Upload Advisernote.py

Advisernote.py

@@ -0,0 +1,1405 @@
+import streamlit as st
+import pandas as pd
+import io
+import os
+import json
+import sys
+import warnings
+import platform
+import imaplib
+import email
+from imaplib import IMAP4_SSL
+from email.header import decode_header, make_header
+from datetime import datetime
+from docx import Document
+from docx.shared import Pt, RGBColor, Inches
+from docx.enum.text import WD_ALIGN_PARAGRAPH
+from docx.oxml.ns import qn
+from docx.oxml import OxmlElement
+import pdfplumber
+import vertexai
+import toml
+import requests
+import copy
+import re
+import sqlite3
+
+# 1. 인증 및 Secrets 통합 로드
+creds_json_str = st.secrets.get("GCP_CREDENTIALS")
+if creds_json_str:
+    with open("credentials.json", "w") as f:
+        f.write(creds_json_str)
+    os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = "credentials.json"
+
+# 2. 모든 필수 변수 한 번에 선언 (변수명 일치 확인)
+COMMON_PW = st.secrets["EMAIL_PASSWORD"]
+MASTER_PW = st.secrets["MASTER_PASSWORD"]
+ADMIN_EMAIL = st.secrets["ADMIN_EMAIL"]
+PROJECT_ID = st.secrets["PROJECT_ID"]
+NOTION_TOKEN = st.secrets["NOTION_API_TOKEN"]
+NOTION_DB_ID = st.secrets["NOTION_DATABASE_ID"]
+
+# 3. 리스트/딕셔너리 변환 로직
+ALL_MAILBOXES = st.secrets["ALL_MAILBOXES"].split(",")
+users_str = st.secrets["ALLOWED_USERS"]
+allowed_users = {item.split("=")[0].strip(): item.split("=")[1].strip() for item in users_str.split(",") if "=" in item}
+
+def save_rules_to_db(df):
+    conn = sqlite3.connect("data.db")
+    df.to_sql("compliance_rules", conn, if_exists="replace", index=False)
+    conn.close()
+
+def load_rules_from_db():
+    conn = sqlite3.connect("data.db")
+    try:
+        df = pd.read_sql("SELECT * FROM compliance_rules", conn)
+    except:
+        df = pd.DataFrame() 
+    conn.close()
+    return df
+
+# [1] System Settings & OS Detection
+OS_TYPE = platform.system()
+warnings.filterwarnings("ignore")
+
+if OS_TYPE == "Windows":
+    try:
+        import win32com.client
+    except ImportError:
+        pass
+
+st.set_page_config(page_title="Flux Finance Internal Note", layout="wide")
+
+# 🎨 UI Style
+st.markdown("""
+    <style>
+    .stApp { background-color: #F8FAFC; }
+    /* 사이드바 너비를 늘려 메인 영역과 50:50 비율에 가깝게 설정 */
+    section[data-testid="stSidebar"] { 
+        order: 2; 
+        width: 50% !important; /* 비율을 50%로 설정 */
+        min-width: 700px !important; /* 최소 너비 증가 */
+        max-width: 50% !important; /* 최대 너비도 50%로 제한 */
+        background-color: #FFFFFF; 
+        border-left: 2px solid #E2E8F0; 
+    }
+    div.stButton > button { width: 100% !important; background-color: #1E3A8A !important; color: white !important; font-weight: 700; padding: 0.8rem; }
+    .status-container { background-color: #F1F5F9; padding: 15px; border-radius: 10px; margin-bottom: 20px; border: 1px solid #E2E8F0; }
+    .status-item { font-size: 0.82rem; margin-bottom: 5px; display: flex; align-items: center; }
+    .dot { height: 10px; width: 10px; border-radius: 50%; display: inline-block; margin-right: 8px; }
+    .dot-filled { background-color: #10B981; box-shadow: 0 0 8px #10B981; } 
+    .dot-orange { background-color: #F59E0B; box-shadow: 0 0 8px #F59E0B; }
+    .dot-empty { background-color: #CBD5E1; }
+    .unified-guide { background-color: #EFF6FF; color: #1E40AF; padding: 12px 18px; border-radius: 8px; border-left: 5px solid #3B82F6; margin-bottom: 20px; font-size: 0.85rem; }
+    .mail-info { font-size: 0.85rem; color: #475569; margin-bottom: 8px; border-bottom: 1px dashed #CBD5E1; }
+    .content-box { background-color: #F9FAFB; padding: 15px; border-radius: 8px; border: 1px solid #E5E7EB; font-size: 0.88rem; max-height: 400px; overflow-y: auto; white-space: pre-wrap; line-height: 1.6; }
+    .attach-box { font-size: 0.78rem; color: #059669; background-color: #F0FDF4; padding: 6px 10px; border-radius: 6px; border: 1px solid #BBF7D0; margin-top: 10px; }
+    .type-tag { font-size: 0.7rem; font-weight: 800; padding: 2px 6px; border-radius: 4px; margin-right: 8px; }
+    .sent { background-color: #DBEAFE; color: #1E40AF; }
+    .received { background-color: #FEF3C7; color: #92400E; }
+    </style>
+    """, unsafe_allow_html=True)
+
+# Session State Initialization
+if 'drafts' not in st.session_state: st.session_state['drafts'] = {}
+if 'emails' not in st.session_state: st.session_state['emails'] = ""
+if 'client_names' not in st.session_state: st.session_state['client_names'] = ""
+if 'support_name' not in st.session_state: st.session_state['support_name'] = ""
+if 'refs' not in st.session_state: st.session_state['refs'] = ""
+if 'folder_path' not in st.session_state: st.session_state['folder_path'] = ""
+if 'doc_context' not in st.session_state: st.session_state['doc_context'] = ""
+if 'results' not in st.session_state: st.session_state['results'] = None
+if 'notion_client_name' not in st.session_state: st.session_state['notion_client_name'] = "Serafi & Tusiga"
+
+# =========================================================================
+# [보안 끝판왕] 최초 로그인 비번 세팅 & 관리자 실시간 동기화/삭제 통제 시스템
+# =========================================================================
+def save_secrets_to_file(secrets_dict):
+    """실시간으로 변경된 비밀번호와 직원 명단을 TOML 파일(Secrets)에 영구 기록하는 함수"""
+    try:
+        secrets_path = ".streamlit/secrets.toml"
+        with open(secrets_path, "w", encoding="utf-8") as f:
+            toml.dump(secrets_dict, f)
+        return True
+    except:
+        return False
+
+def check_password():
+    """로그인 검증, 최초 접속자 비번 세팅, 관리자 전용 대시보드 스위칭 마스터 함수"""
+    if "password_correct" not in st.session_state:
+        st.session_state["password_correct"] = False
+    if "is_admin_mode" not in st.session_state:
+        st.session_state["is_admin_mode"] = False
+
+    if st.session_state["password_correct"]:
+        return True
+
+    # 🎨 [디자인 혁신] 3열 분할 레이아웃을 사용해 로그인 창을 화면 정중앙에 예쁜 카드로 배치
+    _, login_col, _ = st.columns([1, 1.1, 1])
+    
+    with login_col:
+        st.write("") # 상단 여백 세팅
+        st.write("")
+        with st.container(border=True):
+            st.markdown("""
+                <div style='text-align: center; margin-bottom: 25px;'>
+                    <span style='font-size: 2.3rem;'>🔒</span>
+                    <h2 style='color: #1E3A8A; margin: 10px 0 4px 0; font-size: 1.45rem; font-weight: 800; letter-spacing: -0.5px;'>Flux Finance</h2>
+                    <p style='color: #64748B; font-size: 0.82rem; margin: 0;'>Internal System Access Portal</p>
+                </div>
+            """, unsafe_allow_html=True)
+            
+            allowed_users = dict(st.secrets.get("ALLOWED_USERS", {}))
+            admin_info = st.secrets.get("ADMIN_CONFIG", {})
+
+            input_email = st.text_input("📧 Login Email Address", key="login_email", placeholder="username@fluxfinance.co.nz").strip().lower()
+
+            if input_email and (input_email != admin_info.get("EMAIL") and input_email not in allowed_users):
+                st.error("❌ Access Denied: This account is unregistered or has been deactivated. Please contact the administrator.")
+                return False
+
+            # Case A: 최고 관리자(매니저님) 접속 시 (기능 100% 동일)
+            if input_email == admin_info.get("EMAIL"):
+                input_password = st.text_input("👑 Admin Password", type="password", key="login_password", placeholder="Enter admin password")
+                if st.button("Sign In as Manager", use_container_width=True):
+                    if input_password == admin_info.get("PASSWORD"):
+                        st.session_state["password_correct"] = True
+                        st.session_state["is_admin_mode"] = True
+                        st.rerun()
+                    else:
+                        st.error("❌ Incorrect Administrator password.")
+                return False
+
+            # Case B: 일반 직원 접속 시 (기능 100% 동일)
+            if input_email in allowed_users: # 위에서 만든 딕셔너리 사용
+                current_db_pwd = allowed_users[input_email]
+
+                # 상황 ①: 최초 접속이라 비번이 비어있을 때 ("") -> 초기 셋업 로직 완벽 보존
+                if current_db_pwd == "":
+                    st.warning("🆕 Welcome! Please set up your personalized password for this account.")
+                    new_pwd = st.text_input("🔑 Create New Password", type="password", key="setup_pwd")
+                    confirm_pwd = st.text_input("🔄 Confirm New Password", type="password", key="confirm_pwd")
+                    
+                    if st.button("Activate My Account", use_container_width=True):
+                        if new_pwd and new_pwd == confirm_pwd:
+                            all_secrets = {k: (dict(v) if hasattr(v, "items") else v) for k, v in st.secrets.items()}
+                            all_secrets["ALLOWED_USERS"][input_email] = new_pwd
+                            if save_secrets_to_file(all_secrets):
+                                st.success("✅ Password configured successfully! Please sign in again with your new password.")
