app.py

import streamlit as st
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import os
import pickle
import io
import traceback
import sys
import base64
from datetime import datetime

# Import ML libraries with error handling
try:
    from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer
    from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
    from sklearn.preprocessing import LabelEncoder
    st.success("✅ Sklearn imported successfully")
except ImportError as e:
    st.error(f"❌ Sklearn import error: {e}")

# Import custom modules with error handling
try:
    from NoCodeTextClassifier.EDA import Informations, Visualizations
    from NoCodeTextClassifier.preprocessing import process, TextCleaner, Vectorization  
    from NoCodeTextClassifier.models import Models
    st.success("✅ NoCodeTextClassifier imported successfully")
except ImportError as e:
    st.error(f"❌ NoCodeTextClassifier import error: {e}")
    st.info("Please ensure NoCodeTextClassifier package is installed")

# Set page config
st.set_page_config(page_title="Fixed Text Classification", page_icon="🔧", layout="wide")

# Debug section
st.sidebar.header("🔍 Debug Information")
debug_mode = st.sidebar.checkbox("Enable Debug Mode", value=True)

def debug_log(message, level="INFO"):
    """Debug logging function"""
    if debug_mode:
        timestamp = datetime.now().strftime("%H:%M:%S")
        st.sidebar.write(f"**{timestamp} [{level}]:** {message}")

# Alternative file upload methods
def alternative_file_upload():
    """Alternative file upload methods to bypass 403 error"""
    st.subheader("🔧 Alternative File Upload Methods")
    
    # Method 1: Text area paste
    st.markdown("### Method 1: Copy-Paste CSV Content")
    st.info("Copy your CSV content and paste it in the text area below")
    
    csv_content = st.text_area(
        "Paste your CSV content here:",
        height=200,
        placeholder="name,age,city\nJohn,25,New York\nJane,30,London"
    )
    
    if csv_content and st.button("Load from Text Area", type="primary"):
        try:
            df = pd.read_csv(io.StringIO(csv_content))
            st.success("✅ CSV loaded from text area!")
            return df, "text_area"
        except Exception as e:
            st.error(f"Error parsing CSV: {e}")
            return None, None
    
    # Method 2: Base64 upload (for advanced users)
    st.markdown("### Method 2: Base64 Upload")
    with st.expander("For Advanced Users - Base64 Upload"):
        st.info("Convert your CSV to base64 and paste here")
        st.code("""
# Python code to convert CSV to base64:
import base64
with open('your_file.csv', 'rb') as f:
    encoded = base64.b64encode(f.read()).decode()
    print(encoded)
        """)
        
        base64_content = st.text_area("Paste base64 encoded CSV:", height=100)
        
        if base64_content and st.button("Load from Base64"):
            try:
                decoded = base64.b64decode(base64_content)
                df = pd.read_csv(io.BytesIO(decoded))
                st.success("✅ CSV loaded from base64!")
                return df, "base64"
            except Exception as e:
                st.error(f"Error decoding base64: {e}")
                return None, None
    
    # Method 3: Sample data
    st.markdown("### Method 3: Use Sample Data")
    if st.button("Load Sample Text Classification Data"):
        # Create sample data
        sample_data = {
            'text': [
                'I love this product, it works great!',
                'This is terrible, waste of money',
                'Good quality and fast delivery',
                'Not satisfied with the purchase',
                'Excellent service and support',
                'Poor quality, arrived damaged',
                'Amazing product, highly recommend',
                'Disappointed with the results'
            ],
            'label': ['positive', 'negative', 'positive', 'negative', 
                     'positive', 'negative', 'positive', 'negative']
        }
        df = pd.DataFrame(sample_data)
        st.success("✅ Sample data loaded!")
        return df, "sample"
    
    return None, None

def safe_file_uploader_with_fallback():
    """Try normal upload first, then fallback methods"""
    st.markdown("### 📁 Upload Your CSV File")
    
    # Try standard uploader first
    uploaded_file = st.file_uploader(
        "Choose a CSV file", 
        type=['csv'],
        help="If upload fails with 403 error, use alternative methods below"
    )
    
    if uploaded_file is not None:
        try:
            debug_log("📁 File uploaded successfully via standard method")
            df = pd.read_csv(uploaded_file)
            st.success("✅ File uploaded successfully!")
            return df, "standard"
        except Exception as e:
            st.error(f"Error reading uploaded file: {e}")
            debug_log(f"❌ Standard upload failed: {e}", "ERROR")
    
    # If standard upload fails or no file uploaded, show alternatives
    st.markdown("---")
    st.markdown("### 🔄 Alternative Upload Methods")
    st.warning("If you're getting a 403 error, try one of these alternative methods:")
    
    return alternative_file_upload()

