app.py

"""
EcoVision — Smart Sustainability Analyzer
Gradio 6. Background thread trains model at startup.
All click handlers wrapped in try/except.
"""

import os, io, tempfile, warnings, threading
from datetime import datetime
import numpy as np
import pandas as pd
import plotly.graph_objects as go
import gradio as gr

warnings.filterwarnings("ignore")

BASE_DIR = os.path.dirname(os.path.abspath(__file__))

from eco_score import calculate_eco_score, get_score_description
from ml_model import predict_impact, train_model
from llm_analysis import generate_eco_explanation, compare_products_llm
from vision_model import classify_image

# ── Load CSV (instant) ────────────────────────────────────────────────────────
DF = pd.read_csv(os.path.join(BASE_DIR, "dataset", "products.csv"))
ALL_PRODUCTS   = sorted(DF["product_name"].tolist())
ALL_CATS       = sorted(DF["category"].unique().tolist())
PACKAGING_OPTS = [
    "Plastic Bottle","Glass Bottle","Cardboard Box","Compostable",
    "Recycled Plastic","Single-use Plastic","Paper Wrapper","Other",
]

# ── Train model in background thread so UI is never blocked ───────────────────
_model_ready  = threading.Event()
_model_error  = None

def _train_bg():
    global _model_error
    try:
        model_path = os.path.join(BASE_DIR, "models", "eco_classifier.pkl")
        os.makedirs(os.path.join(BASE_DIR, "models"), exist_ok=True)
        if not os.path.exists(model_path):
            train_model()
        _model_ready.set()
        print("✅ Model ready")
    except Exception as e:
        _model_error = str(e)
        _model_ready.set()   # unblock anyway so clicks don't hang forever
        print(f"❌ Model training failed: {e}")

threading.Thread(target=_train_bg, daemon=True).start()

def wait_for_model():
    """Wait up to 120s for model, return error string or None."""
    _model_ready.wait(timeout=120)
    return _model_error


# ── Charts ────────────────────────────────────────────────────────────────────
def _bar_color(v):
    if v > 0.5:  return "#C62828"
    if v > 0.25: return "#E65100"
    return "#2E7D32"

BASE_LAYOUT = dict(
    paper_bgcolor="#FFFFFF",
    plot_bgcolor="#F4FBF4",
    font=dict(color="#1A1A1A", family="sans-serif"),
    margin=dict(t=50, b=30, l=55, r=20),
)

def gauge_chart(score, grade):
    gc = {"A":"#2E7D32","B":"#43A047","C":"#E65100",
          "D":"#C62828","F":"#6A0000"}.get(grade,"#2E7D32")
    fig = go.Figure(go.Indicator(
        mode="gauge+number", value=score,
        title=dict(text=f"Eco Score  •  Grade {grade}",
                   font=dict(size=14, color="#1B5E20")),
        number=dict(font=dict(size=46, color=gc), suffix=" / 10"),
        gauge=dict(
            axis=dict(range=[1,10], tickfont=dict(color="#607D63")),
            bar=dict(color=gc, thickness=0.28),
            bgcolor="#F4FBF4", borderwidth=1, bordercolor="#C8E6C9",
            steps=[dict(range=[1,3], color="#FFEBEE"),
                   dict(range=[3,5], color="#FFF3E0"),
                   dict(range=[5,7], color="#F1F8E9"),
                   dict(range=[7,10],color="#E8F5E9")],
            threshold=dict(line=dict(color=gc,width=4),thickness=0.8,value=score),
        ),
    ))
    fig.update_layout(**BASE_LAYOUT, height=260)
    return fig

def metric_bar(label, value, icon):
    color = _bar_color(value)
    fig = go.Figure(go.Bar(
        x=[label], y=[value], marker_color=[color],
        text=[f"{value:.0%}"], textposition="outside",
        textfont=dict(color="#1A1A1A", size=14), width=[0.4],
    ))
    fig.update_layout(
        **BASE_LAYOUT,
        title=dict(text=f"{icon} {label}", font=dict(color="#1B5E20",size=12)),
        yaxis=dict(range=[0,1.3], gridcolor="#D4E8D4", tickformat=".0%",
                   tickfont=dict(color="#607D63")),
        xaxis=dict(showgrid=False),
        height=220,
    )
    return fig

def prob_bar(probs):
    fig = go.Figure(go.Bar(
        x=list(probs.keys()), y=list(probs.values()),
        marker_color=["#2E7D32","#43A047","#66BB6A"][:len(probs)],
        text=[f"{v:.0%}" for v in probs.values()],
        textposition="outside", textfont=dict(color="#1A1A1A",size=12),
    ))
    fig.update_layout(
        **BASE_LAYOUT,
        title=dict(text="ML Confidence", font=dict(color="#1B5E20",size=12)),
        yaxis=dict(range=[0,1.3], gridcolor="#D4E8D4", tickformat=".0%"),
        xaxis=dict(showgrid=False),
        height=220,
    )
    return fig

