app.py

from llama_cpp import Llama
from typing import Optional, Dict, Union
from huggingface_hub import hf_hub_download
import gradio as gr
import time

# Download the model from Hugging Face
model_path = hf_hub_download(
    repo_id="omeryentur/phi-3-sql", 
    filename="phi-3-sql.Q4_K_M.gguf", 
    use_auth_token=True
)

# Initialize the Llama model
llm = Llama(
    model_path=model_path, 
    n_ctx=512, 
    n_threads=1,
)

def generate_sql_query(text_input_schema: str, text_input_question: str):
    try:
        # Construct the prompt for the model
        prompt = f"""
<|system|>
{text_input_schema}

<|user|>
{text_input_question}

<|sql|>"""
        
        # Generate SQL query
        completion = llm(
            prompt, 
            max_tokens=512, 
            temperature=0, 
            stop=["<end_of_turn>"]
        )
        
        # Extract and return the generated SQL query
        generated = completion['choices'][0]['text'].strip()
        return {"sql_query":generated}
    except Exception as e:
        return {"error": str(e)}

# Create Gradio interface
with gr.Blocks() as demo:
    gr.Markdown("# Sql Query")
    
    with gr.Row():
        with gr.Column():
            text_input_schema = gr.TextArea(label="Schema")
            text_input_question = gr.Textbox(label="question")
            generate_btn = gr.Button("Create Sql Query")
    
    with gr.Row():
        with gr.Column():
            output = gr.JSON(label="Sql Query:")
    
    generate_btn.click(
        fn=generate_sql_query, 
        inputs=[text_input_schema, text_input_question], 
        outputs=[output]
    )

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