Skip to content

feat: structured parsing #190

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
minglet merged 2 commits into from
Jan 9, 2026
Merged

feat: structured parsing #190

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cc2c7ad
feat: structured parsing
Jan 8, 2026
d4046b6
Merge branch 'dev' into feat/structured_parsing
minglet Jan 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
93 changes: 92 additions & 1 deletion backend/main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,9 +5,11 @@
import shutil
import uuid
import os
import json
import traceback

from util.parser import run_parser
from util.resume_structured_parser import parse_resume_to_structured
from get_similarity.main import matching
from openai import OpenAI
import uvicorn
Expand All @@ -33,6 +35,7 @@

# 캐시 저장소
resume_cache = {}
resume_structured_cache = {} # 구조화된 이력서 데이터 캐시
analysis_cache = {}

# 전역 변수로 선언 (함수 내에서 global 키워드 사용)
Expand Down Expand Up @@ -165,10 +168,25 @@ async def run(data: MatchRequest):
resume = run_parser(resume_path)
resume_content_text = resume[0] # 첫 번째 반환값이 텍스트

# 캐시 저장
# 캐시 저장 (기존 문자열 저장)
resume_cache[resume_path] = resume_content_text
logger.info(f"[{trace_id}] parsed_pages={len(resume_content_text)} total_chars={len(resume_content_text)}")

# 구조화된 데이터 생성 및 저장
try:
if resume_path not in resume_structured_cache or resume_structured_cache[resume_path] is None:
structured_resume = await parse_resume_to_structured(resume_content_text)
if structured_resume:
resume_structured_cache[resume_path] = structured_resume
logger.info(f"[{trace_id}] structured resume data saved")
# 구조화된 이력서 결과를 JSON 형식으로 로그 출력
logger.info(f"[{trace_id}] ========== 구조화된 이력서 파싱 결과 ==========")
logger.info(f"[{trace_id}] {json.dumps(structured_resume, ensure_ascii=False, indent=2)}")
logger.info(f"[{trace_id}] ============================================")
except Exception as e:
logger.warning(f"[{trace_id}] failed to parse structured resume: {e}")
# 구조화 파싱 실패해도 기존 로직은 계속 진행

# 채용공고 추천
# res, job_description, job_url, c_name = await matching(
# resume_content_text, location=location_cache, remote=remote_cache, jobtype=job_type_cache
Expand Down Expand Up @@ -277,6 +295,21 @@ async def chat(request: Request):
resume = run_parser(resume_path)
resume_content_text = resume[0]
resume_cache[resume_path] = resume_content_text

# 구조화된 데이터 생성 및 저장
try:
if resume_path not in resume_structured_cache or resume_structured_cache[resume_path] is None:
structured_resume = await parse_resume_to_structured(resume_content_text)
if structured_resume:
resume_structured_cache[resume_path] = structured_resume
# 구조화된 이력서 결과를 JSON 형식으로 로그 출력
logger.info(f"[CHAT] ========== 구조화된 이력서 파싱 결과 ==========")
logger.info(f"[CHAT] resume_path={resume_path}")
logger.info(f"[CHAT] {json.dumps(structured_resume, ensure_ascii=False, indent=2)}")
logger.info(f"[CHAT] ============================================")
except Exception as e:
logger.warning(f"[CHAT] failed to parse structured resume: {e}")
# 구조화 파싱 실패해도 기존 로직은 계속 진행
else:
resume_content_text = resume_cache[resume_path]

Expand Down Expand Up @@ -410,6 +443,64 @@ async def evaluate(request: EvaluateRequest):
raise HTTPException(status_code=500, detail=str(e))


# /resume/structured - 구조화된 이력서 데이터 조회
class StructuredResumeRequest(BaseModel):
resume_path: str


@api_router.post("/resume/structured")
async def get_structured_resume(request: StructuredResumeRequest):
"""
구조화된 이력서 데이터를 조회합니다.

Args:
resume_path: 이력서 파일 경로

Returns:
JSONResponse: 구조화된 이력서 데이터 (스키마에 맞는 형식)
"""
resume_path = request.resume_path

if not resume_path:
raise HTTPException(status_code=400, detail="resume_path is required.")

# 캐시에서 조회
if resume_path in resume_structured_cache:
structured_data = resume_structured_cache[resume_path]
if structured_data:
return JSONResponse(content=structured_data)

# 캐시에 없으면 텍스트에서 생성 시도
if resume_path in resume_cache:
try:
resume_content_text = resume_cache[resume_path]
structured_resume = await parse_resume_to_structured(resume_content_text)
if structured_resume:
resume_structured_cache[resume_path] = structured_resume
return JSONResponse(content=structured_resume)
except Exception as e:
logger.warning(f"failed to parse structured resume: {e}")

# 모두 실패 시 기본 구조 반환
default_structured = {
"basic_info": {
"name": None,
"phone": None,
"email": None,
"address": None
},
"summary": None,
"careers": [],
"educations": [],
"skills": [],
"activities": [],
"languages": [],
"links": [],
"additional_info": {}
}
return JSONResponse(content=default_structured)


