.asText());
|
||||
if (code != null) {
|
||||
regionCodes.add(code);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 7. 写入业务表
|
||||
jobCleanTransactionService.saveJobData(root, data, companyId, sourceId, regionCodes);
|
||||
|
||||
// 拿到刚插入的 job
|
||||
Job insertedJob = jobMapper.selectOne(new LambdaQueryWrapper<Job>().eq(Job::getSourceId, sourceId).last("LIMIT 1"));
|
||||
if (insertedJob == null) {
|
||||
return;
|
||||
}
|
||||
|
||||
// 8. 第二次AI：专业匹配（失败不影响岗位入库）
|
||||
try {
|
||||
String title = root.path("title").asText("");
|
||||
String desc = root.path("description").asText("");
|
||||
String req = root.path("requirement").asText("");
|
||||
matchMajor(insertedJob.getId(), title, desc, req);
|
||||
} catch (Exception ex) {
|
||||
log.warn("专业匹配失败, id={}", data.getId(), ex);
|
||||
}
|
||||
|
||||
// 9. 第三次AI：技能提取（失败不影响岗位入库）
|
||||
try {
|
||||
String title = root.path("title").asText("");
|
||||
String desc = root.path("description").asText("");
|
||||
String req = root.path("requirement").asText("");
|
||||
extractSkillTags(insertedJob.getId(), title, desc, req);
|
||||
} catch (Exception ex) {
|
||||
log.warn("技能提取失败, id={}", data.getId(), ex);
|
||||
}
|
||||
|
||||
} catch (Exception e) {
|
||||
log.error("AI 返回解析失败, id={}, response={}", data.getId(), aiResponse, e);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 第二次AI调用：匹配专业 + 专业敏感度
|
||||
* <p>传入岗位信息和三级专业分类列表，AI返回 requiredMajorIds + majorSensitivity → 更新 bg_job</p>
|
||||
*/
|
||||
private void matchMajor(Long jobId, String title, String description, String requirement) {
|
||||
String systemPrompt = """
|
||||
你是一个岗位专业匹配助手。根据岗位信息，判断该岗位对专业的要求。
|
||||
返回JSON格式：
|
||||
{
|
||||
"requiredMajorIds": [专业ID数组，从专业列表中选择最相关的，最多3个，无明确要求则空数组],
|
||||
"majorSensitivity": 0-2的数字(0=专业不限 1=优先相关专业 2=强制要求专业)
|
||||
}
|
||||
规则：
|
||||
1. 只能从给定专业列表中选择ID
|
||||
2. 根据岗位描述判断专业敏感度：明确写"XX专业"→2，写"相关专业优先"→1，未提及→0
|
||||
3. majorSensitivity为0时，requiredMajorIds应为空数组
|
||||
4. 只返回JSON，不要其他内容
|
||||
""";
|
||||
|
||||
String userMessage = "【岗位信息】\n标题: " + title + "\n职责: " + description + "\n要求: " + requirement +
|
||||
"\n\n【专业分类列表】\n" + dictCacheService.getMajorCategoryText();
|
||||
|
||||
String aiResponse = aiChatAbility.chat(systemPrompt, userMessage);
|
||||
String json = AiResponseCleanTool.clean(aiResponse);
|
||||
|
||||
try {
|
||||
JsonNode root = HttpTool.objectMapper.readTree(json);
|
||||
int majorSensitivity = root.path("majorSensitivity").asInt(0);
|
||||
|
||||
List<Long> majorIds = new ArrayList<>();
|
||||
JsonNode idsNode = root.path("requiredMajorIds");
|
||||
if (idsNode.isArray()) {
|
||||
for (JsonNode node : idsNode) {
|
||||
long id = node.asLong(0);
|
||||
if (id > 0) {
|
||||
majorIds.add(id);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 更新 bg_job
|
||||
jobCleanTransactionService.updateJobMajor(jobId, majorIds.isEmpty() ? null : majorIds, majorSensitivity);
|
||||
} catch (Exception e) {
|
||||
log.warn("专业匹配AI返回解析失败: {}", json, e);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 第三次AI调用：自由提取技能标签
