import gradio as gr
import pandas as pd
import numpy as np
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import matplotlib.patches as mpatches
import io
import re
from datetime import datetime

# ── VADER sentiment (graceful fallback if not installed) ──────────────────────
try:
    from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
    _analyzer = SentimentIntensityAnalyzer()
    def vader_score(text):
        return _analyzer.polarity_scores(str(text))
except ImportError:
    def vader_score(text):
        text = text.lower()
        pos = sum(w in text for w in ["good","strong","growth","positive","gain","profit"])
        neg = sum(w in text for w in ["breach","hack","lawsuit","strike","loss","fine","shut","miss","fraud","attack"])
        compound = round((pos - neg) / max(pos + neg, 1), 3)
        return {"neg": neg / max(pos+neg,1), "neu": 0.5, "pos": pos / max(pos+neg,1), "compound": compound}

# ── Styling constants ─────────────────────────────────────────────────────────
PALETTE = {
    "critical response": "#C0392B",
    "escalate":          "#E67E22",
    "review":            "#2980B9",
    "monitor":           "#27AE60",
}
BG      = "#0D0F14"
CARD_BG = "#141720"
ACCENT  = "#4F8EF7"
TEXT    = "#E8EAF0"

CRISIS_ICONS = {
    "cybersecurity": "🔐",
    "legal":         "⚖️",
    "operations":    "🏭",
    "labor":         "👷",
    "financial":     "📉",
}

# ─────────────────────────────────────────────────────────────────────────────
# HELPERS
# ─────────────────────────────────────────────────────────────────────────────

def infer_crisis_type(text):
    text = text.lower()
    if any(w in text for w in ["hack","breach","cyber","ransomware","malware","data leak","phishing"]):
        return "cybersecurity"
    if any(w in text for w in ["lawsuit","antitrust","regulator","fine","court","SEC","penalty","sanction"]):
        return "legal"
    if any(w in text for w in ["factory","supply chain","shutdown","production","recall","outage"]):
        return "operations"
    if any(w in text for w in ["strike","worker","protest","union","layoff","walkout"]):
        return "labor"
    if any(w in text for w in ["profit warning","earnings miss","downgrade","revenue","loss","debt","bankruptcy"]):
        return "financial"
    return "general"

SEVERITY_BASE = {
    "cybersecurity": 4, "legal": 4, "operations": 3,
    "labor": 3, "financial": 5, "general": 3,
}
URGENCY_BASE = {
    "cybersecurity": 5, "legal": 4, "operations": 4,
    "labor": 3, "financial": 5, "general": 3,
}
MARKET_BASE = {
    "cybersecurity": -2.8, "legal": -2.1, "operations": -1.8,
    "labor": -1.2, "financial": -3.0, "general": -1.5,
}

def assign_priority(severity, urgency):
    score = severity + urgency
    if score >= 9:  return "critical response"
    if score >= 7:  return "escalate"
    if score >= 5:  return "review"
    return "monitor"

def sentiment_label(compound):
    if compound >=  0.05: return "Positive 🟢"
    if compound <= -0.05: return "Negative 🔴"
    return "Neutral ⚪"

# ─────────────────────────────────────────────────────────────────────────────
# TAB 1 — Single headline analyser
# ─────────────────────────────────────────────────────────────────────────────

def analyse_headline(headline: str):
    if not headline.strip():
        return "⚠️ Please enter a headline.", None

    scores     = vader_score(headline)
    compound   = round(scores["compound"], 3)
    crisis     = infer_crisis_type(headline)
    severity   = SEVERITY_BASE[crisis]
    urgency    = URGENCY_BASE[crisis]
    market     = round(MARKET_BASE[crisis] + np.random.normal(0, 0.4), 2)
    priority   = assign_priority(severity, urgency)
    sent_label = sentiment_label(compound)
    icon       = CRISIS_ICONS.get(crisis, "📰")
    color      = PALETTE.get(priority, "#888")

    # ── bar chart ─────────────────────────────────────────────────────────────
    fig, axes = plt.subplots(1, 2, figsize=(9, 3.5))
    fig.patch.set_facecolor(BG)

