Overview

The BasisOS agent is designed as a modular, multi-agent system that leverages a graph of specialized sub-agents, each focusing on a particular domain (e.g., technical documentation retrieval, data warehouse connections, character interaction, etc.). It is built on top of the langgraph framework and uses a multi-supervisor approach, where several supervisory agents coordinate tasks and communication among sub-agents. This structure allows for horizontal scalability and the seamless addition or replacement of specialized sub-agents over time.

Key Concepts

Multi-Agent Systems
A multi-agent system consists of multiple autonomous or semi-autonomous agents that interact and collaborate to achieve a set of goals. In BasisOS, these agents are arranged in a graph structure, enabling them to share information, coordinate on tasks, and divide complex objectives into manageable subtasks.
Agent Graphs
An agent graph defines the relationships between different agents and sub-agents, outlining how messages, data, and tasks flow through the system. In BasisOS, each node in the graph represents a specialized agent (e.g., a retrieval agent, a data agent, or a character agent), while the edges represent communication pathways and supervisory relationships.
Supervisors
Supervisors are specialized agents responsible for overseeing and coordinating the activities of other agents. They manage task delegation, track progress, and handle conflict resolution or prioritization issues. In BasisOS, multiple supervisors may operate in parallel, each focusing on a different layer or subset of agents, which helps ensure scalability and robust orchestration.
ReAct Agents
“ReAct” refers to a pattern of Reasoning and Action. In a ReAct agent:
- Reasoning involves analyzing the current state, the conversation or observations, and the agent’s internal knowledge to plan the next steps.
- Action is the execution of a chosen step, such as calling a tool, updating a database, or sending a message.This cycle of reasoning and action allows agents to dynamically respond to new observations or updates in the environment, making them more adaptive and context-aware.

PreviousExecution Flow NextCore Components

Last updated 5 days ago

Key Concepts

Multi-Agent Systems
A multi-agent system consists of multiple autonomous or semi-autonomous agents that interact and collaborate to achieve a set of goals. In BasisOS, these agents are arranged in a graph structure, enabling them to share information, coordinate on tasks, and divide complex objectives into manageable subtasks.
Agent Graphs
An agent graph defines the relationships between different agents and sub-agents, outlining how messages, data, and tasks flow through the system. In BasisOS, each node in the graph represents a specialized agent (e.g., a retrieval agent, a data agent, or a character agent), while the edges represent communication pathways and supervisory relationships.
Supervisors
Supervisors are specialized agents responsible for overseeing and coordinating the activities of other agents. They manage task delegation, track progress, and handle conflict resolution or prioritization issues. In BasisOS, multiple supervisors may operate in parallel, each focusing on a different layer or subset of agents, which helps ensure scalability and robust orchestration.
ReAct Agents
“ReAct” refers to a pattern of Reasoning and Action. In a ReAct agent:
- Reasoning involves analyzing the current state, the conversation or observations, and the agent’s internal knowledge to plan the next steps.
- Action is the execution of a chosen step, such as calling a tool, updating a database, or sending a message.This cycle of reasoning and action allows agents to dynamically respond to new observations or updates in the environment, making them more adaptive and context-aware.