import gradio as gr import pandas as pd import duckdb import requests import re import os from io import BytesIO # 🔐 Read Together API key from Hugging Face Secrets TOGETHER_API_KEY = os.getenv("TOGETHER_API_KEY") # 🧠 Generate SQL from Prompt def generate_sql_from_prompt(prompt, df): schema = ", ".join([f"{col} ({str(dtype)})" for col, dtype in df.dtypes.items()]) full_prompt = f""" You are a SQL expert. Here is a table called 'df' with the following schema: {schema} User question: "{prompt}" Write a valid SQL query using the 'df' table. Return only the SQL code. """ url = "https://api.together.xyz/v1/chat/completions" headers = { "Authorization": f"Bearer {TOGETHER_API_KEY}", "Content-Type": "application/json" } payload = { "model": "mistralai/Mixtral-8x7B-Instruct-v0.1", "messages": [{"role": "user", "content": full_prompt}], "temperature": 0.2, "max_tokens": 200 } try: response = requests.post(url, headers=headers, json=payload) response.raise_for_status() result = response.json() return result['choices'][0]['message']['content'].strip("```sql").strip("```").strip() except Exception as e: return f"Error in API call: {str(e)}" # 🧽 Clean SQL for DuckDB def clean_sql_for_duckdb(sql, df_columns): sql = sql.replace("`", '"') for col in df_columns: if " " in col and f'"{col}"' not in sql: pattern = r'\b' + re.escape(col) + r'\b' sql = re.sub(pattern, f'"{col}"', sql) return sql # 🔄 Global dataframe holder df_global = {"df": None} # 📂 Upload handler def upload_excel(file): if file is None: return "❌ No file uploaded." try: df = pd.read_excel(BytesIO(file.read())) df_global["df"] = df return f"✅ Uploaded file with shape: {df.shape}" except Exception as e: return f"❌ Failed to load file: {e}" # 💬 Handle prompt def handle_query(prompt): df = df_global.get("df") if df is None: return "❌ Please upload an Excel file first.", pd.DataFrame() try: sql = generate_sql_from_prompt(prompt, df) if sql.startswith("Error"): return sql, pd.DataFrame() cleaned_sql = clean_sql_for_duckdb(sql, df.columns) result_df = duckdb.query(cleaned_sql).to_df() return f"📜 Generated SQL:\n{sql}", result_df except Exception as e: return f"❌ Error: {e}", pd.DataFrame() # 🎨 UI with gr.Blocks() as demo: gr.Markdown("# 🤖 SQL Chatbot with Together API + DuckDB") file_input = gr.File(label="📂 Upload Excel File (.xlsx only)", file_types=[".xlsx"]) upload_status = gr.Textbox(label="Status", interactive=False) with gr.Row(): prompt_box = gr.Textbox(label="💬 Your Question", placeholder="e.g., Show me total sales by region") run_button = gr.Button("🚀 Generate SQL + Run") sql_output = gr.Textbox(label="📜 SQL Query") result_table = gr.Dataframe(label="📊 Query Results") file_input.change(upload_excel, inputs=file_input, outputs=upload_status) run_button.click(handle_query, inputs=prompt_box, outputs=[sql_output, result_table]) if __name__ == "__main__": demo.launch()