Key Takeaways from Guangzhou AI‑Native App Salon: AgentScope, HiMarket, Serverless

Hands‑On Workshop: Building a Financial Stock Expert Agent

Participants used AgentRun together with RocketMQ to implement a "financial stock expert" agent. The workflow consisted of:

Define a stock‑query tool that wraps a third‑party market data API.

Register the tool in the AgentScope runtime.

Deploy the agent as a serverless function on AgentRun.

Configure a RocketMQ LiteTopic for A2A communication between the front‑end chatbot and the back‑end stock agent.

Key commands:

# Create tool implementation (Java)
public class StockTool implements Tool {
    public String call(String ticker) {
        // HTTP request to market data service
        return HttpUtil.get("https://api.example.com/quote?symbol=" + ticker);
    }
}

# Register tool and start agent
Agent agent = new ReActAgent();
agent.addTool(new StockTool());
agent.runAsync();

# RocketMQ producer for user queries
producer.sendMessage("user123", "conv456", "What is the price of AAPL?");

This example demonstrated end‑to‑end integration, real‑time response, and observability via LoongSuite traces.