README.md

metadata

sdk: gradio

🧠 Cho's Risk Scoring App (Powered by DSPy & Gradio)

A real-time risk evaluation tool for retail investors, built with LLM-based Chain-of-Thought (CoT) reasoning.
This app analyzes stock-related questions and visualizes risk scores across 6 dimensions using natural language inputs.

---

📌 How to Use

1. Enter a natural language query like:

   Is Tesla risky these days?
   삼성전자는 지금 고평가됐어?
   

2. Click Submit.
3. The app will automatically:
   • Detect the stock ticker (US or Korea)
   • Fetch recent Yahoo Finance headlines
   • Run CoT-based risk scoring via OpenAI GPT-4o
   • Display results as:
     • 📈 1-Month price trend (Plotly)
     • 📰 Headline list (HTML)
     • 📊 Risk breakdown bar chart (Matplotlib)
     • 🧠 Detailed reasoning per risk category

---

🧠 Risk Factors Evaluated

• Overvaluation
• Poor Earnings
• Financial Instability
• Theme Overheating
• Recurring Negatives
• FII Sell-off

---

🛠️ Tech Stack

• Gradio: frontend UI
• DSPy: CoT prompting & reasoning
• yfinance: stock price and metadata
• Yahoo RSS: finance news headlines
• matplotlib / plotly: data visualization

---

🔐 Environment Variable (required)

This app uses OpenAI's GPT-4o via the DSPy framework.
You must define the following secret in Hugging Face:

OPENAI_API_KEY=sk-...  ← your actual OpenAI API key

Add it in your Hugging Face Space → Settings → Secrets.

---

🏫 About

Developed by HYU researcher Cho
Based on DSPy by Stanford NLP Group