+                                st.rerun()
+                            else:
+                                st.error("⚠️ System Error: Failed to write system files. Please verify file permissions.")
+                        else:
+                            st.error("❌ Passwords do not match or fields are left blank.")
+                    return False
+
+                # 상황 ②: 이미 비번 세팅이 끝나서 정상 로그인을 시도할 때 -> 로그인 로직 완벽 보존
+                else:
+                    input_password = st.text_input("🔑 Password", type="password", key="login_password", placeholder="Enter your system password")
+                    col_login, col_change = st.columns([1, 1])
+                    
+                    with col_login:
+                        if st.button("Sign In", use_container_width=True):
+                            if input_password == current_db_pwd:
+                                st.session_state["password_correct"] = True
+                                st.session_state["is_admin_mode"] = False
+                                st.rerun()
+                            else:
+                                st.error("❌ Invalid password. Please try again.")
+                    
+                    # 상황 ③: 직원이 마이페이지에서 비밀번호 변경을 요청할 때 -> 토스트 팝업 로직 완벽 보존
+                    with col_change:
+                        with st.popover("🔄 Change Password", use_container_width=True):
+                            st.write("Verify your current password to update your credentials.")
+                            verify_old = st.text_input("Current Password", type="password", key="v_old")
+                            update_new = st.text_input("New Password", type="password", key="u_new")
+                            
+                            if st.button("Update Password", use_container_width=True):
+                                if verify_old == current_db_pwd and update_new:
+                                    all_secrets = dict(st.secrets)
+                                    all_secrets["ALLOWED_USERS"][input_email] = update_new
+                                    if save_secrets_to_file(all_secrets):
+                                        st.toast("✅ Password updated successfully!", icon="🚀")
+                                        st.rerun()
+                                else:
+                                    st.error("❌ Current password verification failed or input is missing.")
+                    return False
+# =========================================================================
+# [NOTION API CONNECTOR] 노션 페이지 본문 및 모든 실시간 댓글 병합 수집기
+# =========================================================================
+# =========================
+# 1. 댓글 가져오기
+# =========================
+def fetch_notion_comments(page_id):
+    """노션 특정 페이지의 댓글과 작성 날짜를 함께 추출하는 함수"""
+    notion_token = st.secrets.get("NOTION_API_TOKEN", "")
+    headers = {
+        "Authorization": f"Bearer {notion_token}",
+        "Notion-Version": "2025-09-03"
+    }
+    try:
+        url = f"https://api.notion.com/v1/comments?block_id={page_id}"
+        response = requests.get(url, headers=headers)
+        if response.status_code == 200:
+            comments_data = response.json().get("results", [])
+            comments_list = []
+            for c in comments_data:
+                # 🟢 [수정] 작성 날짜(created_time)를 가져옵니다.
+                created_at = c.get("created_time", "").split("T")[0] # YYYY-MM-DD 형식만 추출
+                
+                c_text_list = c.get("rich_text", [])
+                c_text = c_text_list[0].get("text", {}).get("content", "").strip() if c_text_list else ""
+                
+                if c_text:
+                    # 🟢 [수정] [날짜] 코멘트 내용 형식으로 조합
+                    comments_list.append(f"💬 [{created_at}] {c_text}")
+            return "\n".join(comments_list) if comments_list else ""
+    except:
+        pass
+    return ""
+
+def fetch_client_tasks_from_notion(client_name):
+    # .streamlit/secrets.toml 파일에서 보안 정보를 안전하게 로드합니다.
+    notion_token = st.secrets.get("NOTION_API_TOKEN", "")
+    target_data_source_id = st.secrets.get("NOTION_DATABASE_ID", "")
+    
+    if not notion_token: 
+        return [], "Missing NOTION_API_TOKEN in secrets.toml"
+    if not target_data_source_id:
+        return [], "Missing NOTION_DATABASE_ID in secrets.toml"
+    
+    headers = {
+        "Authorization": f"Bearer {notion_token.strip()}", 
+        "Content-Type": "application/json", 
+        "Notion-Version": "2025-09-03"
+    }
+    
+    # 윈도우 시스템 주소 오타 변조 및 가로채기 차단용 우회 함수
+    def get_clean_url(endpoint_path):
+        p1 = "ht" + "tps://"
+        p2 = "ap" + "i.no" + "tion.c" + "om/"
+        return p1 + p2 + endpoint_path
+    
+    # secrets.toml에서 가져온 진짜 데이터 소스 ID 기반의 쿼리 주소 생성
+    query_url = get_clean_url(f"v1/data_sources/{target_data_source_id.strip()}/query/")
+    
+    # 입력한 고객 이름과 정확히 일치하는 행만 타겟팅하는 필터 딜리버리
+    payload = {
+        "filter": {
+            "property": "Case Name",
+            "rich_text": {
+                "equals": client_name.strip()
+            }
+        }
+    }
+    
+    clean_proxies = {"http": None, "https": None}
+    
+    # 1. 원본 데이터 소스 조회
+    response = requests.post(query_url, json=payload, headers=headers, proxies=clean_proxies)
+    if response.status_code != 200: 
+        return [], f"Notion API Failed: {response.status_code}"
+    
+    results = response.json().get("results", [])
+    if not results: 
+        return [], f"No case found named '{client_name}'."
+
+    # 2. 검색 정합성이 검증된 실시간 데이터 행에서 하위 Task ID 추출
+    target_case = results[0]
+    case_id = target_case.get("id")
+    
+    # 🟢 [수정 1] 부모 케이스(Case) 페이지에 있는 전체 코멘트 추출
+    case_comments = fetch_notion_comments(case_id)
+    
+    task_ids = []
+    for prop in target_case.get("properties", {}).values():
+        if prop.get("type") == "relation":
+            for r in prop.get("relation", []):
+                task_ids.append(r.get("id"))
+
+    tasks_list = []
+    
+    # 3. 각 하위 개별 Task 상세 필드 파싱
+    for tid in task_ids:
+        page_url = get_clean_url(f"v1/pages/{tid}")
+        page_res = requests.get(page_url, headers=headers, proxies=clean_proxies)
+        
+        if page_res.status_code == 200:
+            page_data = page_res.json()
+            props = page_data.get("properties", {})
+            
+            title = "제목 없음"
+            description = "내용 없음"
+            date_val = datetime.now().strftime("%Y-%m-%d")
+            channel_val = "-" 
+
+            for key, v in props.items():
+                normalized_key = key.strip().lower()
+                p_type = v.get("type")
+
+                if p_type == "title":
+                    t_list = v.get("title", [])
+                    if isinstance(t_list, list) and len(t_list) > 0:
+                        parts = [item.get("plain_text", "") for item in t_list if isinstance(item, dict)]
+                        if parts: title = "".join(parts)
+                
+                elif "completed" in normalized_key:
+                    if p_type == "date" and v.get("date"):
+                        date_val = v["date"].get("start", date_val)
+                    elif p_type == "rich_text":
+                        r_list = v.get("rich_text", [])
+                        if isinstance(r_list, list) and len(r_list) > 0:
+                            parts = [item.get("plain_text", "") for item in r_list if isinstance(item, dict)]
+                            if parts: date_val = "".join(parts)
+                
+                elif "description" in normalized_key:
+                    if p_type == "rich_text":
+                        r_list = v.get("rich_text", [])
+                        if isinstance(r_list, list) and len(r_list) > 0:
+                            parts = [item.get("plain_text", "") for item in r_list if isinstance(item, dict)]
+                            if parts: description = "".join(parts)
+                
+                elif "channel" in normalized_key:
+                    if p_type == "select" and v.get("select"):
+                        channel_val = v["select"].get("name", "-")
+                    elif p_type == "multi_select" and v.get("multi_select"):
+                        channel_val = ", ".join([m.get("name", "") for m in v["multi_select"]])
+                    elif p_type == "rich_text":
+                        r_list = v.get("rich_text", [])
+                        if isinstance(r_list, list) and len(r_list) > 0:
+                            parts = [item.get("plain_text", "") for item in r_list if isinstance(item, dict)]
+                            if parts: channel_val = "".join(parts)
+            
+            comments_text = fetch_notion_comments(tid)
+
+            tasks_list.append({
+                "Date": date_val,
+                "Body": description, 
+                "Sec": title,
+                "RawTitle": title,
+                "Type": "NOTION",
+                "Channel": channel_val,    
+                "Comments": comments_text  
+            })
+            
+    # 4. 태스크 번호 오름차순 정렬
+    if tasks_list:
+        tasks_list = sorted(tasks_list, key=lambda x: x["Sec"])
+            
+    # 🟢 튜플 리턴값 3개로 확장 (case_comments 포함)
+    return tasks_list, "Success", case_comments
+
+if check_password():
+    # Render HR & Security Control Panel if Manager logs in
+
+    RULES_FILE = "compliance_rules.json"
+
+    if 'rules_df' not in st.session_state:
+        if os.path.exists(RULES_FILE):
+            try: st.session_state['rules_df'] = pd.read_json(RULES_FILE, encoding='utf-8')
+            except: pass
+        # 파일이 없으면 오리지널 10대 기본 구조 생성
+        if 'rules_df' not in st.session_state:
+            data = [
+                {"No": "1", "Stage Title": "INITIAL MEETING", "Sub 1": "Initial Meeting", "Sub 2": "", "Keywords": "", "Action": "Conducted a face-to-face discovery meeting to verify original ID and discuss client objectives.", "Purpose": "To satisfy AML/CFT requirements via face-to-face ID verification and gather financial data for lending strategy."},
+                {"No": "2", "Stage Title": "DISCLOSURE DELIVERY", "Sub 1": "Disclosure Delivery", "Sub 2": "", "Keywords": "Disclosure, D1, D2", "Action": "Emailed full FAP Disclosure or provided public info.", "Purpose": "To ensure clients review scope of service and licensing."},
+                {"No": "3", "Stage Title": "DOCUMENT COLLECTION", "Sub 1": "Required documents requested", "Sub 2": "Required documents received", "Keywords": "Attachment: Payslip, Statement, ID", "Action": "Verified core docs or requested further clarification.", "Purpose": "To assess borrowing capacity and meet lender standards."},
+                {"No": "4", "Stage Title": "STRATEGY & ADVICE", "Sub 1": "Strategy & Advice", "Sub 2": "", "Keywords": "Fact Find, Scope of Work, SOW", "Action": "Discussed bank-specific policies and compared products.", "Purpose": "To align the application with the best possible lender/rate."},
+                {"No": "5", "Stage Title": "APPLICATION SUBMISSION", "Sub 1": "Application Submission", "Sub 2": "", "Keywords": "", "Action": "Reviewed SOW/Fact-Find and formally lodged application.", "Purpose": "To ensure data accuracy before formal credit assessment."},
+                {"No": "6", "Stage Title": "LENDER QUERY", "Sub 1": "Bank Query Received", "Sub 2": "Query Response Sent", "Keywords": "", "Action": "Received and addressed technical queries from the lender.", "Purpose": "To satisfy credit requirements and progress the file."},
+                {"No": "7", "Stage Title": "CONDITIONAL APPROVAL", "Sub 1": "Conditional approval received", "Sub 2": "Conditional approval provided to client", "Keywords": "Letter of Offer, LOO", "Action": "Reviewed conditional approval and updated the clients.", "Purpose": "To identify and guide clients through remaining conditions."},
+                {"No": "8", "Stage Title": "LOAN STRUCTURE REVIEW", "Sub 1": "Proposed Structure Discussed", "Sub 2": "Final Instruction to Lender", "Keywords": "Statement of Advice, SOA", "Action": "Provided SOA and confirmed final interest rate lock-in.", "Purpose": "To ensure the loan structure meets the client's risk profile."},
+                {"No": "9", "Stage Title": "FINAL APPROVAL", "Sub 1": "Final Approval", "Sub 2": "", "Keywords": "Final, Unconditional", "Action": "Received final approval and instructed the solicitor.", "Purpose": "To confirm all conditions met and initiate legal process."},
+                {"No": "10", "Stage Title": "SETTLEMENT COMPLETION", "Sub 1": "Settlement Completion", "Sub 2": "", "Keywords": "Settlement", "Action": "Coordinated funds and confirmed successful completion.", "Purpose": "To finalize the advice process and transition to service."}
+            ]
+            st.session_state['rules_df'] = pd.DataFrame(data)
+
+    # 1️⃣ [최상단] Admin 전용 통제 구역 (Manager 로그인 시에만 노출)
+    if st.session_state.get("is_admin_mode", False):
+        with st.expander("👑 Admin Control", expanded=False):
+            st.markdown("### 👑 Admin Control")
+            st.caption("Monitor real-time password setups for active team members and revoke access for offboarded employees.")