# Utility functions (same as before but with debug)
def save_artifacts(obj, folder_name, file_name):
    """Save artifacts with debugging"""
    debug_log(f"💾 Saving {file_name} to {folder_name}")
    try:
        os.makedirs(folder_name, exist_ok=True)
        full_path = os.path.join(folder_name, file_name)
        
        with open(full_path, 'wb') as f:
            pickle.dump(obj, f)
        
        debug_log(f"✅ Successfully saved {file_name}")
        return True
        
    except Exception as e:
        debug_log(f"❌ Error saving {file_name}: {str(e)}", "ERROR")
        st.error(f"Save error: {str(e)}")
        return False

def load_artifacts(folder_name, file_name):
    """Load artifacts with debugging"""
    debug_log(f"📂 Loading {file_name} from {folder_name}")
    try:
        full_path = os.path.join(folder_name, file_name)
        
        if not os.path.exists(full_path):
            debug_log(f"❌ File not found: {full_path}", "ERROR")
            return None
            
        with open(full_path, 'rb') as f:
            obj = pickle.load(f)
        
        debug_log(f"✅ Successfully loaded {file_name}")
        return obj
        
    except Exception as e:
        debug_log(f"❌ Error loading {file_name}: {str(e)}", "ERROR")
        return None

def load_model(model_name):
    """Load model with debugging"""
    debug_log(f"🤖 Loading model: {model_name}")
    return load_artifacts("models", model_name)

def predict_text(model_name, text, vectorizer_type="tfidf"):
    """Make prediction with debugging"""
    debug_log(f"🔮 Starting prediction with {model_name}")
    
    try:
        # Load components
        model = load_model(model_name)
        if model is None:
            return None, None
        
        vectorizer_file = f"{vectorizer_type}_vectorizer.pkl"
        vectorizer = load_artifacts("artifacts", vectorizer_file)
        if vectorizer is None:
            return None, None
        
        encoder = load_artifacts("artifacts", "encoder.pkl")
        if encoder is None:
            return None, None
        
        debug_log("🧹 Cleaning text...")
        text_cleaner = TextCleaner()
        clean_text = text_cleaner.clean_text(text)
        
        debug_log("🔢 Vectorizing text...")
        text_vector = vectorizer.transform([clean_text])
        
        debug_log("🎯 Making prediction...")
        prediction = model.predict(text_vector)
        prediction_proba = None
        
        if hasattr(model, 'predict_proba'):
            try:
                prediction_proba = model.predict_proba(text_vector)[0]
            except:
                debug_log("No prediction probabilities available", "WARNING")
        
        predicted_label = encoder.inverse_transform(prediction)[0]
        debug_log(f"✅ Prediction complete: {predicted_label}")
        
        return predicted_label, prediction_proba
        
    except Exception as e:
        debug_log(f"❌ Prediction error: {str(e)}", "ERROR")
        st.error(f"Prediction error: {str(e)}")
        return None, None

# Main App
st.title('🔧 Fixed Text Classification App')
st.write('Workaround version to bypass 403 upload errors')

# Show environment info in sidebar if debug mode
if debug_mode:
    st.sidebar.subheader("🖥️ Environment Info")
    st.sidebar.write(f"Python version: {sys.version}")
    st.sidebar.write(f"Streamlit version: {st.__version__}")
    st.sidebar.write(f"Current directory: {os.getcwd()}")

# Navigation
section = st.sidebar.radio("Choose Section", [
    "Upload Data", "Data Analysis", "Train Model", "Predictions"
])

# Session state
if 'train_df' not in st.session_state:
    st.session_state.train_df = None
if 'upload_method' not in st.session_state:
    st.session_state.upload_method = None
if 'vectorizer_type' not in st.session_state:
    st.session_state.vectorizer_type = "tfidf"

# Upload Data Section
if section == "Upload Data":
    st.subheader("📁 Upload Your Dataset")
    
    df, method = safe_file_uploader_with_fallback()
    
    if df is not None:
        st.session_state.train_df = df
        st.session_state.upload_method = method
        
        st.write("### 📊 Data Preview")
        st.dataframe(df.head())
        
        st.write("### 📈 Basic Info")
        col1, col2, col3 = st.columns(3)
        with col1:
            st.metric("Rows", df.shape[0])
        with col2:
            st.metric("Columns", df.shape[1])
        with col3:
            st.metric("Missing Values", df.isnull().sum().sum())
        
        st.write("### 🏷️ Select Columns")
        columns = df.columns.tolist()
        
        col1, col2 = st.columns(2)
        with col1:
            text_column = st.selectbox("Select text column:", columns)
        with col2:
            target_column = st.selectbox("Select target/label column:", columns)
        
        if text_column and target_column:
            st.session_state.text_column = text_column
            st.session_state.target_column = target_column
            
            # Show sample data
            st.write("### 📝 Sample Data")
            sample_df = df[[text_column, target_column]].head()
            st.dataframe(sample_df)
            
            # Show target distribution
            st.write("### 🎯 Target Distribution")
            target_counts = df[target_column].value_counts()
            st.bar_chart(target_counts)
            
            st.success("✅ Data ready for processing!")

