ModelCubLocal-First MLOps for Computer Vision

Quick Start ​

# Install
pip install modelcub

# Initialize project
modelcub project init my-project
cd my-project

# Import dataset
modelcub dataset add --source ./data --name v1

# Launch UI
modelcub ui

Why ModelCub? ​

The Problem ​

Cloud platforms lock you in: Roboflow charges $500-$8k/month. Your data lives on their servers. Switching is painful.

DIY is fragmented: Label Studio + Ultralytics + custom scripts = integration hell. No version control. Hard to reproduce.

Privacy is compromised: Medical imaging, defense projects, and proprietary data can't use cloud platforms.

The Solution ​

ModelCub gives you a professional, integrated platform that runs entirely on your infrastructure:

• ✅ Complete workflow (import → annotate → train → deploy)
• ✅ 100% local, 100% private
• ✅ Git-like version control for datasets
• ✅ Clean Python SDK + CLI + Web UI
• ✅ Free and open source

Comparison ​

Python SDK ​

from modelcub import Project, Dataset

# Initialize project
project = Project.init("my-project")

# Import dataset
dataset = Dataset.from_yolo("./data", name="v1")

# Get statistics
print(f"Images: {dataset.num_images}")
print(f"Classes: {dataset.classes}")

# Access splits
train_images = dataset.splits['train'].images
print(f"Training images: {len(train_images)}")

Use Cases ​

Medical Imaging ​

HIPAA-compliant, on-premise training. Your patient data never leaves your servers. Perfect for hospitals and research institutions.

Startups & Indie Developers ​

Save $96k/year in cloud costs. Use that budget to hire engineers instead of paying SaaS fees.

Research Labs ​

Reproducible experiments with full audit trails. Version datasets alongside code. Perfect for academic papers.

Defense & Government ​

Air-gapped deployments. Zero external dependencies. Complete control over your data and infrastructure.

Architecture ​

ModelCub is built on clean, layered architecture:

┌─────────────────────────────────┐
│    CLI  │  SDK  │  Web UI      │
├─────────────────────────────────┤
│      FastAPI Backend            │
├─────────────────────────────────┤
│      Core Services              │
├─────────────────────────────────┤
│   File System State             │
│  (.modelcub directory)          │
└─────────────────────────────────┘

Key Principles:

• API-First: Everything is composable
• Stateless: No hidden database, all state in files
• Format-Agnostic: YOLO internally, import/export anything
• Git-Friendly: Human-readable text files

Get Started ​

# Install
pip install modelcub

# Verify
modelcub --version

# Create your first project
modelcub project init demo-project

Read the Installation Guide or jump to the Quick Start.