What it does

daily_stock_analysis is a Python automation that pulls A-share, Hong Kong, and US stock data from sources like AkShare, Tushare, and YFinance, mixes in news from SerpAPI or Tavily, and feeds the bundle to an LLM. The model returns a scored decision dashboard—buy, hold, sell—with risk alerts and catalysts, delivered daily via GitHub Actions to WeChat, Feishu, Telegram, Discord, Slack, or plain email. There is also a FastAPI web UI for manual analysis, backtests, and an agent chat that claims 15 built-in strategies from moving-average crossovers to Elliott Wave.

The interesting bit

The project treats GitHub Actions as a zero-cost scheduler: fork, paste API keys into Secrets, set a stock list, and the workflow fires every weekday at 18:00 Beijing time. It is essentially a glue-and-prompt engineering layer around existing data APIs and LLMs, but the orchestration is thorough—fallback data sources, retry logic, and a long list of notification channels.

Key highlights

• Supports A-shares, HK, US stocks, and ETFs with multi-source data aggregation (price, technicals, fundamentals, news, fund flows)
• LLM-agnostic: works with OpenAI-compatible APIs, Gemini, Claude, DeepSeek, Qwen, Ollama, and two sponsored providers (Anspire, AIHubMix)
• Push notifications to six channels including enterprise WeChat and Feishu bots
• Optional web UI with dark mode, backtesting, portfolio tracking, and an agent chat for strategy Q&A
• Docker and local Python deployment supported alongside the GitHub Actions path

Caveats

• Heavy reliance on sponsored or third-party APIs; free tiers will likely throttle serious use
• The README is enthusiastic about AI decision-making but offers no verified backtest results or performance metrics
• “Zero cost” claim assumes free API quotas and GitHub Actions minutes; at scale, neither is guaranteed

Verdict

Worth a look if you want a quick, scheduled LLM stock digest pushed to your team chat. Skip it if you need audited quant strategies or are allergic to API key sprawl.