Create app.py

app.py

@@ -0,0 +1,150 @@
+import gradio as gr
+import pandas as pd
+import plotly.graph_objects as go
+
+# ── load data ──────────────────────────────────────────────
+df = pd.read_csv("restaurant_master_scored.csv")
+
+CUISINES    = ["All"] + sorted(df["cuisine_type"].unique().tolist())
+PRICES      = ["All"] + ["€", "€€", "€€€", "€€€€"]
+ARRONDS     = ["All"] + sorted(df["arrondissement"].unique().tolist())
+
+LABEL_COLOR = {"Worth It": "#1A6B3C", "Maybe": "#7D5A00", "Skip It": "#8B1A1A"}
+LABEL_BG    = {"Worth It": "#EAF3DE", "Maybe": "#FAEEDA",  "Skip It": "#FCEBEB"}
+
+# ── score component chart ──────────────────────────────────
+def make_chart(row):
+    components = {
+        "Sentiment adj":  round(row["sentiment_adj_norm"] * 0.35 * 100, 1),
+        "Price fairness": round(row["price_fairness"]     * 0.25 * 100, 1),
+        "Consistency":    round(row["consistency_index"]  * 0.20 * 100, 1),
+        "Trust":          round(row["trust_score"]        * 0.10 * 100, 1),
+        "Divergence −":   round(-row["star_sentiment_divergence"] * 0.10 * 100, 1),
+    }
+    colors = ["#3D2B8E", "#7B6FCA", "#1A6B3C", "#7D5A00", "#8B1A1A"]
+    fig = go.Figure(go.Bar(
+        x=list(components.values()),
+        y=list(components.keys()),
+        orientation="h",
+        marker_color=colors,
+        text=[f"{v:+.1f}" for v in components.values()],
+        textposition="outside",
+    ))
+    fig.update_layout(
+        margin=dict(l=10, r=60, t=10, b=10),
+        height=220,
+        xaxis=dict(showgrid=False, zeroline=True,
+                   zerolinecolor="#cccccc", title="Points"),
+        yaxis=dict(showgrid=False),
+        paper_bgcolor="rgba(0,0,0,0)",
+        plot_bgcolor="rgba(0,0,0,0)",
+        font=dict(size=12),
+    )
+    return fig
+
+# ── AI explanation (rule-based, no API key needed) ─────────
+def explain(row):
+    name   = row["restaurant_name"]
+    label  = row["worth_it_label"]
+    score  = row["true_score"]
+    price  = row["price_category"]
+    sent   = row["sentiment_score"]
+    stars  = row["star_rating"]
+    vols   = row["sentiment_volatility"]
+    pf     = row["price_fairness"]
+
+    sentiment_word = "positive" if sent > 0.2 else ("neutral" if sent > -0.2 else "negative")
+    volatility_note = (
+        " Reviews are polarised — strong opinions on both sides."
+        if vols > 0.5 else
+        " Reviews are consistent."
+    )
+    value_note = (
+        f" At {price}, it delivers excellent value for money."
+        if pf > 0.6 else
+        f" At {price}, the price-to-experience ratio is fair."
+        if pf > 0.3 else
+        f" At {price}, some diners feel it falls short of expectations."
+    )
+
+    if label == "Worth It":
+        verdict = f"This restaurant scores {score:.0f}/100 and is genuinely worth a visit."
+    elif label == "Maybe":
+        verdict = f"This restaurant scores {score:.0f}/100 — a solid option depending on your preferences."
+    else:
+        verdict = f"This restaurant scores {score:.0f}/100 — consider alternatives before booking."
+
+    return (
+        f"**{name}** — _{label}_\n\n"
+        f"{verdict} "
+        f"Customer reviews are broadly {sentiment_word} (raw sentiment: {sent:+.2f}), "
+        f"and the star rating is {stars:.1f}/5.{volatility_note}{value_note}"
+    )
+
+# ── main function ──────────────────────────────────────────
+def search(cuisine, price, arrond, top_n):
+    filtered = df.copy()
+    if cuisine != "All":
+        filtered = filtered[filtered["cuisine_type"] == cuisine]
+    if price != "All":
+        filtered = filtered[filtered["price_category"] == price]
+    if arrond != "All":
+        filtered = filtered[filtered["arrondissement"] == arrond]
+
+    if filtered.empty:
+        return "No restaurants match your filters.", None, ""
+
+    top = filtered.nlargest(int(top_n), "true_score").reset_index(drop=True)
+    best = top.iloc[0]
+
+    # Results table
+    display = top[[
+        "restaurant_name", "cuisine_type", "arrondissement",
+        "price_category", "star_rating", "true_score", "worth_it_label"
+    ]].rename(columns={
+        "restaurant_name": "Name",
+        "cuisine_type": "Cuisine",
+        "arrondissement": "Area",
+        "price_category": "Price",
+        "star_rating": "Stars",
+        "true_score": "True Score",
+        "worth_it_label": "Verdict",
+    })
+
+    chart = make_chart(best)
+    explanation = explain(best)
+
+    return display, chart, explanation
+
+# ── interface ──────────────────────────────────────────────
+with gr.Blocks(title="Restaurant Worth-It Score — Paris") as demo:
+    gr.Markdown("## Restaurant Worth-It Score — Paris\n"
+                "AI-powered ratings that go beyond the stars. "
+                "Select your filters to find restaurants that truly deliver.")
+
+    with gr.Row():
+        cuisine_dd = gr.Dropdown(CUISINES, value="All", label="Cuisine type")
+        price_dd   = gr.Dropdown(PRICES,   value="All", label="Price tier")
+        arrond_dd  = gr.Dropdown(ARRONDS,  value="All", label="Arrondissement")
+        top_n_sl   = gr.Slider(3, 20, value=10, step=1, label="Results to show")
+
+    search_btn = gr.Button("Find restaurants", variant="primary")
+
+    gr.Markdown("### Top results")
+    results_table = gr.Dataframe(label="Ranked restaurants", interactive=False)
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            gr.Markdown("#### Score breakdown — top result")
+            score_chart = gr.Plot(label="")
+        with gr.Column(scale=1):
+            gr.Markdown("#### AI explanation — top result")
+            explanation_box = gr.Markdown()
+
+    search_btn.click(
+        fn=search,
+        inputs=[cuisine_dd, price_dd, arrond_dd, top_n_sl],
+        outputs=[results_table, score_chart, explanation_box],
+    )
+
+demo.launch()