ai-finance

AI Finance is a 24/7 automated market analysis and monitoring system built on GitHub Actions. The README clearly states that it is intended only for simulation and analysis, prohibits automated trading, and notes that actual investment decisions require the user’s own judgment. Results are delivered through a GitHub Pages dashboard and real-time Telegram alerts. Core features are divided into several areas. Market analysis covers more than 22 stocks across seven sectors every four hours from HMM regime, sentiment, and portfolio perspectives. Breaking news monitoring checks seven RSS feeds and six categories of keywords every hour. For overseas markets, 13 global indices and futures are tracked hourly while KOSPI is closed. The Japan Crisis Monitor tracks yen carry-trade unwinding pressure and BOJ rate increases daily at 09:30 and 22:00. The Bitcoin Standard Monitor covers ETF fund flows, mining, and the ARMA bill daily at 09:00 and 22:00. Value-analysis features are also included. The Value Screener calculates value scores and a WATCHLIST for 47 KOSPI stocks daily at 06:00, using relative valuation against the KOSPI average. The Value Impairment Monitor monitors deterioration in fundamentals, including EPS, BPS, and profitability, rather than share prices alone. DCF Fair Value estimates cash flows for 15 large-cap stocks, and Quarterly Earnings Review performs earnings analysis based on the DART API. AutoResearch evolves value weights and evaluates forward tests on daily and monthly schedules. Installation and operation are presented in the following order: clone the repository, install requirements.txt, configure the Telegram Bot token, Chat ID, and DART API key, run local test scripts, and push to GitHub to trigger workflows automatically. The project structure consists of core utilities for environment-variable loading, Telegram, a yfinance wrapper, and state storage; docs for GitHub Pages output; config for sector definitions; .github/workflows containing 11 workflows; tests; and various monitoring and analysis scripts. Automation is operated through 11 workflows: Daily Finance Analysis every four hours, Breaking News Monitor every hour, Overseas Market Monitor every hour while KOSPI is closed, Japan Crisis Monitor, Bitcoin Standard Monitor, Value Screener daily at 06:00, Auto-Research Value at 07:00, Auto-Research Portfolio at 03:00, DCF Valuator at 07:30, Earnings Reviewer at 09:00, and Forward Test Eval on the first day of every month at 08:00. For security, API keys are stored in GitHub Secrets, committing config/.env is prohibited, XXE attacks are defended against with defusedxml, external calls use HTTPS, workflow concurrency control prevents race conditions, and subprocess calls use list arguments. A threat-model document and regression tests are also provided. The license and disclaimer emphasize that the project is for personal learning and analysis purposes only.