|
||||
* <p>AI返回技能名数组 → INSERT IGNORE 入 bg_skill_tag → 查ID → 写关联表</p>
|
||||
*/
|
||||
private void extractSkillTags(Long jobId, String title, String description, String requirement) {
|
||||
String systemPrompt = """
|
||||
你是一个技能提取助手。根据岗位信息，提取该岗位要求的核心专业能力和工具技能。
|
||||
返回JSON数组格式，如：["java", "spring boot", "mysql", "redis"]
|
||||
规则：
|
||||
1. 统一使用小写字母
|
||||
2. 只保留核心词，去掉多余修饰（如"plc编程"→"plc"，"c语言"→"c"，"cad制图"→"cad"）
|
||||
3. 同一技能只保留最具体的表述，不要同时出现上位词和下位词（如有"机械设计"就不要再出"机械"）
|
||||
4. 提取范围包括：技术栈、专业领域知识、行业工具、专业资质能力等
|
||||
5. 不提取纯软技能（如沟通能力、团队协作、学习能力、积极主动）
|
||||
6. 不提取过于宽泛的标签（如"办公软件"、"windows"）
|
||||
7. 如果岗位完全没有专业能力要求（纯看态度和素质），返回空数组 []
|
||||
8. 最多15个，按重要性排序
|
||||
9. 只返回JSON数组，不要其他内容
|
||||
示例1（技术岗）：
|
||||
输入：需要熟悉Java、Spring Boot框架，了解MySQL数据库和Redis缓存，会PLC编程
|
||||
输出：["java", "spring boot", "mysql", "redis", "plc"]
|
||||
示例2（财务岗）：
|
||||
输入：负责费用管理与审核，月度经营利润分析，要求财务管理、会计学相关专业
|
||||
输出：["财务管理", "会计", "经营分析"]
|
||||
示例3（制造岗）：
|
||||
输入：负责模具开发，熟悉CAD制图，了解注塑成型工艺，有SolidWorks经验
|
||||
输出：["模具", "cad", "注塑", "solidworks"]
|
||||
示例4（纯素质岗）：
|
||||
输入：具备较强的沟通能力和创新意识，积极主动，专业不限
|
||||
输出：[]
|
||||
""";
|
||||
|
||||
String userMessage = "【岗位信息】\n标题: " + title + "\n职责: " + description + "\n要求: " + requirement;
|
||||
|
||||
String aiResponse = aiChatAbility.chat(systemPrompt, userMessage);
|
||||
String json = AiResponseCleanTool.clean(aiResponse);
|
||||
|
||||
try {
|
||||
JsonNode arrayNode = HttpTool.objectMapper.readTree(json);
|
||||
if (!arrayNode.isArray() || arrayNode.isEmpty()) {
|
||||
return;
|
||||
}
|
||||
|
||||
List<Long> skillTagIds = new ArrayList<>();
|
||||
for (JsonNode node : arrayNode) {
|
||||
String skillName = node.asText("").trim().toLowerCase();
|
||||
if (skillName.isBlank() || skillName.length() > 50) {
|
||||
continue;
|
||||
}
|
||||
|
||||
// INSERT IGNORE + SELECT 获取ID
|
||||
Long tagId = findOrCreateSkillTag(skillName);
|
||||
if (tagId != null && !skillTagIds.contains(tagId)) {
|
||||
skillTagIds.add(tagId);
|
||||
}
|
||||
}
|
||||
|
||||
if (!skillTagIds.isEmpty()) {
|
||||
jobCleanTransactionService.saveSkillTagRelations(jobId, skillTagIds);
|
||||
}
|
||||
} catch (Exception e) {
|
||||
log.warn("技能提取AI返回解析失败: {}", json, e);
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 查找或创建技能标签（依靠数据库唯一索引保证并发安全）
|
||||
* <p>INSERT IGNORE 后 SELECT，避免加锁</p>
|
||||
*/
|
||||
private Long findOrCreateSkillTag(String name) {
|
||||
// 先尝试插入（忽略重复），ID 由 IdWorker 生成
|
||||
skillTagMapper.insertIgnore(IdWorker.getId(), name);
|
||||
|
||||
// 再查询拿ID
|
||||
SkillTag tag = skillTagMapper.selectOne(new LambdaQueryWrapper<SkillTag>().eq(SkillTag::getName, name).last("LIMIT 1"));
|
||||
return tag != null ? tag.getId() : null;
|
||||
}
|
||||
|
||||
/** 构建第一次AI的系统提示词 */
|
||||
private String buildSystemPrompt() {
|
||||
return """
|
||||
你是一个岗位数据清洗助手。请根据提供的原始岗位数据，提取并结构化为JSON格式。
|
||||
|
||||
返回JSON格式要求：
|
||||
{
|
||||
"valid": true/false,
|
||||
"title": "岗位名称",
|
||||
"salary": "标准化薪资，如10-20K、面议，无效则null",
|
||||
"education": 0-4的数字(0=不限 1=大专 2=本科 3=硕士 4=博士),
