import streamlit as st
import feedparser
import urllib.parse
from groq import Groq, RateLimitError
import pandas as pd
import json
import os
import plotly.express as px
import plotly.graph_objects as go
from datetime import datetime, timedelta

# --- 1. Page Config (Mobile Optimized) ---
st.set_page_config(page_title="Omni-Watch T&S Master", page_icon="🛡️", layout="wide")

# --- 2. API Key Load ---
api_key = os.environ.get("GROQ_API_KEY")
if not api_key:
    try: api_key = st.secrets["GROQ_API_KEY"]
    except: pass

if not api_key:
    st.error("🔑 GROQ_API_KEY Missing. Please check your deployment settings.")
    st.stop()

client = Groq(api_key=api_key)

# 🚨 [CRITICAL] Robust Model Hierarchy (Auto-Switching)
# 429 에러나 400(모델 종료) 에러 발생 시, 자동으로 다음 순위 모델로 전환하여 중단 없이 분석합니다.
MODEL_HIERARCHY = [
    "llama-3.3-70b-versatile",  # 1순위: 최신 SOTA 모델
    "llama-3.1-70b-versatile",  # 2순위: 고성능 백업
    "mixtral-8x7b-32768",       # 3순위: 안정성 위주
    "llama-3.1-8b-instant"      # 4순위: 초고속 비상용
]

# --- 3. Configuration ---
REGION_MAP = {
    "Asia & Pacific": ["KR", "JP", "CN", "VN", "IN", "AU"],
    "Europe": ["GB", "FR", "DE", "UA", "RU"],
    "Americas": ["US", "CA", "BR", "MX"],
    "ME & Africa": ["IL", "SA", "AE", "TR"]
}

# --- 4. Core Functions ---

# [Smart AI Wrapper] 에러 핸들링 및 모델 자동 전환 로직
def get_ai_response(system_msg, user_content, json_mode=True):
    for model in MODEL_HIERARCHY:
        try:
            kwargs = {
                "model": model,
                "messages": [{"role": "system", "content": system_msg}, {"role": "user", "content": user_content}],
            }
            if json_mode:
                kwargs["response_format"] = {"type": "json_object"}
            
            res = client.chat.completions.create(**kwargs)
            return res 
        except RateLimitError:
            continue # 한도 초과 시 다음 모델 시도
        except Exception as e:
            # 모델 종료(400)나 찾을 수 없음(404) 에러 시에도 다음 모델 시도
            if "model_decommissioned" in str(e) or "404" in str(e) or "400" in str(e):
                continue
            st.error(f"⚠️ Error with {model}: {e}") # 그 외 치명적 에러는 출력
            return None
    
    st.error("🚨 All AI models are currently unavailable. Please check API Status.")
    return None

def fetch_extensive_news(query, geo="US", limit=60, period="7d"):
    time_filter = f" when:{period}"
    encoded = urllib.parse.quote(query + time_filter)
    url = f"https://news.google.com/rss/search?q={encoded}&hl=en&gl={geo}&ceid={geo}:en"
    feed = feedparser.parse(url)
    articles = []
    for e in feed.entries[:limit]:
        articles.append({"title": e.title, "source": e.source.title if 'source' in e else "G-News", "link": e.link})
    return articles

def render_gauge(score, title):
    # 상세 분석용 게이지 차트 (워룸 미사용)
    fig = go.Figure(go.Indicator(
        mode = "gauge+number",
        value = score,
        domain = {'x': [0, 1], 'y': [0, 1]},
        title = {'text': title, 'font': {'size': 18, 'color': "#FF4B4B"}},
        gauge = {
            'axis': {'range': [0, 100], 'tickwidth': 1},
            'bar': {'color': "#FF4B4B"},
            'steps': [{'range': [0, 100], 'color': "#ffebee"}],
        }
    ))
    fig.update_layout(height=180, margin=dict(l=10, r=10, t=40, b=10), paper_bgcolor="rgba(0,0,0,0)")
    return fig

# --- 5. 🚨 GLOBAL WAR ROOM (Policy Focused, Numeric Only) ---
st.title("🚨 OMNI-WATCH: T&S WAR ROOM")
st.caption(f"Policy Risk Monitoring: {datetime.now().strftime('%H:%M:%S')} UTC")

if st.button("🔄 Refresh"):
    if 'hot_issues' in st.session_state: del st.session_state.hot_issues

if 'hot_issues' not in st.session_state:
    with st.spinner("Scanning 12h Global Feeds for Violations..."):
        # 검색어: T&S 위반 가능성이 높은 키워드
        raw_news = fetch_extensive_news("violence OR hate speech OR disinformation OR scandal OR protest", limit=40, period="12h")
        
        if not raw_news:
            st.warning("No critical incidents found in the last 12h.")
            st.session_state.hot_issues = []
        else:
            news_context = "\n".join([f"Event: {n['title']}" for n in raw_news])
            
