From 310d99cfc32ae4edb0da53ad7f3fae1e15d99fd8 Mon Sep 17 00:00:00 2001
From: zk <sim18502043706@163.com>
Date: Fri, 5 Jun 2026 15:06:59 +0800
Subject: [PATCH] =?UTF-8?q?AI=E5=B2=97=E4=BD=8D=E6=8E=A8=E8=8D=90,?=
 =?UTF-8?q?=E4=BF=AE=E6=94=B9=E5=80=99=E9=80=89=E5=B2=97=E4=BD=8D=E6=B1=A0?=
 =?UTF-8?q?2000?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .../java/org/jiayunet/service/JobService.java | 30 ++++++++++++-------
 1 file changed, 19 insertions(+), 11 deletions(-)

diff --git a/client-api/src/main/java/org/jiayunet/service/JobService.java b/client-api/src/main/java/org/jiayunet/service/JobService.java
index f9bf592..3619438 100644
--- a/client-api/src/main/java/org/jiayunet/service/JobService.java
+++ b/client-api/src/main/java/org/jiayunet/service/JobService.java
@@ -614,7 +614,7 @@ public class JobService {
 
     /**
      * 求职助手岗位推荐
-     * <p>1. 查求职意向构造筛选条件（无意向则不设条件） 2. 排除已推荐的+已投递的 3. 取前35条候选 4. AI精筛返回8-10个</p>
+     * <p>1. 查求职意向构造筛选条件（无意向则不设条件） 2. 排除已推荐的+已投递的 3. 取2000条候选池 4. 随机取100条 5. AI精筛返回8-10个</p>
      */
     public JobAgentRecommendDto recommendJobs(JobAgentRecommendParam param, Long userId) {
         StopWatch sw = new StopWatch("recommendJobs");
@@ -627,12 +627,13 @@ public class JobService {
         // 2. 构造查询参数
         JobQueryParam queryParam = new JobQueryParam();
         queryParam.setPageNum(1);
-        queryParam.setPageSize(100);
+        queryParam.setPageSize(2000);
         if (intention != null) {
             queryParam.setCategoryIds(intention.getCategoryIds());
             queryParam.setRegionCodes(intention.getRegionCodes());
             queryParam.setIndustryIds(intention.getIndustryIds());
             queryParam.setEmploymentType(intention.getEmploymentType());
+            queryParam.setRecruitCategory(intention.getRecruitCategory());
         }
 
         // 3. 合并排除列表：入参传的 + 已投递/待投递的
@@ -646,7 +647,7 @@ public class JobService {
         applications.forEach(a -> excludeIds.add(a.getJobId()));
         queryParam.setExcludeJobIds(excludeIds);
 
-        // 4. 查询候选岗位（完整列表数据）
+        // 4. 查询候选池（2000条）
         sw.start("查候选岗位");
         PageResult<JobDto> candidates = listJobs(queryParam, userId);
         sw.stop();
@@ -657,27 +658,34 @@ public class JobService {
             return dto;
         }
 
-        // 5. 批量查岗位详情（title + description + requirement）
+        // 5. 随机取100条作为AI候选
+        List<JobDto> candidateList = new ArrayList<>(candidates.getList());
+        Collections.shuffle(candidateList);
+        if (candidateList.size() > 100) {
+            candidateList = candidateList.subList(0, 100);
+        }
+
+        // 6. 批量查岗位详情（title + description + requirement）
         sw.start("查岗位详情");
-        List<Long> candidateJobIds = candidates.getList().stream().map(JobDto::getId).collect(Collectors.toList());
+        List<Long> candidateJobIds = candidateList.stream().map(JobDto::getId).collect(Collectors.toList());
         List<Job> jobDetails = jobMapper.selectList(new LambdaQueryWrapper<Job>().in(Job::getId, candidateJobIds).select(Job::getId, Job::getTitle, Job::getDescription, Job::getRequirement));
         Map<Long, Job> jobDetailMap = jobDetails.stream().collect(Collectors.toMap(Job::getId, j -> j));
         sw.stop();
 
-        // 6. 构造别名映射（短序号 → 真实ID）和候选岗位映射
+        // 7. 构造别名映射（短序号 → 真实ID）和候选岗位映射
         Map<Integer, Long> aliasToRealId = new HashMap<>();
         Map<Long, JobDto> candidateMap = new HashMap<>();
         int seq = 1;
-        for (JobDto dto : candidates.getList()) {
+        for (JobDto dto : candidateList) {
             aliasToRealId.put(seq, dto.getId());
             candidateMap.put(dto.getId(), dto);
             seq++;
         }
 
-        // 7. 构造AI输入（用短序号替代19位雪花ID，减少token消耗）
+        // 8. 构造AI输入（用短序号替代19位雪花ID，减少token消耗）
         StringBuilder jobInfo = new StringBuilder();
         seq = 1;
-        for (JobDto dto : candidates.getList()) {
+        for (JobDto dto : candidateList) {
             Job detail = jobDetailMap.get(dto.getId());
             jobInfo.append("ID:").append(seq++)
                     .append("\n标题:").append(dto.getTitle())
@@ -695,7 +703,7 @@ public class JobService {
                 "只返回JSON，不要其他内容。";
         String userMessage = "【用户偏好（仅供参考）】\n" + preferenceInfo + "\n\n【候选岗位】\n" + jobInfo;
 
-        // 8. 调用AI
+        // 9. 调用AI
         sw.start("AI调用");
         String aiResponse = aiChatAbility.chat("job-recommend", systemPrompt, userMessage);
         String json = AiResponseCleanTool.clean(aiResponse);
@@ -703,7 +711,7 @@ public class JobService {
 
         log.info("recommendJobs耗时统计:\n{}", sw.prettyPrint());
 
-        // 9. 解析AI返回，别名映射回真实ID，过滤出选中的岗位
+        // 10. 解析AI返回，别名映射回真实ID，过滤出选中的岗位
         try {
             JsonNode root = HttpTool.objectMapper.readTree(json);
             String summary = root.path("summary").asText("为你精选了合适的岗位");