    # sentiment bars
    ax = axes[0]
    ax.set_facecolor(CARD_BG)
    cats   = ["Negative", "Neutral", "Positive"]
    vals   = [scores["neg"], scores["neu"], scores["pos"]]
    colors = ["#C0392B", "#7F8C8D", "#27AE60"]
    bars   = ax.barh(cats, vals, color=colors, height=0.5)
    ax.set_xlim(0, 1)
    ax.set_title("Sentiment Breakdown", color=TEXT, fontsize=11, pad=8)
    ax.tick_params(colors=TEXT, labelsize=9)
    for spine in ax.spines.values(): spine.set_visible(False)
    ax.xaxis.label.set_color(TEXT)
    ax.set_xlabel("Score", color=TEXT, fontsize=8)
    for bar, val in zip(bars, vals):
        ax.text(val + 0.01, bar.get_y() + bar.get_height()/2,
                f"{val:.2f}", va="center", color=TEXT, fontsize=8)

    # severity / urgency gauge
    ax2 = axes[1]
    ax2.set_facecolor(CARD_BG)
    metrics = ["Severity", "Urgency"]
    mvals   = [severity, urgency]
    mcols   = [ACCENT, color]
    b2      = ax2.barh(metrics, mvals, color=mcols, height=0.5)
    ax2.set_xlim(0, 5)
    ax2.set_title("Risk Scores (out of 5)", color=TEXT, fontsize=11, pad=8)
    ax2.tick_params(colors=TEXT, labelsize=9)
    for spine in ax2.spines.values(): spine.set_visible(False)
    ax2.set_xlabel("Score", color=TEXT, fontsize=8)
    ax2.xaxis.label.set_color(TEXT)
    for bar, val in zip(b2, mvals):
        ax2.text(val + 0.05, bar.get_y() + bar.get_height()/2,
                 str(val), va="center", color=TEXT, fontsize=9, fontweight="bold")

    plt.tight_layout(pad=1.5)

    # ── markdown result card ───────────────────────────────────────────────────
    md = f"""
### {icon} Analysis Result

| Field | Value |
|---|---|
| **Crisis Type** | `{crisis.upper()}` |
| **Sentiment** | {sent_label} (compound: `{compound}`) |
| **Severity Score** | `{severity} / 5` |
| **Response Urgency** | `{urgency} / 5` |
| **Est. Market Impact** | `{market:+.2f}%` |
| **Priority Action** | <span style='color:{color};font-weight:bold'>{priority.upper()}</span> |

---
**Recommended Response:** {"🚨 Immediate leadership escalation and cross-team crisis coordination required." if priority == "critical response" else "⚡ Escalate to risk and communications teams for coordinated response." if priority == "escalate" else "🔍 Schedule a structured review in the next reporting cycle." if priority == "review" else "👁️ Routine monitoring — flag if coverage increases."}
"""
    return md, fig


# ─────────────────────────────────────────────────────────────────────────────
# TAB 2 — CSV Dashboard
# ─────────────────────────────────────────────────────────────────────────────

def build_dashboard(file):
    if file is None:
        return "⚠️ Please upload **crisis_news_enriched.csv** (output of Notebook 1).", None

    try:
        df = pd.read_csv(file.name)
    except Exception as e:
        return f"❌ Could not read file: {e}", None

    required = {"crisis_type", "priority_action", "severity_score",
                "estimated_market_impact_pct", "company"}
    missing  = required - set(df.columns)
    if missing:
        return f"❌ Missing columns: {missing}. Please upload `crisis_news_enriched.csv`.", None

    # ── 4-panel figure ────────────────────────────────────────────────────────
    fig, axes = plt.subplots(2, 2, figsize=(13, 9))
    fig.patch.set_facecolor(BG)
    fig.suptitle("Crisis Intelligence Dashboard", color=TEXT,
                 fontsize=16, fontweight="bold", y=0.98)

    # 1. Crisis type distribution
    ax = axes[0, 0]
    ax.set_facecolor(CARD_BG)
    ct = df["crisis_type"].value_counts()
    bar_colors = [ACCENT] * len(ct)
    bars = ax.bar(ct.index, ct.values, color=bar_colors, width=0.6)
    ax.set_title("Headlines by Crisis Type", color=TEXT, fontsize=11)
    ax.tick_params(colors=TEXT, labelsize=8)
    ax.set_xlabel("Crisis Type", color=TEXT, fontsize=9)
    ax.set_ylabel("Count", color=TEXT, fontsize=9)
    for spine in ax.spines.values(): spine.set_color("#2A2D3A")
    for bar in bars:
        ax.text(bar.get_x() + bar.get_width()/2, bar.get_height() + 0.3,
                str(int(bar.get_height())), ha="center", color=TEXT, fontsize=8)

