app.py

import streamlit as st
import pandas as pd
import os
import json
import requests
from ctgan import CTGAN
from sklearn.preprocessing import LabelEncoder

def generate_schema(prompt):
    """Fetches schema from Hugging Face Spaces API."""
    API_URL = "https://infinitymatter-synthetic-data-generator.hf.space/"
    
    # Fetch API token securely
    hf_token = st.secrets["hf_token"]
    headers = {"Authorization": f"Bearer {hf_token}"}

    payload = {"data": [prompt]}

    try:
        response = requests.post(API_URL, headers=headers, json=payload)
        response.raise_for_status()
        schema = response.json()

        if 'columns' not in schema or 'types' not in schema or 'size' not in schema:
            raise ValueError("Invalid schema format!")

        return schema
    except requests.exceptions.RequestException as e:
        st.error(f"❌ API request failed: {e}")
        return None
    except json.JSONDecodeError:
        st.error("❌ Failed to parse JSON response.")
        return None


def train_and_generate_synthetic(real_data, schema, output_path):
    """Trains a CTGAN model and generates synthetic data."""
    categorical_cols = [col for col, dtype in zip(schema['columns'], schema['types']) if dtype == 'string']
    
    # Store label encoders
    label_encoders = {}
    for col in categorical_cols:
        le = LabelEncoder()
        real_data[col] = le.fit_transform(real_data[col])
        label_encoders[col] = le
    
    # Train CTGAN
    gan = CTGAN(epochs=300)
    gan.fit(real_data, categorical_cols)
    
    # Generate synthetic data
    synthetic_data = gan.sample(schema['size'])
    
    # Decode categorical columns
    for col in categorical_cols:
        synthetic_data[col] = label_encoders[col].inverse_transform(synthetic_data[col])
    
    # Save to CSV
    os.makedirs('outputs', exist_ok=True)
    synthetic_data.to_csv(output_path, index=False)
    st.success(f"✅ Synthetic data saved to {output_path}")

def fetch_data(domain):
    """Fetches real data for the given domain and ensures it's a valid DataFrame."""
    data_path = f"datasets/{domain}.csv"
    if os.path.exists(data_path):
        df = pd.read_csv(data_path)
        if not isinstance(df, pd.DataFrame) or df.empty:
            raise ValueError("❌ Loaded data is invalid!")
        return df
    else:
        st.error(f"❌ Dataset for {domain} not found.")
        return None

st.title("✨ AI-Powered Synthetic Dataset Generator")
st.write("Give a short description of the dataset you need, and AI will generate it for you using real data + GANs!")

# User input
user_prompt = st.text_input("Describe the dataset (e.g., 'Create dataset for hospital patients')", "")
domain = st.selectbox("Select Domain for Real Data", ["healthcare", "finance", "retail", "other"])

data = None
if st.button("Generate Schema"):
    if user_prompt.strip():
        with st.spinner("Generating schema..."):
            schema = generate_schema(user_prompt)

        if schema is None:
            st.error("❌ Schema generation failed. Please check API response.")
        else:
            st.success("✅ Schema generated successfully!")
            st.json(schema)
            data = fetch_data(domain)
    else:
        st.warning("⚠️ Please enter a dataset description before generating the schema.")

if data is not None and schema is not None:
    output_path = "outputs/synthetic_data.csv"
    if st.button("Generate Synthetic Data"):
        with st.spinner("Training GAN and generating synthetic data..."):
            train_and_generate_synthetic(data, schema, output_path)
        with open(output_path, "rb") as file:
            st.download_button("Download Synthetic Data", file, file_name="synthetic_data.csv", mime="text/csv")