dev → main: ABM 영구저장 + PDF 빌더 정합성 + Emerging 재설계 + 팀 문서 (2026-05-11)

.audit/reverted_patches/bdbd5754_jwt_integration.patch

@@ -0,0 +1,260 @@
+From bdbd575406179282a4a2a435c5b96fb5624e0ce2 Mon Sep 17 00:00:00 2001
+From: bat1120 <bat1120@naver.com>
+Date: Tue, 5 May 2026 00:49:54 +0900
+Subject: [PATCH] =?UTF-8?q?A1:=20corp=5Fbrand=5Fresolver=20JWT=20=EC=9E=90?=
+ =?UTF-8?q?=EB=8F=99=20=ED=86=B5=ED=95=A9=20(frontend=20biz=5Fnumber=20?=
+ =?UTF-8?q?=EB=88=84=EB=9D=BD=20=EC=8B=9C=20=ED=86=A0=ED=81=B0=EC=97=90?=
+ =?UTF-8?q?=EC=84=9C=20=EC=B6=94=EC=B6=9C)?=
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+PR #188 의 corp_brand_resolver 가 frontend 가 biz_number 명시 안 보내면
+검증 skip 되는 회귀 fix. JWT 토큰에서 자동 추출.
+
+추가:
+- _resolve_user_biz_number(user) — JWT UserContext → users.biz_number 조회
+  * master role: user_id → users.id 매칭
+  * manager role: owner_id → users.id 매칭 (소속 owner)
+- _validate_and_resolve_brand(input_data, current_user) 시그니처 확장
+- biz_number 우선순위:
+  1. input_data.biz_number (frontend 명시)
+  2. JWT 토큰의 user.user_id 또는 owner_id → users.biz_number
+  3. 없으면 검증 skip (비회원/개인사업자 호환)
+
+7 endpoint 시그니처에 Depends(get_optional_user) 추가:
+- /analyze, /analyze/llm, /analyze/llm/async, /analyze/quick
+- /predict, /predict/async, /simulate
+
+import:
+- from fastapi import Depends 추가
+- from src.services.jwt_auth import UserContext, get_optional_user
+
+기존 동작 유지:
+- 비로그인 사용자 → optional auth, current_user=None → skip (기존과 동일)
+- frontend biz_number 명시 → 토큰 무시 (frontend 우선)
+- frontend biz_number 없음 + 로그인 → 토큰에서 자동 추출 (이번 fix)
+
+DB 변경: 0
+
+Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
+---
+ backend/src/main.py | 98 ++++++++++++++++++++++++++++++++++-----------
+ 1 file changed, 75 insertions(+), 23 deletions(-)
+
+diff --git a/backend/src/main.py b/backend/src/main.py
+index e0a06478..15046217 100644
+--- a/backend/src/main.py
++++ b/backend/src/main.py
+@@ -47,7 +47,7 @@ import redis.asyncio as aioredis
+ # LangSmith 트레이싱: langchain import 전에 os.environ 주입 필수
+ # (langchain SDK는 import 시점에 LANGCHAIN_TRACING_V2를 읽으므로 순서가 중요)
+ from dotenv import load_dotenv
+-from fastapi import FastAPI, HTTPException, Request, Response
++from fastapi import Depends, FastAPI, HTTPException, Request, Response
+ from fastapi.concurrency import run_in_threadpool
+ from fastapi.middleware.cors import CORSMiddleware
+ from fastapi.responses import JSONResponse
+@@ -78,6 +78,7 @@ from src.schemas.simulation_input import SimulationInput
+ from src.services.auth import AuthService
+ from src.services.biz_mapper import BizMapper
+ from src.services.corp_brand_resolver import resolve_brand_for_industry
++from src.services.jwt_auth import UserContext, get_optional_user
+
+ from models.explainability.shap_analysis import explain_tcn_prediction
+ from models.explainability.simulation import (
+@@ -226,28 +227,56 @@ def _pipeline_key(input_data: Any) -> str:
+     return f"{input_data.target_district}:{input_data.business_type}:{input_data.brand_name}:{rent}:{area}:{radius}:{pop_w}"
+
+
+-def _validate_and_resolve_brand(input_data: SimulationInput) -> None:
++def _resolve_user_biz_number(user: UserContext | None) -> str | None:
++    """JWT user → users.biz_number 조회. master 는 본인, manager 는 owner 의 biz_number."""
