Retrieval Augmented Generation models have the constant necessity to be updated when there are changes in the source of information. In order to keep the model delivering good answers it is important to set a up procedure to refresh its knowledge. This can be done by hand for a proof of concept project but for bigger projects we should set up a pipeline to automate this process. In this lighting talk I am going to explain how it is possible to create a Tekton pipeline to feed your RAG model with new updates, as well as, make this pipeline reusable and use it to ingest data into different RAG models.

Create delivery pipelines can be a hard task for non devops people. There is a lot of tools out there, each one with your own syntax and many of them claims that is simpler and easier than the others. A common concern between software engineers is to reuse components that execute the same task in order to avoid the same thing again. This is possible in Kubernetes environment with Tekton, where we can describe the task in a YAML file. The aim of this lighting talk is to introduce the Artifact Hub to developers and show that you don't need to be a devops master to create a simple pipeline able to clone, build, push and deploy your app, instead you can just fetch these pieces from Artifact Hub and put them together to have your pipeline. Let me explain how we can do this by creating a reusable pipeline in the talk.