app.py

import os
import gradio as gr
import base64
from llama_index.core import StorageContext, load_index_from_storage
from dotenv import load_dotenv
from retrieve import get_latest_dir, get_latest_html_file
from graph_handler import query_graph_qa, plot_subgraph
from embed_handler import query_rag_qa
from evaluate import evaluate_llm, reasoning_graph, get_coupon
import base64
import time

load_dotenv()

KG_INDEX_PATH = get_latest_dir(os.getenv("GRAPH_DIR"))
KG_PLOT_PATH = get_latest_html_file(os.getenv("GRAPH_VIS"))
RAG_INDEX_PATH = get_latest_dir(os.getenv("EMBEDDING_DIR"))

# Load Graph-RAG index
graph_rag_index = load_index_from_storage(
    StorageContext.from_defaults(persist_dir=KG_INDEX_PATH)
)

# Load RAG index
rag_index = load_index_from_storage(
    StorageContext.from_defaults(persist_dir=RAG_INDEX_PATH)
)


def query_tqa(query, search_level):
    """
    Query the Graph-RAG and RAG models for a given query.

    Args:
    query (str): The query to ask the RAGs.
    search_level (int): The max search level to use for the Graph RAG.

    Returns:
    tuple: The response, reference, and reference text for the Graph-RAG and RAG models.
    """

    if not query.strip():
        raise gr.Error("Please enter a query before asking.")

    grag_response, grag_reference, grag_reference_text = query_graph_qa(
        graph_rag_index, query, search_level
    )
    # rag_response, rag_reference, rag_reference_text = query_rag_qa(
    #     rag_index, query, search_level
    # )
    return (
        grag_response,
        # grag_reference,
        # grag_reference_text,
        # rag_response,
        # rag_reference,
        # rag_reference_text,
    )

def show_graph():
    """
    Show the latest graph visualization in an iframe.

    Returns:
    str: The HTML content to display the graph visualization in an iframe.
    """

    graph_vis_dir = os.getenv("GRAPH_VIS", "graph_vis")
    try:
        latest_graph = get_latest_html_file(graph_vis_dir)
        if latest_graph:
            with open(latest_graph, "r", encoding="utf-8") as f:
                html_content = f.read()

            encoded_html = base64.b64encode(html_content.encode()).decode()
            iframe_html = f'<iframe src="data:text/html;base64,{encoded_html}" width="100%" height="1000px" frameborder="0"></iframe>'
            return iframe_html
        else:
            return "No graph visualization found."
    except Exception as e:
        return f"Error: {str(e)}"


with gr.Blocks() as demo:
    gr.Markdown("# Comfy Virtual Assistant")
    chatbot = gr.Chatbot()
    msg = gr.Textbox(label="Input Your Query")
    clear = gr.ClearButton([msg, chatbot])

    def respond(message, chat_history):
            bot_message = query_tqa(message, 2)
            chat_history.append((message, bot_message))
            time.sleep(1)
            return "", chat_history
    
    msg.submit(respond, [msg, chatbot], [msg, chatbot])


    with gr.Row():
        plot_button = gr.Button("Plot Knowledge Graph", variant="secondary")

    kg_output = gr.HTML()


demo.launch(auth=(os.getenv("ID"), os.getenv("PASS")), share=False)