def comparison_bar(n1, n2, eco1, eco2):
    labels = ["Deforestation","Pollution","Biodiversity"]
    v1 = [eco1["metrics"]["deforestation_risk"],
          eco1["metrics"]["pollution_level"],
          eco1["metrics"]["biodiversity_impact"]]
    v2 = [eco2["metrics"]["deforestation_risk"],
          eco2["metrics"]["pollution_level"],
          eco2["metrics"]["biodiversity_impact"]]
    fig = go.Figure(data=[
        go.Bar(name=n1[:18], x=labels, y=v1,
               text=[f"{v:.0%}" for v in v1], textposition="outside",
               marker_color="#2E7D32"),
        go.Bar(name=n2[:18], x=labels, y=v2,
               text=[f"{v:.0%}" for v in v2], textposition="outside",
               marker_color="#C62828"),
    ])
    fig.update_layout(
        **BASE_LAYOUT, barmode="group", height=280,
        title=dict(text="Side-by-Side Metrics",font=dict(color="#1B5E20",size=13)),
        yaxis=dict(range=[0,1.4], gridcolor="#D4E8D4", tickformat=".0%"),
        xaxis=dict(showgrid=False),
        legend=dict(bgcolor="#F4FBF4", bordercolor="#C8E6C9", borderwidth=1),
        margin=dict(t=55,b=35,l=55,r=20),
    )
    return fig


# ── HTML helpers ──────────────────────────────────────────────────────────────
def err_html(msg):
    return (f"<div style='background:#FFEBEE;border:1px solid #EF9A9A;"
            f"border-radius:10px;padding:14px;font-family:sans-serif;"
            f"color:#C62828;font-size:14px;'>❌ {msg}</div>")

def ok_html(msg):
    return (f"<div style='background:#E8F5E9;border:1px solid #A5D6A7;"
            f"border-radius:10px;padding:12px;font-family:sans-serif;"
            f"color:#2E7D32;font-size:14px;'>✅ {msg}</div>")

def score_card(name, eco, ml):
    gc = {"A":"#2E7D32","B":"#43A047","C":"#E65100",
          "D":"#C62828","F":"#6A0000"}.get(eco["grade"],"#2E7D32")
    impact = ml["predicted_impact"]
    ibg = {"Low Impact":"#E8F5E9","Medium Impact":"#FFF3E0",
           "High Impact":"#FFEBEE"}.get(impact,"#F5F5F5")
    ifg = {"Low Impact":"#2E7D32","Medium Impact":"#E65100",
           "High Impact":"#C62828"}.get(impact,"#333")

    badges = []
    if eco["metrics"]["deforestation_risk"]  < 0.2: badges.append("🌲 Forest Protector")
    if eco["metrics"]["pollution_level"]      < 0.3: badges.append("💧 Clean Waters")
    if eco["metrics"]["biodiversity_impact"]  < 0.2: badges.append("🦋 Biodiversity Guard")
    if eco["eco_score"] >= 7:                         badges.append("⭐ Eco Star")
    if eco["eco_score"] >= 9:                         badges.append("🌟 Green Leader")

    badge_html = "".join(
        f"<span style='display:inline-block;margin:3px;padding:4px 12px;background:#E8F5E9;"
        f"color:#2E7D32;border:1px solid #A5D6A7;border-radius:20px;font-size:12px;'>{b}</span>"
        for b in badges
    ) or "<span style='color:#607D63;font-size:13px;'>No badges yet</span>"

    def pill(lbl, val):
        bg = "#FFEBEE" if val>0.5 else "#FFF8E1" if val>0.3 else "#E8F5E9"
        fg = "#C62828" if val>0.5 else "#E65100" if val>0.3 else "#2E7D32"
        return (f"<div style='flex:1;min-width:80px;background:{bg};border-radius:10px;"
                f"padding:10px;text-align:center;'>"
                f"<div style='font-size:20px;font-weight:700;color:{fg};'>{val:.0%}</div>"
                f"<div style='font-size:10px;color:#607D63;font-weight:600;"
                f"text-transform:uppercase;letter-spacing:.5px;margin-top:2px;'>{lbl}</div></div>")

