app.py

import gradio as gr
import pandas as pd
import difflib

# Load your CSV data with error handling
try:
    df = pd.read_csv("dialect_data.csv")
    print("CSV loaded successfully!")
    print("Available columns:", df.columns.tolist())
except Exception as e:
    print(f"Error loading CSV: {e}")
    # Create empty dataframe as fallback
    df = pd.DataFrame(columns=['Dialect Bengali', 'Translation in Ac English', 'Meaning'])

# Function to translate using your CSV data
def translate_text(text):
    if not text.strip():
        return "Please enter a phrase or question"
    
    # Check if dataframe is empty
    if df.empty:
        return "Database not loaded. Please check your CSV file."
    
    # Normalize text
    def normalize_phrase(phrase):
        return ''.join(char for char in phrase.lower() if char.isalpha() or char.isspace())
    
    user_clean = normalize_phrase(text)
    
    # Get all phrases from CSV for matching
    all_phrases = []
    for index, row in df.iterrows():
        try:
            csv_phrase = str(row['Dialect Bengali'])
            clean_csv = normalize_phrase(csv_phrase)
            translation = str(row['Translation in Ac English'])
            meaning = str(row['Meaning'])
            all_phrases.append((clean_csv, translation, meaning, csv_phrase))
        except:
            continue
    
    if not all_phrases:
        return "No phrases found in database. Check CSV format."
    
    # 1. First try exact match
    for clean_csv, translation, meaning, original in all_phrases:
        if user_clean == clean_csv:
            return f"✅ EXACT MATCH:\n💬 {translation}\n📖 {meaning}"
    
    # 2. Try close matches with high probability
    close_matches = difflib.get_close_matches(
        user_clean, 
        [clean_csv for clean_csv, trans, meaning, orig in all_phrases], 
        n=3,
        cutoff=0.6
    )
    
    if close_matches:
        suggestions = []
        for match in close_matches:
            for clean_csv, translation, meaning, original in all_phrases:
                if clean_csv == match:
                    similarity = int(difflib.SequenceMatcher(None, user_clean, clean_csv).ratio() * 100)
                    if similarity >= 60:
                        suggestions.append(f"🎯 '{original}' ({similarity}% match)\n💬 {translation}\n📖 {meaning}")
        
        if suggestions:
            return "🔍 SIMILAR PHRASES FOUND:\n\n" + "\n\n".join(suggestions)
    
    # 3. Show sample available phrases
    sample_phrases = [orig for clean_csv, trans, meaning, orig in all_phrases[:5]]
    return "❓ ASK DIFFERENT QUESTION\n\n" + \
           "📋 Try these phrases:\n" + \
           "\n".join([f"• '{phrase}'" for phrase in sample_phrases]) + \
           "\n\n💡 Example: 'Kita kobor?', 'Goto kali', 'gesle ni'"

# Create the app interface
demo = gr.Interface(
    fn=translate_text,
    inputs=gr.Textbox(label="Type phrase in our language", 
                     placeholder="Example: Kita kobor?, Goto kali, gesle ni..."),
    outputs=gr.Textbox(label="Translation Result"),
    title="🌍 Smart Dialect Translator",
    description="Translates with smart matching - finds similar phrases",
    examples=[["gesle ni"], ["Kita kobor?"], ["Goto kali"]]
)

# Launch the app
demo.launch()