++    if user is None:
++        return None
++    target_id = user.owner_id if user.role == "manager" else user.user_id
++    if not target_id:
++        return None
++    try:
++        import sqlalchemy as sa
++
++        engine = sa.create_engine(settings.postgres_url)
++        with engine.connect() as conn:
++            row = conn.execute(
++                sa.text("SELECT biz_number FROM users WHERE id = :id"),
++                {"id": target_id},
++            ).first()
++        return row._mapping["biz_number"] if row else None
++    except Exception as ex:
++        logger.warning(f"[brand_resolver] biz_number 조회 실패: {ex}")
++        return None
++
++
++def _validate_and_resolve_brand(
++    input_data: SimulationInput,
++    current_user: UserContext | None = None,
++) -> None:
+     """biz_number 입력 시 corp 검증 + 다업종 corp 의 brand auto-resolve.
+
+-    동작 (input_data.biz_number 가 입력됐을 때만):
++    biz_number 우선순위:
++    1. ``input_data.biz_number`` (frontend 명시 입력)
++    2. JWT ``current_user`` 토큰에서 자동 추출 (master.user_id 또는 manager.owner_id)
++    3. 없으면 검증 skip (개인사업자 / 비회원 호환)
++
++    동작:
+     1. business_type 이 사용자 corp 의 운영 업종인지 검증.
+     2. 운영 외 업종 → HTTPException(400) + 운영 가능 업종 list 응답.
+     3. 운영 내 업종 + corp 의 해당 업종 brand 가 다른 brand 면 brand_name override.
+-
+-    biz_number 미입력 (개인사업자 / 비회원) → 검증 skip, 사용자 brand_name 그대로.
+-    FTC 미등록 corp → 검증 skip + 경고 로그.
+     """
+-    if not input_data.biz_number:
++    biz_number = input_data.biz_number or _resolve_user_biz_number(current_user)
++    if not biz_number:
+         return
+
+-    result = resolve_brand_for_industry(input_data.biz_number, input_data.business_type)
++    result = resolve_brand_for_industry(biz_number, input_data.business_type)
+
+     if result.get("error") == "INDUSTRY_NOT_OPERATED":
+         raise HTTPException(status_code=400, detail=result)
+
+     if result.get("error") in {"USER_NOT_FOUND", "CORP_NOT_IN_FTC", "INVALID_COMPANY_NAME"}:
+         # 비회원 / FTC 미등록 → 검증 skip, 사용자 brand_name 그대로
+-        logger.warning(f"[brand_resolver] {result['error']} biz={input_data.biz_number} — fallback to input.brand_name")
++        logger.warning(f"[brand_resolver] {result['error']} biz={biz_number} — fallback to input.brand_name")
+         return
+
+     # 성공: brand_name override (사용자가 다른 brand 입력했어도 corp 정합 brand 로 교체)
+@@ -963,14 +992,18 @@ async def get_status(job_id: str):
+
+
+ @app.post("/analyze")
+-async def analyze_location(input_data: SimulationInput, response: Response):
++async def analyze_location(
++    input_data: SimulationInput,
++    response: Response,
++    current_user: UserContext | None = Depends(get_optional_user),
++):
+     """[DEPRECATED] 풀파이프 상권 분석 — 전환 기간 동안만 유지.
+
+     IM3-259로 endpoint를 분리(/predict + /analyze/llm)했으므로 신규 호출은
+     그쪽으로 옮길 것. 이 endpoint는 기존 프론트/테스트 호환을 위해 유지하다가
+     충분히 검증되면 제거 예정.
