After almost 2 years of work and delaying the release several times, we are super excited that today we can say.. we have officially released version 5!
The changes we've made address challenges we've encountered over years of real-world deployments, and they enable a more flexible, extensible platform that better serves all parts of our community: researchers, algorithm developers, system administrators, and data organizations.
Why This Release Matters
Over the past years, working with our growing user base, we encountered several critical limitations. Algorithms were tightly coupled to specific data sources, making them difficult to share across projects. Researchers couldn't easily explore their data before running computations. System administrators asked for better security and standard tooling. The vantage6 infrastructure was custom-built, which meant maintaining features that Kubernetes or other platforms already solved well.
Rather than continue patching individual problems, we took the opportunity to redesign some of vantage6 core mechanisms. This release is the result of that effort.
Three Major Changes
Sessions: Decoupling Data from Computation
We've introduced sessions, a new way of working that separates data extraction, preprocessing, and compute into distinct, reusable steps. This might sound technical, but it impacts the entire workflow.
In the old model, an algorithm function had to handle everything: retrieving data, preparing it, and computing results. This meant algorithms were bound to specific databases and had to be rewritten for different projects. Now, these concerns are separated. A data extraction function can be written once and reused. Preprocessing steps can be shared. Compute functions remain generic and database-agnostic. For the first time, algorithms written by researchers can easily be adapted for use in different collaborations without modification.
For users, this means faster iteration. You can extract data once and run multiple analyses on it without expensive re-queries. The UI understands your data structure before you begin, so you can preview columns and validate your setup upfront.
Read our deep dive on sessions →
From Docker to Kubernetes
When vantage6 began in 2017, Docker was the clear choice for containerization. Since then, the ecosystem has evolved. We found ourselves building custom tooling for networking, resource management, security isolation, and deployment—features that Kubernetes provides out of the box, maintained by thousands of contributors worldwide.
The transition to Kubernetes is not just a technical upgrade. It brings better security (no more mounted Docker sockets with root access), proper resource management, support for diverse hardware configurations (GPUs, HPC clusters), and simplified deployment through Helm charts. System administrators can now use familiar Kubernetes tools instead of learning vantage6-specific solutions.
Read our detailed explanation →
Keycloak: A Professional Identity Layer
One of the most significant but less visible changes in v5 is the move from a custom authentication system to Keycloak, an open-source identity and access management platform used by thousands of organizations worldwide. This makes the platform more secure and opens new possibilities.
Getting Started
If you're running algorithms in v4, start with our migration guide. If you're setting up new infrastructure, see the updated installation documentation.
The vantage6 team is super proud of what version 5 enables. We believe this release will serve our community better as the platform continues to grow.