OpenModel Registry

The official public model registry for OpenModelStudio.

Browse, discover, and install community-contributed models into any OpenModelStudio instance.

How It Works

Each model lives in its own folder under models/. A GitHub Actions workflow aggregates all model metadata into registry/index.json on every push, providing a single endpoint for fast lookups.

Installing Models

# Install a model from the registry
openmodelstudio install titanic-rf

# Search for models
openmodelstudio search "image classification"

# List installed models
openmodelstudio list

Or from a notebook / Python script:

import openmodelstudio as oms

# Install from registry
oms.registry_install("titanic-rf")

# Search the registry
results = oms.registry_search("classification")

# Load an installed model and train it
model = oms.use_model("titanic-rf")
handle = oms.register_model("my-titanic", model=model)
job = oms.start_training(handle.model_id, wait=True)

Submitting a Model

1. Fork this repository
2. Create a folder under models/ with your model name (e.g., models/my-model/)
3. Add a model.json with metadata (see schema below)
4. Add your model code as model.py with train(ctx) and infer(ctx) functions
5. Open a Pull Request

model.json Schema

{
  "name": "my-model",
  "version": "1.0.0",
  "description": "Short description of what the model does",
  "author": "your-github-username",
  "framework": "pytorch",
  "category": "classification",
  "tags": ["tabular", "binary-classification"],
  "license": "MIT",
  "files": ["model.py"],
  "dependencies": ["scikit-learn>=1.3"],
  "homepage": "https://github.com/you/my-model"
}

Required Fields

Field	Description
name	Unique model identifier (lowercase, hyphens)
version	Semantic version string
description	What the model does
author	GitHub username or org
framework	One of: pytorch, sklearn, tensorflow, jax, python, rust
category	One of: classification, regression, nlp, computer-vision, generative, time-series, clustering, anomaly-detection, reinforcement-learning, multimodal
tags	Array of searchable tags
license	SPDX license identifier
files	Array of files to include (relative to model folder)

Custom Registries

You can point your OpenModelStudio instance at a different registry:

import openmodelstudio as oms
oms.set_registry("https://raw.githubusercontent.com/your-org/your-registry/main/registry/index.json")

Or set the environment variable:

export OPENMODELSTUDIO_REGISTRY_URL="https://raw.githubusercontent.com/your-org/your-registry/main/registry/index.json"

License

This registry and all contributed models are subject to their individual licenses. The registry infrastructure itself is licensed under GPL-3.0.