AI‑Based Stock Analysis & Portfolio Optimisation App

A Streamlit app that fetches market data, computes technical indicators (MA, RSI, MACD, Bollinger Bands), generates short AI insights, and produces a clear Buy / Hold / Sell signal per ticker. It can also run a simple portfolio optimisation over selected assets.

Demo: If you deployed this on Hugging Face Spaces, open your Space URL:
https://huggingface.co/spaces/<your-username>/<your-space-name>
Replace the placeholders with your actual Space path.

✨ What you can do

• Enter 1..N tickers (e.g. TSLA, AAPL or RELIANCE.NS, TCS.NS)
• Choose date range and indicator windows and risk tolerance
• See charts + indicator values
• Read AI one‑line insights for MA/BB/RSI/MACD (optional; requires OpenAI key)
• Get a color‑coded investment signal (Buy / Hold / Sell) with reasons
• Optimise a simple equal‑risk / target‑risk allocation across your selection

Disclaimer: This app is for educational use only. Not financial advice. Markets are risky; do your own research.

🧱 Project structure

.
├── app.py                          # Streamlit UI
├── data_loader.py                  # Robust price downloader using yfinance API
├── indicators.py                   # Indicator calculations and plots using Plotly
├── ai_insights.py                  # get_ai_insight() wrapper for OpenAI one‑liners
├── investment_signal.py            # get_investment_signal() – rule‑based Buy/Hold/Sell
├── portfolio_optimiser.py          # optimise weights from closing price data
└── requirements.txt                # dependencies (see variants below)

🛠️ Local setup

1) Create & activate a virtual environment

python -m venv .venv
# Windows
.venv\Scripts\activate
# macOS/Linux
source .venv/bin/activate

2) Install dependencies

If your Python is 3.11+ (Windows/macOS/Linux):

# requirements.txt
streamlit
yfinance
pandas
numpy==1.23.5
scipy
plotly
openai

Then install:

pip install -r requirements.txt

3) (Optional) Add your OpenAI API key

• Env var (works everywhere):

  # Windows (PowerShell)
  setx OPENAI_API_KEY "sk-..."
  # macOS/Linux (bash/zsh)
  export OPENAI_API_KEY="sk-..."

• Streamlit secrets (local): create .streamlit/secrets.toml

  OPENAI_API_KEY = "sk-..."

Without a key, the app still runs – it just skips AI text.

4) Run the app

python -m streamlit run app.py

Then open the local URL (usually http://localhost:8501).

🚀 Hugging Face Spaces (demo & reproduce)

1. Create a new Space → SDK: Streamlit.
2. Upload your project files and requirements.txt.
3. For Spaces’ default Python 3.10 image, use this requirements.txt (pinned to stable builds):

    streamlit
    yfinance
    pandas
    numpy==1.23.5
    scipy
    plotly
    openai

4. Set your OPENAI_API_KEY in Settings → Variables & secrets (optional).
5. Click Restart / Rebuild. Your app will be live at:
   https://huggingface.co/spaces/<your-username>/<your-space-name>

📈 Indicators & Signal

Indicators:

• MA – Simple moving average (window configurable)
• RSI – Momentum (14 default)
• MACD – Trend / momentum (12, 26, 9 default)
• Bollinger Bands – Volatility bands (20, 2 default)
• Risk Tolerance – User Risk Tolerance between (0, 1)

Signal (rule‑based):

• Points for RSI (<30 oversold → +2; >70 overbought → −2)
• Points for MACD (>0 bullish → +1; <0 bearish → −1)
• Points for Trend (Close vs MA: above → +1; below → −1)
• Points for Bollinger extremes (below lower band → +1; above upper → −1)
• Total ≥ +2 → Buy; ≤ −2 → Sell; else Hold

This is a transparent heuristic. Improve realism by adding regime filters, confirmations, ATR‑based risk mgmt, and a walk‑forward backtest.

🧪 Reproducibility checklist

• Use the pinned requirements.txt above.
• Fix a random seed where applicable (optimiser, if stochastic).
• Record the exact tickers and date range in your test.
• Export the final results (weights/metrics) with a timestamp for audit.

🩹 Troubleshooting

• streamlit: command not found (Windows)
  Activate your venv and run via module:

  .venv\Scripts\activate
  python -m streamlit run app.py

• OpenAI key not found
  Set OPENAI_API_KEY as an env var or in .streamlit/secrets.toml. The app handles missing keys gracefully.

🔐 Security & data usage

• Do not commit secrets (API keys). Use environment variables or platform secrets.
• This project fetches public market data for research and demo purposes only.

📜 License

MIT — free to use for learning and personal projects. If you build a commercial product, make sure your data sources comply with their respective terms.