A collection of ready-to-use Python sample agents, built with the AgentScope multi-agent framework and its AgentScope Runtime deployment solution. It targets developers looking to rapidly build and deploy LLM-based intelligent applications, offering practical examples that range from simple CLI tools to complex, full-stack applications, thereby accelerating the adoption and experimentation with multi-agent systems.

How It Works

AgentScope facilitates the creation of LLM-powered applications by enabling developers to define agents, integrate tools, manage conversational context, and orchestrate collaboration among multiple agents. AgentScope Runtime extends this by providing a framework for deploying these agents as scalable API services, featuring secure sandbox execution for tools and robust deployment management for multi-agent systems.

Quick Start & Requirements

Installation and execution instructions are specific to each example and can be found in their respective README.md files. All examples are Python-based. Users should consult individual sample documentation for detailed prerequisites. Links to AgentScope Documentation and AgentScope Runtime Documentation are available.

Highlighted Details

• DataJuicer Agent: A multi-agent system for intelligent data processing, leveraging over 200 Data-Juicer operators, supporting natural language to YAML configuration generation, custom operator creation, and multiple retrieval modes.
• Alias-Agent: An adaptable intelligent assistant designed for diverse real-world tasks, offering "Simple React," "Planner-Worker," and "Built-in Agents" (e.g., Deep Research, Browser-use) operational modes, serving as a foundational template.
• EvoTraders: A self-evolving financial trading system framework employing multi-agent collaboration, memory systems for reflection and style evolution, real-time data integration, backtesting capabilities, and a visualized dashboard.
• Full-Stack Runtime Examples: Several samples demonstrate the integration of AgentScope Runtime for deployable applications, featuring both backend API services and frontend user interfaces.

Maintenance & Community

Community discussions are hosted on Discord and DingTalk. Contributions, including bug reports, feature requests, and code, are welcomed via GitHub Issues, following guidelines in CONTRIBUTING.md.

Licensing & Compatibility

This project is licensed under the Apache 2.0 License. No specific restrictions for commercial use or closed-source linking are mentioned.

Limitations & Caveats

The provided README does not explicitly detail any project limitations, alpha status, or known bugs.