Interactive Galaxy Tools: Combining real-time user interaction with scientific workflows

Galaxy [1] is a GUI-based scientific data management and analysis platform designed to make scientific computing accessible to users without advanced technical skills. Typically, Galaxy runs asynchronous workflows where user actions trigger jobs executed on a remote system, with results returned upon completion. However, some scientific tools—particularly interactive visualization software—require synchronous execution with user input. Galaxy's Interactive Tools (ITs) [2] address this by combining Galaxy's tool and job systems with modern containerization. ITs allow dynamic applications (e.g., R Shiny, Python Voilà, Ruby on Rails) to run interactively within Galaxy. Developers supply a container image and a Galaxy tool wrapper, and the system handles execution, user access via proxy, and dataset integration. Users interact with the tool via an auto-generated URL, can use Galaxy datasets as input, and save outputs as standard Galaxy datasets. ITs can be used within workflows, with downstream steps triggered after finalizing interactive results. All activity is recorded in the user's history, ensuring reproducibility, sharing, and full use of Galaxy's compute and storage infrastructure. A compelling use case that highlights the power of integrating custom functions into IT applications is the development of a machine learning model within Galaxy using an interactive Jupyter Notebook [3]. Galaxy and the Jupyter Notebook IT enable seamless data exchange through integrated put and get functions, allowing users to transfer data between the notebook and Galaxy without manual uploads or downloads. For example, a user might generate training data using a Galaxy workflow—such as one of the metagenomic workflows provided by the microGalaxy community [4]—which performs a computationally intensive task that would be impractical to run locally. This training data can then be sent to the Jupyter Notebook IT, potentially in an environment preconfigured with access to GPU resources. The user can train a custom machine learning model within the notebook, and the resulting predictions can be sent back to Galaxy for further processing, such as statistical analysis with other tools. Results may then be returned to the notebook for custom analysis or visualization. This setup supports dynamic and exploratory research workflows while overcoming common technical and security limitations, such as data transfer bottlenecks, hardware constraints, and institutional restrictions. Similar integrations are already in use—for instance, in the Shiny app phyloseq [5], where the put and get functions enable real-time figure generation and data transfer back to Galaxy—demonstrating the versatility of custom function integration across diverse platforms. To date, the 4 public galaxy servers usegalaxy.eu/.fr/.org/org.au, have already had 140,000+ IT jobs executed. ITs can provide an opportunity to integrate more interactive scientific use-cases and visualizations and will increasingly gain popularity in the community. In this contribution, we will walk through the integration and practical usage of the ITs to gradually increase interest in the deployment of new ITs and support more interesting and diverse use cases.