"""
NeoSyn: AI-Powered Synthetic Data & Insights

This Gradio app hosts multiple MCP tools for different databases (sales, customers, products, transactions),
and allows users to chat naturally for generating synthetic data, visualizing insights, or fetching summaries.
All interactions are stored in ChromaDB memory.

Author: Saivivek Katkuri
Date: June 2025
"""

import gradio as gr
import os
from langchain_community.vectorstores import Chroma
from langchain_community.chat_models import ChatOpenAI
from db_connector import fetch_table_data
from data_processor import generate_synthetic_data
from plotly_tools import generate_plot

# Setup ChromaDB for chat memory (local storage)
memory = Chroma(persist_directory="./chroma_memory")

# Define database paths
DB_PATHS = {
    "sales": "sales.db",
    "customers": "customers.db",
    "products": "products.db",
    "transactions": "transactions.db"
}

# Core function to handle queries
def process_query(user_query, user_id):
    try:
        query_lower = user_query.lower()
        response = ""

        # Generate synthetic data
        if "generate" in query_lower and "records" in query_lower:
            for db_name, db_path in DB_PATHS.items():
                if db_name in query_lower:
                    table_name = db_name
                    df = fetch_table_data(db_path, table_name)
                    num = int(''.join([c for c in query_lower if c.isdigit()]) or 10)
                    synthetic_df = generate_synthetic_data(df, num)
                    file_path = f"generated/{db_name}_synthetic.csv"
                    synthetic_df.to_csv(file_path, index=False)
                    response = f"✅ Generated {num} synthetic records for {db_name}. [Download CSV]({file_path})"
                    memory.add_texts([f"{user_id}: {response}"])
                    return response

        # Generate plot
        elif "plot" in query_lower or "distribution" in query_lower or "graph" in query_lower:
            for db_name, db_path in DB_PATHS.items():
                if db_name in query_lower:
                    desc, html_path, png_path = generate_plot(user_query, db_path)
                    response = f"{desc} [View Plot (HTML)]({html_path}) or [Download PNG]({png_path})"
                    memory.add_texts([f"{user_id}: {response}"])
                    return response

        # Fetch table summary
        else:
            for db_name, db_path in DB_PATHS.items():
                if db_name in query_lower:
                    df = fetch_table_data(db_path, db_name)
                    summary = f"📊 {db_name.upper()} Data Summary:\n{df.describe().to_string()}"
                    response = summary
                    memory.add_texts([f"{user_id}: {response}"])
                    return response

        response = "❓ Sorry, I couldn't understand. Please specify the table or ask for synthetic data or plots."
        memory.add_texts([f"{user_id}: {response}"])
        return response

    except Exception as e:
        return f"❌ Error: {str(e)}"

# Gradio ChatInterface
with gr.Blocks() as demo:
    gr.Markdown("# NeoSyn: AI-Powered Synthetic Data & Insights")
    gr.Markdown("Chat naturally to generate synthetic data, get summaries, or plots from sales, customers, products, or transactions.")

    chat = gr.ChatInterface(
        fn=process_query,
        chatbot=gr.Chatbot(type='messages'),  # Use OpenAI-style messages format
        additional_inputs=[gr.Textbox(label="User ID (optional)", value="default_user")],
        title="NeoSyn Chat",
        description="Ask me to generate synthetic data, get insights, or plots!"
    )

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