|
||||
"minExperience": 最低工作年限数字(不要求则0),
|
||||
"employmentType": 0或1(0=全职 1=兼职，默认0),
|
||||
"categoryId": 岗位分类ID(必选，从分类列表中选最接近的),
|
||||
"requiredIndustryId": 行业ID(仅当明确提到行业经验要求时设置，列表中无完全匹配则选最相似的，未提到则null),
|
||||
"description": "岗位职责，保持原文风格，格式化展示",
|
||||
"requirement": "任职要求，保持原文风格，格式化展示",
|
||||
"bonus": "加分项，无则null",
|
||||
"tags": ["核心职能标签，最多5个，如数据分析、产品策略"],
|
||||
"skillTags": ["技能关键词，最多8个，如Java、Spring Boot"],
|
||||
"companyShortName": "简洁的公司简称，如字节跳动、中国平安",
|
||||
"cities": ["工作城市列表，精确到市"]
|
||||
}
|
||||
|
||||
规则：
|
||||
1. description/requirement/bonus 均从原始的 description+experience+education 内容中提取，保持原文风格
|
||||
2. 岗位标题不存在时，从描述中归纳生成
|
||||
3. 薪资标准化为 10-20K、20K、面议 等格式，无效或空则null
|
||||
4. categoryId 必须从分类列表中选一个，不允许为null
|
||||
5. requiredIndustryId 仅当描述中明确提到行业经验要求时设置
|
||||
6. tags 是核心职能标签（如数据分析、团队协作），最多5个
|
||||
7. skillTags 是技能关键词（如Java、MySQL），最多8个
|
||||
8. companyShortName 去掉地区后缀、招聘后缀、括号内容，保持简洁
|
||||
9. 字符串值中不允许出现Tab、换行等控制字符，用空格或中文标点替代
|
||||
10. 只返回JSON，不要其他内容
|
||||
""";
|
||||
}
|
||||
|
||||
/** 构建第一次AI的用户消息 */
|
||||
private String buildUserMessage(AppJobData data) {
|
||||
StringBuilder sb = new StringBuilder();
|
||||
sb.append("【原始数据】\n");
|
||||
sb.append("岗位名称: ").append(nullToEmpty(data.getJobTitle())).append("\n");
|
||||
sb.append("薪资: ").append(nullToEmpty(data.getSalary())).append("\n");
|
||||
sb.append("工作地点: ").append(nullToEmpty(data.getLocation())).append("\n");
|
||||
sb.append("公司: ").append(nullToEmpty(data.getCompany())).append("\n");
|
||||
sb.append("经验要求: ").append(nullToEmpty(data.getExperience())).append("\n");
|
||||
sb.append("学历要求: ").append(nullToEmpty(data.getEducation())).append("\n");
|
||||
sb.append("岗位详情: ").append(nullToEmpty(data.getDescription())).append("\n\n");
|
||||
sb.append("【岗位分类列表】\n").append(dictCacheService.getJobCategoryText()).append("\n\n");
|
||||
sb.append("【行业列表】\n").append(dictCacheService.getIndustryText());
|
||||
return sb.toString();
|
||||
}
|
||||
|
||||
private String nullToEmpty(String s) {
|
||||
return s == null ? "" : s;
|
||||
}
|
||||
}
|
||||
@@ -1,193 +0,0 @@
|
||||
package org.jiayunet.service;
|
||||
|
||||
import com.baomidou.mybatisplus.core.conditions.query.LambdaQueryWrapper;
|
||||
import com.baomidou.mybatisplus.core.conditions.update.LambdaUpdateWrapper;
|
||||
import com.fasterxml.jackson.databind.JsonNode;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import org.jiayunet.mapper.AppJobDataMapper;
|
||||
import org.jiayunet.mapper.CompanyMapper;
|
||||
import org.jiayunet.mapper.JobMapper;
|
||||
import org.jiayunet.mapper.JobRegionRelationMapper;
|
||||
import org.jiayunet.mapper.JobSkillTagRelationMapper;
|
||||
import org.jiayunet.pojo.po.AppJobData;
|
||||
import org.jiayunet.pojo.po.Company;
|
||||
import org.jiayunet.pojo.po.Job;
|
||||
import org.jiayunet.pojo.po.JobRegionRelation;
|
||||
import org.jiayunet.pojo.po.JobSkillTagRelation;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.springframework.transaction.annotation.Transactional;
|
||||
|
