ARGOS

ARGOS: Autonomous Infrastructure Agent

ARGOS, which stands for Autonomous Resilient Guardian Orchestration System, is a self-hosted AI-driven infrastructure agent designed to effectively monitor, manage, and operate a complete server fleet using natural language. This solution is particularly beneficial for system administrators seeking AI capabilities without the dependence on cloud services.

Core Features

• Chat Interface: Interact with your infrastructure using natural language. ARGOS can execute commands, read files, and manage services across all machines through SSH.
• Fleet Management: Gain a unified view of all systems, whether physical servers, virtual machines, or containers, with real-time online/offline detection facilitated by heartbeat and ping methods.
• Health Monitoring: Monitor key metrics such as CPU and memory usage, database latency, and container status, with a live SSE activity stream for quick insights.
• Job System: Execute multi-step operations with built-in risk assessment and approval workflows. Critical operations will require explicit authorization to enhance security.
• Agent Loop (Commander): An autonomous task execution system incorporating a phased state machine (executing, verifying, fixing), complete with safety guards and a fix loop counter.
• Skills System: Leverages over 110 skills that are loaded contextually via BM25 scoring and can auto-learn new systems through web search capabilities.
• Multi-LLM Support: Utilizes Claude (primary) and Ollama (local fallback) through a provider abstraction layer for enhanced performance.

Architecture Overview

Browser (Alpine.js + htmx)  
    |  
    v  SSE + REST  
ARGOS API (FastAPI, port 666)  
    |  
    +---> PostgreSQL (40+ tables)  
    +---> SSH to all fleet nodes  
    +---> Claude API / Ollama  

• Frontend: Utilizes Alpine.js 3.14 and htmx 2.0 to maintain a zero build pipeline.
• Backend: Powered by Python 3.13, FastAPI, asyncpg, and the Anthropic SDK.
• Database: Based on PostgreSQL 16.
• Deployment Options: Supports deployment via Docker Swarm for production environments or Docker Compose for single-node setups.

Quick Start Guide

For detailed installation instructions, refer to INSTALL.md.

git clone git@github.com:DarkAngel-agents/argos.git  
cd argos  
cp env.example .argos.env   # Customize your API keys  
docker compose -f docker/docker-compose-standalone.yml up -d  
# Access the interface at http://localhost:666  

System Requirements

• Docker and Docker Compose
• Anthropic API key
• PostgreSQL 16 (included in the standalone compose setup)

Project Structure

argos/  
+-- argos-core/  
|   +-- api/          # FastAPI endpoints  
|   +-- agent/        # Commander agent loop  
|   +-- llm/          # LLM provider abstraction  
|   +-- skills/       # Contextual skill files  
|   +-- tools/        # Audit tools  
|   +-- ui/           # Frontend  
|   +-- heartbeat.py  # Node health daemon  
+-- docker/           # Docker configs  
+-- argos-nanite/     # Bootable agent ISO (Work in Progress)  

Current Status

Alpha: The system is functional for single-user self-hosted use cases.

Author

Developed by DarkAngel, a seasoned sysadmin with over 20 years of experience, based in France.