import torch
from transformers import FinBERTTokenizer, FinBERTForSequenceClassification
import yfinance as yf
import gradio as gr

# Load the FinBERT tokenizer and model
tokenizer = FinBERTTokenizer.from_pretrained('yiyanghkust/finbert-tone', model_name_or_path='yiyanghkust/finbert-tone')
model = FinBERTForSequenceClassification.from_pretrained('yiyanghkust/finbert-tone', model_name_or_path='yiyanghkust/finbert-tone')

def predict_stock_price(day, month, year, input_text):
    # Get the stock symbol from the input text
    stock_symbol = extract_stock_symbol(input_text)

    # Format the date for yfinance
    start_date = f"{year}-{month:02d}-{day:02d}"
    end_date = f"{year}-{month:02d}-{day:02d}"

    # Get the historical stock data using yfinance
    stock_data = yf.download(stock_symbol, start=start_date, end=end_date)

    # Perform any necessary data preprocessing

    # Tokenize the input text with appropriate keyword arguments
    inputs = tokenizer(input_text, return_tensors="pt")

    # Forward pass through the model
    outputs = model(**inputs)

    # Get the predicted stock price
    predicted_price = outputs.logits.item()

    return predicted_price

def extract_stock_symbol(input_text):
    # Implement the logic to extract the stock symbol from the input text
    # For example, you can use regular expressions or a custom parsing algorithm
    # For simplicity, let's assume the stock symbol is the first word in the input text
    return input_text.split()[0]

# Define the Gradio interface
inputs = [
    gr.inputs.Number(label="Day", min_value=1, max_value=31),
    gr.inputs.Number(label="Month", min_value=1, max_value=12),
    gr.inputs.Number(label="Year", min_value=2000, max_value=2022),
    gr.inputs.Textbox(label="Input Text")
]
output = gr.outputs.Textbox(label="Predicted Stock Price")

gr.Interface(fn=predict_stock_price, inputs=inputs, outputs=output).launch()