    return f"""
<div style='background:#fff;border:1px solid #C8E6C9;border-radius:14px;
            padding:20px;font-family:sans-serif;margin-top:12px;'>
  <h2 style='margin:0 0 10px;color:#1B5E20;font-size:18px;'>{name}</h2>
  <div style='display:flex;align-items:center;gap:10px;flex-wrap:wrap;margin-bottom:14px;'>
    <span style='background:{ibg};color:{ifg};border:2px solid {ifg};
                 border-radius:20px;padding:4px 14px;font-size:13px;font-weight:700;'>
      {impact}</span>
    <span style='color:#607D63;font-size:13px;'>
      ML confidence: <strong style='color:#1A1A1A;'>{ml["confidence"]:.0%}</strong></span>
  </div>
  <div style='display:flex;align-items:center;gap:16px;flex-wrap:wrap;margin-bottom:16px;'>
    <div style='text-align:center;'>
      <div style='font-size:60px;font-weight:800;color:{gc};line-height:1;'>{eco["eco_score"]}</div>
      <div style='color:#607D63;font-size:11px;'>out of 10</div>
    </div>
    <div style='border-left:3px solid #C8E6C9;padding-left:14px;flex:1;min-width:150px;'>
      <div style='font-size:26px;font-weight:800;color:{gc};'>Grade {eco["grade"]}</div>
      <div style='font-size:14px;color:#3E4E3F;'>{eco["grade_label"]}</div>
      <div style='font-size:13px;color:#607D63;margin-top:4px;line-height:1.5;'>
        {get_score_description(eco["eco_score"])}</div>
    </div>
  </div>
  <div style='display:flex;gap:8px;flex-wrap:wrap;margin-bottom:16px;'>
    {pill("Deforestation", eco["metrics"]["deforestation_risk"])}
    {pill("Pollution",     eco["metrics"]["pollution_level"])}
    {pill("Biodiversity",  eco["metrics"]["biodiversity_impact"])}
  </div>
  <div style='border-top:1px solid #E8F5E9;padding-top:12px;'>
    <div style='color:#607D63;font-size:11px;font-weight:700;text-transform:uppercase;
                letter-spacing:1px;margin-bottom:8px;'>Badges</div>
    {badge_html}
  </div>
</div>"""

def alts_card(category, current=""):
    rows = DF[(DF["category"]==category) &
              (DF["eco_label"]=="Low Impact") &
              (DF["product_name"]!=current)].head(3)
    if rows.empty:
        return ""
    items = ""
    for _, r in rows.iterrows():
        ae = calculate_eco_score(r["deforestation_risk"],
                                  r["pollution_level"], r["biodiversity_impact"])
        gc = {"A":"#2E7D32","B":"#43A047","C":"#E65100",
              "D":"#C62828"}.get(ae["grade"],"#2E7D32")
        items += (
            f"<div style='background:#F1F8F1;border-left:4px solid #2E7D32;"
            f"border-radius:0 10px 10px 0;padding:12px 16px;margin:6px 0;"
            f"display:flex;justify-content:space-between;align-items:center;flex-wrap:wrap;gap:8px;'>"
            f"<div><div style='font-weight:700;color:#1B5E20;font-size:14px;'>{r['product_name']}</div>"
            f"<div style='color:#607D63;font-size:12px;'>📦 {r['packaging_type']}</div></div>"
            f"<div style='font-size:20px;font-weight:800;color:{gc};'>{ae['eco_score']}"
            f"<span style='font-size:12px;color:#607D63;'>/10</span></div></div>"
        )
    return (f"<div style='font-family:sans-serif;margin-top:16px;'>"
            f"<h3 style='color:#1B5E20;font-size:15px;margin:0 0 8px;'>"
            f"🌱 Eco-Friendly Alternatives</h3>{items}</div>")

def llm_card(text, model_name=""):
    out = ""
    for line in text.split("\n"):
        s = line.strip()
        if not s: continue
        if s.startswith("## "):
            out += (f"<h3 style='color:#1B5E20;font-size:15px;font-weight:700;"
                    f"margin:14px 0 5px;border-bottom:2px solid #C8E6C9;"
                    f"padding-bottom:3px;'>{s[3:]}</h3>")
        elif len(s)>2 and s[0].isdigit() and s[1]==".":
            c = s[3:].replace("**","<strong>",1).replace("**","</strong>",1)
            out += (f"<div style='background:#F1F8F1;border-left:4px solid #43A047;"
                    f"border-radius:0 8px 8px 0;padding:10px 14px;margin:5px 0;"
                    f"color:#1A1A1A;font-size:13px;'>#{s[0]} {c}</div>")
        elif s.startswith("- "):
            out += f"<li style='color:#1A1A1A;font-size:13px;margin:3px 0;'>{s[2:]}</li>"
        else:
            s2 = s.replace("**","<strong>",1).replace("**","</strong>",1)
            out += (f"<p style='color:#1A1A1A;font-size:13px;line-height:1.7;"
                    f"margin:5px 0;'>{s2}</p>")
    badge = (f"<span style='background:#F1F8F1;border:1px solid #C8E6C9;border-radius:20px;"
             f"padding:2px 10px;font-size:11px;color:#607D63;'>{model_name}</span>"
             ) if model_name else ""
    return (f"<div style='background:#fff;border:1px solid #C8E6C9;border-radius:14px;"
            f"padding:20px;font-family:sans-serif;margin-top:16px;'>"
            f"<h3 style='color:#1B5E20;font-size:16px;margin:0 0 12px;'>"
            f"🧠 AI Environmental Analysis {badge}</h3><div>{out}</div></div>")