    # 2. Priority action distribution (donut)
    ax2 = axes[0, 1]
    ax2.set_facecolor(CARD_BG)
    pa   = df["priority_action"].value_counts()
    cols = [PALETTE.get(k, "#888") for k in pa.index]
    wedges, texts, autotexts = ax2.pie(
        pa.values, labels=pa.index, colors=cols,
        autopct="%1.0f%%", startangle=140,
        wedgeprops=dict(width=0.55),
        textprops={"color": TEXT, "fontsize": 8}
    )
    for at in autotexts: at.set_color(BG); at.set_fontweight("bold")
    ax2.set_title("Priority Action Distribution", color=TEXT, fontsize=11)

    # 3. Avg market impact by crisis type
    ax3 = axes[1, 0]
    ax3.set_facecolor(CARD_BG)
    mi = df.groupby("crisis_type")["estimated_market_impact_pct"].mean().sort_values()
    bar_c = ["#C0392B" if v < -2.5 else "#E67E22" if v < -1.5 else "#2980B9" for v in mi.values]
    bars3 = ax3.barh(mi.index, mi.values, color=bar_c, height=0.5)
    ax3.axvline(0, color=TEXT, linewidth=0.5, alpha=0.4)
    ax3.set_title("Avg Market Impact % by Crisis Type", color=TEXT, fontsize=11)
    ax3.tick_params(colors=TEXT, labelsize=8)
    ax3.set_xlabel("Est. Impact (%)", color=TEXT, fontsize=9)
    for spine in ax3.spines.values(): spine.set_color("#2A2D3A")
    for bar, val in zip(bars3, mi.values):
        ax3.text(val - 0.05, bar.get_y() + bar.get_height()/2,
                 f"{val:.2f}%", va="center", ha="right", color=TEXT, fontsize=8)

    # 4. Severity heatmap (crisis type × priority)
    ax4 = axes[1, 1]
    ax4.set_facecolor(CARD_BG)
    pivot = df.groupby(["crisis_type", "priority_action"])["severity_score"].mean().unstack(fill_value=0)
    im = ax4.imshow(pivot.values, cmap="RdYlGn_r", aspect="auto", vmin=1, vmax=5)
    ax4.set_xticks(range(len(pivot.columns)))
    ax4.set_yticks(range(len(pivot.index)))
    ax4.set_xticklabels(pivot.columns, color=TEXT, fontsize=7, rotation=20, ha="right")
    ax4.set_yticklabels(pivot.index, color=TEXT, fontsize=8)
    ax4.set_title("Avg Severity: Crisis Type × Priority", color=TEXT, fontsize=11)
    for i in range(pivot.shape[0]):
        for j in range(pivot.shape[1]):
            val = pivot.values[i, j]
            if val > 0:
                ax4.text(j, i, f"{val:.1f}", ha="center", va="center",
                         color="white", fontsize=8, fontweight="bold")
    cbar = fig.colorbar(im, ax=ax4, fraction=0.03)
    cbar.ax.tick_params(colors=TEXT, labelsize=7)

    plt.tight_layout(rect=[0, 0, 1, 0.96])

    # ── summary stats ─────────────────────────────────────────────────────────
    n          = len(df)
    n_critical = (df["priority_action"] == "critical response").sum()
    top_type   = df["crisis_type"].value_counts().idxmax()
    top_co     = df["company"].value_counts().idxmax()
    avg_impact = df["estimated_market_impact_pct"].mean()

    md = f"""
### 📊 Dataset Summary

| Metric | Value |
|---|---|
| Total headlines | `{n}` |
| Critical response alerts | `{n_critical}` ({100*n_critical/n:.0f}%) |
| Most common crisis type | `{top_type.upper()}` |
| Most exposed company | `{top_co}` |
| Avg estimated market impact | `{avg_impact:+.2f}%` |

Upload `crisis_news_enriched.csv` (generated by Notebook 1) to refresh.
"""
    return md, fig


# ─────────────────────────────────────────────────────────────────────────────
# BUILD UI
# ─────────────────────────────────────────────────────────────────────────────

CUSTOM_CSS = """
body, .gradio-container { background: #0D0F14 !important; color: #E8EAF0 !important; font-family: 'IBM Plex Mono', monospace; }
.gr-button-primary { background: #4F8EF7 !important; border: none !important; color: #0D0F14 !important; font-weight: 700 !important; }
.gr-button-primary:hover { background: #6FA3FA !important; }
h1, h2, h3 { color: #E8EAF0 !important; }
.gr-panel, .gr-box { background: #141720 !important; border-color: #2A2D3A !important; }
textarea, input[type=text] { background: #1C1F2B !important; color: #E8EAF0 !important; border-color: #2A2D3A !important; }
.gr-tab-nav button { color: #9AA0B4 !important; }
.gr-tab-nav button.selected { color: #4F8EF7 !important; border-bottom: 2px solid #4F8EF7 !important; }
"""

ABOUT_MD = """
# 🔍 Crisis Monitor — AI-Powered Business Risk Intelligence

## What this app does
This tool is the interactive front-end of an end-to-end crisis monitoring pipeline built for the **AI for Big Data Management** course at ESCP Business School.

