📁 Subsystem Deep Dive

utils/swarm/

📁 utils/swarm/ 📄 7 files

Tutorial: swarm

This project implements a hierarchical multi-agent system ("swarm") where a leader agent can spawn and coordinate autonomous teammate agents to perform tasks in parallel. It abstracts the physical execution environment, allowing teammates to run as isolated in-process tasks, tmux panes, or iTerm2 splits through a unified interface. The system ensures safe collaboration via a distributed permission system that routes tool approval requests from workers to the leader, while persisting team state to disk to maintain continuity across sessions.

flowchart TD A0["Execution Backends"] A1["Teammate Executor Adapter"] A2["Distributed Permission System"] A3["Team State Persistence"] A4["In-Process Teammate Runtime"] A5["Environment Layout Management"] A1 -->|"Adapts and controls"| A0 A1 -->|"Spawns internal agents"| A4 A5 -->|"Configures terminal panes"| A0 A4 -->|"Requests tool approvals"| A2 A2 -->|"Reads team configuration"| A3 A4 -->|"Updates member state"| A3

Chapters

  1. In-Process Teammate Runtime
  2. Teammate Executor Adapter
  3. Execution Backends
  4. Environment Layout Management
  5. Distributed Permission System
  6. Team State Persistence

Generated by Code IQ

Files in this section

01_in_process_teammate_runtime.md
Chapter 1: In-Process Teammate Runtime
Welcome to the Swarm project! In this first chapter, we are going to explore the foundation of how our AI agents live a…
02_teammate_executor_adapter.md
Chapter 2: Teammate Executor Adapter
In the previous chapter, In-Process Teammate Runtime, we learned how to run efficient agents inside our main applicatio…
03_execution_backends.md
Chapter 3: Execution Backends
In the previous chapter, the Teammate Executor Adapter, we built a "Universal Remote" that allows us to send commands l…
04_environment_layout_management.md
Chapter 4: Environment Layout Management
In the previous chapter, Execution Backends, we learned how to physically create terminal panes using tools like tmux o…
05_distributed_permission_system.md
Chapter 5: Distributed Permission System
In the previous chapter, Environment Layout Management, we organized our agents into neat, colorful windows.
06_team_state_persistence.md
Chapter 6: Team State Persistence
In the previous chapter, Distributed Permission System, we learned how to make our swarm safe by requiring permission f…