+            
+            # [🔥 FIXED] Deep copy all nested layers to allow dictionary item deletion without Streamlit restrictions
+            import copy
+            raw_secrets = {k: (dict(v) if hasattr(v, "items") else v) for k, v in st.secrets.items()}
+            all_secrets = copy.deepcopy(raw_secrets)
+            
+            current_users = all_secrets.get("ALLOWED_USERS", {})
+            admin_data = []
+            
+            for u_email, u_pwd in current_users.items():
+                status_tag = "🔴 Pending Setup" if u_pwd == "" else "🟢 Active (Password Set)"
+                
+                # [🔒 SECURITY MASKING] Convert plain text passwords (e.g., "dong9ri1" -> "d*******")
+                if u_pwd and len(u_pwd) > 1:
+                    masked_pwd = u_pwd[0] + "*" * (len(u_pwd) - 1)
+                elif u_pwd:
+                    masked_pwd = "*"
+                else:
+                    masked_pwd = "(None)"
+                    
+                admin_data.append({
+                    "Employee Email": u_email, 
+                    "Recorded Password": masked_pwd,  # Displays secure masked version on screen
+                    "Status": status_tag
+                })
+            
+            st.table(pd.DataFrame(admin_data))
+            
+            st.markdown("**Deactivate & Revoke User Permissions**")
+            target_del = st.selectbox("Select email address to revoke access", ["-"] + list(current_users.keys()))
+            
+            if target_del != "-" and st.button("🚨 Revoke Access"):
+                if target_del in all_secrets["ALLOWED_USERS"]:
+                    # Successfully deletes the user without throwing __delitem__ error
+                    del all_secrets["ALLOWED_USERS"][target_del]
+                    if save_secrets_to_file(all_secrets):
+                        st.success(f"⚠️ Access permanently revoked for {target_del}. This user can no longer sign in.")
+                        st.rerun()
+        
+        # (2) 규칙 편집 관리 (Rules Configuration - Admin Control 바로 밑으로 배치)
+        with st.expander("⚙️ Rules Configuration", expanded=False):
+            st.caption("📝 웹 화면에서 수정·추가하면 시스템 파일에 영구 저장되어 연동됩니다.")
+            edited_rules = st.data_editor(st.session_state['rules_df'], num_rows="dynamic", use_container_width=True, hide_index=True, key="main_rules_editor")
+            if not edited_rules.equals(st.session_state['rules_df']):
+                st.session_state['rules_df'] = edited_rules
+                # 하드디스크 JSON 파일로 덮어쓰기 박제 (영구 저장의 핵심)
+                edited_rules.to_json(RULES_FILE, orient='records', force_ascii=False, indent=4)
+                st.toast("대출 매칭 규칙이 영구 저장되었습니다! 💾", icon="✅")
+        
+        st.divider()
+
+    # Compliance Rules Data
+    if 'rules_df' not in st.session_state:
+        try:
+            st.session_state['rules_df'] = pd.read_excel('rules.xlsx')
+        except:
+            st.session_state['rules_df'] = pd.DataFrame(columns=['Stage Title', 'Sub 1', 'Sub 2', 'Keyword 1', 'Keyword 2'])
+
+    if 'status' not in st.session_state:
+        st.session_state['status'] = ["⚪"] * len(st.session_state['rules_df'])
+    # --- Helper Functions ---
+    def add_line_to_docx(paragraph):
+        p_ptr = paragraph._p.get_or_add_pPr()
+        p_bdr = OxmlElement('w:pBdr')
+        bottom = OxmlElement('w:bottom')
+        bottom.set(qn('w:val'), 'single'); bottom.set(qn('w:sz'), '4')
+        bottom.set(qn('w:space'), '1'); bottom.set(qn('w:color'), 'E2E8F0')
+        p_bdr.append(bottom); p_ptr.append(p_bdr)
+
+    def load_docs_from_path(path):
+        combined_text = ""
+        path = path.strip().replace('"', '')
+        if not os.path.exists(path): return None, f"Path not found: {path}"
+        files = [f for f in os.listdir(path) if f.lower().endswith(('.pdf', '.docx'))]
+        if not files: return None, "No reference docs found in the folder."
+        for f_name in files:
+            f_path = os.path.join(path, f_name)
+            try:
+                if f_name.lower().endswith('.pdf'):
+                    with pdfplumber.open(f_path) as pdf:
+                        for page in pdf.pages: combined_text += (page.extract_text() or "") + "\n"
+                elif f_name.lower().endswith('.docx'):
+                    doc = Document(f_path)
+                    for para in doc.paragraphs: combined_text += para.text + "\n"
+            except: continue
+        return combined_text, f"✅ Analyzed {len(files)} reference document(s)."
+
+   # 🟢 [콜백 삭제] 악명 높은 에러를 일으키는 assign_callback은 아예 삭제합니다!
+
+    # 🟢 메인 연산 함수: 상태 리스트를 '새 리스트 통째로 덮어쓰기'로 깔끔하게 유지
+    def update_status():
+        if st.session_state.get('results') is None: return
+        
+        df = st.session_state['results']
+        rules = st.session_state['rules_df']
+        rules_list = rules.to_dict('records')
+        
+        new_status = ["⚪"] * len(rules_list)
+        
+        if st.session_state.get('active_mode') == "NOTION":
+            assigned = [re.sub(r'^\d+\.\s*', '', str(x)).lower().strip() for x in df[df['Type'] == "NOTION"]['Sec'].tolist()]
+            
+            for idx, r in enumerate(rules_list):
+                stage_val = str(r.get('Stage Title', '')).strip().lower()
+                s1 = str(r.get('Sub 1', '')).strip().lower()
+                s2 = str(r.get('Sub 2', '')).strip().lower()
+                
+                has_stage = stage_val not in ["", "nan", "x", "none", "-"]
+                has_s1 = s1 not in ["", "nan", "x", "none", "-"]
+                has_s2 = s2 not in ["", "nan", "x", "none", "-"]
+                
+                is_stage_mapped = (stage_val in assigned) if has_stage else False
+                is_s1_mapped = (s1 in assigned) if has_s1 else False
+                is_s2_mapped = (s2 in assigned) if has_s2 else False
+
+                if is_stage_mapped:
+                    new_status[idx] = "🟢"
+                else:
+                    if has_s1 and not has_s2:
+                        new_status[idx] = "🟢" if is_s1_mapped else "⚪"
+                    elif has_s1 and has_s2:
+                        if is_s1_mapped and is_s2_mapped: new_status[idx] = "🟢"
+                        elif is_s1_mapped or is_s2_mapped: new_status[idx] = "🟠"
+                        else: new_status[idx] = "⚪"
+                    else:
+                        new_status[idx] = "⚪"
+        else:
+            assigned = [str(x).lower().strip() for x in df[df['Type'] != "NOTION"]['Sec'].tolist()]
+            
+            for idx, r in enumerate(rules_list):
+                s1 = str(r.get('Sub 1', '')).strip().lower()
+                s2 = str(r.get('Sub 2', '')).strip().lower()
+                has_s1 = s1 not in ["", "nan", "x", "None", "-"]
+                has_s2 = s2 not in ["", "nan", "x", "None", "-"]
+                
+                is_s1_mapped = (s1 in assigned) if has_s1 else False
+                is_s2_mapped = (s2 in assigned) if has_s2 else False
+
+                if has_s1 and not has_s2:
+                    new_status[idx] = "🟢" if is_s1_mapped else "⚪"
+                elif has_s1 and has_s2:
+                    if is_s1_mapped and is_s2_mapped: new_status[idx] = "🟢"
+                    elif is_s1_mapped or is_s2_mapped: new_status[idx] = "🟠"
+                    else: new_status[idx] = "⚪"
+                else:
+                    new_status[idx] = "⚪"
+                    
+        st.session_state['status'] = new_status
+    # --- Helper Functions ---
+    # (기존 add_line_to_docx, load_docs_from_path, update_status 등은 그대로 유지)
+
+    def ask_ai_note(stage_key, items_data, row_info, manual_note=""):
+        # 🚨 [파이썬 강력 통제 1] 매핑된 이메일 데이터가 단 한 개도 없고, 추가 수기 메모도 완전히 비어있다면
+        # AI 호출 자체를 생략하고 즉시 완벽한 3줄 공란을 리턴하여 비용과 오류를 원천 차단합니다.
