OxyGent 多智能体协作框架新版本发布

Core Insights - OxyGent has released a new version of its multi-agent collaboration framework, introducing features such as multi-modal information transfer, fine-grained message control, MCP reconnection, and front-end streaming output [1] - The framework has been effectively utilized in various business scenarios within JD.com and external communities, supporting the implementation of agent technology and driving AI application development [1] Group 1: OxyGent Framework Features - OxyGent allows developers to flexibly combine and build multi-agent systems by abstracting Agents, Tools, LLMs, and Flows into pluggable atomic AI components [2] - The framework follows a Stateless design principle and incorporates AOP design concepts, providing extreme scalability and full-link decision traceability [2] - OxyGent enables dynamic planning based on defined permission relationships between agents, allowing real-time generation of actual call flow diagrams [2] Group 2: OxyGent Execution Lifecycle - The execution lifecycle of Oxy includes several steps to manage and coordinate different components within the multi-agent system, ensuring data is processed, recorded, saved, and sent at the right time [3][4][5][6][7][8][9] - Key steps include data preprocessing, logging tool calls, saving data, formatting input, executing main logic, and post-processing results [3][4][5][6][7][8][9] Group 3: Data Scopes in OxyGent - OxyGent provides four data scopes: Application, SessionGroup, Request, and Node, allowing flexible read/write operations to enhance data management efficiency and development convenience [11][12] Group 4: Practical Applications of OxyGent - OxyGent has been effectively implemented in JD.com's internal business scenarios, such as SOP processes, data analysis, tool invocation, and multi-level classification [14][15][16][17][18] - In SOP scenarios, the framework supports breaking down business processes into multiple agents, improving execution efficiency and facilitating process tracking [15] - In data analysis, OxyGent automates data collection, preprocessing, analysis, and visualization, enhancing the efficiency of data-driven scenarios [16] Group 5: Community Feedback and Use Cases - OxyGent has demonstrated usability and scalability through community developer feedback, showcasing applications in metric queries, automated tool invocation, flowchart generation, and slow SQL governance [19] - Notable use cases include natural language to SQL conversion, web information extraction, and automated slow SQL diagnosis, validating OxyGent's practicality as an enterprise-level open-source agent platform [19] Group 6: Developer Support and Community Engagement - The company has provided detailed tutorials for developers to quickly get started with OxyGent, covering the entire process from environment setup to distributed agent deployment [20] - Community engagement has led to the resolution of common issues and the emergence of innovative external cases and code contributions through competitions [21]