+     """
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.config.constants import MAPO_DISTRICTS
+
+     # IM3-259: deprecation 헤더 — 클라이언트가 /predict + /analyze/llm 으로 옮길 것을 알림
+@@ -1021,12 +1054,15 @@ async def analyze_location(input_data: SimulationInput, response: Response):
+
+
+ @app.post("/analyze/llm")
+-async def analyze_llm(input_data: SimulationInput):
++async def analyze_llm(
++    input_data: SimulationInput,
++    current_user: UserContext | None = Depends(get_optional_user),
++):
+     """AI 분석 전용 endpoint — slow_graph 실행 (~80-140초).
+
+     /predict와 독립 병렬 호출 가능. winner는 ranking 단계에서 자체 결정.
+     """
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.config.constants import MAPO_DISTRICTS
+     from src.schemas.simulation_output import AnalysisOutput
+
+@@ -1094,9 +1130,12 @@ _SLOW_GRAPH_NODE_TOTAL = 4
+
+
+ @app.post("/analyze/llm/async")
+-async def analyze_llm_async(input_data: SimulationInput) -> dict[str, Any]:
++async def analyze_llm_async(
++    input_data: SimulationInput,
++    current_user: UserContext | None = Depends(get_optional_user),
++) -> dict[str, Any]:
+     """AI 분석 비동기 시작 — 즉시 job_id 반환. LangGraph 노드별 진행률 추적."""
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.config.constants import MAPO_DISTRICTS
+     from src.schemas.simulation_output import AnalysisOutput
+     from src.services.job_progress_store import (
+@@ -1208,7 +1247,10 @@ async def analyze_llm_job_status(job_id: str) -> dict[str, Any]:
+
+
+ @app.post("/analyze/quick")
+-async def analyze_quick(input_data: SimulationInput):
++async def analyze_quick(
++    input_data: SimulationInput,
++    current_user: UserContext | None = Depends(get_optional_user),
++):
+     """
+     LLM 없는 경량 랭킹 엔드포인트 (district_ranking 에이전트만 실행).
+
+@@ -1217,7 +1259,7 @@ async def analyze_quick(input_data: SimulationInput):
+
+     응답: { district_rankings, winner_district, top_3_candidates }
+     """
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.agents.nodes.district_ranking import district_ranking_node
+     from src.agents.nodes.market_analyst import db_client
+
+@@ -1692,7 +1734,10 @@ def _mock_simulation_response(target_district: str, request_id: str) -> dict:
+
+
+ @app.post("/predict")
+-async def predict_districts(input_data: SimulationInput):
++async def predict_districts(
++    input_data: SimulationInput,
++    current_user: UserContext | None = Depends(get_optional_user),
++):
+     """
+     선택 동 1~4개 ML 예측 전용 엔드포인트 (LangGraph 미사용)
+
+@@ -1700,7 +1745,7 @@ async def predict_districts(input_data: SimulationInput):
+     - target_districts 전체에 대해 TCN/BEP/폐업률/폐업위험도/SHAP 병렬 실행
+     - 응답: 동별 예측 결과 리스트 (프론트 멀티라인 차트용)
+     """
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.config.constants import MAPO_DISTRICTS
+
+     target_districts = getattr(input_data, "target_districts", None) or [input_data.target_district]
+@@ -1762,9 +1807,12 @@ async def predict_districts(input_data: SimulationInput):
+ # 단계: 동별 _predict_single_district 가 끝날 때마다 progress = done/total.
+ # ---------------------------------------------------------------------------
+ @app.post("/predict/async")
+-async def predict_districts_async(input_data: SimulationInput) -> dict[str, Any]:
++async def predict_districts_async(
++    input_data: SimulationInput,
++    current_user: UserContext | None = Depends(get_optional_user),
++) -> dict[str, Any]:
+     """ML 예측 비동기 시작 — 즉시 job_id 반환. 진행률은 status endpoint 폴링."""