+        if not items_data and not manual_note.strip(): 
+            return "Date: \nAction: \nPurpose: \n"
+
+        try:
+            #vertexai.init(project=st.secrets["PROJECT_ID"], location="us-west1")
+            from vertexai.generative_models import GenerativeModel
+            model = GenerativeModel('gemini-2.5-flash')
+            
+            # 디지털 이메일/노션 소스 로그 포맷팅
+            context_str = ""
+            for i, m in enumerate(items_data):
+                record_type = m.get('Type', 'EMAIL')
+                d = m.get('Date', 'N/A')
+                b = m.get('Body', '')[:1500]
+                
+                if record_type == "NOTION":
+                    ch = m.get('Channel', '-')
+                    cm = m.get('Comments', '')
+                    context_str += f"Record {i+1} [Type: NOTION Task]\nDate: {d}\nChannel: {ch}\nDescription: {b}\nComments: {cm}\n\n"
+                else:
+                    context_str += f"Record {i+1} [Type: EMAIL]\nDate: {d}\nBody: {b}\n\n"
+        
+            # 규칙 데이터 안전 변환
+            rule_dict = dict(row_info) if row_info is not None else {}
+            stage_title = rule_dict.get('Stage Title', stage_key)
+            rule_action = rule_dict.get('Action', "").strip()
+            rule_purpose = rule_dict.get('Purpose', "").strip()
+
+            email_count = len(items_data)
+
+            # 🚨 [파이썬 강력 통제 2] 만약 디지털 매핑 데이터는 아예 없는데 '수기 메모'만 존재하는 경우,
+            # 이 수기 메모가 현재 순회 중인 단계와 연관이 있는 단어(예: 세미나, 첫 미팅 등)를 포함하는지 하드코딩으로 1차 검증합니다.
+            # 1번 INITIAL MEETING이 아닌 다른 단계에 미팅 정보가 침범하는 것을 완벽하게 막아냅니다.
+            if email_count == 0 and manual_note.strip():
+                lower_note = manual_note.lower()
+                # 1번 초기 미팅과 관련된 핵심 단어들 리스트
+                initial_keywords = ["initial", "meeting", "seminar", "remax", "처음", "만났", "미팅", "신분증", "얼굴"]
+                
+                # 현재 단계가 1번 초기 미팅이 '아닌데', 수기 메모 내용이 초기 미팅 관련 내용뿐이라면 즉시 공란 리턴하고 차단
+                if "1." not in stage_key and any(kw in lower_note for kw in initial_keywords):
+                    return "Date: \nAction: \nPurpose: \n"
+
+            # 🚨 [프롬프트 가드레일 개조]
+            prompt = f"""
+            You are a strict compliance auditor representing the New Zealand Financial Markets Authority (FMA).
+            Your job is to draft a professional File Note ONLY for the specific mortgage process stage: [{stage_title}]
+
+            [STRICT COMPLIANCE FILTERING MATRIX]
+            - Current Target Stage: {stage_title}
+            - Authorized Compliance Actions: {rule_action}
+            - Strategic Compliance Objective: {rule_purpose}
+
+            [REAL CLIENT DATA SETS (GROUND TRUTH)]
+            - Number of Digital Logs Matched: {email_count}
+            - Digital Logs (Matched Emails/Notion): {context_str if items_data else "NO DIGITAL LOGS FOR THIS STAGE"}
+            - User's Additional Memo: "{manual_note if manual_note.strip() else "NONE PROVIDED"}"
+
+            [⚠️ CRITICAL GENERATION RULES - DO NOT VIOLATE]
+            1. MANDATORY OVERRIDE FOR MATCHED LOGS: If 'Number of Digital Logs Matched' is 1 or more, it means the user has manually verified and mapped these emails to [{stage_title}]. You MUST trust this mapping and generate a summary based on these logs.
+            2. NO DATA = COMPLETE BLANK: If the provided digital logs are empty, and the User's Additional Memo does not explicitly describe a unique event for [{stage_title}], you are strictly forbidden from writing ANYTHING. You MUST output exactly this 3-line block and STOP. Do NOT copy the 'Strategic Compliance Objective' text into Purpose if there is no matching fact:
+               Date: 
+               Action: 
+               Purpose: 
+            3. NEVER COPY TEMPLATE TEXT: Do not copy-paste or mimic the phrase "{rule_action}" or "{rule_purpose}" into the note. Always write unique, professional descriptions based on the client logs.
+
+            [🚨 CRITICAL FORMAT RULE: NUMBER OF ITEMS MATCHING]
+            - CASE A (Single Source): If there is only 1 email/record matched, OR only the User's Memo is valid for this stage, output EXACTLY ONE pair of "Date:" and "Action:". (Do NOT use numbers like Date 1 or Action 1).
+            - CASE B (Multiple Sources): If there are 2 or more distinct emails/records matched, group them chronologically and use "Date 1:", "Action 1:", "Date 2:", "Action 2:" etc.
+
+            [PROPER FORMAT EXAMPLES]
+            Example for CASE A:
+            Date: 15 May 2026, 10:32 AM
+            Action: Sent an email to the client confirming the next steps of disclosure and requested verification docs.
+
+            Purpose: Written custom compliant rationale based on the real event.
+            """
+            
+            response = model.generate_content(prompt)
+            res_text = response.text.strip()
+            
+            if "Date: \nAction: \nPurpose:" in res_text or not res_text:
+                return "Date: \nAction: \nPurpose: \n"
+                
+            return res_text
+            
+        except Exception as e:
+            return f"⚠️ Vertex AI Error: {str(e)}"
+
+
+
+    # ─────────────────────────────────────────────────────────────────
+    # 🟢 [실시간 동기화 마법의 엔진] 사이드바를 그리기 전에 드롭다운 값을 먼저 가로채서 계산!
+    # ─────────────────────────────────────────────────────────────────
+    def sync_and_update_status():
+        if st.session_state.get('results') is None: return
+        
+        # 1. 화면 맨 밑의 드롭다운(selectbox) 최신 값을 미리 가로채서 데이터프레임에 덮어씀
+        df = st.session_state['results']
+        for row_idx in df.index:
+            sel_key = f"sel_{row_idx}"
+            if sel_key in st.session_state:
+                df.at[row_idx, 'Sec'] = st.session_state[sel_key]
+                
+        # 2. 전구(status) 즉시 계산
+        rules = st.session_state['rules_df']
+        rules_list = rules.to_dict('records')
+        new_status = ["⚪"] * len(rules_list)
+        
+        if st.session_state.get('active_mode') == "NOTION":
+            assigned = [re.sub(r'^\d+\.\s*', '', str(x)).lower().strip() for x in df[df['Type'] == "NOTION"]['Sec'].tolist()]
+            for idx, r in enumerate(rules_list):
+                stage_val = str(r.get('Stage Title', '')).strip().lower()
+                s1 = str(r.get('Sub 1', '')).strip().lower()
+                s2 = str(r.get('Sub 2', '')).strip().lower()
+                has_stage = stage_val not in ["", "nan", "x", "none", "-"]
+                has_s1 = s1 not in ["", "nan", "x", "none", "-"]
+                has_s2 = s2 not in ["", "nan", "x", "none", "-"]
+                
+                is_stage_mapped = (stage_val in assigned) if has_stage else False
+                is_s1_mapped = (s1 in assigned) if has_s1 else False
+                is_s2_mapped = (s2 in assigned) if has_s2 else False
+
+                if is_stage_mapped: new_status[idx] = "🟢"
+                else:
+                    if has_s1 and not has_s2: new_status[idx] = "🟢" if is_s1_mapped else "⚪"
+                    elif has_s1 and has_s2:
+                        if is_s1_mapped and is_s2_mapped: new_status[idx] = "🟢"
+                        elif is_s1_mapped or is_s2_mapped: new_status[idx] = "🟠"
+                        else: new_status[idx] = "⚪"
+                    else: new_status[idx] = "⚪"
+        else:
+            assigned = [str(x).lower().strip() for x in df[df['Type'] != "NOTION"]['Sec'].tolist()] if 'Type' in df.columns else [str(x).lower().strip() for x in df['Sec'].tolist()]
+            for idx, r in enumerate(rules_list):
+                s1 = str(r.get('Sub 1', '')).strip().lower()
+                s2 = str(r.get('Sub 2', '')).strip().lower()
+                has_s1 = s1 not in ["", "nan", "x", "None", "-"]
+                has_s2 = s2 not in ["", "nan", "x", "None", "-"]
+                
+                is_s1_mapped = (s1 in assigned) if has_s1 else False
+                is_s2_mapped = (s2 in assigned) if has_s2 else False
+
+                if has_s1 and not has_s2: new_status[idx] = "🟢" if is_s1_mapped else "⚪"
+                elif has_s1 and has_s2:
+                    if is_s1_mapped and is_s2_mapped: new_status[idx] = "🟢"
+                    elif is_s1_mapped or is_s2_mapped: new_status[idx] = "🟠"
+                    else: new_status[idx] = "⚪"
+                else: new_status[idx] = "⚪"
+                
+        st.session_state['status'] = new_status
+
+    # 🚀 사이드바를 그리기 직전에 이 엔진을 무조건 가동시킵니다!
