 Hello, my name's Catherine and I'm a software developer at the University of Melbourne. I'm going to talk about Galaxy Australia's tool installation system, which has been built by myself, Simon Gladman and Tom Coddy. We've designed a process to install tools conditionally on the outcome of tool tests. All of the code for this is in our tools GitHub repository and we use a Jenkins server to run it. Tool installation can be triggered at any time through the merging of a pull request that will send a webhook to the Jenkins server. Tools are updated once a week, subject to their owners being on a list of trusted owners. Testing is an integral part of this process. A tool that does not pass its tests upon installation is uninstalled and resubmitted as a pull request. Each tool installation is recorded as a row in a searchable TSV file in the GitHub repository. Now each new tool shared repository on Galaxy Australia is added as a result of a pull request. You can make a pull request for a tool on Galaxy Australia by adding a file in the requests directory. We use the ephemeris format for these files since we are using ephemeris for installing and testing. The file is a YAML list containing at minimum name, owner and section label. We have a linting script that will check the validity of tools with regard to correct YAML, correct keys, a section label from an allowed list and also that the tool is installable. And there is also a simpler way to request a tool in Galaxy Australia. On our site we have this button, request tool or data set which links to a request form for a Galaxy user to request a new tool or reference data set. The form collects some use case information, a toolshed link if the request is for a tool and creates a GitHub issue for us that use Galaxy AU tools so that we can easily convert this toolshed link into a pull request. Because of the automation we can have a fairly fast turnaround time for user tool requests. And this is a picture of all of the steps in tool installation. There are two processes, tool installation which runs from the merging of a pull request and tool updates which run once a week. The steps in orange are the steps that require human action and everything else is automatic. There is a list of tools to install either from pull request files or from a list of tools with updates available. This is divided into one file per unique repository. Each repository one by one is installed on our staging server, tested on staging, installed on production and tested on production. If anything goes wrong, the installation is wound back and the file is saved for a follow-up pull request. More often it ends up in this green state, everything has passed and the file is removed. After this is run for all of the tool files, if any of the files have failed, a pull request is made. Then logs and lists of installed tools are updated in the repository. And a tiny bit about the code, it's written in bash with some Python scripts as well. Everything runs from the main.sh file including building the Python environment, installing and testing with Ephemeris, uninstalling tools with Bioblend, creating planamoutest reports, committing changes back to the repository and creating follow-up pull requests. Main.sh takes one positional argument which could be install or update. We use Jenkins but the script can be run locally and has also been adapted by Galaxy Norway to use GitHub Actions instead of Jenkins. And for tools that do not pass the tests, we review the planamoutest reports for these. Sometimes it doesn't matter. Sometimes there's a condo problem that we need to resolve. A couple of times we have had to reject tools because they would never run. And if the text force in square brackets is in the pull request title, the installation script knows to just install anyway and forget about tests. We've been running this since February 2020 and we've used it to install 1,500 repositories, 400 of which are new to Galaxy Australia. And that's all. I'd like to acknowledge the whole Galaxy Australia team but particularly my co-author and boss, Simon Gladman, who envisaged this project long before my involvement with Galaxy. Also Melbourne Bioinformatics, Australian Biocommons and the whole Galaxy community. Thanks for your time.