data_ai/extract/extract.go

package extract

import (
	"data_ai/clean"
	"data_ai/prompt"
	"data_ai/ul"
	log "github.com/donnie4w/go-logger/logger"
	qu "jygit.jydev.jianyu360.cn/data_processing/common_utils"
	"strings"
	"sync"
	"unicode/utf8"
)

// 识别结构化字段
func ExtractFieldInfo(sid string, eid string) {
	q := map[string]interface{}{
		"_id": map[string]interface{}{
			"$gt":  ul.StringTOBsonId(sid),
			"$lte": ul.StringTOBsonId(eid),
		},
	}
	//先查询抽取表-确定大模型需要识别到范围
	dict := ConfrimExtractInfo(q)
	log.Debug("查询语句...", q, "~", len(dict))
	if len(dict) >= ul.MaxUdp { //根据数量限制使用具体模型
		log.Debug("数量超过限制临时使用：glm-4-flashx")
		ul.FlashModel = "glm-4-flashx"
	} else {
		ul.FlashModel = "glm-4-flash"
	}
	pool_mgo := make(chan bool, ul.Reading)
	wg_mgo := &sync.WaitGroup{}

	sess := ul.SourceMgo.GetMgoConn()
	defer ul.SourceMgo.DestoryMongoConn(sess)
	total, isok := 0, 0
	it := sess.DB(ul.SourceMgo.DbName).C(ul.Bid_Name).Find(&q).Sort("_id").Iter()
	for tmp := make(map[string]interface{}); it.Next(&tmp); total++ {
		if total%200 == 0 {
			log.Debug("cur ai index ", total)
		}
		tmpid := ul.BsonTOStringId(tmp["_id"])
		infoformat := qu.IntAll(tmp["infoformat"])
		if infoformat > 1 || dict[tmpid] == nil {
			tmp = make(map[string]interface{})
			continue
		}
		isok++
		pool_mgo <- true
		wg_mgo.Add(1)
		go func(tmp map[string]interface{}) {
			defer func() {
				<-pool_mgo
				wg_mgo.Done()
			}()
			u_id := ul.BsonTOStringId(tmp["_id"])
			data := ResolveInfo(tmp)
			if len(data) > 0 || u_id == "" {
				ul.SourceMgo.UpdateById(ul.Ext_Name, u_id, map[string]interface{}{
					"$set": map[string]interface{}{"ai_zhipu": data},
				})
			}
		}(tmp)
		tmp = make(map[string]interface{})
	}
	wg_mgo.Wait()
	log.Debug("ai is over ...", sid, "~", eid)

}

// 获取处理数据...
func ResolveInfo(v map[string]interface{}) map[string]interface{} {
	tmpid := ul.BsonTOStringId(v["_id"])
	detail := getDetailText(v, tmpid) //获取正文文本
	title := qu.ObjToString(v["title"])
	dl := utf8.RuneCountInString(detail) //文本长度
	//过滤数据···
	if strings.Contains(title, "开标记录") || v["jyfb_data"] != nil {
		return map[string]interface{}{}
	}
	if dl < 20 {
		return map[string]interface{}{}
	}
	//识别结构，短文本结构
	f_data, shorText := map[string]interface{}{}, false
	if dl < 100 {
		shorText = true
	}
	//文本格式转换
	detail = ul.HttpConvertToMarkdown(detail)
	//短文本判断是否有效性
	if shorText {
		if info := prompt.AcquireJudgeShortInfo(detail); info["结果"] != "是" {
			return map[string]interface{}{}
		}
	}
	//获取外围字段数据-拆分合并字段
	f_info_1 := prompt.AcquireExtractFieldInfoFirst(detail)
	f_info_2 := prompt.AcquireExtractFieldInfoFirst(detail)
	f_info := MergeInfo([]map[string]interface{}{f_info_1, f_info_2})

	if !shorText {
		//获取分包信息
		if pkg := prompt.AcquireNewMultiplePackageInfo(detail); len(pkg) > 0 {
			f_info["s_pkg"] = pkg
		}
		//获取分类字段数据
		s_toptype, s_subtype := prompt.AcquireClassInfo(detail, title, qu.ObjToString(v["toptype"]))
		f_info["s_toptype"] = s_toptype
		f_info["s_subtype"] = s_subtype

		//调用标的物识别
		if p_list := getPurList(v, detail, f_info); len(p_list) > 0 {
			f_info["purchasinglist"] = p_list
		}
	}

	//字段清洗
	fns := getpnsinfo(v) //获取附件名字
	f_data = clean.CleanFieldInfo(f_info, fns)

	//采购单位二级校验
	CheckOutBuyerInfo(f_data)

	//强制逻辑判断-
	ForcedLogicDecideInfo(f_data)
	return f_data
}

// 确认抽取范围
func ConfrimExtractInfo(q map[string]interface{}) map[string]interface{} {
	dict := map[string]interface{}{}
	sess := ul.SourceMgo.GetMgoConn()
	defer ul.SourceMgo.DestoryMongoConn(sess)
	total := 0
	it := sess.DB(ul.SourceMgo.DbName).C(ul.Ext_Name).Find(&q).Select(map[string]interface{}{"_id": 1, "ai_zhipu": 1}).Iter()
	for tmp := make(map[string]interface{}); it.Next(&tmp); total++ {
		if total%1000 == 0 {
			log.Debug("cur index ", total)
		}
		if tmp["ai_zhipu"] == nil { //已经识别的数据-不再识别
			tmpid := ul.BsonTOStringId(tmp["_id"])
			dict[tmpid] = tmpid
		}
		tmp = make(map[string]interface{})
	}
	return dict
}