+    sync_and_update_status()
+
+    # --- [Sidebar Area: 통합 워드 스타일 에디터] ---
+    with st.sidebar:
+        
+        # [1] Progress Status (상단 고정)
+        st.markdown("### 📊 Progress Status")
+        for row_idx in range(0, 10, 5):
+            cols = st.columns(5)
+            for i in range(5):
+                idx = row_idx + i
+                if idx < len(st.session_state['status']):
+                    with cols[i]:
+                        icon = st.session_state['status'][idx]
+                        dt = "🟢" if icon == "🟢" else ("🟠" if icon == "🟠" else "⚪")
+                        full_title = st.session_state['rules_df'].iloc[idx]['Stage Title']
+                        st.markdown(
+                            f"""<div style='text-align: center; line-height: 1.2; margin-bottom: 12px; min-height: 60px;'>
+                                <div style='font-size: 1.1rem;'>{dt}</div>
+                                <div style='font-size: 0.75rem; font-weight: bold; color: #1E3A8A;'>{idx+1}</div>
+                                <div style='font-size: 0.65rem; color: #334155; word-break: keep-all; white-space: normal;'>{full_title}</div>
+                            </div>""", unsafe_allow_html=True)
+
+        # ✨ [가이드 안내 문구 추가] 전구 상태 설명 및 AI 실행 안내 텍스트 이식
+        st.markdown("""
+            <div style='font-size: 0.73rem; color: #64748B; background-color: #F8FAFC; border: 1px solid #E2E8F0; padding: 10px 12px; border-radius: 6px; margin-top: 2px; margin-bottom: 18px; line-height: 1.5;'>
+                ⚪ <b>Unmapped</b> &nbsp;|&nbsp; 
+                🟠 <b>Partially Mapped (1/2)</b> &nbsp;|&nbsp; 
+                🟢 <b>Fully Mapped</b><br>
+                <div style='margin-top: 5px; color: #334155; font-weight: 600;'></div>
+            </div>
+        """, unsafe_allow_html=True)
+        
+        # [수정 핵심] 변수 정의를 if문 밖으로 꺼내어 에러 방지
+        display_client = st.session_state.get('client_names', '').replace(';', ' & ').strip()
+        support_person = st.session_state.get('support_name', 'Edward Lee (BDM)')
+
+        # [2] 작업 버튼 영역 (Sync 완료 시에만 나타남)
+        if st.session_state.get('results') is not None:
+            
+            # 🟢 [수정 2] UI 명칭 변경 및 프롬프트 역할 강조
+            ai_manual_note = st.text_area(
+                "💡 AI Additional Prompt (Optional)", 
+                placeholder="AI에게 전달할 지시사항 (예: 전체 수집된 이메일 내역을 보고 Adviser note 알아서 생성 등)", 
+                height=100
+            )
+
+            # 🟢 1. 현재 작업 중인 탭이 '이메일'인지 확인합니다.
+            current_mode = st.session_state.get('active_mode', 'NOTION')
+            is_email = (current_mode == "EMAIL")
+
+            # 🟢 2. 이메일 탭일 경우 체크박스를 강제로 끄고(False), 클릭 못 하게 잠급니다(disabled).
+            use_ai = st.checkbox(
+                "🤖 AI로 자동 작성하기 (API 비용 발생)", 
+                value=False, 
+                disabled=is_email, 
+                help="🚫 이메일 모드에서는 AI 요약이 제한되며 빈 템플릿만 생성됩니다." if is_email else "체크를 끄면 비용 없이 날짜만 추출된 빈 템플릿이 즉시 생성됩니다."
+            )
+
+            col_ai, col_dl = st.columns(2)
+
+            with col_ai:
+
+                if st.button("✨ Run AI Summary", use_container_width=True):
+                    with st.spinner("Adviser AI Generating Report..."):
+                        
+                        # 🔄 [실시간 탭 감지 엔진] 현재 활성화된 모드에 따라 화면 입력값을 다이렉트로 추적
+                        current_mode = st.session_state.get('active_mode', 'NOTION')
+                        
+                        if current_mode == "NOTION":
+                            # 노션 탭 입력창 값 가져와서 세미콜론 치환
+                            raw_name = st.session_state.get('notion_client_name', '').strip()
+                            final_client_name = raw_name.replace(';', ' & ') if raw_name else "N/A"
+                        else:
+                            # 이메일 탭 입력창 값 가져와서 세미콜론 치환
+                            raw_name = st.session_state.get('client_names', '').strip()
+                            final_client_name = raw_name.replace(';', ' & ') if raw_name else "N/A"
+                            
+                        # 전역 변수에도 최신 상태 최종 박제 (워드 파일명 연동용)
+                        st.session_state['global_client_name'] = final_client_name
+
+                        # 상단 메타 헤더 정보 생성 (동적 추출된 이름 주입)
+                        combined_content = f"ADVISER FILE NOTES LOG\n\nClient Name: {final_client_name}\nAdviser: Tim Park\nSupport: {st.session_state.get('support_name', 'N/A')}\nStatus: Completed / Final Approval Issued\n"
+                        combined_content += "="*67 + "\n"
+
+                        full_txt = st.session_state.get('full_note', '').strip()
+                        if full_txt:
+                            doc_bio = io.BytesIO()
+                            final_doc = Document()
+
+                        # [기존 기능 유지] 상단 메타 헤더 정보 생성
+                        combined_content = f"ADVISER FILE NOTES LOG\n\nClient Name: {final_client_name}\nAdviser: Tim Park\nSupport: {st.session_state.get('support_name', 'N/A')}\nStatus: Completed / Final Approval Issued\n"
+                        combined_content += "="*67 + "\n"
+                        
+                        # [기존 기능 유지] 1. 화면상의 수동 매핑 상태(드롭다운 값)를 세션 데이터프레임에 동기화
+                        if st.session_state.get('results') is not None:
+                            for row_idx in st.session_state['results'].index:
+                                sel_key = f"sel_{row_idx}"
+                                if sel_key in st.session_state:
+                                    st.session_state['results'].at[row_idx, 'Sec'] = st.session_state[sel_key]
+                        
+                        # [기존 기능 유지] 2. 현재 작업 모드(EMAIL 또는 NOTION)에 맞춰 소스 데이터 필터링
+                        full_df = st.session_state['results'].copy() if st.session_state.get('results') is not None else pd.DataFrame()
+                        if not full_df.empty:
+                            if st.session_state.get('active_mode') == "NOTION":
+                                df = full_df[full_df['Type'] == "NOTION"].copy()
+                            else:
+                                df = full_df[full_df['Type'] != "NOTION"].copy()
+                        else:
+                            df = pd.DataFrame(columns=['Sec', 'Type', 'Date', 'Body', 'Subject'])
+                        
+                        # [기존 기능 유지] 실시간 규칙 세션 데이터셋 불러오기
+                        rules_list = st.session_state['rules_df'].to_dict('records')
+                        
+                        # 🚨 [동적 대응 업데이트] 고정된 10개가 아니라 웹에서 추가/삭제된 모든 단계를 실시간으로 유연하게 순회
+                        for row in rules_list:
+                            # 항목이 웹에서 동적으로 추가되더라도 에러가 나지 않도록 안전하게 변수 추출
+                            r_no = row.get('No', '-')
+                            r_title = row.get('Stage Title', 'Unknown Stage')
+                            sk = f"{r_no}. {r_title}"
+                            
+                            sub1 = str(row.get('Sub 1', '')).strip().lower()
+                            sub2 = str(row.get('Sub 2', '')).strip().lower()
+                            stage_title_lower = str(r_title).strip().lower()
+                            
+                            # 🟢 [수정] 노션 데이터의 번호표("02. ", "03. " 등)를 떼어내고 정규화
+                            temp_df = df.copy()
+                            temp_df['Sec_Clean'] = temp_df['Sec'].astype(str).str.replace(r'^\d+\.\s*', '', regex=True).str.strip().str.lower()
+                            
+                            # 🟢 [핵심] 수동 매핑(Sub 1, 2)과 노션 자동 매핑(Stage Title)을 모두 인정해 줍니다!
+                            valid_targets = [sub1, sub2, stage_title_lower]
+                            valid_targets = [x for x in valid_targets if x not in ["", "nan", "x", "none", "-"]]
+                            
+                            targets = temp_df[temp_df['Sec_Clean'].isin(valid_targets)] if not temp_df.empty else pd.DataFrame()
+
+                            combined_content += f"\n[{sk}]\n"
+                            
+                            # [기존 기능 유지] 매핑된 이메일이 있거나, 혹은 이메일이 없더라도 사용자가 AI Additional Prompt를 적은 경우에만 AI 작동
+                            if not targets.empty or ai_manual_note.strip():
+                                e_data = targets.to_dict('records') if not targets.empty else []
+                                
+                                # [기존 기능 유지] Ambiguous 에러 방지를 위해 row를 dict형으로 안전하게 변환하여 AI 호출
+                                note_text = ask_ai_note(sk, e_data, dict(row), ai_manual_note)
+                                combined_content += f"{note_text}\n\n"
+                            else:
+                                # [기존 기능 유지] 매핑된 이메일도 없고 추가 프롬프트 내용도 없는 구역은 깨끗하게 공란 폼만 출력
+                                combined_content += "Date: \nAction: \nPurpose: \n\n"
+                        
+                        # [기존 기능 유지] 최종 결과물을 사이드바 에디터 세션에 주입하고 화면 리프레시
+                        st.session_state['full_note'] = combined_content
+                        st.session_state['editor_version'] = datetime.now().strftime("%H%M%S")
+                        st.rerun()
+
+
+            with col_dl:
+
+                # 🟢 전역 변수로 확정된 이름을 가져옵니다.
