import gradio as gr
from huggingface_hub import InferenceClient
import gspread
from google.oauth2.service_account import Credentials

# Connecting to Google Sheets
SERVICE_ACCOUNT_FILE = "sheet-integration-agent@bold-physics-445007-g4.iam.gserviceaccount.com.json"
scopes = ["https://www.googleapis.com/auth/spreadsheets"]
credentials = Credentials.from_service_account_file(SERVICE_ACCOUNT_FILE, scopes=scopes)
client_gs = gspread.authorize(credentials)
spreadsheet = client_gs.open_by_key("1LHgb4eYDM2ImFY3e3h4kPUnDh6oRHpVz2PA29lQ3Xxg")
worksheet = spreadsheet.sheet1

# Connecting to the zephyr-7b-beta model via Inference API
client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")

# Function to retrieve data from Google Sheets
def get_apartment_data(query):
    rows = worksheet.get_all_records()
    results = []
    for row in rows:
        # Basic check if the query is found in "district" or "developer"
        if query.lower() in row["district"].lower() or query.lower() in row["developer"].lower():
            results.append(row)
    if results:
        return "\n".join(
            [f"District: {row['district']}, Developer: {row['developer']}, Area: {row['apt_area']} m²"
             for row in results]
        )
    else:
        return "Sorry, no matching options found."

# Function to handle queries
def respond(message, history):
    """
    message (str): latest user message
    history (list): the entire conversation so far, as a list of dicts like:
        [
          {"role": "user", "content": "..."},
          {"role": "assistant", "content": "..."},
          ...
        ]
    Returns a single string (the assistant's answer).
    """

    print("Query received:", message)
    print("Chat history:", history)

    try:
        # If the query is about district or developer, fetch from Sheets
        if "district" in message.lower() or "developer" in message.lower():
            response = get_apartment_data(message)
        else:
            # Build messages from history for context
            conversation = []
            for turn in history:
                conversation.append({"role": turn["role"], "content": turn["content"]})

            # Add the new user message
            conversation.append({"role": "user", "content": message})

            # Query Zephyr
            generated_response = client.chat_completion(
                messages=conversation,
                max_tokens=512,
                temperature=0.7,
                top_p=0.95,
            )
            response = generated_response["choices"][0]["message"]["content"]

        # Just return the assistant's text
        return response

    except Exception as e:
        print(f"Error processing query: {e}")
        return "An error occurred while processing your query."

# Gradio Interface
chat_interface = gr.ChatInterface(
    fn=respond,
    title="Apartment Agent",
    description="This chatbot can provide apartment-related information and answer your questions."
)

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