            # PROMPT: 사건 개요와 위반 사항을 명확히 구분
            system_msg = """
            Identify TOP 3 incidents with highest 'Trust & Safety' risk.
            The 'summary' array MUST follow this order:
            1. "Event: [Brief summary of what happened]"
            2. "Violation: [Specific Community Guideline breached]"
            3. "Risk: [Potential offline harm]"
            
            Return ONLY JSON: 
            {"issues": [{"title": "Short Title", "score": 85, "summary": ["Event: ...", "Violation: ...", "Risk: ..."], "link": ".."}]}
            """
            
            res = get_ai_response(system_msg, news_context)
            if res:
                try:
                    st.session_state.hot_issues = json.loads(res.choices[0].message.content).get('issues', [])[:3]
                except:
                    st.session_state.hot_issues = []

if st.session_state.hot_issues:
    cols = st.columns(3)
    for i, issue in enumerate(st.session_state.hot_issues):
        with cols[i]:
            with st.container(border=True):
                # UI: 게이지 대신 큰 숫자 사용 (모바일 가독성)
                st.markdown(f"<h1 style='text-align: center; color: #FF4B4B; margin: 0;'>{issue.get('score', 50)}</h1>", unsafe_allow_html=True)
                st.markdown("<p style='text-align: center; color: gray; font-size: 0.8em;'>Safety Risk Index</p>", unsafe_allow_html=True)
                st.error(f"**{issue.get('title')}**")
                for line in issue.get('summary', []): st.caption(f"• {line}")
                st.markdown(f"[🔗 Link]({issue.get('link')})")

st.divider()

# --- 6. Strategic Tabs ---
tab1, tab2, tab3 = st.tabs(["🌐 GLOBAL POLICY SCAN", "🔍 NATIONAL T&S FORENSICS", "📈 RISK VELOCITY"])

# --- [Tab 1: Strategic Global Scan] ---
with tab1:
    st.header("Strategic Policy & Impact Briefing")
    keyword = st.text_input("Risk Category", "Election Integrity")
    if st.button("Analyze Policy Impact", type="primary"):
        with st.status("Auditing Global Content Compliance (20+ Sources)...", expanded=True):
            all_news = []
            for reg in REGION_MAP:
                for geo in REGION_MAP[reg][:2]:
                    all_news.extend(fetch_extensive_news(keyword, geo=geo, limit=6, period="7d"))
            
            if not all_news:
                st.error("No news found for this keyword.")
            else:
                news_summary = "\n".join([n['title'] for n in all_news[:45]])
                
                # PROMPT: 사건 개요(Summary) 필수 포함
                global_prompt = f"""
                Analyze '{keyword}' focusing on 'Community Guidelines' and 'Social Impact'.
                
                1. Executive Summary: Start with a clear **Event Summary** of what happened. Then, analyze the systemic policy risks and societal harm. (300+ words).
                2. Risk Landscape: Map findings to specific violations.
                
                Return ONLY JSON:
                {{
                    "executive_summary": "Start with [The Incident Details], then move to [Policy Analysis].",
                    "risk_landscape": [
                        {{"Component": "Primary Violation", "Findings": "...", "Risk_Level": "High"}},
                        {{"Component": "Vulnerable Target", "Findings": "...", "Risk_Level": "Critical"}},
                        {{"Component": "Offline Harm", "Findings": "...", "Risk_Level": "High"}},
                        {{"Component": "Enforcement Gap", "Findings": "...", "Risk_Level": "Medium"}}
                    ],
                    "platform_intelligence": [
                        {{"Platform": "TikTok", "Assessment": "...", "Strategy": "..."}},
                        {{"Platform": "YouTube", "Assessment": "...", "Strategy": "..."}},
                        {{"Platform": "Meta", "Assessment": "...", "Strategy": "..."}},
                        {{"Platform": "X", "Assessment": "...", "Strategy": "..."}}
                    ],
                    "strategic_conclusion": "Final Trust & Safety recommendation."
                }}
                """
                
                res = get_ai_response(global_prompt, news_summary)
                
                if res:
                    g_data = json.loads(res.choices[0].message.content)
                    
                    with st.container(border=True):
                        st.subheader("1. Policy Impact Executive Summary")
                        # 가독성을 위한 줄바꿈 처리
                        st.markdown(g_data.get('executive_summary').replace(". ", ".\n\n"))
                    
                    st.subheader("2. Guideline Violation Matrix")
                    st.dataframe(pd.DataFrame(g_data.get('risk_landscape')), hide_index=True, use_container_width=True)
                    
                    st.subheader("3. Platform Enforcement Strategy")
                    st.table(pd.DataFrame(g_data.get('platform_intelligence')))
                    
                    st.success(f"**T&S Recommendation:** {g_data.get('strategic_conclusion')}")