+                current_mode = st.session_state.get('active_mode', 'NOTION')
+
+                if current_mode == "NOTION":
+                    raw_name = st.session_state.get('notion_client_name', '').strip()
+                    final_client_name = raw_name.replace(';', ' & ') if raw_name else "N/A"
+                else:
+                    raw_name = st.session_state.get('client_names', '').strip()
+                    final_client_name = raw_name.replace(';', ' & ') if raw_name else "N/A"
+                
+                full_txt = st.session_state.get('full_note', '').strip()
+                if full_txt:
+                    doc_bio = io.BytesIO()
+                    final_doc = Document()
+                    
+                    # 워드 문서 전체 기본 폰트 설정
+                    style = final_doc.styles['Normal']
+                    font = style.font
+                    font.name = 'Arial'
+                    font.size = Pt(11)
+
+                    lines = full_txt.split('\n')
+                    for i, line in enumerate(lines):
+                        # 1. 문서 제목 (가운데 정렬, 굵게, 크기 키움)
+                        if line.strip() == "ADVISER FILE NOTES LOG":
+                            p = final_doc.add_paragraph()
+                            p.alignment = WD_ALIGN_PARAGRAPH.CENTER
+                            run = p.add_run(line.strip())
+                            run.bold = True
+                            run.font.size = Pt(16)
+                            continue
+
+                        # 2. 메타 정보 (이름, 어드바이저 등 항목 이름만 굵게 처리)
+                        is_meta = False
+                        for meta_key in ["Client Name:", "Adviser:", "Support:", "Status:"]:
+                            if line.startswith(meta_key):
+                                p = final_doc.add_paragraph()
+                                parts = line.split(meta_key, 1)
+                                run_bold = p.add_run(meta_key)
+                                run_bold.bold = True
+                                if len(parts) > 1:
+                                    p.add_run(parts[1])
+                                is_meta = True
+                                break
+                        if is_meta:
+                            continue
+
+                        # 3. 최상단 실선 구분선 유지 (====)
+                        if line.startswith('===='):
+                            p_line = final_doc.add_paragraph(line)
+                            p_line.paragraph_format.space_after = Pt(0) 
+                            continue
+
+                        # 4. 각 섹션 제목 ([1. INITIAL MEETING] 등 파란색, 굵게)
+                        if line.strip().startswith('[') and line.strip().endswith(']'):
+                            if i > 10: 
+                                p_dash = final_doc.add_paragraph()
+                                p_dash.add_run("-" * 117)
+
+                            p = final_doc.add_paragraph()
+                            run = p.add_run(line.strip())
+                            run.bold = True
+                            run.font.color.rgb = RGBColor(30, 58, 138)
+                            run.font.size = Pt(12)
+                            continue
+
+                        # 5. 일반 텍스트 내 Date, Action, Purpose 라벨만 진하게 처리하는 구역
+                        stripped_line = line.strip()
+                        is_label_line = False
+                        
+                        for target_label in ["Date", "Action", "Purpose"]:
+                            if stripped_line.startswith(target_label):
+                                p = final_doc.add_paragraph()
+                                if ":" in stripped_line:
+                                    label_part, content_part = stripped_line.split(":", 1)
+                                    run_label = p.add_run(f"{label_part}:")
+                                    run_label.bold = True
+                                    p.add_run(content_part)
+                                else:
+                                    run_all = p.add_run(stripped_line)
+                                    run_all.bold = True
+                                is_label_line = True
+                                break
+                        
+                        if is_label_line:
+                            continue
+
+                        final_doc.add_paragraph(line)
+
+                    final_doc.save(doc_bio)
+                    
+                    # 🛠️ 수정: 안전하게 동기화된 final_client_name으로 파일명 빌드 및 다운로드 활성화
+                    st.download_button("💾 Save Word", data=doc_bio.getvalue(), file_name=f"Adviser_File_Notes_Log_{final_client_name}.docx", use_container_width=True)
+                else:
+                    st.button("💾 Save Word", use_container_width=True, disabled=True, help="Run AI Summary first")
+            st.divider()
+
+        # [3] 통합 텍스트 에디터 (중복 박스 제거, 텍스트 창만 유지)
+        if 'full_note' not in st.session_state: st.session_state['full_note'] = ""
+        ver = st.session_state.get('editor_version', 'default')
+
+        st.session_state['full_note'] = st.text_area(
+            label="Integrated Note Editor",
+            value=st.session_state['full_note'],
+            height=800, 
+            key=f"final_editor_{ver}", 
+            label_visibility="collapsed"
+        )
+    # --- [Main Area: Configuration] ---
+    st.title("📑 Flux Finance Internal Note")
+
+    # 💡 [세션 동기화 보강] 앱 초기 구동 시 액티브 모드가 비어있으면 기본값으로 이메일을 지정하여 락을 해제합니다.
+    if 'active_mode' not in st.session_state:
+        st.session_state['active_mode'] = "Notion"
+
+    # 💡 [탭 순서 혁신] 노션 탭을 배열의 맨 앞으로 배치하여 웹 접속 시 첫 화면에 무조건 노션 입력창이 먼저 뜨도록 제어합니다.
+    tab_notion, tab_email = st.tabs(["📝 Pull Notion", "✉️ Scan Emails"])
+
+    # TAB 1: 노션 검색창
+    with tab_notion:
+        with st.container(border=True):
+            st.caption("💡 **Notion Sync:** Pull matching compliance tracks and thread messages.")
+            c_notion1, c_notion2 = st.columns([5, 1])
+            search_client = c_notion1.text_input("👤 Case Name (Applicants Name)", value=st.session_state['notion_client_name'], key="notion_search_bar_fixed_tab")
+            st.session_state['notion_client_name'] = search_client
+            notion_sync_btn = c_notion2.button("🔍 Pull", use_container_width=True, key="notion_sync_trigger")
+            
+            if notion_sync_btn and search_client:
+                st.session_state['active_mode'] = "NOTION"
+                st.session_state['global_client_name'] = search_client.replace(';', ' & ').strip()
+
+                with st.status("🔍 Syncing Notion Workspace...", expanded=True) as status:
+                    # 🟢 리턴받은 case_comments를 세션에 저장
+                    tasks_data, msg, case_comments = fetch_client_tasks_from_notion(search_client)
+                    
+                    if tasks_data:
+                        st.session_state['results'] = pd.DataFrame(tasks_data)
+                        st.session_state['results']['Type'] = "NOTION"
+                        st.session_state['case_comments'] = case_comments # 코멘트 세션 보관
+                        
+                        update_status() 
+                        status.update(label="✅ Notion Tasks Synced.", state="complete")
+                        st.rerun()
+
+                    elif tasks_data is not None and not tasks_data:
+                        status.update(label="⚠️ 매칭 데이터 없음", state="error")
+                    else:
+                        status.update(label=f"❌ Failed: {msg}", state="error")
+    # TAB 2: 이메일 검색창
+    with tab_email:                    
+        with st.container(border=True):
+            
+            st.session_state['folder_path'] = st.text_input("📂 G: Drive Folder Path", value=st.session_state['folder_path'], key="fixed_mail_path")
+            
+            # 🎯 [추가] secrets에서 주소만 쏙 발라내어 멀티 셀렉트 박스 생성
+            raw_mailbox_list = st.secrets["company"].get("ALL_MAILBOXES", [])
+            mailbox_options = []
+            mailbox_map = {}
+            for entry in raw_mailbox_list:
+                entry_clean = entry.strip()
+                if not entry_clean: continue
+                cred_part = entry_clean.split("|")[0] if "|" in entry_clean else entry_clean
+                clean_email = cred_part.split(":")[0].strip() if ":" in cred_part else cred_part.strip()
+                mailbox_options.append(clean_email)
+                mailbox_map[clean_email] = entry_clean
+            
+            selected_boxes = st.multiselect("📬 Select Mailboxes to Scan", options=mailbox_options, default=[], key="active_mailboxes_select")
+            
+            
+            c1, c2, c3, c4 = st.columns([2, 2, 1.5, 1])
+            st.session_state['emails'] = c1.text_input("Emails", value=st.session_state['emails'], key="fixed_mail_addrs")
+            st.session_state['client_names'] = c2.text_input("Names", value=st.session_state['client_names'], key="fixed_mail_names")
+            st.session_state['refs'] = c3.text_input("Keywords", value=st.session_state['refs'], key="fixed_mail_refs")
+            
+            email_sync_btn = c4.button("🚀 Sync Emails", use_container_width=True, key="email_sync_trigger")
+            # ✨ [디자인 혁신] 투박한 안내 문구를 엔터프라이즈급 가이드 대시보드로 변경
+            st.markdown("""
+                <div style='background-color: #F8FAFC; border: 1px solid #E2E8F0; padding: 14px 18px; border-radius: 8px; margin-top: 15px;'>
+                    <div style='font-size: 0.85rem; font-weight: 700; color: #1E3A8A; margin-bottom: 4px;'>💡 System Usage Instructions</div>
+                    <div style='font-size: 0.8rem; color: #475569; margin-bottom: 12px;'>
+                        Use a semicolon (;) to search multiple items.</b>.
