app.py

import pandas as pd
import gradio as gr
import matplotlib.pyplot as plt
import io
from PIL import Image
from transformers import pipeline

# ----------------------------
# Load AI Models
# ----------------------------
sentiment_model = pipeline(
    "sentiment-analysis",
    model="distilbert-base-uncased-finetuned-sst-2-english"
)

theme_model = pipeline(
    "zero-shot-classification",
    model="valhalla/distilbart-mnli-12-3"
)

THEMES = [
    "product quality",
    "delivery",
    "price",
    "packaging",
    "taste",
    "customer service"
]

# ----------------------------
# Analyze One Review
# ----------------------------
def analyze_review(text):
    text = str(text)[:512]

    sentiment_result = sentiment_model(text)[0]
    label = sentiment_result["label"]
    confidence = round(sentiment_result["score"] * 100, 1)

    theme_result = theme_model(text, THEMES)
    top_theme = theme_result["labels"][0]
    theme_score = round(theme_result["scores"][0] * 100, 1)

    return label, confidence, top_theme, theme_score


# ----------------------------
# Build Charts
# ----------------------------
def build_chart(df):
    fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 4))

    sentiment_counts = df["Sentiment"].value_counts()

    color_map = {
        "POSITIVE": "#22c55e",
        "NEGATIVE": "#ef4444",
        "NEUTRAL": "#94a3b8"
    }

    colors = [color_map.get(label, "#94a3b8") for label in sentiment_counts.index]

    ax1.pie(
        sentiment_counts.values,
        labels=sentiment_counts.index,
        autopct="%1.0f%%",
        colors=colors,
        startangle=90,
        wedgeprops=dict(width=0.5)
    )

    ax1.set_title("Sentiment Split")

    theme_counts = df["Top Theme"].value_counts()

    ax2.barh(theme_counts.index, theme_counts.values)
    ax2.set_title("Top Themes")
    ax2.invert_yaxis()

    plt.tight_layout()

    buf = io.BytesIO()
    plt.savefig(buf, format="png")
    plt.close(fig)

    buf.seek(0)

    return Image.open(buf)


# ----------------------------
# Analyze Uploaded Excel File
# ----------------------------
def analyze_excel(file):

    df = pd.read_excel(file.name)

    if "review_text" not in df.columns:

        if "Text" in df.columns:
            df = df.rename(columns={"Text": "review_text"})
        else:
            return (
                None,
                "❌ Excel file must contain a 'review_text' or 'Text' column.",
                None
            )

    df = df.dropna(subset=["review_text"])

    results = []

    for _, row in df.iterrows():

        label, confidence, theme, theme_confidence = analyze_review(
            row["review_text"]
        )

        results.append(
            {
                "Review": str(row["review_text"])[:100] + "...",
                "Sentiment": label,
                "Confidence %": confidence,
                "Top Theme": theme,
                "Theme Confidence %": theme_confidence
            }
        )

    output_df = pd.DataFrame(results)

    positive_pct = round(
        (output_df["Sentiment"] == "POSITIVE").mean() * 100,
        1
    )

    summary = (
        f"✅ Successfully analyzed {len(output_df)} reviews.\n"
        f"Positive Reviews: {positive_pct}%"
    )

    chart = build_chart(output_df)

    return output_df, summary, chart


# ----------------------------
# Gradio UI
# ----------------------------
with gr.Blocks(title="Customer Feedback Analyzer") as app:

    gr.Markdown(
        """
# 🧠 AI Customer Feedback Analyzer

Upload an Excel (.xlsx) file containing customer reviews.

The app automatically predicts:

- Sentiment
- Confidence
- Top Theme
- Theme Confidence
"""
    )

    file_input = gr.File(
        label="Upload Excel File (.xlsx)",
        file_types=[".xlsx"]
    )

    analyze_button = gr.Button("Analyze")

    summary_output = gr.Textbox(label="Summary")

    chart_output = gr.Image(label="Charts")

    table_output = gr.Dataframe(label="Analysis Results")

    analyze_button.click(
        analyze_excel,
        inputs=file_input,
        outputs=[
            table_output,
            summary_output,
            chart_output
        ]
    )


app.launch()