app.py

import os
import streamlit as st
from dotenv import load_dotenv  # Import load_dotenv to load environment variables
from langchain import HuggingFaceHub

# Load environment variables from the .env file
load_dotenv()

# Set your Hugging Face API token from the environment variable
HUGGINGFACE_API_TOKEN = os.getenv("HUGGINGFACE_API_TOKEN")

# Function to return the SQL query from natural language input
def load_sql_query(question):
    try:
        # Initialize the Hugging Face model using LangChain's HuggingFaceHub class
        llm = HuggingFaceHub(
            repo_id="Salesforce/grappa_large_jnt",  # Hugging Face model repo for text-to-SQL
            huggingfacehub_api_token=HUGGINGFACE_API_TOKEN,  # Pass your API token
            model_kwargs={"temperature": 0.3}  # Optional: Adjust response randomness
        )
        
        # Call the model with the user's question and get the SQL query
        sql_query = llm.predict(question)
        return sql_query
    except Exception as e:
        # Capture and return any exceptions or errors
        return f"Error: {str(e)}"

# Streamlit App UI starts here
st.set_page_config(page_title="Text-to-SQL Demo", page_icon=":robot:")
st.header("Text-to-SQL Demo")

# Function to get user input
def get_text():
    input_text = st.text_input("Ask a question (related to a database):", key="input")
    return input_text

# Get user input
user_input = get_text()

# Create a button for generating the SQL query
submit = st.button('Generate SQL')

# If the generate button is clicked and user input is not empty
if submit and user_input:
    response = load_sql_query(user_input)
    st.subheader("Generated SQL Query:")
    st.write(response)
elif submit:
    st.warning("Please enter a question.")  # Warning for empty input