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     from src.config.constants import MAPO_DISTRICTS
+     from src.services.job_progress_store import (
+         create_job,
+@@ -1878,9 +1926,13 @@ async def predict_job_status(job_id: str) -> dict[str, Any]:
+
+
+ @app.post("/simulate", deprecated=True)
+-async def run_simulation(input_data: SimulationInput, response: Response):
++async def run_simulation(
++    input_data: SimulationInput,
++    response: Response,
++    current_user: UserContext | None = Depends(get_optional_user),
++):
+     """기본 시뮬레이션 엔드포인트"""
+-    _validate_and_resolve_brand(input_data)
++    _validate_and_resolve_brand(input_data, current_user)
+     response.headers["Deprecation"] = "true"
+     response.headers["Link"] = '</predict>; rel="successor-version", </analyze/llm>; rel="successor-version"'
+
+-- 
+2.53.0.windows.2
+

.audit_post_session.txt

@@ -0,0 +1,57 @@
+### POST-SESSION DB 실측 (PR #178~#197 머지 후) ###
+
+=== alembic head ===
+91b66e68ec18
+
+=== 테이블 + row counts (top 30) ===
+  ('living_population_grid', 10538127)
+  ('bus_boarding_daily', 3710508)
+  ('seoul_ttareungi_usage_daily', 985653)
+  ('living_population', 961071)
+  ('seoul_adstrd_stor', 849552)
+  ('district_sales_seoul', 475334)
+  ('sgis_population', 224517)
+  ('seoul_subway_passenger_daily', 199340)
+  ('golmok_commercial', 178840)
+  ('jeonse_monthly_rent', 168342)
+  ('sgis_business', 137356)
+  ('seoul_district_stores', 100587)
+  ('seoul_training_dataset', 87938)
+  ('seoul_district_sales', 87938)
+  ('kakao_store_menu', 81037)
+  ('seoul_signgu_stor', 69704)
+  ('seoul_signgu_selng', 43043)
+  ('ftc_brand_franchise', 34708)
+  ('naver_trend_monthly', 33985)
+  ('store_info', 30488)
+  ('sgis_household', 25550)
+  ('golmok_stores', 15800)
+  ('seoul_resident_pop_quarterly', 13508)
+  ('seoul_adstrd_flpop', 11900)
+  ('seoul_golmok_rent', 11900)
+  ('seoul_adstrd_change_ix', 11900)
+  ('mart_brand_territory', 11849)
+  ('langchain_pg_embedding', 10255)
+  ('seoul_population_quarterly', 10176)
+  ('small_store_rent_q', 10020)
+
+=== 100% NULL 컬럼 재측정 (master 메타 backfill 후) ===
+  district_sales_seoul: 100% NULL = ['raw_json']
+  dong_mapping: 100% NULL = ['trdar_codes']
+  ecos_timeseries: 100% NULL = ['item_name2', 'cycle']
+  invite_codes: 100% NULL = ['expires_at']
+  kakao_store_hours: 100% NULL = ['mon_hours', 'tue_hours', 'wed_hours', 'thu_hours', 'fri_hours', 'sat_hours', 'sun_hours']
+  living_population: 100% NULL = ['male_70_74', 'female_70_74']
+  mart_brand_territory: 100% NULL = ['extraction_confidence']
+  master_ttareungi_station: 100% NULL = ['dong_code', 'opened_at']
+  molit_nrg_trade: 100% NULL = ['realty_type']
+  rent_cost: 100% NULL = ['transaction_date', 'price', 'floor_area', 'floor']
+  seoul_dong_master: 100% NULL = ['comment']
+
+=== orphan FK ===
+  ✓ orphan 0
+
+=== ORM ↔ DB 정합 ===
+  ORM: 77, DB: 87
+  ORM only (zombie): []
+  DB only (raw SQL): ['alembic_version', 'langchain_pg_collection', 'langchain_pg_embedding', 'living_population_grid', 'mapo_schools', 'password_reset_tokens', 'seoul_district_sales_imputed_v4', 'seoul_district_sales_imputed_v4_detail', 'seoul_resident_pop_quarterly', 'user_usage']