def compare_card(p, eco, ml, winner):
    gc = {"A":"#2E7D32","B":"#43A047","C":"#E65100",
          "D":"#C62828","F":"#6A0000"}.get(eco["grade"],"#2E7D32")
    border = "#2E7D32" if winner else "#E0E0E0"
    pills  = "".join(
        f"<div style='background:#F5F5F5;border-radius:8px;padding:7px;text-align:center;'>"
        f"<div style='font-size:13px;font-weight:700;color:{_bar_color(v)};'>{v:.0%}</div>"
        f"<div style='font-size:10px;color:#607D63;'>{l}</div></div>"
        for l,v in [("🌲 Deforest.",eco["metrics"]["deforestation_risk"]),
                    ("💨 Pollution",eco["metrics"]["pollution_level"]),
                    ("🦋 Biodiv.",eco["metrics"]["biodiversity_impact"])]
    )
    return (f"<div style='background:#fff;border:2px solid {border};border-radius:14px;"
            f"padding:16px;font-family:sans-serif;'>"
            f"<div style='font-size:15px;font-weight:700;color:#1B5E20;margin-bottom:4px;'>"
            f"{'🏆 ' if winner else ''}{p['product_name']}</div>"
            f"<div style='color:#607D63;font-size:12px;margin-bottom:12px;'>"
            f"{p['category']} · {p['packaging_type']}</div>"
            f"<div style='display:flex;gap:12px;margin-bottom:12px;'>"
            f"<div style='text-align:center;'>"
            f"<div style='font-size:48px;font-weight:800;color:{gc};line-height:1;'>"
            f"{eco['eco_score']}</div>"
            f"<div style='color:#607D63;font-size:10px;'>SCORE</div></div>"
            f"<div style='text-align:center;'>"
            f"<div style='font-size:40px;font-weight:800;color:{gc};line-height:1;'>"
            f"{eco['grade']}</div>"
            f"<div style='color:#607D63;font-size:10px;'>GRADE</div></div></div>"
            f"<div style='display:grid;grid-template-columns:1fr 1fr 1fr;gap:6px;"
            f"margin-bottom:10px;'>{pills}</div>"
            f"<div style='font-size:12px;color:#3E4E3F;background:#F9FBF9;"
            f"border-radius:8px;padding:8px 12px;'>"
            f"ML: <strong>{ml['predicted_impact']}</strong> · "
            f"Conf: <strong>{ml['confidence']:.0%}</strong></div></div>")


# ── PDF ───────────────────────────────────────────────────────────────────────
def save_pdf(name, category, eco, ml, explanation, packaging=""):
    try:
        from reportlab.lib.pagesizes import letter
        from reportlab.lib.styles import ParagraphStyle
        from reportlab.lib.colors import HexColor
        from reportlab.platypus import (SimpleDocTemplate, Paragraph,
                                        Spacer, Table, TableStyle)
        from reportlab.lib.units import inch
        buf = io.BytesIO()
        doc = SimpleDocTemplate(buf, pagesize=letter, topMargin=inch,
                                bottomMargin=inch, leftMargin=inch, rightMargin=inch)
        TT = ParagraphStyle("TT",fontSize=20,fontName="Helvetica-Bold",
                             textColor=HexColor("#1B5E20"),spaceAfter=6)
        HH = ParagraphStyle("HH",fontSize=13,fontName="Helvetica-Bold",
                             textColor=HexColor("#2E7D32"),spaceAfter=4,spaceBefore=8)
        BB = ParagraphStyle("BB",fontSize=10,fontName="Helvetica",
                             textColor=HexColor("#1A1A1A"),spaceAfter=3,leading=14)
        content = [
            Paragraph("EcoVision Sustainability Report", TT),
            Paragraph(f"Generated: {datetime.now().strftime('%B %d, %Y  %H:%M')}", BB),
            Spacer(1,0.15*inch), Paragraph("Product Overview", HH),
        ]
        rows = [["Product",name],["Category",category],
                ["Eco Score",f"{eco['eco_score']}/10  Grade {eco['grade']}"],
                ["Impact",ml.get("predicted_impact","N/A")],["Packaging",packaging]]
        tbl = Table(rows, colWidths=[1.8*inch, 4.7*inch])
        tbl.setStyle(TableStyle([
            ("BACKGROUND",(0,0),(0,-1),HexColor("#E8F5E9")),
            ("TEXTCOLOR",(0,0),(0,-1),HexColor("#1B5E20")),
            ("TEXTCOLOR",(1,0),(1,-1),HexColor("#1A1A1A")),
            ("FONTNAME",(0,0),(-1,-1),"Helvetica"),
            ("FONTNAME",(0,0),(0,-1),"Helvetica-Bold"),
            ("FONTSIZE",(0,0),(-1,-1),10),
            ("GRID",(0,0),(-1,-1),0.5,HexColor("#C8E6C9")),
            ("ROWBACKGROUNDS",(0,0),(-1,-1),[HexColor("#FFFFFF"),HexColor("#F7F9F4")]),
            ("TOPPADDING",(0,0),(-1,-1),5),("BOTTOMPADDING",(0,0),(-1,-1),5),
        ]))
        content += [tbl, Spacer(1,0.12*inch), Paragraph("Analysis", HH)]
        for line in explanation.replace("**","").replace("## ","").split("\n")[:30]:
            if line.strip():
                content.append(Paragraph(line.strip(), BB))
        doc.build(content)
        data = buf.getvalue()
    except Exception:
        data = (f"EcoVision Report\nProduct: {name}\n"
                f"Eco Score: {eco['eco_score']}/10\nGrade: {eco['grade']}\n"
                f"Impact: {ml.get('predicted_impact')}\n\n{explanation}").encode("utf-8")
    ext  = "pdf" if data[:4]==b"%PDF" else "txt"
    path = os.path.join(tempfile.gettempdir(),
                        f"eco_{name[:20].replace(' ','_')}.{ext}")
    with open(path,"wb") as f: f.write(data)
    return path


