How xtreme1-io/xtreme1 works
Project type and purpose
• This is a full-stack, open-source platform (“Xtreme1”) for multimodal ML data annotation, curation, and ontology management.
• Backend: Java Spring Boot, hexagonal (ports/adapters) architecture.
• Frontend: a Vite-based monorepo with several Vue applications (main dashboard, image-tool, pc-tool, text-tool).
• Infrastructure as code under “.ops” (Dockerfiles, Kubernetes manifests via Kustomize).
File structure
• Top-level directories:
• backend/ (Spring Boot service with adapter/api, usecase, port, entity layers)
• frontend/ (multiple Vue apps: main, image-tool, pc-tool, text-tool)
• .ops/ (Dockerfiles + k8s/base, dev, test overlays)
• deploy/ (docker-compose, nginx, MySQL migration scripts)
• Patterns:
• Backend follows hexagonal layering: adapter → usecase → port → entity
• CI/CD in .github & .ops, containerization in Docker and Kubernetes
README for additional insights
• Services: MySQL, Redis, MinIO, plus application containers (backend, frontend).
• Built-in “model” services deployed via Docker profiles (requires GPU + CUDA).
• Docker Compose orchestrates pull vs. build, volumes, networks.
Detail guidance
• MySQL, Redis, MinIO around backend to show direct connections.
• External model inference containers alongside backend, show HTTP calls from adapter/api/job or rpc callers.
• UI apps under “Frontend” and show they all call the same backend API endpoint (or different endpoints if segregated).
• Keep labels concise but specific (e.g., “AuthController (JWT Filter)”, “DatasetUseCase”, “MinIO Object Store (S3 API)”).