agentUniverse
A large-language-model multi-agent framework, originating from AntGroup's financial business practice, providing reusable multi-agent collaboration patterns such as plan-execute-express-review and letting a team fold its own domain expertise into how agents work together.
agentUniverse is a multi-agent framework built on large language models, providing flexible, extensible building blocks for individual agents plus a set of ready-made multi-agent collaboration patterns for different kinds of problems. It grew out of AntGroup's own financial-services use of multi-agent systems and is aimed at helping teams fold their own domain experience into how agents collaborate.
Two collaboration patterns are documented: PEER, which splits a complex task across agents responsible for planning, execution, expression and review, iterating on feedback, suited to tasks such as event interpretation and industry analysis; and DOE, which uses data-mining, opinion-injection and expression agents for data-intensive tasks that also need expert judgment, such as financial report generation. Model integration is done through configuration, and the framework already supports models including the Qwen family.