ml_stock_sentiment.py

import streamlit as st
import threading
from datetime import datetime

# LumiBot imports v1
from lumibot.backtesting import YahooDataBacktesting
from lumibot.brokers import Alpaca
from lumibot.traders import Trader
from lumibot.credentials import ALPACA_CREDS
# Import your strategy (example)
from lumibot.example_strategies.stock_sentiment import StockSentiment

st.set_page_config(
    page_title='Finance ML',
    layout="wide",
    page_icon=':bar_chart:',
)

# Streamlit UI
st.title("📊 FinRAG Financial Assistant")
# st.title("Sentiment-Based Trading Bot with Live Trading")
st.subheader("FinBERT pre-trained NLP model to analyze sentiment of financial text.")
st.write("""
Automated sentiment or polarity analysis of texts produced by financial actors using NLP methods.
This strategy uses FinBERT to analyze the sentiment of financial news and make trading decisions based on the sentiment score.
""")
st.write("Ask any finance-related question, and I'll fetch the best answer for you!")

# Store results in session_state so we can display them from the main thread
if "backtest_results" not in st.session_state:
    st.session_state["backtest_results"] = None
if "tear_sheet_figure" not in st.session_state:
    st.session_state["tear_sheet_figure"] = None
if "backtest_running" not in st.session_state:
    st.session_state["backtest_running"] = False

# UI Elements for user interaction in the main view
st.header("Strategy Parameters")

# Select backtest period
col1, col2 = st.columns(2)
with col1:
    start_date = st.date_input("Start Date", datetime(2020, 1, 1))
with col2:
    end_date = st.date_input("End Date", datetime(2020, 11, 12))

# Select stock symbol
symbol = st.selectbox("Select Stock Symbol", ["SPY", "AAPL", "GOOGL", "AMZN", "TSLA"])

# Set cash at risk
cash_at_risk = st.slider("Cash at Risk (%)", 0.1, 1.0, 0.5)

# Button to run sentiment analysis
if st.button("Run Sentiment Analysis"):
    st.write(f"Running sentiment analysis for {symbol} from {start_date} to {end_date} with {cash_at_risk * 100}% cash at risk...")
    # Here you would call the sentiment analysis function and store the results

# Button to view sentiment results and trading decisions
if st.button("View Sentiment Results"):
    if st.session_state["backtest_results"] is not None:
        st.write("### Sentiment Results")
        st.write(st.session_state["backtest_results"])
    else:
        st.warning("No sentiment results available. Please run the sentiment analysis first.")

def run_backtest():
    """Long-running backtest & live trading in a background thread."""
    try:
        st.session_state["backtest_running"] = True

        # Define the backtest period
        start_date = datetime(2020,1,1)
        end_date   = datetime(2020,11,12)
        print(f"[Thread] Starting backtest from {start_date} to {end_date}...")

        # Set up broker and strategy
        broker = Alpaca(ALPACA_CREDS)
        strategy = StockSentiment(
            name="mlstrat",
            broker=broker,
            parameters={
                "symbol": symbol,
                "cash_at_risk": cash_at_risk
            }
        )

        # Run the backtest
        results = strategy.backtest(
            YahooDataBacktesting,
            start_date,
            end_date,
            parameters={"symbol": symbol, "cash_at_risk": cash_at_risk},
        )
        st.session_state["backtest_results"] = results

        # OPTIONAL: Generate a tear sheet if LumiBot (or you) provide a function
        if hasattr(results, "create_tear_sheet"):
            fig = results.create_tear_sheet()
            st.session_state["tear_sheet_figure"] = fig
        elif hasattr(results, "display_tear_sheet"):
            # If display_tear_sheet returns a figure
            fig = results.display_tear_sheet(return_fig=True)
            st.session_state["tear_sheet_figure"] = fig
        else:
            # If there's no tear sheet method, you'll need custom logic:
            print("[Thread] No tear sheet method found in results.")
            st.session_state["tear_sheet_figure"] = None

        print("[Thread] Backtest finished. Starting live trading...")

        # Start live trading (This is a blocking call)
        trader = Trader()
        trader.add_strategy(strategy)
        # Start the trading in a separate thread to avoid blocking Streamlit
        trader_thread = threading.Thread(target=trader.run_all)
        trader_thread.start()

        # Display a progress bar in the Streamlit app
        while trader_thread.is_alive():
            st.progress(100 * (1 - trader_thread._stop.__self__.is_alive()))
            st.write("Running live trading...")
        
    finally:
        st.session_state["backtest_running"] = False

def start_background_backtest():
    # Reset old results
    st.session_state["backtest_results"] = None
    st.session_state["tear_sheet_figure"] = None

    # Start a thread to run the backtest
    worker_thread = threading.Thread(target=run_backtest, args=())
    worker_thread.start()

# Button to trigger backtest
if st.button("Start Backtest"):
    # Prevent multiple backtest runs
    if not st.session_state["backtest_running"]:
        st.write("Backtest is starting in the background...")
        start_background_backtest()
    else:
        st.warning("A backtest is already running!")

# Status / Results display
if st.session_state["backtest_running"]:
    st.info("Backtest is running... (this might take a while).")

elif st.session_state["backtest_results"] is not None:
    st.success("Backtest completed!")
    st.write("### Backtest Results")
    st.write(st.session_state["backtest_results"])

    # Display the tear sheet if we have it
    if st.session_state["tear_sheet_figure"] is not None:
        st.write("### Tear Sheet")
        st.pyplot(st.session_state["tear_sheet_figure"])
    else:
        st.warning("No tear sheet figure available.")