Mathematics - Multi-agent

Key features

• Multi-agent automatic solutionthe main agent hands the problem to the scheduler, and the scheduler dispatches explorer/solver/verifier (v1/v2/v3) and planner (planning agent, v3), method-keeper (method sorting agent, v3) and other sub-agents to solve the problem collaboratively. You do not need to do it manually node by node.
• The whole process status is placed in VibeMath_State/, resume breaks and continues running, and manual mode suspends manual decision-making at dispatch/judgment/promotion.
• The whole process status is written to the disk, resume is used to resume running at breakpoints, and reportMode can report via file/push/both.
• One-sentence pipelineN resident sub-agents (continuable, persistent context) are initially generated and each brainstorms and produces initial insights/directions → After that, all task arrangements are decided independently by leaving messages to each other + collective meetings (the framework only does message bus/meeting/task board/product precipitation, and never allocates tasks)
• Thesis-style Markdown knowledge base (v3): problem list (including inter-question dependencies, causes and plans of epigenetic problems), research logs, propositions, and method libraries are all written and continued in the md paper/research report style (the logs of the old direction are automatically archived and retained when the direction is re-derived)