src/streamlit_app.py

import streamlit as st
import warnings

# --- 1. SILENCE ALL WARNINGS & ERRORS ---
warnings.filterwarnings("ignore", category=RuntimeWarning)
warnings.filterwarnings("ignore", category=FutureWarning)

import os
import sys
import logging
import contextlib
from dotenv import load_dotenv
import yfinance as yf
from duckduckgo_search import DDGS
from langchain_core.language_models.llms import LLM
from huggingface_hub import InferenceClient
from typing import Any, List, Optional
from pydantic import Field

# Silence yfinance logger
logging.getLogger('yfinance').setLevel(logging.CRITICAL)

@contextlib.contextmanager
def suppress_output():
    """Redirects stdout/stderr to devnull to hide yfinance errors."""
    with open(os.devnull, "w") as devnull:
        old_stdout, old_stderr = sys.stdout, sys.stderr
        try:
            sys.stdout, sys.stderr = devnull, devnull
            yield
        finally:
            sys.stdout, sys.stderr = old_stdout, old_stderr

# --- CONFIG ---
st.set_page_config(page_title="Stock Scout", page_icon="📈", layout="centered")
load_dotenv()
hf_token = os.getenv("HF_TOKEN")

if not hf_token:
    st.error("Missing HF_TOKEN in .env file.")
    st.stop()

# --- MODEL (Unbreakable Cluster) ---
class HFChatModel(LLM):
    token: str = Field(...)
    models: List[str] = [
        "microsoft/Phi-3.5-mini-instruct", 
        "Qwen/Qwen2.5-7B-Instruct",         
        "HuggingFaceH4/zephyr-7b-beta"
    ]
    
    def _call(self, prompt: str, stop: Optional[List[str]] = None, **kwargs: Any) -> str:
        client = InferenceClient(token=self.token)
        messages = [{"role": "user", "content": prompt}]
        for model_id in self.models:
            try:
                response = client.chat_completion(
                    model=model_id, messages=messages, max_tokens=600, temperature=0.5
                )
                return response.choices[0].message.content
            except:
                continue
        return "System Busy."

    @property
    def _llm_type(self) -> str:
        return "custom_hf_chat_cluster"

llm = HFChatModel(token=hf_token)

# --- FUNCTIONS ---
def get_stock_data(ticker):
    try:
        with suppress_output():
            stock = yf.Ticker(ticker)
            hist = stock.history(period="5d")
        if len(hist) < 2: return None, None
        curr = hist['Close'].iloc[-1]
        prev = hist['Close'].iloc[-2]
        change = ((curr - prev) / prev) * 100
        return curr, change
    except:
        return None, None

def analyze_news(ticker):
    try:
        with suppress_output():
            results = DDGS().news(f"{ticker} stock news", max_results=5)
        if not results: return "No recent news found."
        
        news_context = []
        sources_txt = "\n\n**Sources:**\n"
        for i, res in enumerate(results, 1):
            title = res.get('title', '?')
            link = res.get('url', '#')
            news_context.append(f"[Source {i}]: {title}")
            sources_txt += f"{i}. [{title}]({link})\n"
            
        full_text = "\n".join(news_context)
        prompt = f"""
        Analyze news for {ticker}:
        {full_text}
        1. Why is it moving? (Cite [Source X])
        2. Sentiment: BULLISH/BEARISH/NEUTRAL
        """
        return llm.invoke(prompt) + sources_txt
    except:
        return "Analysis failed."

# --- UI (SIMPLE DROPDOWN ONLY) ---
st.title("📈 Stock Scout")

# 1. The Safe List
POPULAR_TICKERS = [
    "AAPL", "NVDA", "TSLA", "AMD", "AMZN", "MSFT", "GOOGL", "META",
    "JPM", "BAC", "WMT", "DIS", "NFLX", "KO", "PEP", "BA", "INTC", "PYPL"
]

# 2. The Selector
selected_ticker = st.selectbox("Select Stock:", options=POPULAR_TICKERS)

# 3. Execution (Instant)
if selected_ticker:
    price, change = get_stock_data(selected_ticker)
    
    if price:
        st.metric(selected_ticker, f"${price:.2f}", f"{change:.2f}%")
        st.divider()
        with st.spinner(f"Analyzing {selected_ticker}..."):
            st.info(analyze_news(selected_ticker))
    else:
        st.error("Data unavailable.")