Skip to main content

Administration Guide

Welcome to the AtlasML Administration Guide! This guide provides comprehensive documentation for deploying, configuring, and maintaining AtlasML in production environments.

About AtlasML​

AtlasML is a FastAPI-based microservice that provides AI-powered competency management features for the Artemis learning platform. It requires a centralized Weaviate vector database, Azure OpenAI for embeddings, and is deployed exclusively via Docker Compose for production workloads.

Prerequisites​

Before deploying AtlasML, ensure you have:

  • Docker and Docker Compose installed on your server
  • A centralized Weaviate instance (see Weaviate Setup Guide)
  • Azure OpenAI API credentials for embedding generation
  • API keys for securing AtlasML endpoints
  • Basic knowledge of Docker, environment variables, and reverse proxies

Quick Start​

Follow this checklist to deploy AtlasML to production:

Deployment Checklist:

Documentation Sections​

Installation​

Step-by-step guide to deploy AtlasML using Docker Compose, including Weaviate setup, environment configuration, and initial deployment.

Configuration​

Complete reference for all environment variables, including Weaviate connection settings, Azure OpenAI credentials, API keys, and optional Sentry integration.

Deployment​

Production best practices, CI/CD workflows with GitHub Actions, secrets management, and deployment strategies.

Monitoring​

Health check endpoints, log management, container monitoring, and Sentry error tracking for production observability.

Troubleshooting​

Common issues and solutions for startup failures, Weaviate connection problems, API errors, and performance issues.

Architecture Overview​

AtlasML follows a microservice architecture:

  • AtlasML Service: FastAPI application serving REST endpoints
  • Centralized Weaviate: Shared vector database with HTTPS and API key authentication
  • Azure OpenAI: Embedding generation service
  • Artemis: Primary client consuming AtlasML's competency management features

Communication is unidirectional—Artemis calls AtlasML, and AtlasML never initiates requests back to Artemis.

Support​