Automating Stock Analysis with GitHub Actions

The Core Mechanism

The project functions by leveraging GitHub Actions as a free, serverless compute environment. Users fork the repository, input their own LLM API keys into GitHub Secrets, and enable the scheduled workflow. The system executes a Python-based cron job every weekday after market close, which fetches financial data, processes it through an LLM, and pushes a summary report to platforms like Telegram, Discord, or Slack. Because the logic resides entirely within the user's fork, there is no centralized server to maintain or monthly SaaS fee to pay.

Data Processing and Architecture

The system utilizes LiteLLM for model routing, allowing users to swap between providers like Claude, Gemini, or local models via OpenAI-compatible endpoints without modifying the core codebase. Data ingestion is handled through a rotation of providers, including Yahoo Finance, Finnhub, and Alpha Vantage, featuring automatic failover to ensure reliability. The analysis engine supports 15 built-in strategies, ranging from simple moving average crossovers to complex event-driven setups. Version 3.18 introduced an upgraded alert engine that manages cooldowns to prevent notification spam and allows users to assign specific strategies to individual tickers rather than applying a generic analysis to the entire watchlist.

Trade-offs and Limitations

While the project is highly popular, it is primarily optimized for Chinese markets (A-shares and Hong Kong stocks), with American market support being a secondary priority. The "free" nature of the tool refers to the compute provided by GitHub Actions, but users remain responsible for the token costs associated with their chosen LLM provider. The project is explicitly a research tool and not a financial advisor, as the underlying models are prone to hallucinations and incorrect market assessments.