// 获取附件名字信息
func getpnsinfo(tmp map[string]interface{}) []string {
	arr := []string{}
	if projectinfo := qu.ObjToMap(tmp["projectinfo"]); projectinfo != nil {
		if attachments := qu.ObjToMap((*projectinfo)["attachments"]); attachments != nil {
			for _, v := range *attachments {
				if info := qu.ObjToMap(v); info != nil {
					if filename := qu.ObjToString((*info)["filename"]); filename != "" {
						arr = append(arr, filename)
					}
				}
			}
		}
	}
	return arr
}

// 获取正文数据
func getDetailText(v map[string]interface{}, tmpid string) string {
	detail := qu.ObjToString(v["detail"])
	if ul.IsTool {
		detail = qu.ObjToString(v["details"])
		filetext := qu.ObjToString(v["filetext"])
		if utf8.RuneCountInString(detail) < 100 && filetext != "" {
			detail = filetext
		}
	} else {
		//if bs := ul.OssGetObject(tmpid); bs != "" {
		//	detail = bs
		//}
	}
	return detail
}

// 获取标的物-过滤产权-拟建
func getPurList(v map[string]interface{}, detail string, f_info map[string]interface{}) []map[string]interface{} {
	if qu.ObjToString(v["toptype"]) == "拟建" || qu.ObjToString(v["toptype"]) == "产权" {
		return []map[string]interface{}{}
	}
	p_data := map[string]interface{}{}
	p_data["detail"] = qu.ObjToString(v["title"]) + "\n" + detail
	p_data["site"] = v["site"]
	p_data["attach_text"] = v["attach_text"]
	p_data["toptype"] = v["toptype"]
	if f_info["s_toptype"] != nil {
		p_data["toptype"] = f_info["s_toptype"]
	}
	if p_info := ul.PostPurchasingList(p_data); len(p_info) > 0 {
		if qu.IntAll(p_info["status"]) == 200 {
			p_list := ul.IsMarkInterfaceMap(p_info["purchasinglist"])
			return p_list
		}
	}
	return []map[string]interface{}{}
}

// 二次校验采购单位
func CheckOutBuyerInfo(f_data map[string]interface{}) {
	if s_buyer := qu.ObjToString(f_data["s_buyer"]); s_buyer != "" {
		if zp_buyer := prompt.AcquireBuyerInfo(s_buyer); zp_buyer["实体单位"] != nil {
			if ns_buyer := clean.CleanBuyer(qu.ObjToString(zp_buyer["实体单位"])); ns_buyer != "" {
				f_data["s_buyer"] = ns_buyer
			}
		}
	}
}

// 合并字段
func MergeInfo(infos []map[string]interface{}) map[string]interface{} {
	info := map[string]interface{}{}
	for _, v := range infos {
		for k1, v1 := range v {
			info[k1] = v1
		}
	}
	return info
}

// 强制逻辑判断数据
func ForcedLogicDecideInfo(f_data map[string]interface{}) {
	//原则大模型
	//多单位不能一致
	s_buyer := qu.ObjToString(f_data["s_buyer"])
	s_winner := qu.ObjToString(f_data["s_winner"])
	if s_buyer == s_winner && s_buyer != "" {
		/*
			1、若单位名称-不含公司保留采购单位
			2、若单位名称-含公司保留中标单位
		*/
		if strings.Contains(s_buyer, "公司") {
			f_data["s_buyer"] = ""
		} else {
			f_data["s_winner"] = ""
		}
	}

	//代理机构
	if s_agency := qu.ObjToString(f_data["s_agency"]); s_agency != "" {
		if s_agency == s_buyer || s_agency == s_winner {
			f_data["s_agency"] = ""
		}
	}
}

// 暂时不启用...无限重试
func RunResetUpdateFieldInfo(arr []string, name string, s_name string) {
	//log.Debug("开始重置更新...", len(arr))
	//reset := []string{}
	//for k, v := range arr {
	//	log.Debug("...", k, "...", v)
	//	data := ul.SourceMgo.FindById(name, v)
	//	content := PromptFieldText(qu.ObjToString(data["detail"]))
	//	zp, ok := map[string]interface{}{}, 0
	//	for {
	//		ok++
	//		if zp = ai.PostZhiPuAI(content); len(zp) > 0 {
	//			break
	//		}
	//		if ok >= 5 {
	//			log.Debug("请求数据失败...", v)
	//			reset = append(reset, v)
	//			break
	//		}
	//	}
	//	ul.SourceMgo.UpdateById(s_name, v, map[string]interface{}{
	//		"$set": map[string]interface{}{
	//			"zhipu": zp,
	//		},
	//	})
	//}
	//if len(reset) > 0 { //无限尝试
	//	RunResetUpdateFieldInfo(reset, name, s_name)
	//}
	//log.Debug("本轮重置更新结束......")
}