sql_api.py

"""
SQL API Endpoint - Gradio-based API for remote SQL execution.
Deploy alongside Streamlit or as standalone Space.
"""
import gradio as gr
import duckdb
import pandas as pd
import json

def execute_sql(sql_query: str, max_rows: int = 1000) -> str:
    """
    Execute SQL query via DuckDB with HTTPFS.
    Returns JSON string with results or error.
    """
    try:
        con = duckdb.connect(':memory:')
        con.execute("INSTALL httpfs; LOAD httpfs;")
        
        # Add LIMIT if not present to prevent OOM
        sql_lower = sql_query.lower().strip()
        if 'limit' not in sql_lower:
            sql_query = f"{sql_query.rstrip(';')} LIMIT {max_rows}"
        
        result = con.execute(sql_query).fetchdf()
        
        return json.dumps({
            "status": "success",
            "rows": len(result),
            "columns": list(result.columns),
            "data": result.to_dict(orient='records')
        }, default=str)
    except Exception as e:
        return json.dumps({
            "status": "error",
            "message": str(e)
        })

# Gradio Interface
demo = gr.Interface(
    fn=execute_sql,
    inputs=[
        gr.Textbox(label="SQL Query", lines=5, placeholder="SELECT * FROM read_parquet('URL')"),
        gr.Number(label="Max Rows", value=1000, precision=0)
    ],
    outputs=gr.JSON(label="Result"),
    title="🦆 DuckDB SQL API",
    description="Execute SQL queries on remote HF Parquet files. Results returned as JSON.",
    examples=[
        ["SELECT COUNT(*) as total FROM read_parquet('https://huggingface.co/datasets/gionuibk/hyperliquid-l4-data/resolve/main/data/l4_universal_data_1765167177_1.parquet')", 10],
    ]
)

if __name__ == "__main__":
    demo.launch(server_port=7860)