MedOmni
The ecosystem for omni-modality medical intelligence
MedOmni is open infrastructure for medical AI on any data modality: tables, waveforms, images, volumes and notes.
Goal
A clinician reads lab tables, ECGs, sleep recordings, X-rays, retinal scans and notes side by side. MedOmni aims to be the shared, open infrastructure where agents and models learn to do the same: take in any medical data modality, be tested on it, and improve. A dataset or encoder added to MedOmni once can be used by every agent and model tested on it.
Overview
MedOmni is a shared Core with projects built on it. Each project depends on Core, and the projects exchange data only through Core's contracts, so each can be used and changed on its own.
| Project | What you get | Command |
|---|
| Bench | Test your agent, or a foundation model behind one, on built medical tasks, and score it. | medomni bench |
| Perception | Give your agent perception across medical modalities: encoders behind one HTTP gateway, each reading one kind of input, such as a 12-lead ECG. | medomni perception |
| Studio | The rest of the workflow: survey a dataset, preview and select cohorts, build tasks, and audit them. | medomni studio |
| Core | What every project shares: the contracts, the medomni command, plug-in loading, dataset loading, the cache and the testing kit. | medomni |
Anyone can add a dataset, an encoder, a task or an agent without changing Core: by pull request, or as a plug-in package in their own repository. See Contributing.
Data use and patient privacy
MedOmni works in a sensitive medical domain, and we respect the data use agreement of every dataset it supports. No patient data ships with MedOmni: you build tasks from generated dummy data, open datasets such as VitalDB, or your own licensed copy of a dataset such as MIMIC-IV, and what is built from patient data stays on your machine.
Install
Python 3.11 or later, Docker and uv.
git clone --branch v0.3.0 https://github.com/medomni/MedOmni medomni && cd medomni
uv venv && source .venv/bin/activate && uv pip install -e .
medomni --version
Then follow Get started to build, run and score a first task set on dummy data.
Repository
src/medomni/ the package: core/, bench/, perception/, studio/
core/families/ one file per input family, and aliases.yaml
bench/ benchmarks/ (benchmark cards and task sets), agents/
perception/ encoders/<family>/<encoder>/, deployed.yaml
studio/ datasets/<dataset>/ (dataset packages, templates, cohorts)
tests/ one folder per project
docs/ the documentation
Documentation
The documentation is at medomni.ai. To start:
Licence
MedOmni's code is under the Apache License 2.0; see LICENSE. Each encoder, dataset package and task set states its own licence on its card or policy.
Releases
This is v0.3.0, the first public release, a beta. Releases are tags on github.com/medomni/MedOmni; there is no PyPI release.
