⭐ New features

Here is a quick summary of the major updates in the platform that are relevant for users. Minor updates to the documentation will still be performed continuously but not notified here.

📘 docs, 📈 dashboard, 📌 others

• 📈 new section in the Dashboard with usage stats (2024-05-14)

• 📘 support for token authentication in Federated Learning trainings (2024-04-16)

• 📘 new New features page (meta!) (2024-04-03)

• 📘 new tutorials for using AI workflows tools (Elyra, Flowfuse) (2024-03-18)

• 📘 new MLflow tutorials (2024-02-09)

• 📘 new Frequently Asked Questions (FAQ) page (2024-02-05)

• 📌 MLflow available under new domain (2024-01-29)

• 📈 we enforce a quota of 2 GPUs per user for better resource distribution (2024-01-25)

• 📘 updated Quickstart page (2024-01-24)

• 📈 users are able to select specific GPU models (2024-01-18)

• 📌 new Nextcloud storage available (2024-01-15)

🚀 Upcoming features

In this section, we will offer some insight on to where the AI4OS stack is heading.

First, the are a number of mature features that are in the process of being integrated in the stack:

• 📌 new unified AI module structure

• 📈 usage stats (per user and per VO) are displayed in the Dashboard

• 📈 support for secrets in the Dashboard (eg. Federated Learning server)

• 📈 inference endpoints with OSCAR

• 📈 short-lived endpoints to try the modules

• 📈 email notifications for deployments that take a longer to deploy

• 📈 new tool for image annotation (CVAT)

• 📈 new tool for efficient video streaming (Kafka)

Along with these upcoming features, we are exploring more improvements. These are in an experimental status and might not finally get integrated in the stack.

• 📈 carbon footprint estimator for your deployments

• 📈 show map with datacenter usage stats

• 📈 graph visualizations of your module/deployment metadata

• 📈 ability to pause deployments

• 📌 using volumes in deployments to better share data across deployments

Do you find something missing? Please check how to suggest a new feature.