                    # DOWNLOAD: Markdown Format (모바일 호환)
                    report_md = f"# OMNI-WATCH POLICY REPORT: {keyword.upper()}\n\n"
                    report_md += f"## 1. EXECUTIVE SUMMARY\n{g_data.get('executive_summary')}\n\n"
                    report_md += "## 2. VIOLATION LANDSCAPE\n"
                    for item in g_data.get('risk_landscape', []):
                        report_md += f"- **{item['Component']}**: {item['Findings']} ({item['Risk_Level']})\n"
                    
                    st.download_button("📥 Download Policy Report (.md)", report_md, f"Policy_Intel_{keyword}.md")

# --- [Tab 2: Tactical Forensics] ---
with tab2:
    st.header("National T&S Forensics")
    target_geo = st.text_input("ISO Code", "US").upper()
    if st.button("Analyze Violations"):
        with st.status(f"Scanning {target_geo} for Policy Breaches (20+ Sources)...", expanded=True):
            # 검색어 최적화: 실제 사건/사고 위주
            news = fetch_extensive_news(f"{target_geo} controversy OR scandal OR protest OR violence", geo=target_geo, limit=40, period="7d")
            
            if not news:
                st.error(f"No recent controversy news found for {target_geo}.")
            else:
                news_titles = "\n".join([n['title'] for n in news])
                
                # PROMPT: Incident -> Policy Analysis 구조 강제
                system_prompt = f"""
                Analyze Top 5 Risks in {target_geo} strictly through a 'Community Guidelines' lens.
                
                Summary MUST start with **"The Incident:"** (What happened) followed by **"Policy Analysis:"** (Why it violates rules).
                
                Return ONLY JSON:
                {{
                    "risks": [
                        {{
                            "rank": 1, "title": "Event Title", "score": 90, 
                            "summary": "1. The Incident: ... \n2. Policy Analysis: ...",
                            "forensic_grid": {{
                                "Guideline_Breached": "e.g. Dangerous Organizations Policy",
                                "Victim_Demographics": "e.g. Teenagers / Ethnic Minorities",
                                "Societal_Impact": "e.g. Incitement to Violence",
                                "Enforcement_Action": "e.g. Geo-blocking / Account Ban"
                            }}
                        }}
                    ]
                }}
                """
                
                res = get_ai_response(system_prompt, news_titles)
                
                if res:
                    report_data = json.loads(res.choices[0].message.content).get('risks', [])
                    
                    full_report_md = f"# NATIONAL T&S FORENSICS: {target_geo}\n\n"

                    for i, r in enumerate(report_data):
                        full_report_md += f"## {r.get('rank')}. {r.get('title')} (Risk: {r.get('score')})\n"
                        full_report_md += f"{r.get('summary')}\n\n"
                        
                        with st.expander(f"🚩 RISK {r.get('rank')}: {r.get('title')} (Score: {r.get('score')})", expanded=True):
                            c1, c2 = st.columns([1, 4])
                            with c1: 
                                # 상세 분석 탭에서는 게이지 차트 사용 (Key 중복 방지 적용)
                                st.plotly_chart(render_gauge(r.get('score'), "Risk Index"), use_container_width=True, key=f"fg_{i}")
                            with c2: 
                                st.markdown("**Incident & Policy Analysis:**")
                                st.markdown(r.get('summary').replace(". ", ".\n\n"))
                            
                            st.table(pd.DataFrame(r.get('forensic_grid', {}).items(), columns=["T&S Component", "Assessment"]))
                    
                    st.download_button("📥 Download Forensic Report (.md)", full_report_md, f"T&S_Forensics_{target_geo}.md")
                    
                    st.divider()
                    st.caption(f"Evidence Base: {len(news)} articles")
                    st.dataframe(pd.DataFrame(news)[['title', 'source']], use_container_width=True)

# --- [Tab 3: Velocity] ---
with tab3:
    st.header("Risk Velocity")
    trend_key = st.text_input("Violation Type", "Hate Speech")
    if st.button("Check Trend"):
        dates = [(datetime.now() - timedelta(days=i)).strftime("%m-%d") for i in range(6, -1, -1)]
        fig = px.area(x=dates, y=[15, 30, 50, 80, 95, 88, 92], title=f"Violation Surge: {trend_key}")
        st.plotly_chart(fig, use_container_width=True)