# ── Core pipeline ─────────────────────────────────────────────────────────────
def run_eco(name, category, deforest, pollution, biodiversity,
            packaging, ingredients, groq_key):
    if groq_key:
        os.environ["GROQ_API_KEY"] = groq_key.strip()
    eco = calculate_eco_score(deforest, pollution, biodiversity)
    ml  = predict_impact(category, deforest, pollution, biodiversity)
    llm = generate_eco_explanation(
        name, category, deforest, pollution, biodiversity,
        eco["eco_score"], ml["predicted_impact"], packaging, ingredients,
    )
    return eco, ml, llm

EMPTY9 = ("", None, None, None, None, None, "", "", None)

# ── Tab handlers ──────────────────────────────────────────────────────────────
def search_fn(query, groq_key):
    try:
        model_err = wait_for_model()
        if model_err:
            return (err_html(f"Model training failed: {model_err}"),) + EMPTY9[1:]
        if not query.strip():
            return (err_html("Please enter a product name."),) + EMPTY9[1:]

        q = query.lower()
        m = DF[DF["product_name"].str.lower().str.contains(q, na=False)]
        if m.empty:
            m = DF[DF["category"].str.lower().str.contains(q, na=False)]
        if m.empty:
            return (err_html(f"'{query}' not found. Try: Patagonia · Lays · "
                             f"Head &amp; Shoulders · Shein · Ethique"),) + EMPTY9[1:]

        row = m.iloc[0]
        eco, ml, llm = run_eco(
            row["product_name"], row["category"],
            float(row["deforestation_risk"]), float(row["pollution_level"]),
            float(row["biodiversity_impact"]), row["packaging_type"],
            row["ingredients"], groq_key,
        )
        pdf = save_pdf(row["product_name"], row["category"], eco, ml,
                       llm["explanation"], row["packaging_type"])
        return (
            score_card(row["product_name"], eco, ml),
            gauge_chart(eco["eco_score"], eco["grade"]),
            metric_bar("Deforestation Risk", eco["metrics"]["deforestation_risk"], "🌲"),
            metric_bar("Pollution Level",    eco["metrics"]["pollution_level"],    "💨"),
            metric_bar("Biodiversity Impact",eco["metrics"]["biodiversity_impact"],"🦋"),
            prob_bar(ml["all_probabilities"]),
            llm_card(llm["explanation"], llm.get("model","")),
            alts_card(row["category"], row["product_name"]),
            pdf,
        )
    except Exception as e:
        return (err_html(f"Error: {str(e)}"),) + EMPTY9[1:]


def manual_fn(name, category, packaging, ingredients,
              deforest, pollution, biodiversity, groq_key):
    try:
        model_err = wait_for_model()
        if model_err:
            return (err_html(f"Model training failed: {model_err}"),) + EMPTY9[1:]
        if not name.strip():
            return (err_html("Please enter a product name."),) + EMPTY9[1:]

        eco, ml, llm = run_eco(name, category, deforest, pollution, biodiversity,
                                packaging, ingredients, groq_key)
        pdf = save_pdf(name, category, eco, ml, llm["explanation"], packaging)
        return (
            score_card(name, eco, ml),
            gauge_chart(eco["eco_score"], eco["grade"]),
            metric_bar("Deforestation Risk", eco["metrics"]["deforestation_risk"], "🌲"),
            metric_bar("Pollution Level",    eco["metrics"]["pollution_level"],    "💨"),
            metric_bar("Biodiversity Impact",eco["metrics"]["biodiversity_impact"],"🦋"),
            prob_bar(ml["all_probabilities"]),
            llm_card(llm["explanation"], llm.get("model","")),
            alts_card(category, ""),
            pdf,
        )
    except Exception as e:
        return (err_html(f"Error: {str(e)}"),) + EMPTY9[1:]