It lets you:
- **Analyse any news headline** instantly — detecting crisis type, sentiment (VADER), severity, urgency, estimated market impact, and recommended action
- **Upload your enriched dataset** (`crisis_news_enriched.csv`) for a full visual dashboard

## How the pipeline works
```
Google News RSS  ──► Notebook 1 ──► Enriched CSV ──► Notebook 2 ──► Analysis
                      (scraping,        (severity,       (VADER,
                       synthetic         urgency,         ARIMA,
                       enrichment)       market           decision
                                        impact,           support)
                                        priority)
                                             │
                                             ▼
                                      This Hugging Face App
                                      (real-time scanning +
                                       dashboard visualisation)
```

## Crisis types monitored
| Type | Signal keywords |
|---|---|
| 🔐 Cybersecurity | breach, hack, ransomware, data leak |
| ⚖️ Legal | lawsuit, antitrust, regulator, fine |
| 🏭 Operations | factory, supply chain, shutdown, recall |
| 👷 Labor | strike, worker protest, union, layoff |
| 📉 Financial | profit warning, earnings miss, downgrade |

## Priority framework
| Priority | Trigger |
|---|---|
| 🔴 Critical Response | Severity + Urgency ≥ 9 |
| 🟠 Escalate | Score 7–8 |
| 🔵 Review | Score 5–6 |
| 🟢 Monitor | Score < 5 |

---
*Built with Python · Gradio · VADER Sentiment · Matplotlib*
"""

with gr.Blocks(css=CUSTOM_CSS, title="Crisis Monitor") as demo:

    gr.Markdown("""
# 🚨 Crisis Monitor
### AI-Powered Business Risk Intelligence · ESCP Business School
""")

    with gr.Tabs():

        # ── TAB 1 ─────────────────────────────────────────────────────────────
        with gr.Tab("🔐 Headline Scanner"):
            gr.Markdown("Paste any business news headline to get an instant risk assessment.")
            with gr.Row():
                with gr.Column(scale=2):
                    headline_input = gr.Textbox(
                        label="News Headline",
                        placeholder="e.g. Apple faces major data breach exposing 50M user records...",
                        lines=3,
                    )
                    scan_btn = gr.Button("⚡ Analyse Headline", variant="primary")
                with gr.Column(scale=3):
                    result_md  = gr.Markdown()
                    result_fig = gr.Plot()

            scan_btn.click(
                fn=analyse_headline,
                inputs=headline_input,
                outputs=[result_md, result_fig],
            )

            gr.Examples(
                examples=[
                    ["Tesla hit with major ransomware attack, customer data leaked online"],
                    ["Amazon faces antitrust fine from EU regulators over pricing practices"],
                    ["Boeing factory workers go on strike, halting 737 MAX production"],
                    ["Intel issues profit warning, shares drop 12% after earnings miss"],
                    ["Apple supply chain disruption forces iPhone production cuts in China"],
                ],
                inputs=headline_input,
                label="Try an example",
            )

        # ── TAB 2 ─────────────────────────────────────────────────────────────
        with gr.Tab("📊 Crisis Dashboard"):
            gr.Markdown("Upload `crisis_news_enriched.csv` (generated by Notebook 1) for a full portfolio view.")
            with gr.Row():
                csv_input = gr.File(label="Upload crisis_news_enriched.csv", file_types=[".csv"])
                dash_btn  = gr.Button("📈 Generate Dashboard", variant="primary")
            with gr.Row():
                dash_md  = gr.Markdown()
            with gr.Row():
                dash_fig = gr.Plot()

            dash_btn.click(
                fn=build_dashboard,
                inputs=csv_input,
                outputs=[dash_md, dash_fig],
            )

        # ── TAB 3 ─────────────────────────────────────────────────────────────
        with gr.Tab("ℹ️ About"):
            gr.Markdown(ABOUT_MD)

demo.launch()