operations.md · 4.0 KB

Agent Hierarchy and Operations

The mindX system is structured as a multi-tiered hierarchy of agents, each with a distinct level of abstraction and responsibility. This architecture allows for a clear separation of concerns, from long-term system evolution down to tactical task execution.

Tier 1: MastermindAgent - The System Orchestrator

At the apex of the operational hierarchy sits the MastermindAgent. It is the singleton orchestrator for the mindX Augmentic Intelligence system, responsible for its long-term evolution and strategic direction. While it directs the agents within the mindX system, it can be invoked by higher-level intelligences or user directives.

Role: The "Chief Operating Officer" or "System Orchestrator." Responsibilities: System Evolution: Its core function is to manage the lifecycle of the system's components based on strategic directives. This includes: Tool Management: Assessing the current tool suite, identifying strategic gaps, and conceptualizing new tools. Agent Management: Creating, deleting, and triggering the evolution of other agents via the CoordinatorAgent. How it Works: MastermindAgent uses its own internal BDIAgent to translate abstract strategic goals into concrete, multi-step campaigns.

Tier 2: AGInt - The Brain

The AGInt agent is the brain of an operational AI entity. It is the central cognitive core that perceives the environment, makes strategic decisions, and directs the system's actions.

Role: The "Brain" or "Mission Commander." Responsibilities: Cognition and Decision: It operates a Perception-Orient-Decide-Act (P-O-D-A) loop, effectively a thought process for deciding on the next major action. Directive Management: It takes a high-level goal and determines the overall strategy for achieving it. Delegation: The AGInt does not perform tasks itself. Instead, it directs the "hands" of the system—the BDIAgent—to carry out its decisions.

Tier 3: BDIAgent - The Hand

The BDIAgent is the hand of the operational AI. It is the tactical executor that receives commands from the brain (AGInt) and interacts with the environment to get things done.

Role: The "Hand" or "Specialist Executor." Responsibilities: Task Execution: It receives a well-defined goal from a higher-level agent. Planning: It uses an LLM to break the goal down into a detailed, step-by-step plan. This is akin to the hand figuring out the precise movements needed to follow the brain's command. Tool Manipulation: It executes the plan by using the system's available tools, directly interacting with files, APIs, and other resources.

Complete Interaction Flow Example

  • System Evolution Directive: A developer gives the MastermindAgent a directive: "Create a tool for sentiment analysis."
  • Mastermind's BDI Plans: The MastermindAgent gives this directive as a goal to its internal BDIAgent. This BDI agent creates and executes a plan to build the new tool.
  • Operational Directive: A user gives the AGInt (the brain) a separate directive: "Analyze the sentiment of recent news articles about our company."
  • AGInt's Delegation: The AGInt processes this and decides to delegate the task. It sends a command to its BDIAgent (the hand).
  • Operational BDI Plans: The BDIAgent receives the command and formulates a precise plan: [{"type": "web_search", ...}, {"type": "sentiment_analysis_tool", ...}, {"type": "GENERATE_REPORT", ...}].
  • Execution & Reporting: The BDIAgent executes this plan, using the necessary tools. It reports the result back to the AGInt, which then presents the final result to the user.
  • This hierarchical structure creates a robust and scalable system where responsibilities are clearly defined, allowing the AI to manage both its own evolution and complex, user-driven tasks simultaneously.


    All DocumentsDocument IndexThe Book of mindXImprovement JournalAPI Reference