EMPTY_IMG = ("", None, "", None, None, None, None, "", None)

def image_fn(image_path, groq_key):
    try:
        model_err = wait_for_model()
        if model_err:
            return (err_html(f"Model training failed: {model_err}"),) + EMPTY_IMG[1:]
        if not image_path:
            return EMPTY_IMG

        vision   = classify_image(image_path)
        detected = vision["predicted_category"]
        cat_rows = DF[DF["category"]==detected]
        if cat_rows.empty:
            return EMPTY_IMG

        avg = cat_rows[["deforestation_risk","pollution_level","biodiversity_impact"]].mean()
        rep = cat_rows.iloc[0]
        eco, ml, llm = run_eco(
            f"Detected: {detected}", detected,
            float(avg["deforestation_risk"]), float(avg["pollution_level"]),
            float(avg["biodiversity_impact"]), rep["packaging_type"], "", groq_key,
        )
        scores = vision["all_scores"]
        fig_v  = go.Figure(go.Bar(
            x=list(scores.keys()), y=list(scores.values()),
            marker_color=["#2E7D32" if k==detected else "#A5D6A7" for k in scores],
            text=[f"{v:.0%}" for v in scores.values()],
            textposition="outside", textfont=dict(color="#1A1A1A", size=12),
        ))
        fig_v.update_layout(
            **BASE_LAYOUT, height=220,
            title=dict(text=f"Vision — '{detected}' detected",
                       font=dict(color="#1B5E20",size=12)),
            yaxis=dict(range=[0,1.3], gridcolor="#D4E8D4", tickformat=".0%"),
            xaxis=dict(showgrid=False),
        )
        vhtml = (f"<div style='background:#fff;border:1px solid #C8E6C9;"
                 f"border-radius:12px;padding:16px;font-family:sans-serif;'>"
                 f"<div style='color:#607D63;font-size:11px;font-weight:700;"
                 f"text-transform:uppercase;letter-spacing:1px;'>Detected</div>"
                 f"<div style='font-size:26px;font-weight:800;color:#1B5E20;'>"
                 f"📷 {detected}</div>"
                 f"<div style='color:#607D63;font-size:13px;margin-top:4px;'>"
                 f"Confidence: <strong style='color:#2E7D32;'>"
                 f"{vision['confidence']:.0%}</strong> · {vision['method']}</div></div>")
        pdf = save_pdf(f"Detected {detected}", detected, eco, ml,
                       llm["explanation"], rep["packaging_type"])
        return (vhtml, fig_v, score_card(f"Detected {detected}", eco, ml),
                gauge_chart(eco["eco_score"], eco["grade"]),
                metric_bar("Deforestation Risk",eco["metrics"]["deforestation_risk"],"🌲"),
                metric_bar("Pollution Level",   eco["metrics"]["pollution_level"],   "💨"),
                metric_bar("Biodiversity Impact",eco["metrics"]["biodiversity_impact"],"🦋"),
                llm_card(llm["explanation"], llm.get("model","")), pdf)
    except Exception as e:
        return (err_html(f"Error: {str(e)}"),) + EMPTY_IMG[1:]


EMPTY_CMP = ("","",None,None,None,"")

def compare_fn(n1, n2, groq_key):
    try:
        model_err = wait_for_model()
        if model_err:
            return (err_html(f"Model training failed: {model_err}"),) + EMPTY_CMP[1:]
        if groq_key:
            os.environ["GROQ_API_KEY"] = groq_key.strip()

        p1   = DF[DF["product_name"]==n1].iloc[0]
        p2   = DF[DF["product_name"]==n2].iloc[0]
        eco1 = calculate_eco_score(float(p1["deforestation_risk"]),
                                   float(p1["pollution_level"]),
                                   float(p1["biodiversity_impact"]))
        eco2 = calculate_eco_score(float(p2["deforestation_risk"]),
                                   float(p2["pollution_level"]),
                                   float(p2["biodiversity_impact"]))
        ml1  = predict_impact(p1["category"],float(p1["deforestation_risk"]),
                              float(p1["pollution_level"]),float(p1["biodiversity_impact"]))
        ml2  = predict_impact(p2["category"],float(p2["deforestation_risk"]),
                              float(p2["pollution_level"]),float(p2["biodiversity_impact"]))
        d1   = dict(name=n1, category=p1["category"],
                    eco_score=eco1["eco_score"], **eco1["metrics"])
        d2   = dict(name=n2, category=p2["category"],
                    eco_score=eco2["eco_score"], **eco2["metrics"])
        comp = compare_products_llm(d1, d2)
        w1   = eco1["eco_score"] >= eco2["eco_score"]
        return (compare_card(p1,eco1,ml1,w1), compare_card(p2,eco2,ml2,not w1),
                gauge_chart(eco1["eco_score"],eco1["grade"]),
                gauge_chart(eco2["eco_score"],eco2["grade"]),
                comparison_bar(n1,n2,eco1,eco2),
                llm_card(comp["comparison"],comp.get("model","")))
    except Exception as e:
        return (err_html(f"Error: {str(e)}"), "",None,None,None,"")


