import streamlit as st
import pandas as pd
import numpy as np
import plotly.express as px
import requests
import os

# ── Page config ──────────────────────────────────────────────
st.set_page_config(
    page_title="Portfolio Monitoring Dashboard",
    page_icon="📈",
    layout="wide"
)

# ── Hugging Face AI (FinBERT sentiment) ──────────────────────
HF_API_KEY = os.environ.get("HF_API_KEY", "")

def analyze_sentiment(text: str) -> str:
    """Calls FinBERT on Hugging Face to get financial sentiment."""
    if not HF_API_KEY:
        return "⚠️ No API key"
    url = "https://api-inference.huggingface.co/models/ProsusAI/finbert"
    headers = {"Authorization": f"Bearer {HF_API_KEY}"}
    try:
        r = requests.post(url, headers=headers,
                          json={"inputs": text}, timeout=10)
        result = r.json()
        if isinstance(result, list) and result:
            top = max(result[0], key=lambda x: x["score"])
            emoji = {"positive": "🟢", "negative": "🔴",
                     "neutral": "🟡"}.get(top["label"].lower(), "⚪")
            return f"{emoji} {top['label'].capitalize()} ({top['score']:.0%})"
    except Exception:
        return "❌ Error"
    return "❓ Unknown"

# ── Load data ─────────────────────────────────────────────────
@st.cache_data
def load_data():
    portfolio     = pd.read_csv("portfolio_output.csv")
    risk_metrics  = pd.read_csv("risk_metrics_output.csv")
    daily_returns = pd.read_csv("portfolio_daily_returns_output.csv",
                                parse_dates=["date"])
    return portfolio, risk_metrics, daily_returns

try:
    portfolio, risk_metrics, daily_returns = load_data()
    data_loaded = True
except FileNotFoundError:
    data_loaded = False

# ── Sidebar ───────────────────────────────────────────────────
st.sidebar.title("Portfolio Monitor")
st.sidebar.caption("ESCP — Applied Data Science Workshop")

# ── Main title ────────────────────────────────────────────────
st.title("📈 Portfolio Monitoring Dashboard")
st.caption("Real-time portfolio performance, risk alerts & AI-powered news sentiment")

st.markdown("""
<div style="
    background-color: #1A3C6E;
    padding: 10px 20px;
    border-radius: 8px;
    margin-bottom: 10px;
">
    <p style="color: white; font-size: 13px; margin: 0; text-align: center;">
        👥 <b>Group Project</b> — 
        Clément De Ceukeleire · Laure Dumont · Matéo François · Romain Prudhon
    </p>
    <p style="color: #A9CCE3; font-size: 12px; margin: 4px 0 0 0; text-align: center;">
        ESCP Business School — Applied Data Science Workshop
    </p>
</div>
""", unsafe_allow_html=True)

st.divider()

# ── Demo mode if no CSV ───────────────────────────────────────
if not data_loaded:
    st.warning(
        "⚠️ No data files found. Showing demo data. "
        "Upload your CSV files to see real results."
    )
    np.random.seed(42)
    portfolio = pd.DataFrame({
        "Ticker":                 ["AAPL", "MSFT", "NVDA", "GOOGL", "AMZN"],
        "Friendly name":          ["Apple", "Microsoft", "Nvidia",
                                   "Alphabet", "Amazon"],
        "market_value":           [12000, 9500, 8200, 6100, 5400],
        "invested_amount":        [10000, 8000, 5000, 5500, 6000],
        "unrealized_pnl":         [2000, 1500, 3200, 600, -600],
        "cumulative_realized_pnl":[500, 300, 200, 100, 50],
        "total_pnl":              [2500, 1800, 3400, 700, -550],
        "weight":                 [0.29, 0.23, 0.20, 0.15, 0.13],
        "asset_concentration_flag": [False, False, False, False, False],
        "stressed_value":         [10200, 8075, 6970, 5185, 4590],
        "stress_test_loss":       [-1800,-1425,-1230, -915, -810],
        "alert_level":            ["Normal","Normal","Normal",
                                   "Normal","Warning loss"],
    })
    portfolio["unrealized_return_pct"] = (
        portfolio["unrealized_pnl"] / portfolio["invested_amount"]
    )

    dates = pd.date_range(end=pd.Timestamp.today(), periods=120, freq="B")
    daily_returns = pd.DataFrame({
        "date": dates,
        "portfolio_daily_returns": np.random.normal(0.0005, 0.012, 120)
    })
    risk_metrics = pd.DataFrame({
        "Metric": ["Mean daily return", "Daily volatility",
                   "Annualized volatility", "Worst daily return",
                   "Best daily return", "Sharpe ratio"],
        "Value":  [0.0005, 0.012, 0.190, -0.032, 0.028, 0.66]
    })

# ── Ticker filter ─────────────────────────────────────────────
ticker_list = ["All"] + sorted(portfolio["Ticker"].dropna().unique().tolist())
selected = st.sidebar.selectbox("Filter by asset", ticker_list)
pv = portfolio if selected == "All" else portfolio[portfolio["Ticker"] == selected]

# ── KPI Cards ─────────────────────────────────────────────────
st.subheader("Portfolio Summary")
c1, c2, c3, c4 = st.columns(4)
c1.metric("💰 Invested",       f"{pv['invested_amount'].sum():,.0f} €")
c2.metric("📊 Market Value",   f"{pv['market_value'].sum():,.0f} €")
c3.metric("📈 Total P&L",      f"{pv['total_pnl'].sum():,.0f} €")
c4.metric("🏦 Positions",      int(pv["Ticker"].nunique()))
st.divider()

