Four Bots and a Stock Ticker: A-Share Analysis by Committee

What it does This project orchestrates four CrewAI agents to research Chinese equities using data from AKShare. Each agent plays a specialist role—market technician, financial analyst, sentiment researcher, or investment advisor—and they collaborate to produce a composite report on a given stock. It supports both mainland A-shares and Hong Kong tickers, and defaults to local Ollama inference rather than cloud APIs.

The interesting bit The authors are admirably upfront about the project’s purpose: the README repeats that everything is “for educational use only” and explicitly disclaims investment advice. The A-share-specific customizations—policy impact analysis, retail-investor psychology, and limit-up/limit-down mechanics—are genuinely localized touches that distinguish it from generic stock-analysis bots.

Key highlights

• Four configurable agents defined in YAML with distinct goals, backstories, and tool sets
• AKShare integration provides real-time quotes, historical K-lines, financial statements, and capital-flow data
• A-share tailored logic: policy effects, retail sentiment, and the unique mainland limit-up/limit-down rules
• Supports standard ticker formats for Shanghai, Shenzhen, and Hong Kong exchanges
• Local-first architecture using Ollama (llama3.1 by default), with an optional switch to OpenAI GPT-4

Caveats

• The default demo analyzes Tencent (00700.HK), a Hong Kong stock, which slightly undercuts the “A-share” branding
• The codebase is largely integration glue between AKShare and CrewAI; most of the novelty sits in the prompt engineering and task decomposition
• The authors plaster the README with risk disclaimers, which is a candid signal that LLM-generated financial analysis should not be mistaken for professional research

Verdict A useful reference if you are building multi-agent workflows around Chinese financial data, but approach it as a CrewAI tutorial with a specialized data source rather than a production research platform. Traders looking for alpha should look elsewhere—exactly as the README suggests.