+                    </div>
+                    <div style='display: grid; grid-template-columns: repeat(3, 1fr); gap: 16px; border-top: 1px solid #E2E8F0; padding-top: 12px;'>
+                        <div>
+                            <span style='font-size: 0.72rem; font-weight: 800; background-color: #DBEAFE; color: #1E40AF; padding: 3px 6px; border-radius: 4px; letter-spacing: 0.5px;'>EMAILS SCOPE</span>
+                            <div style='font-size: 0.78rem; color: #64748B; margin-top: 6px; line-height: 1.4;'>Scans <b>From / To / CC</b> </div>
+                        </div>
+                        <div>
+                            <span style='font-size: 0.72rem; font-weight: 800; background-color: #E0F2FE; color: #0369A1; padding: 3px 6px; border-radius: 4px; letter-spacing: 0.5px;'>NAMES SCOPE</span>
+                            <div style='font-size: 0.78rem; color: #64748B; margin-top: 6px; line-height: 1.4;'>Scans <b>From / To / CC / Subject</b>.<br><span style='font-size: 0.72rem; color: #94A3B8;'>※ Order-agnostic full AND match</span></div>
+                        </div>
+                        <div>
+                            <span style='font-size: 0.72rem; font-weight: 800; background-color: #F1F5F9; color: #334155; padding: 3px 6px; border-radius: 4px; letter-spacing: 0.5px;'>KEYWORDS SCOPE</span>
+                            <div style='font-size: 0.78rem; color: #64748B; margin-top: 6px; line-height: 1.4;'>Scans <b>Subject / Body</b></div>
+                        </div>
+                    </div>
+                </div>
+            """, unsafe_allow_html=True)
+
+            if email_sync_btn:
+                st.session_state['active_mode'] = "EMAIL"
+                # 🟢 [여기에 추가] 이메일 이름을 전역 변수로 확정 박아버림! (여러 명이면 &로 치환)
+                raw_names = st.session_state.get('client_names', '').strip()
+                st.session_state['global_client_name'] = raw_names.replace(';', ' & ') if raw_names else ''
+
+                with st.status("🔍 Syncing...", expanded=True) as status:
+                    progress_text = st.empty()
+                    progress_text.info("Connecting and scanning emails... ⏳")
+
+                    if st.session_state['folder_path']:
+                        txt, msg = load_docs_from_path(st.session_state['folder_path'])
+                        st.session_state['doc_context'] = txt if txt else ""; st.write(msg)
+
+                    all_mails = []
+                    e_list = [e.strip().lower() for e in st.session_state['emails'].replace(";", ",").split(",") if "@" in e]
+                    name_groups = [[w.strip().lower() for w in n.replace(",", " ").split() if len(w.strip()) > 1] for n in st.session_state['client_names'].split(";")]
+                    ref_list = [r.strip().lower() for r in st.session_state['refs'].replace(";", ",").split(",") if len(r.strip()) > 3]
+
+                    stats = {"count": 0}
+                    # 🎯 [교체 완료] 전체를 다 도는 대신, 멀티 셀렉트에서 선택된 사서함들만 추출
+                    target_mailbox_entries = [mailbox_map[m] for m in selected_boxes if m in mailbox_map]
+
+                    if target_mailbox_entries:
+                        try:
+                            import email.utils
+                            
+                            for entry in target_mailbox_entries:
+                                entry_clean = entry.strip()
+                                if not entry_clean: continue
+                                
+                                if "|" in entry_clean:
+                                    credential_part, imap_host = entry_clean.split("|", 1)
+                                    imap_host = imap_host.replace("|", "").strip()
+                                else:
+                                    credential_part = entry_clean
+                                    imap_host = "arrow.mxrouting.net"
+                                
+                                if ":" in credential_part:
+                                    my_mail_addr, custom_pw = credential_part.split(":", 1)
+                                    my_mail_addr = my_mail_addr.strip()
+                                    target_pw = custom_pw.strip()
+                                else:
+                                    my_mail_addr = credential_part.strip()
+                                    target_pw = COMMON_PW
+
+                                status.write(f"🔐 Scanning Mailbox: '{my_mail_addr}'...")
+                                
+                                try:
+                                    mail = IMAP4_SSL(imap_host, 993)
+                                    mail.login(my_mail_addr, target_pw)
+                                    
+                                    res_list, folder_list = mail.list()
+                                    all_target_folders = []
+                                    
+                                    if res_list == "OK":
+                                        for f_info in folder_list:
+                                            f_str = f_info.decode("utf-8", errors="ignore")
+                                            # 🎯 [하위 폴더 공백/특수문자 파싱 버그 완벽 교정]
+                                            # 폴더 전체 경로를 찢어 먹지 않고 쌍따옴표 구역을 역추적해 통째로 가져옵니다.
+                                            if f_str.endswith('"'):
+                                                last_quote = f_str.rfind('"')
+                                                prev_quote = f_str.rfind('"', 0, last_quote)
+                                                raw_folder_name = f_str[prev_quote+1 : last_quote]
+                                            else:
+                                                raw_folder_name = f_str.split()[-1].strip().strip('"')
+                                                
+                                            if any(k in raw_folder_name.lower() for k in ["junk", "trash", "deleted", "sync", "spam"]):
+                                                continue
+                                            if raw_folder_name:
+                                                all_target_folders.append(raw_folder_name)
+                                    
+                                    if not all_target_folders:
+                                        all_target_folders = ["INBOX", "Sent"]
+
+                                    for folder in all_target_folders:
+                                        try:
+                                            mail.select(f'"{folder}"', readonly=True)
+                                            matched_ids = set()
+                                            search_queries = []
+                                            
+                                            # (A) Applicants Emails 쿼리 후보 수집
+                                            for e_addr in e_list:
+                                                clean_e = e_addr.strip()
+                                                if clean_e:
+                                                    search_queries.append(f'(OR FROM "{clean_e}" (OR TO "{clean_e}" CC "{clean_e}"))')
+
+                                            # (B) Applicants Names 쿼리 후보 수집
+                                            for group in name_groups:
+                                                valid_words = [w.strip().replace(",", "") for w in group if len(w.strip()) > 1]
+                                                for word in valid_words:
+                                                    if word:
+                                                        search_queries.append(f'(OR SUBJECT "{word}" (OR HEADER From "{word}" (OR HEADER To "{word}" HEADER Cc "{word}")))')
+                                            
+                                            # (C) Ref Numbers 쿼리 후보 수집
+                                            for r_num in ref_list:
+                                                clean_r = r_num.strip()
+                                                if clean_r:
+                                                    search_queries.append(f'TEXT "{clean_r}"')
+
+                                            # ⚡ [속도 혁신] 수집된 모든 개별 쿼리를 하나로 묶어 폴더당 단 1번만 서버에 요청 (1분 -> 3초 단축)
+                                            if search_queries:
+                                                combined_query = search_queries[0]
+                                                for q in search_queries[1:]:
+                                                    combined_query = f'(OR {combined_query} {q})'
+                                                    
+                                                res, msg_ids = mail.search(None, combined_query)
+                                                if msg_ids[0]: 
+                                                    matched_ids.update(msg_ids[0].split())
+
+                                            if not matched_ids:
+                                                continue
+                                                
+                                            # 획득한 메일 번호 기반 본문 추출 엔진 정위치 안착
+                                            for m_id in matched_ids:
+                                                stats["count"] += 1
+                                                res_body, body_data = mail.fetch(m_id, "(RFC822)")
+                                                msg_raw = email.message_from_bytes(body_data[0][1])
+                                                
+                                                subj = str(make_header(decode_header(msg_raw.get("Subject", ""))))
+                                                snd = str(make_header(decode_header(msg_raw.get("From", ""))))
+                                                to_header = str(msg_raw.get("To", ""))
+                                                
+                                                # 💡 대소문자 통합 방어막 구축: Cc든 CC, cc든 대소문자 구분 없이 무조건 포획합니다.
+                                                cc_raw = ""
+                                                for header_key in msg_raw.keys():
+                                                    if header_key.lower() == "cc":
+                                                        cc_raw = msg_raw.get(header_key, "")
+                                                        break
+                                                
+                                                cc_header = str(make_header(decode_header(str(cc_raw)))) if cc_raw else ""
+
+                                                body = ""
+                                                if msg_raw.is_multipart():
+                                                    for b_part in msg_raw.walk():
+                                                        if b_part.get_content_type() == "text/plain":
+                                                            try:
+                                                                body = b_part.get_payload(decode=True).decode(errors='ignore')
+                                                                break
+                                                            except: pass
+                                                else:
+                                                    try: body = msg_raw.get_payload(decode=True).decode(errors='ignore')
+                                                    except: body = "본문을 읽어올 수 없습니다."
+                                                    
+                                                attachments = []
+                                                if msg_raw.is_multipart():
+                                                    for b_part in msg_raw.walk():
+                                                        filename = b_part.get_filename()
+                                                        if filename:
+                                                            decoded_filename = str(make_header(decode_header(filename)))
+                                                            attachments.append(decoded_filename)
+                                                            
+                                                dt = email.utils.parsedate_to_datetime(msg_raw.get("Date"))
+                                                
+                                                # 🎯 [타임존 변환 버그 교정] 서버 시간(UTC)을 뉴질랜드(Auckland) 현지 시간으로 정확하게 강제 변환합니다.
+                                                if dt and dt.tzinfo:
+                                                    try:
+                                                        from zoneinfo import ZoneInfo
+                                                        dt = dt.astimezone(ZoneInfo("Pacific/Auckland"))
+                                                    except:
+                                                        dt = dt.astimezone(None)
+
+                                                # 🎯 각 항목별 검색 구역을 완벽하게 분리
+                                                email_area = (snd + " " + to_header + " " + cc_header).lower()
+                                                name_area = (subj + " " + snd + " " + to_header + " " + cc_header).lower()
+                                                ref_area = (subj + " " + body).lower()
+                                                
+                                                # 1. 이메일 주소는 오직 From, To, CC 영역에서만 매칭 (OR)
+                                                email_matched = any(e_addr in email_area for e_addr in e_list) if e_list else False
+                                                
+                                                # 2. 대출 참조 번호는 오직 제목(Subject)과 본문(Body) 영역에서만 매칭 (OR)
+                                                ref_matched = any(r_num in ref_area for r_num in ref_list) if ref_list else False
+                                                
+                                                # 3. 이름은 From, To, CC, Subject 영역에서 검사하되, 성과 이름 단어가 '순서 상관없이 모두(AND)' 존재해야 함!