# ── Charts row 1 ──────────────────────────────────────────────
col_l, col_r = st.columns(2)

with col_l:
    st.subheader("🥧 Portfolio Allocation")
    fig_pie = px.pie(pv, names="Ticker", values="market_value",
                     hole=0.35)
    fig_pie.update_traces(textinfo="percent+label")
    st.plotly_chart(fig_pie, use_container_width=True)

with col_r:
    st.subheader("📊 Market Value by Asset")
    color_col = "alert_level" if "alert_level" in pv.columns else "Ticker"
    fig_bar = px.bar(
        pv.sort_values("market_value", ascending=False),
        x="Ticker", y="market_value", color=color_col,
        color_discrete_map={
            "Normal": "#2ecc71",
            "Warning loss": "#f39c12",
            "Critical loss": "#e74c3c"
        }
    )
    st.plotly_chart(fig_bar, use_container_width=True)

# ── Cumulative return ─────────────────────────────────────────
st.subheader("📉 Cumulative Portfolio Return")
daily_returns["cumulative_return"] = (
    (1 + daily_returns["portfolio_daily_returns"]).cumprod() - 1
)
fig_line = px.line(daily_returns, x="date", y="cumulative_return",
                   labels={"cumulative_return": "Cumulative Return",
                           "date": "Date"})
fig_line.add_hline(y=0, line_dash="dash", line_color="black")
fig_line.update_traces(line_color="#3498db")
st.plotly_chart(fig_line, use_container_width=True)

# ── Unrealized return scatter ──────────────────────────────────
st.subheader("⚡ Unrealized Return by Asset")
if "unrealized_return_pct" in pv.columns:
    fig_ret = px.bar(
        pv.sort_values("unrealized_return_pct"),
        x="Ticker", y="unrealized_return_pct",
        color=pv["unrealized_return_pct"].apply(
            lambda x: "Gain" if x >= 0 else "Loss"
        ),
        color_discrete_map={"Gain": "#2ecc71", "Loss": "#e74c3c"},
        labels={"unrealized_return_pct": "Unrealized Return %"}
    )
    fig_ret.add_hline(y=0, line_dash="dash", line_color="black")
    st.plotly_chart(fig_ret, use_container_width=True)

# ── Risk metrics ──────────────────────────────────────────────
st.subheader("🔬 Risk Metrics")
st.dataframe(risk_metrics, use_container_width=True, hide_index=True)

# ── Stress test ───────────────────────────────────────────────
if "stressed_value" in pv.columns:
    st.subheader("💥 Stress Test (−15% market shock)")
    sk1, sk2, sk3 = st.columns(3)
    sk1.metric("Current Value",  f"{pv['market_value'].sum():,.0f} €")
    sk2.metric("Stressed Value", f"{pv['stressed_value'].sum():,.0f} €",
               delta=f"{pv['stress_test_loss'].sum():,.0f} €")
    sk3.metric("Estimated Loss", f"{pv['stress_test_loss'].sum():,.0f} €")

# ── Alert table ───────────────────────────────────────────────
st.subheader("🚨 Risk Alert Table")
alert_cols = [c for c in ["Ticker", "market_value", "weight",
                           "unrealized_return_pct", "alert_level",
                           "asset_concentration_flag"] if c in pv.columns]
st.dataframe(pv[alert_cols].sort_values("market_value", ascending=False),
             use_container_width=True, hide_index=True)

st.divider()

# ── AI Sentiment Analysis ─────────────────────────────────────
st.subheader("🤖 AI Sentiment Analysis (FinBERT)")
st.caption("Powered by Hugging Face — ProsusAI/finbert")

news_examples = {
    "AAPL": "Apple reports record quarterly earnings driven by iPhone sales",
    "MSFT": "Microsoft faces antitrust investigation in European markets",
    "NVDA": "Nvidia surges on strong AI chip demand forecast",
    "GOOGL": "Alphabet announces major layoffs amid cost-cutting efforts",
    "AMZN": "Amazon expands logistics network with new warehouse openings",
}

st.info(
    "Enter a news headline below and click Analyze to get "
    "an AI-powered sentiment score using FinBERT, "
    "a model trained specifically on financial text."
)

col_input, col_btn = st.columns([4, 1])
with col_input:
    headline = st.text_input(
        "News headline",
        value="Apple reports record quarterly earnings driven by iPhone sales"
    )
with col_btn:
    st.write("")
    st.write("")
    run_sentiment = st.button("🔍 Analyze")

if run_sentiment and headline:
    with st.spinner("Calling FinBERT model..."):
        sentiment = analyze_sentiment(headline)
    st.success(f"**Sentiment result:** {sentiment}")

# Pre-loaded examples
if st.checkbox("Show sentiment for example headlines"):
    results = []
    for ticker, text in news_examples.items():
        with st.spinner(f"Analyzing {ticker}..."):
            sent = analyze_sentiment(text)
        results.append({"Ticker": ticker, "Headline": text, "Sentiment": sent})
    st.dataframe(pd.DataFrame(results), use_container_width=True, hide_index=True)

st.divider()

# ── Download ──────────────────────────────────────────────────
st.download_button(
    label="⬇️ Download Portfolio Table (CSV)",
    data=portfolio.to_csv(index=False).encode("utf-8"),
    file_name="portfolio_monitoring_output.csv",
    mime="text/csv"
)

st.caption("ESCP Business School — Applied Data Science Workshop | Group Project")