# /scrape-jd - JD URL 크롤링
class ScrapeJDRequest(BaseModel):
url: str
Expand Down
186 changes: 186 additions & 0 deletions backend/util/resume_structured_parser.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,186 @@
"""
이력서 텍스트를 구조화된 JSON 형식으로 변환하는 모듈
"""
import json
import re
import logging
from typing import Dict, Any, Optional
from langchain_upstage import ChatUpstage
from configs import UPSTAGE_API_KEY, RAG_MODEL

logger = logging.getLogger("jobpt")


def parse_json_loose(text: str) -> dict:
"""
LLM 출력에서 JSON을 추출하는 함수 (코드펜스, 불완전한 JSON 처리)
"""
# 코드펜스 제거
cleaned = re.sub(r"```[a-zA-Z]*\n?|```", "", text).strip()
# 첫 번째 {...} 블록만 추출
m = re.search(r"\{[\s\S]*\}", cleaned)
if not m:
raise ValueError("No JSON object found in LLM output")
obj = m.group(0)
# 따옴표 없는 키에 따옴표 부여: {sequence: "x"} -> {"sequence": "x"}
obj = re.sub(r"(\{|,)\s*([A-Za-z_][A-Za-z0-9_]*)\s*:", r'\1 "\2":', obj)
return json.loads(obj)


async def parse_resume_to_structured(resume_text: str) -> Optional[Dict[str, Any]]:
"""
이력서 텍스트를 구조화된 JSON 형식으로 변환

Args:
resume_text: 파싱된 이력서 텍스트

Returns:
구조화된 이력서 데이터 (스키마에 맞는 형식) 또는 None (실패 시)
"""
if not resume_text or not resume_text.strip():
return None

llm = ChatUpstage(model=RAG_MODEL, temperature=0, api_key=UPSTAGE_API_KEY)

prompt = f"""다음 이력서 텍스트를 구조화된 JSON 형식으로 변환해주세요.

이력서 텍스트:
{resume_text}

다음 JSON 스키마에 맞춰서 변환해주세요:
{{
"basic_info": {{
"name": "string | null",
"phone": "string | null",
"email": "string | null",
"address": "string | null"
}},
"summary": "string | null",
"careers": [
{{
"company_name": "string",
"period": "string | null",
"employment_type": "string | null",
"role": "string | null",
"achievements": []
}}
],
"educations": [
{{
"school_name": "string",
"period": "string | null",
"graduation_status": "string | null",
"major_and_degree": "string | null",
"content": "string | null"
}}
],
"skills": ["string"],
"activities": [
{{
"activity_name": "string",
"period": "string | null",
"activity_type": "string | null",
"content": "string | null"
}}
],
"languages": [
{{
"language_name": "string",
"level": "string | null",
"certification": "string | null",
"acquisition_date": "string | null"
}}
],
"links": ["string"],
"additional_info": {{}}
}}

주의사항:
- 정보가 없는 필드는 null로 설정
- 배열이 비어있으면 빈 배열 []로 설정
- 정확한 JSON 형식으로만 응답 (추가 설명 없이)
- company_name, school_name, activity_name, language_name은 필수 필드
- skills는 최대 50개까지
- links는 중복 제거된 URL 배열

JSON만 응답해주세요:"""

try:
logger.info("[ResumeParser] LLM을 통한 구조화 파싱 시작")
response = await llm.ainvoke(prompt)
content = response.content if hasattr(response, 'content') else str(response)

logger.info(f"[ResumeParser] LLM 응답 수신 완료 (길이: {len(content)} 문자)")

# JSON 파싱
structured_data = parse_json_loose(content)
logger.info("[ResumeParser] JSON 파싱 완료")

# 빈 문자열을 null로 변환하는 헬퍼 함수
def normalize_value(value):
"""빈 문자열이나 공백만 있는 문자열을 null로 변환"""
if value is None:
return None
if isinstance(value, str):
return value.strip() if value.strip() else None
return value

# basic_info 정규화
basic_info_raw = structured_data.get("basic_info", {})
if isinstance(basic_info_raw, dict):
basic_info = {
"name": normalize_value(basic_info_raw.get("name")),
"phone": normalize_value(basic_info_raw.get("phone")),
"email": normalize_value(basic_info_raw.get("email")),
"address": normalize_value(basic_info_raw.get("address")),
}
else:
basic_info = {
"name": None,
"phone": None,
"email": None,
"address": None
}

# 기본 구조 검증 및 기본값 설정
result = {
"basic_info": basic_info,
"summary": normalize_value(structured_data.get("summary")),
"careers": structured_data.get("careers", []),
"educations": structured_data.get("educations", []),
"skills": structured_data.get("skills", []),
"activities": structured_data.get("activities", []),
"languages": structured_data.get("languages", []),
"links": structured_data.get("links", []),
"additional_info": structured_data.get("additional_info", {})
}

# skills 최대 50개 제한
if len(result["skills"]) > 50:
result["skills"] = result["skills"][:50]

logger.info(f"[ResumeParser] 구조화 파싱 완료 - 기본정보: name={result['basic_info'].get('name')}, phone={result['basic_info'].get('phone')}, email={result['basic_info'].get('email')}")
logger.info(f"[ResumeParser] 경력: {len(result['careers'])}개, 학력: {len(result['educations'])}개, 스킬: {len(result['skills'])}개")

return result

except Exception as e:
logger.error(f"[ResumeParser] 이력서 구조화 파싱 실패: {e}")
# 실패 시 기본 구조 반환
return {
"basic_info": {
"name": None,
"phone": None,
"email": None,
"address": None
},
"summary": None,
"careers": [],
"educations": [],
"skills": [],
"activities": [],
"languages": [],
"links": [],
"additional_info": {}
}

Loading