# Data Analysis Section
elif section == "Data Analysis":
    if st.session_state.train_df is not None:
        df = st.session_state.train_df
        text_col = st.session_state.get('text_column')
        target_col = st.session_state.get('target_column')
        
        if text_col and target_col:
            st.subheader("📊 Data Analysis")
            
            try:
                # Process data using custom classes
                info = Informations(df, text_col, target_col)
                df['clean_text'] = info.clean_text()
                df['text_length'] = info.text_length()
                
                # Update session state
                st.session_state.train_df = df
                
                # Show analysis
                st.write("**Data Shape:**", info.shape())
                st.write("**Class Distribution:**", info.class_imbalanced())
                st.write("**Missing Values:**", info.missing_values())
                
                # Text length analysis
                st.write("**Text Length Analysis:**")
                st.write(info.analysis_text_length('text_length'))
                
                # Visualizations
                vis = Visualizations(df, text_col, target_col)
                
                col1, col2 = st.columns(2)
                with col1:
                    st.write("**Class Distribution:**")
                    vis.class_distribution()
                
                with col2:
                    st.write("**Text Length Distribution:**")
                    vis.text_length_distribution()
                
            except Exception as e:
                st.error(f"Error in analysis: {e}")
                debug_log(f"Analysis error: {e}", "ERROR")
        else:
            st.warning("Please select text and target columns in the Upload Data section.")
    else:
        st.warning("Please upload data first.")

# Train Model Section  
elif section == "Train Model":
    if st.session_state.train_df is not None:
        df = st.session_state.train_df
        text_col = st.session_state.get('text_column')
        target_col = st.session_state.get('target_column')
        
        if text_col and target_col and 'clean_text' in df.columns:
            st.subheader("🤖 Train Model")
            
            col1, col2 = st.columns(2)
            
            with col1:
                model_choice = st.selectbox("Choose Model:", [
                    "Logistic Regression", "Decision Tree", "Random Forest",
                    "Linear SVC", "SVC", "Multinomial Naive Bayes"
                ])
            
            with col2:
                vectorizer_choice = st.selectbox("Choose Vectorizer:", 
                    ["Tfidf Vectorizer", "Count Vectorizer"])
            
            if st.button("🚀 Train Model", type="primary"):
                with st.spinner("Training model..."):
                    try:
                        # Prepare data
                        if vectorizer_choice == "Tfidf Vectorizer":
                            vectorizer = TfidfVectorizer(max_features=10000)
                            st.session_state.vectorizer_type = "tfidf"
                        else:
                            vectorizer = CountVectorizer(max_features=10000)
                            st.session_state.vectorizer_type = "count"
                        
                        # Label encoding
                        label_encoder = LabelEncoder()
                        y = label_encoder.fit_transform(df[target_col])
                        X = vectorizer.fit_transform(df['clean_text'])
                        
                        # Split data
                        X_train, X_test, y_train, y_test = process.split_data(X, y)
                        
                        # Save artifacts
                        save_artifacts(vectorizer, "artifacts", f"{st.session_state.vectorizer_type}_vectorizer.pkl")
                        save_artifacts(label_encoder, "artifacts", "encoder.pkl")
                        
                        # Train model
                        models = Models(X_train=X_train, X_test=X_test, y_train=y_train, y_test=y_test)
                        
                        if model_choice == "Logistic Regression":
                            models.LogisticRegression()
                        elif model_choice == "Decision Tree":
                            models.DecisionTree()
                        elif model_choice == "Random Forest":
                            models.RandomForestClassifier()
                        elif model_choice == "Linear SVC":
                            models.LinearSVC()
                        elif model_choice == "SVC":
                            models.SVC()
                        elif model_choice == "Multinomial Naive Bayes":
                            models.MultinomialNB()
                        
                        st.success("🎉 Model trained successfully!")
                        
                    except Exception as e:
                        st.error(f"Training error: {e}")
                        debug_log(f"Training error: {e}", "ERROR")
        else:
            st.warning("Please complete data analysis first to process the text data.")
    else:
        st.warning("Please upload data first.")

# Predictions Section
elif section == "Predictions":
    st.subheader("🔮 Make Predictions")
    
    # Check for models
    if os.path.exists("models") and os.listdir("models"):
        available_models = [f for f in os.listdir("models") if f.endswith('.pkl')]
        
        if available_models:
            selected_model = st.selectbox("Choose trained model:", available_models)
            
            # Single prediction
            st.write("### Single Text Prediction")
            text_input = st.text_area("Enter text to classify:", height=100)
            
            if st.button("🎯 Predict") and text_input:
                prediction, probabilities = predict_text(
                    selected_model, 
                    text_input, 
                    st.session_state.get('vectorizer_type', 'tfidf')
                )
                
                if prediction is not None:
                    st.success(f"**Prediction:** {prediction}")
                    
                    if probabilities is not None:
                        encoder = load_artifacts("artifacts", "encoder.pkl")
                        if encoder is not None:
                            prob_df = pd.DataFrame({
                                'Class': encoder.classes_,
                                'Probability': probabilities
                            }).sort_values('Probability', ascending=False)
                            
                            st.bar_chart(prob_df.set_index('Class'))
        else:
            st.info("No trained models found. Train a model first.")
    else:
        st.info("No models directory found. Train a model first.")

# Show upload method used in sidebar
if st.session_state.upload_method:
    st.sidebar.success(f"✅ Data loaded via: {st.session_state.upload_method}")