Home

Nebulus Atom Wiki

Nebulus Atom is a professional, autonomous AI engineer CLI that interacts with a local LLM server (Nebulus/Ollama) to assist with software engineering tasks. It includes Nebulus Swarm, a multi-agent orchestration system that autonomously processes GitHub issues at scale.

Quick Links

Getting Started	Core System	Swarm System
Installation	Architecture	Nebulus Swarm
Quick Start	CLI Reference	Swarm Overlord
Configuration	Features	Swarm Minion
Deployment	Testing	Swarm Dashboard
Troubleshooting	Contributing	Model Router

What is Nebulus Atom?

Nebulus Atom is an AI-powered CLI agent that:

• Reads, writes, and edits code autonomously
• Manages context with pinned files and semantic search (RAG)
• Creates checkpoints and supports smart undo
• Builds and executes persistent skills
• Runs in autonomous execution mode with TDD support
• Provides a TUI dashboard for real-time monitoring

What is Nebulus Swarm?

Nebulus Swarm extends Nebulus Atom into a multi-agent system:

• Overlord orchestrates work via Slack commands or cron schedules
• Minions are containerized agents that clone repos, work on issues, and create PRs
• Model Router routes simple tasks to lightweight models and complex tasks to heavyweight models
• Dashboard provides a Streamlit web UI for monitoring swarm activity
• PR Reviewer automatically reviews minion-created pull requests

Project Status

• Version: 0.1.0
• Python: 3.12+
• License: See repository
• 376 tests passing

Architecture Overview

nebulus-atom/
├── nebulus_atom/          # Core CLI agent (MVC architecture)
│   ├── models/            # Data structures
│   ├── views/             # UI (Rich TUI)
│   ├── controllers/       # Orchestration
│   ├── services/          # Integrations (LLM, RAG, Skills)
│   └── main.py            # Entry point
│
├── nebulus_swarm/          # Multi-agent swarm system
│   ├── overlord/           # Control plane (Slack, Docker, State)
│   ├── minion/             # Worker agents (GitHub, LLM, Tools)
│   ├── dashboard/          # Streamlit monitoring UI
│   └── reviewer/           # Automated PR review
│
└── tests/                  # 376 tests (pytest)