# ── Build UI ──────────────────────────────────────────────────────────────────
with gr.Blocks(title="EcoVision — Smart Sustainability Analyzer") as demo:

    gr.HTML("""
    <div style='text-align:center;padding:28px 16px 18px;
                background:linear-gradient(135deg,#E8F5E9,#F1F8E9);
                border-bottom:2px solid #C8E6C9;'>
      <div style='font-size:44px;font-weight:800;color:#1B5E20;'>🌿 EcoVision</div>
      <p style='color:#3E4E3F;font-size:14px;margin:8px 0 12px;'>
        Smart Sustainability Analyzer &nbsp;·&nbsp; ML · Computer Vision · LLM
      </p>
      <div style='display:flex;justify-content:center;gap:8px;flex-wrap:wrap;'>
        <span style='background:#fff;border:1px solid #C8E6C9;border-radius:20px;
             padding:3px 12px;font-size:12px;font-weight:600;color:#2E7D32;'>
             ✅ Rule-based Scoring</span>
        <span style='background:#fff;border:1px solid #C8E6C9;border-radius:20px;
             padding:3px 12px;font-size:12px;font-weight:600;color:#2E7D32;'>
             🤖 RandomForest ML</span>
        <span style='background:#fff;border:1px solid #C8E6C9;border-radius:20px;
             padding:3px 12px;font-size:12px;font-weight:600;color:#2E7D32;'>
             📷 MobileNetV2 CV</span>
        <span style='background:#fff;border:1px solid #C8E6C9;border-radius:20px;
             padding:3px 12px;font-size:12px;font-weight:600;color:#2E7D32;'>
             🧠 Groq LLM</span>
      </div>
    </div>""")

    with gr.Accordion("🔑 Groq API Key (optional)", open=False):
        gr.HTML("<p style='color:#3E4E3F;font-size:13px;margin:4px 0 8px;'>"
                "Free at <a href='https://console.groq.com' target='_blank' "
                "style='color:#2E7D32;font-weight:600;'>console.groq.com</a> — "
                "works without it using template explanations.</p>")
        groq_key = gr.Textbox(label="Groq API Key", type="password", placeholder="gsk_…")

    with gr.Tabs():

        # ── Tab 1: Search ─────────────────────────────────────────────────────
        with gr.Tab("🔍 Product Search"):
            gr.HTML("<p style='color:#3E4E3F;font-size:14px;margin:10px 0;'>"
                    "Search any of our <strong style='color:#1B5E20;'>53+ products</strong>"
                    " — shampoos, snacks, clothing, cosmetics, plastics.</p>")
            with gr.Row():
                s_in  = gr.Textbox(
                    label="Product name or category",
                    placeholder="e.g.  Head & Shoulders  |  Patagonia  |  Lays  |  Shein",
                    scale=4,
                )
                s_btn = gr.Button("🔬 Analyze", variant="primary", scale=1)

            with gr.Accordion("📋 Browse all products", open=False):
                gr.Dataframe(
                    value=DF[["product_name","category","eco_label","packaging_type"]].rename(
                        columns={"product_name":"Product","category":"Category",
                                 "eco_label":"Impact","packaging_type":"Packaging"}),
                    interactive=False,
                )

            s_card = gr.HTML()
            with gr.Row():
                s_gauge = gr.Plot()
                s_prob  = gr.Plot()
            with gr.Row():
                s_b1 = gr.Plot()
                s_b2 = gr.Plot()
                s_b3 = gr.Plot()
            s_llm  = gr.HTML()
            s_alts = gr.HTML()
            s_pdf  = gr.File(label="📄 Download Report")

            s_btn.click(
                fn=search_fn,
                inputs=[s_in, groq_key],
                outputs=[s_card, s_gauge, s_b1, s_b2, s_b3, s_prob,
                         s_llm, s_alts, s_pdf],
            )