+                                                name_matched = False
+                                                if name_groups and any(len(g) > 0 for g in name_groups):
+                                                    for group in name_groups:
+                                                        # 예: 'misun'과 'lim'이 이름 검사 영역에 동시에 다 들어있어야만 true! (Misun Kim 차단)
+                                                        if group and all(word in name_area for word in group):
+                                                            name_matched = True
+                                                            break
+                                                            
+                                                # 4. 세 카테고리(이메일 주소 vs 대출번호 vs 이름) 간의 최종 관계는 완전한 OR
+                                                if email_matched or name_matched or ref_matched:
+                                                    all_mails.append({
+                                                        "Date": dt.strftime("%d %b %Y, %I:%M %p"),
+                                                        "Subject": subj,
+                                                        "From": snd,
+                                                        "To": to_header,
+                                                        "Cc": cc_header,
+                                                        "Type": "SENT" if any(dom in snd.lower() for dom in ["fluxfinance.co.nz", "ewfinance.co.nz"]) else "RECEIVED",
+                                                        "Body": body,
+                                                        "Sec": "-",
+                                                        "Attach": attachments,
+                                                        "RawDate": dt.replace(tzinfo=None)
+                                                    })
+                                        except: continue
+                                    mail.logout()
+                                except: pass
+                        except Exception as e:
+                            st.error(f"IMAP Engine Error: {e}")                
+
+                    if all_mails:
+                        all_mails.sort(key=lambda x: x.get("RawDate", datetime.min))
+                        df = pd.DataFrame(all_mails).drop_duplicates(subset=['Subject', 'Date'])
+                        st.session_state['results'] = df.reset_index(drop=True)
+                        
+                        assigned = st.session_state['results']['Sec'].tolist()
+                        for idx, r in st.session_state['rules_df'].iterrows():
+                            s1 = str(r.get('Sub 1', '')).strip()
+                            s2 = str(r.get('Sub 2', '')).strip()
+                            has_s1 = s1 not in ["", "nan", "x", "None", "-"]
+                            has_s2 = s2 not in ["", "nan", "x", "None", "-"]
+                            is_s1_mapped = s1 in assigned if has_s1 else False
+                            is_s2_mapped = s2 in assigned if has_s2 else False
+                            if has_s1 and not has_s2: st.session_state['status'][idx] = "🟢" if is_s1_mapped else "⚪"
+                            elif has_s1 and has_s2:
+                                if is_s1_mapped and is_s2_mapped: st.session_state['status'][idx] = "🟢"
+                                elif is_s1_mapped or is_s2_mapped: st.session_state['status'][idx] = "🟠"
+                                else: st.session_state['status'][idx] = "⚪"
+                            else: st.session_state['status'][idx] = "⚪"
+                        
+                        status.update(label=f"✅ Scan complete! Found {len(all_mails)} matching email(s).", state="complete", expanded=False)
+                    else:
+                        st.session_state['results'] = None
+                        progress_text.error(f"Checked {stats['count']} emails. No matches found.")
+
+ # ─────────────────────────────────────────────────────────────────
+# 🎯 [4단계] 통합 결과창 드로잉 레이아웃
+# ─────────────────────────────────────────────────────────────────
+if st.session_state['results'] is not None:
+    update_status()
+    for idx, row in st.session_state['results'].iterrows():
+        # 모드 필터링
+        if st.session_state['active_mode'] == "NOTION" and row.get('Type') != "NOTION": continue
+        if st.session_state['active_mode'] == "EMAIL" and row.get('Type') == "NOTION": continue
+        
+        with st.container(border=True):
+            # [1] 타입별 상단 헤더 출력 (유지)
+            if row['Type'] == "NOTION":
+                st.markdown(f"<span class='type-tag notion-tag'></span> <b>📌 Stage: {row['Sec']}</b>",
+                    unsafe_allow_html=True
+                )
+                # 🟢 날짜 옆에 Channel 값을 함께 출력합니다.
+                channel_info = str(row.get('Channel', '-')).strip()
+                st.markdown(f"""
+                    <div style='font-size: 0.88rem; font-weight: 600; color: #1E3A8A; margin-top: 6px; margin-bottom: 4px;'>
+                        📅 Completed On: {row['Date']} &nbsp;&nbsp;|&nbsp;&nbsp; 📡 Channel: <span style='color: #059669;'>{channel_info}</span>
+                    </div>
+                """, unsafe_allow_html=True)
+            else:
+                # 1. 태그와 제목 출력
+                st.markdown(f"<span class='type-tag {'sent' if row['Type']=='SENT' else 'received'}'>{row['Type']}</span> **{row['Subject']}**", unsafe_allow_html=True)
+                
+                # 2. 날짜를 상단에 단독 배치 
+                st.markdown(f"<div style='font-size: 0.88rem; font-weight: 600; color: #1E3A8A; margin-top: 6px; margin-bottom: 4px;'>📅 {row['Date']}</div>", unsafe_allow_html=True)
+                
+                # 3. From, To, CC 각각 개별 줄바꿈
+                cc_value = str(row.get('Cc', '')).strip()
+                cc_display = cc_value if cc_value and cc_value.lower() != 'nan' else "<span style='color: #94a3b8; font-style: italic;'>None</span>"
+                    
+                st.markdown(f"""
+                    <div style='font-size: 0.85rem; color: #64748b; line-height: 1.6; margin-bottom: 12px;'>
+                        <b>From:</b> {row['From']}<br>
+                        <b>To:</b> {row['To']}<br>
+                        <b>CC:</b> {cc_display}
+                    </div>
+                """, unsafe_allow_html=True)
+                
+                        # 💡 [두 번째 if NOTION]: 매핑 상태 (배지 vs 드롭다운) 처리 구역
+            rules = st.session_state['rules_df']
+            rules_list = rules.to_dict('records')
+
+            # 🚨 [기능 추가] 웹에서 실시간으로 추가/수정된 모든 단계를 드롭다운 옵션에 자동 반영
+            opts = ["-"]
+            for _, r in rules.iterrows():
+                s1 = str(r.get('Sub 1', '')).strip()
+                s2 = str(r.get('Sub 2', '')).strip()
+                if s1 and s1 not in ['nan', 'x', '', 'None', '-'] and s1 not in opts: opts.append(s1)
+                if s2 and s2 not in ['nan', 'x', '', 'None', '-'] and s2 not in opts: opts.append(s2)
+
+            # 웹 설정에서 새로 수정한 값이나 기존 값이 opts 배열에 누수되었을 경우를 방지하는 실시간 안전망
+            current_sec_val = str(row.get('Sec', '-')).strip()
+            if current_sec_val not in opts and current_sec_val != "-":
+                opts.insert(1, current_sec_val)
+
+            if row['Type'] == "NOTION":
+                clean_raw = re.sub(r'^\d+\.\s*', '', str(row['RawTitle'])).strip().lower()
+                stage_titles = [str(x).strip().lower() for x in rules['Stage Title'].tolist()]
+
+                if clean_raw in stage_titles:
+                    # 🟢 [기존 기능 유지] 일치함: 자동 매핑 배지 표시
+                    matched_stage_idx = stage_titles.index(clean_raw)
+                    matched_stage_name = rules['Stage Title'].iloc[matched_stage_idx]
+                    st.markdown(f"""
+                        <div style='background-color: #F0F9FF; padding: 6px 12px; border-radius: 6px; 
+                                    border: 1px solid #BAE6FD; font-size: 0.85rem; color: #0369A1; font-weight: 600; margin: 10px 0;'>
+                            ✅ Auto-Mapped: {matched_stage_name}
+                        </div>
+                    """, unsafe_allow_html=True)
+                else:
+                    # ⚠️ [기존 기능 유지] 일치하지 않음: 수동 할당을 위한 경고 배지 및 드롭다운 활성화
+                    st.markdown("<div style='font-size: 0.75rem; color: #EF4444; margin-bottom: 4px; font-weight: bold;'>⚠️ Unmapped (Assign Manually)</div>", unsafe_allow_html=True)
+                    st.selectbox("Assign Action", opts, index=opts.index(current_sec_val) if current_sec_val in opts else 0, key=f"sel_{idx}", label_visibility="collapsed")
+            else:
+                # 🟢 [기존 기능 유지] 이메일 전용: 수동 할당 드롭다운
+                st.selectbox("Assign Action", opts, index=opts.index(current_sec_val) if current_sec_val in opts else 0, key=f"sel_{idx}", label_visibility="collapsed")
+            
+            # [기존 기능 유지] 본문(Body) 출력
+            st.markdown(f"<div class='content-box'>{row.get('Body', '')}</div>", unsafe_allow_html=True)
+            
+            # 🟢 [기존 기능 유지] 코멘트(Comments)가 있을 경우, 본문 아래에 별도의 말풍선 박스로 예쁘게 렌더링
+            comments_data = row.get('Comments', '').strip()
+            if comments_data:
+                html_comments = comments_data.replace('\n', '<br>')
+                st.markdown(f"""
+                    <div style='background-color: #FEF3C7; padding: 12px 15px; border-radius: 8px; border: 1px solid #FCD34D; font-size: 0.85rem; margin-top: 12px; color: #92400E; line-height: 1.6;'>
+                        <b>💬 Notion Comments:</b><br><br>{html_comments}
+                    </div>
+                """, unsafe_allow_html=True)
+
+            # [기존 기능 유지] 첨부파일 출력 구역
+            if row.get('Attach'):
+                st.markdown(f"<div class='attach-box'>📎 <b>Attachments:</b> {', '.join(row['Attach'])}</div>", unsafe_allow_html=True)