What Is daily_stock_analysis and How to Use It: LLM Driven Multi Market Stock Analysis

Checking the news, prices, and sentiment around a position before the open means several browser tabs and a good chance of missing something. A script fixes that, but then you own a whole chain of data sources, model calls, and notifications. daily_stock_analysis packages the chain into one system that produces a scored decision dashboard on a schedule and pushes it to the chat tools you already use. It is a popular Chinese language quant project on GitHub, MIT licensed, and free sources let it run at zero cost.

What Is daily_stock_analysis

daily_stock_analysis is an LLM driven analysis system covering A shares, Hong Kong, US, Japanese, Korean, and Taiwan stocks plus ETFs. It runs daily and delivers a decision dashboard to WeCom, Feishu, Telegram, Discord, Slack, and email. The repository was created on 2026-01-10, had active commits on 2026-11-09, and shows roughly 69,085 stars and 56,443 forks under MIT. That unusually high fork count tells you people treat it as a template. It combines market data, news sentiment, and fundamentals into a report with a conclusion, score, trend, price levels, risk alerts, catalysts, and an action list. A web and desktop workbench adds manual analysis, backtesting, positions, settings, and light and dark themes. The dashboard is the part most people fork first.

Features

• Market coverage. A shares, Hong Kong, US, Japanese, Korean, and Taiwan stocks plus ETFs in one configuration.
• Decision dashboard. Reports include a conclusion, score, trend, price levels, risk alerts, and an action list.
• Agent strategy chat. Fifteen built in strategies cover moving averages, trend, momentum, events, growth, and expectations, with follow up questions.
• Web and desktop workbench. Manual analysis, backtesting, positions, and light and dark themes all live in the interface.
• Zero cost scheduling. Choose GitHub Actions, Docker, a local cron job, or a FastAPI service.
• Swappable models. Gemini, DeepSeek, Qwen, Claude, Ollama local models, and OpenAI compatible endpoints are supported.
• Rich data sources. Market data comes from AkShare, Tushare, Baostock, YFinance, and Longbridge, with news from Tavily, SerpAPI, and Brave.
• Portable configuration. Free sources work with no keys, and token based providers can be added later.

For your own web data collection, see What Is Scrapling and How to Use It. A comparable approach to AI content generation is What Is MoneyPrinterTurbo and How to Use It.

How to Use

1. Clone the repository and follow docs to create a Python environment and install dependencies.
2. Run it once with no keys. The built in free sources produce a report with zero configuration.
3. Copy the configuration template, add an API key for your chosen model, or point it at a local Ollama address.
4. Add credentials for token based providers such as Tushare or Tavily if you want steadier market and news data.
5. Configure delivery with webhook and recipient details for WeCom, Feishu, Telegram, or email.
6. Start the FastAPI service with server.py, or run one manual analysis, and check that the report arrives.
7. Choose GitHub Actions, Docker, or a system cron job and set the trigger time from the documentation.

Trading hours differ, so confirm the schedule against the market you follow.

Supported Platforms

Free sources cover the common markets, and token based commercial providers can be added when stability matters. Both local and container setups have modest system requirements, so an old mini PC is enough to start. The same config file works in a container and on a desktop.

Pricing

The project is MIT licensed. Code and self hosted services are free, with no cap on analyses. Two things cost money. Model API calls are the first, and running a local model through Ollama brings that to zero. The second is any market or news provider that needs a token or a paid quota. GitHub Actions free minutes usually cover one daily run, and a local cron job removes even that limit. WeCom, Feishu, and Telegram delivery are free.

Use Cases

• Pre market briefing. Receive one daily report combining prices, news, and a score before the session starts.
• Multi market tracking. Follow A shares and US stocks without maintaining two separate scripts.
• Strategy research. Ask the agent using one of fifteen strategies, or push back with follow up questions.
• Backtesting and review. Run a historical window in the workbench and compare the conclusion with what followed.
• Team sharing. Send the same dashboard to a group so nobody repeats the same information gathering.
• Forking. The high fork count makes it a practical template for your own sources and strategies.

Scores and price levels are model output, not investment advice, so verify anything before acting on it. Free data can also be wrong, so check the raw quote when a stock is suspended or a definition changes.

How It Works

ZhuLinsen started the project and the community keeps it moving through forks and patches. The repository ships both CLAUDE.md and SKILL.md, which says something about how the maintainer works with AI assistance. MIT licensing lets anyone rebuild it into a private analysis system, and that explains the fork count more than any feature does. The design separates data, model, and delivery layers. Free sources are the default, token based ones are optional, which gives beginners a zero cost start and serious users room to upgrade. The trade for such broad coverage is a long configuration file, so this suits people willing to read the docs. That openness is why the fork count keeps climbing.

Conclusion

If a daily market summary in your group chat sounds useful, this system handles the tedious parts of aggregation and delivery. Start with the default free sources and a local model, confirm the report format works for you, then swap in providers you trust. It organizes information. It does not make the decision for you. A report is a starting point for your own reading, nothing more.

Project https://github.com/ZhuLinsen/daily_stock_analysis