||||
import java.time.Instant;
|
||||
import java.util.List;
|
||||
|
||||
/**
|
||||
* 岗位清洗事务服务
|
||||
* <p>独立出来解决 @Transactional 同类自调用失效问题</p>
|
||||
* <p>依赖：JobMapper、CompanyMapper、JobRegionRelationMapper、JobSkillTagRelationMapper、AppJobDataMapper</p>
|
||||
* <p>使用表：bg_job（写入/更新）、bg_company（查询/创建）、bg_job_region_relation（写入）、bg_job_skill_tag_relation（写入）、app_job_data（更新状态）</p>
|
||||
*
|
||||
* @author zk
|
||||
*/
|
||||
@Slf4j
|
||||
@Service
|
||||
public class JobCleanTransactionService {
|
||||
|
||||
@Autowired
|
||||
private JobMapper jobMapper;
|
||||
|
||||
@Autowired
|
||||
private CompanyMapper companyMapper;
|
||||
|
||||
@Autowired
|
||||
private JobRegionRelationMapper jobRegionRelationMapper;
|
||||
|
||||
@Autowired
|
||||
private AppJobDataMapper appJobDataMapper;
|
||||
|
||||
@Autowired
|
||||
private JobSkillTagRelationMapper jobSkillTagRelationMapper;
|
||||
|
||||
/**
|
||||
* 写入 bg_job + bg_job_region_relation + 更新 clean_status（短事务）
|
||||
*/
|
||||
@Transactional(rollbackFor = Exception.class)
|
||||
public void saveJobData(JsonNode root, AppJobData data, Long companyId, String sourceId, List<String> regionCodes) {
|
||||
Job job = new Job();
|
||||
job.setTitle(root.path("title").asText(""));
|
||||
job.setCompanyId(companyId);
|
||||
job.setCategoryId(root.path("categoryId").asLong(0));
|
||||
job.setEmploymentType(root.path("employmentType").asInt(0));
|
||||
job.setDescription(root.path("description").asText(""));
|
||||
job.setRequirement(root.path("requirement").asText(""));
|
||||
job.setBonus(root.path("bonus").asText(null));
|
||||
|
||||
// 解析 JSON 数组为 List<String>
|
||||
JsonNode tagsNode = root.path("tags");
|
||||
if (tagsNode.isArray()) {
|
||||
List<String> tags = new java.util.ArrayList<>();
|
||||
tagsNode.forEach(node -> tags.add(node.asText()));
|
||||
job.setTags(tags);
|
||||
}
|
||||
|
||||
JsonNode skillTagsNode = root.path("skillTags");
|
||||
if (skillTagsNode.isArray()) {
|
||||
List<String> skillTags = new java.util.ArrayList<>();
|
||||
skillTagsNode.forEach(node -> skillTags.add(node.asText()));
|
||||
job.setSkillTags(skillTags);
|
||||
}
|
||||
|
||||
String salary = root.path("salary").asText(null);
|
||||
job.setSalary("null".equals(salary) ? null : salary);
|
||||
|
||||
job.setEducation(root.path("education").asInt(0));
|
||||
job.setMinExperience(root.path("minExperience").asInt(0));
|
||||
|
||||
Long requiredIndustryId = root.path("requiredIndustryId").asLong(0);
|
||||
job.setRequiredIndustryId(requiredIndustryId == 0 ? null : requiredIndustryId);
|
||||
|
||||
job.setSourceUrl(data.getDetailUrl());
|
||||
job.setSourceId(sourceId);
|
||||
job.setStatus(0);
|
||||
job.setCreateTime(Instant.now());
|
||||
job.setUpdateTime(Instant.now());
|
||||
|
||||
jobMapper.insert(job);
|
||||
|
||||
// 写入岗位-地区关联（批量插入）
|
||||
if (!regionCodes.isEmpty()) {
|
||||
List<JobRegionRelation> relations = regionCodes.stream().map(regionCode -> {
|
||||
JobRegionRelation relation = new JobRegionRelation();
|
||||