        # ── Tab 2: Manual ─────────────────────────────────────────────────────
        with gr.Tab("✏️ Manual Input"):
            gr.HTML("<p style='color:#3E4E3F;font-size:14px;margin:10px 0;'>"
                    "Analyse any custom product by entering details manually.</p>")
            with gr.Row():
                with gr.Column():
                    m_name = gr.Textbox(label="Product Name",
                                         placeholder="e.g. My Eco Shampoo")
                    m_cat  = gr.Dropdown(ALL_CATS, label="Category", value="Shampoo")
                    m_pkg  = gr.Dropdown(PACKAGING_OPTS, label="Packaging",
                                          value="Plastic Bottle")
                    m_ingr = gr.Textbox(label="Ingredients / Materials",
                                         placeholder="e.g. Organic Cotton, Natural Dyes",
                                         lines=3)
                with gr.Column():
                    m_defo = gr.Slider(0.0,1.0,value=0.30,step=0.05,label="🌲 Deforestation Risk")
                    m_poll = gr.Slider(0.0,1.0,value=0.50,step=0.05,label="💨 Pollution Level")
                    m_bio  = gr.Slider(0.0,1.0,value=0.40,step=0.05,label="🦋 Biodiversity Impact")
                    m_btn  = gr.Button("🔬 Analyse Product", variant="primary")

            m_card = gr.HTML()
            with gr.Row():
                m_gauge = gr.Plot()
                m_prob  = gr.Plot()
            with gr.Row():
                m_b1 = gr.Plot()
                m_b2 = gr.Plot()
                m_b3 = gr.Plot()
            m_llm  = gr.HTML()
            m_alts = gr.HTML()
            m_pdf  = gr.File(label="📄 Download Report")

            m_btn.click(
                fn=manual_fn,
                inputs=[m_name, m_cat, m_pkg, m_ingr,
                        m_defo, m_poll, m_bio, groq_key],
                outputs=[m_card, m_gauge, m_b1, m_b2, m_b3, m_prob,
                         m_llm, m_alts, m_pdf],
            )

        # ── Tab 3: Image ──────────────────────────────────────────────────────
        with gr.Tab("📷 Image Scan"):
            gr.HTML("<p style='color:#3E4E3F;font-size:14px;margin:10px 0;'>"
                    "Upload a product photo — MobileNetV2 detects the category.</p>")
            with gr.Row():
                with gr.Column():
                    i_img = gr.Image(type="filepath", label="Upload Product Image",
                                      height=250)
                    i_btn = gr.Button("🔭 Scan & Analyse", variant="primary")
                with gr.Column():
                    i_detect = gr.HTML()
                    i_vchart = gr.Plot()

            i_card = gr.HTML()
            with gr.Row():
                i_gauge = gr.Plot()
            with gr.Row():
                i_b1 = gr.Plot()
                i_b2 = gr.Plot()
                i_b3 = gr.Plot()
            i_llm = gr.HTML()
            i_pdf = gr.File(label="📄 Download Report")

            i_btn.click(
                fn=image_fn,
                inputs=[i_img, groq_key],
                outputs=[i_detect, i_vchart, i_card, i_gauge,
                         i_b1, i_b2, i_b3, i_llm, i_pdf],
            )

        # ── Tab 4: Compare ────────────────────────────────────────────────────
        with gr.Tab("⚖️ Compare Products"):
            gr.HTML("<p style='color:#3E4E3F;font-size:14px;margin:10px 0;'>"
                    "Compare any two products side-by-side.</p>")
            with gr.Row():
                c_p1  = gr.Dropdown(ALL_PRODUCTS, label="🅰️ Product A",
                                     value=ALL_PRODUCTS[0], scale=5)
                gr.HTML("<div style='padding:22px 6px 0;font-size:18px;"
                        "font-weight:800;color:#2E7D32;'>VS</div>")
                c_p2  = gr.Dropdown(ALL_PRODUCTS, label="🅱️ Product B",
                                     value=ALL_PRODUCTS[5], scale=5)
                c_btn = gr.Button("⚖️ Compare", variant="primary", scale=2)

            with gr.Row():
                c_card1 = gr.HTML()
                c_card2 = gr.HTML()
            with gr.Row():
                c_g1 = gr.Plot()
                c_g2 = gr.Plot()
            c_bar = gr.Plot()
            c_llm = gr.HTML()

            c_btn.click(
                fn=compare_fn,
                inputs=[c_p1, c_p2, groq_key],
                outputs=[c_card1, c_card2, c_g1, c_g2, c_bar, c_llm],
            )

    gr.HTML("<div style='text-align:center;color:#607D63;font-size:12px;"
            "padding:14px;border-top:1px solid #C8E6C9;margin-top:16px;"
            "background:#F1F8F1;'>"
            "🌿 <strong>EcoVision</strong> · Gradio · scikit-learn · "
            "PyTorch MobileNetV2 · Groq LLM · Plotly</div>")


if __name__ == "__main__":
    demo.launch(server_name="0.0.0.0", server_port=7860)