Memfit AI

Recursive Dual-Engine Hybrid Architecture for Intelligent Agent Systems

English | 中文

Memfit AI breaks through the limitations of single paradigms by innovatively adopting a Recursive Coupling of ReAct and Plan-Execute hybrid architecture. This architecture organically fuses macroscopic Strategic Planning with microscopic Tactical Execution through a Coordinator, achieving a unity of determinism and flexibility in task processing.

Recursive Coupling Mechanism: When the ReAct engine encounters complex sub-problems, it dynamically triggers the Plan engine for Sub-planning. This recursive mechanism supports fractal expansion of tasks, adapting to business scenarios of arbitrary complexity.

1. Coordinator Entry (Plan-Execute Mode)

   • For vague or complex tasks
   • Flow: User → Coordinator → Planning Loop → Review → Runtime → ReAct Loops

2. ReAct Entry (Instant Execution Mode)

   • For clear, atomic instructions
   • Flow: User → ReAct Loop → Thought → Action → Observation

• Coordinator - System bus and lifecycle manager
• ReActLoop - Minimal execution unit with "Reasoning-Acting" cycle
• Tools & Forges - Standardized atomic capabilities and scenario-based AI Blueprints
• Memory Triage - Intelligent long-term memory using C.O.R.E. P.A.C.T. Framework
• RAG System - Active, agentic knowledge service with hybrid indexing

• Self-Reflection - Analyzes failures and generates correction plans
• Spin Detection - Prevents infinite loops in actions or logic
• Memory-Augmented Recovery - Retrieves historical error cases for decision support

• Full Documentation
• Yaklang - The underlying language powering Memfit AI
• SSA Analysis - Static Single Assignment analysis platform

Memfit AI 突破了单一范式的局限，创新性地采用了 ReAct 与 Plan-Execute 递归耦合 的混合架构。该架构通过 Coordinator（协调器）将宏观的 战略规划 与微观的 战术执行 有机融合，实现了任务处理中确定性与灵活性的统一。

递归耦合机制： 当 ReAct 引擎遇到复杂子问题时，可动态触发 Plan 引擎进行 次级规划。这种递归调用机制支持任务的分形扩展，可适应任意复杂度的业务场景。

1. Coordinator 入口 (Plan-Execute 模式)

   • 适用于模糊或复杂的任务
   • 流程：用户 → Coordinator → 规划循环 → 审查 → 运行时 → ReAct Loops

2. ReAct 入口 (即时执行模式)

   • 适用于明确的原子化指令
   • 流程：用户 → ReAct Loop → 思考 → 行动 → 观察

• Coordinator - 系统总线与生命周期管理器
• ReActLoop - 最小执行单元，驱动 "Reasoning-Acting" 循环
• Tools & Forges - 标准化原子能力与场景化 AI 蓝图
• Memory Triage - 基于 C.O.R.E. P.A.C.T. 框架的智能长期记忆
• RAG System - 主动式代理化知识服务，支持混合索引

• 自我反思 - 分析失败原因并生成修正计划
• 自旋检测 - 防止动作或逻辑死循环
• 记忆增强恢复 - 检索历史相似错误案例辅助决策

• 完整文档
• Yaklang - 驱动 Memfit AI 的底层语言
• SSA 分析平台 - 静态单赋值分析平台