relation.setJobId(job.getId());
|
||||
relation.setRegionCode(regionCode);
|
||||
relation.setCreateTime(Instant.now());
|
||||
return relation;
|
||||
}).toList();
|
||||
jobRegionRelationMapper.batchInsert(relations);
|
||||
}
|
||||
|
||||
// 更新清洗状态
|
||||
updateCleanStatus(data.getId(), 2);
|
||||
}
|
||||
|
||||
/**
|
||||
* 查找或创建公司（加锁防并发重复）
|
||||
* <p>按 short_name 查询，不存在则创建一条待完善记录</p>
|
||||
*/
|
||||
public synchronized Long findOrCreateCompany(String shortName) {
|
||||
Company company = companyMapper.selectOne(
|
||||
new LambdaQueryWrapper<Company>()
|
||||
.eq(Company::getShortName, shortName)
|
||||
.last("LIMIT 1"));
|
||||
|
||||
if (company != null) {
|
||||
return company.getId();
|
||||
}
|
||||
|
||||
Company newCompany = new Company();
|
||||
newCompany.setName(shortName);
|
||||
newCompany.setShortName(shortName);
|
||||
newCompany.setStatus(0);
|
||||
newCompany.setCreateTime(Instant.now());
|
||||
newCompany.setUpdateTime(Instant.now());
|
||||
companyMapper.insert(newCompany);
|
||||
return newCompany.getId();
|
||||
}
|
||||
|
||||
/**
|
||||
* 写入岗位-技能标签关联（批量插入）
|
||||
*/
|
||||
@Transactional(rollbackFor = Exception.class)
|
||||
public void saveSkillTagRelations(Long jobId, List<Long> skillTagIds) {
|
||||
List<JobSkillTagRelation> relations = skillTagIds.stream().map(skillTagId -> {
|
||||
JobSkillTagRelation relation = new JobSkillTagRelation();
|
||||
relation.setJobId(jobId);
|
||||
relation.setSkillTagId(skillTagId);
|
||||
relation.setCreateTime(Instant.now());
|
||||
return relation;
|
||||
}).toList();
|
||||
jobSkillTagRelationMapper.batchInsert(relations);
|
||||
}
|
||||
|
||||
/**
|
||||
* 更新岗位的专业要求和专业敏感度
|
||||
*/
|
||||
public void updateJobMajor(Long jobId, List<Long> requiredMajorIds, Integer majorSensitivity) {
|
||||
Job job = new Job();
|
||||
job.setId(jobId);
|
||||
job.setRequiredMajorIds(requiredMajorIds);
|
||||
job.setMajorSensitivity(majorSensitivity);
|
||||
job.setUpdateTime(Instant.now());
|
||||
jobMapper.updateById(job);
|
||||
}
|
||||
|
||||
/** 更新清洗状态 */
|
||||
public void updateCleanStatus(Long id, int status) {
|
||||
appJobDataMapper.update(null,
|
||||
new LambdaUpdateWrapper<AppJobData>()
|
||||
.set(AppJobData::getCleanStatus, status)
|
||||
.eq(AppJobData::getId, id));
|
||||
}
|
||||
|
||||
/**
|
||||
* 原子锁定一批待清洗数据（事务内 SELECT FOR UPDATE + UPDATE 状态）
|
||||
* <p>行锁保证并发安全，其他线程会阻塞直到事务提交</p>
|
||||
*
|
||||
* @return 锁定成功的数据列表，可能为空
|
||||
*/
|
||||
@Transactional(rollbackFor = Exception.class)
|
||||
public List<AppJobData> lockBatch(int batchSize) {
|
||||
List<AppJobData> dataList = appJobDataMapper.selectForUpdate(batchSize);
|
||||
if (dataList.isEmpty()) {
|
||||
return dataList;
|
||||
}
|
||||
|
||||
List<Long> ids = dataList.stream().map(AppJobData::getId).toList();
|
||||
appJobDataMapper.update(null,
|
||||
new LambdaUpdateWrapper<AppJobData>()
|
||||
.set(AppJobData::getCleanStatus, 1)
|
||||
.in(AppJobData::getId, ids));
|
||||
|
||||
return dataList;
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user