app.py

import gradio as gr
import pandas as pd
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
import sqlite3

# Load model and tokenizer
model_name = "mrm8488/t5-base-finetuned-wikiSQL"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)

def generate_sql_query(natural_language, data):
    # Load uploaded CSV
    df = pd.read_csv(data.name)
    
    # Create in-memory SQLite DB
    conn = sqlite3.connect(":memory:")
    df.to_sql("data_table", conn, index=False, if_exists="replace")
    
    # Create schema description
    schema = ", ".join([f"{col}" for col in df.columns])
    
    # Combine user query and schema
    input_text = f"translate English to SQL: {natural_language} | table columns: {schema}"
    
    # Generate SQL query
    inputs = tokenizer(input_text, return_tensors="pt")
    outputs = model.generate(**inputs, max_length=256)
    sql_query = tokenizer.decode(outputs[0], skip_special_tokens=True)
    
    try:
        # Execute the generated SQL query
        result_df = pd.read_sql_query(sql_query, conn)
    except Exception as e:
        result_df = pd.DataFrame({"Error": [str(e)]})
    
    conn.close()
    return sql_query, result_df.head()

# Gradio UI
iface = gr.Interface(
    fn=generate_sql_query,
    inputs=[
        gr.Textbox(label="Enter your question (Natural Language)", placeholder="e.g., Show customers with age > 30"),
        gr.File(label="Upload CSV dataset")
    ],
    outputs=[
        gr.Textbox(label="Generated SQL Query"),
        gr.Dataframe(label="Query Result")
    ],
    title="🧠 Natural Language to SQL Generator",
    description="Upload a CSV file and ask questions in plain English. The app converts them into SQL and shows the result.",
    allow_flagging="never"
)

if __name__ == "__main__":
    iface.launch()