AI Stock Analysis System
An LLM-powered multi-market stock analysis system that generates daily decision dashboards and pushes them to your preferred channels.
Evidence: Workflow uses GitHub Secrets for API keys, following least privilege; but no user confirmation mechanism (e.g., before push), data flow transparency partial (README describes data sources and push channels but not data flow), sensitive data handling partial (API keys via Secrets but no encryption details), dependency security partial (requirements.txt has version ranges but no vulnerability scanning), external effects partial (push notifications and external API calls but no side effects), rollback absent, source attribution clear (README and LICENSE). Deductions: missing user confirmation, rollback, insufficient data flow and sensitive data handling details.
Evidence: README and code structure consistent, features correspond to files; dependencies have version ranges but not pinned, affecting reproducibility; failure messages partial (README mentions logs but no specific error handling examples). Deductions: dependencies not pinned, failure messages evidence insufficient.
Evidence: README clearly defines target users (individual investors) and scenarios (daily analysis, push); capability boundaries partial (README lists supported markets and data sources but not limitations); trigger precision partial (GitHub Actions scheduled and manual triggers but no details); environment fit partial (supports GitHub Actions, Docker, local but no system requirements). Deductions: capability boundaries and trigger precision details insufficient.
Evidence: README well-structured with TOC and doc links; install notes detailed (quick start, Secrets configuration); naming stability partial (project name and main commands stable but no API stability); examples and FAQ partial (README has push effect examples but no FAQ); known limitations partial (README mentions free source instability but not comprehensive); license clear (MIT); versioning and changelog partial (auto-tag workflow but no CHANGELOG); maintenance responsibility clear (contact email and Issue link). Deductions: missing FAQ, CHANGELOG, known limitations not comprehensive.
Evidence: Output usability (README shows decision dashboard and push effects, clear format); marginal value (multi-market analysis, multiple push channels but no comparison with existing tools); cost benefit (README emphasizes zero-cost but no actual cost analysis). Deductions: marginal value and cost benefit evidence insufficient.
Evidence: Claim traceability (README describes features but no implementation details); cross-source verification (no data source cross-validation mechanism); fact-inference separation (AI analysis may mix facts and inferences but not clearly separated). Deductions: lack of implementation details and verification mechanisms.
- Dependencies are not pinned to exact versions, which may affect reproducibility and security.
- Lack of user confirmation mechanism; push notifications may be sent without user consent.
- Insufficient details on data flow and sensitive data handling; further review needed.
- No rollback mechanism; configuration errors may be difficult to recover.
- AI analysis may mix facts and inferences; handle with caution.
What does this agent do, and when should you use it?
DSA (daily_stock_analysis) is an LLM-powered multi-market stock analysis system supporting A-shares, Hong Kong, US, Japanese, Korean, and Taiwanese stocks. It aggregates multi-source market data, real-time news, and fundamentals, generates AI-driven decision reports, and pushes them to WeCom, Feishu, Telegram, Discord, Slack, and email. The system offers a Web workspace, Agent-based strategy Q&A, backtesting, portfolio management, and supports deployment via GitHub Actions, Docker, or local scheduled tasks. It defaults to free data sources for zero-config operation, but paid sources are recommended for long-term stability.
The system reads the configured watchlist, fetches quotes, candlesticks, and technical indicators from sources like AkShare, Tushare, YFinance, and TickFlow, pulls real-time news and announcements via search APIs, and invokes AI models (Anthropic, OpenAI, DeepSeek, etc.) to generate decision reports including scores, buy/sell points, risk alerts, and catalysts. Reports are pushed to configured channels via webhooks. Users can run one-off analyses via CLI (e.g., python main.py --stocks 600519,hk00700) or interact through the Web UI (python main.py --webui). The system also supports Agent-based strategy Q&A with 15 built-in strategies and can run as a scheduled GitHub Action.
- Retail investors check AI-generated decision reports for their watchlist every morning.
- Quant enthusiasts use Agent strategy Q&A to ask multi-turn questions about specific stocks.
- Enterprises manage multiple watchlists and backtest strategies through the Web workspace.
- Developers integrate stock analysis into CI/CD via GitHub Actions for scheduled, zero-cost runs.
- Analysts use the market review feature to quickly grasp major indices and sector performance.
What are this agent's strengths and limitations?
- Supports multiple markets: A-shares, HK, US, Japan, Korea, Taiwan
- Rich Web workspace and Agent Q&A feature
- Multiple notification channels (WeCom, Feishu, Telegram, Discord, Slack, email)
- Zero-cost deployment via GitHub Actions without a server
- Free data sources can be unstable; paid sources recommended for heavy use
- Requires multiple API keys, which may incur costs
- Skips non-trading days; only runs on weekdays by default
- Project is evolving rapidly; APIs may change
How do you install or deploy this agent?
Option 1: GitHub Actions. Fork the repo, add required secrets (e.g., ANSPIRE_API_KEYS, GEMINI_API_KEY) in Settings, set STOCK_LIST, enable Actions, and manually trigger the workflow. Option 2: Locally, clone the repo, run pip install -r requirements.txt, copy .env.example to .env and fill in env vars, then run python main.py.
How do you use this agent?
After installation, run one-off analyses with python main.py --stocks 600519,hk00700; add --debug for verbose output. To use the Web UI, run python main.py --webui and visit http://127.0.0.1:8000. For automation, refer to the docs for scheduling; the default schedule is weekdays at 18:00 Beijing time.