Join us for an exciting workshop where we explore the powerful combination of OpenShift AI and Azure AI—showing how they truly are better together! This hands-on session is designed specifically to work off an use-case (insurance), providing participants with guided labs and practical exercises to help uncover and implement AI-driven solutions.
Get ready to dive into the future of insurance with Parasol, an innovative company using AI to supercharge the efficiency of claim processing! We're harnessing cutting-edge models and technologies to revolutionize how things are done.
In this workshop, you'll see firsthand how Natural Language Processing (NLP) and Large Language Models (LLMs) are transforming the user experience. But we’re not stopping there! Fine-tuning and comparing models? We’ve got that covered, too.
Throughout this hands-on exercise, you’ll explore how AI can simplify your workflows. From consulting models for information to performing image recognition and comparing model outputs, you'll experience the incredible capabilities AI has to offer.
And the best part? While we’re working with a predefined insurance use case, we’re confident the knowledge and skills you gain will be invaluable for your own AI projects.
Ready to collaborate with Red Hat and Microsoft on your use case? We’re here to support you every step of the way—reach out, and let’s make it happen!
This repository contains the code, instructions, resources and materials associated with the Lab called OpenShift AI + Azure AI Workshop.
To consult the static version of the instructions, please use this URL
If you want to participate in the creation and update of this content, please consult the sections below.
Display Development-centric information
mainbranch is the one used for production. That's where the Prod and Test catalog items from demo.redhat.com point to (instructions, materials used,...).devbranch is for development. That's where the Dev catalog item points to.- Branches are made from
dev(hot fixes could be made frommainif really needed). - When ready, PRs should be made to
dev. Once all features, bug fixes,... are checked in and tested for a new release, another PR will be made fromdevtomain. - Branches must be prefixed with
/feature(examplefeature/new-pipeline-instructions),bugfix, or other meaningful info. - Add your name/handle in the branch name if needed to avoid confusion.
- If your development relates to an Issue or a Feature Request, add its reference in the branch name.
- Try to stash your changes before submitting a PR.
Useful link: https://redhat-scholars.github.io/build-course/rhs-build-course/develop.html
- Podman or Docker
- Add/Modify/Delete content in content/modules/ROOT.
- Navigation is handled in
nav.adoc. - Content pages are in the
pagesfolder. - To build the site, from the root of the repo, run
./content/utilities/lab-build. - To serve the site for previewing, from the root of the repo, run
./content/utilities/lab-serve. - The site will be visible at http://localhost:8443/
- When finished, you can stop serving the site by running from the root of the repo
./content/utilities/lab-stop.
- Python 3.11
- Nodejs > 18
- An existing instance of Hugging Face TGI with a loaded model available at
INFERENCE_SERVER_URL. This application is based on Mistral-TB Prompt format. You will need to modify this format if you are using a different model.
Run npm install from the main folder.
If you want to install packages manually:
- In the
frontendfolder, install the node modules withnpm install. - In the
backendfolder, create a venv and install packages with the provided Pipfile/Pipfile.lock files. - In the
backendfolder, create the file.envbase on the example.env.exampleand enter the configuration for the Inference server.
From the main folder, launch npm run dev or ./start-dev.sh. This will launch both backend and frontend.
- Frontend is accessible at
http://localhost:9000 - Backend is accessible at
http://localhost:5000, with Swagger API doc athttp://localhost:5000/docs
#!/bin/bash
# Script to restart all showroom pods - You must be logged in as a cluster admin to run this script
# Get all namespaces
namespaces=$(oc get namespaces -o jsonpath='{.items[*].metadata.name}' \
| tr ' ' '\n' \
| grep '^showroom')
# Stop all the pods
for namespace in $namespaces; do
# Check if the deployment "showroom" exists in the namespace
if oc -n $namespace get deployment showroom &> /dev/null; then
# If it exists, restart the rollout
# oc -n $namespace rollout restart deployment/showroom
oc -n $namespace scale deploy showroom --replicas=0
fi
done
# wait for them all to fully stop
# start all the pods
for namespace in $namespaces; do
# Check if the deployment "showroom" exists in the namespace
if oc -n $namespace get deployment showroom &> /dev/null; then
# If it exists, restart the rollout
# oc -n $namespace rollout restart deployment/showroom
oc -n $namespace scale deploy showroom --replicas=1
fi
done
- From
dev, create a new branch, likefeature/prepare-for-main-merge. - Modify the following files to make their relevant content point to
main:bootstrap/applicationset/applicationset-bootstrap.yamlcontent/antora.ymlcontent/modules/ROOT/pages/05-03-web-app-deploy-application.adoc
- Make a pull request from this branch to
main, review and merge
Links for event environment assignment
- URL for all labs: Provided during the day
- Login and username: Provided during the day
Let’s begin with an in-depth look at the architecture of our solution.
The entire platform is hosted on Azure, where we harness the combined power of OpenShift AI running on Azure Red Hat OpenShift along with a suite of Azure AI services. This powerful combination enables us to create an advanced, scalable solution designed to meet all of your AI needs.
Azure Red Hat OpenShift (ARO) is an exceptional, fully managed application platform that has been co-engineered and is jointly supported by both Microsoft and Red Hat. It provides the perfect foundation for running enterprise-grade applications. When you integrate OpenShift AI on top of ARO, you can build, train, and deploy AI models seamlessly within a robust, cloud-native environment.
On top of that, Azure offers an array of fully managed AI services like Azure OpenAI, Azure AI Search, and many others. These services allow you to focus on driving business value instead of worrying about the complexities of implementing and maintaining infrastructure. With Azure's managed offerings, you're able to accelerate innovation without the overhead of managing an entire AI platform.
For Parasol's solution, we’ve chosen the following combination of services:
- Azure Red Hat OpenShift (ARO): The foundation for hosting both frontend and backend services.
- OpenShift AI: Providing the AI capabilities on top of ARO.
- SLM Deployment on ARO to compare results with GPT4o
- Azure OpenAI GPT-4o: Powering the natural language interface for efficient interaction.
- Azure AI Search: Serving as a highly capable vector database, with added features beyond just vector storage
- Azure Managed PostgreSQL: Supporting our backend database needs, specifically for handling claims.
In this architecture, the frontend and backend services are running on ARO, with OpenShift AI fully integrated. We leverage Azure OpenAI GPT-4o for a natural language processing interface, making it easier for users to interact with the system. For data storage and management, we rely on Azure Managed PostgreSQL for the claims database, while Azure AI Search serves as our vector database, offering enhanced features to improve data retrieval and search functionalities.
By combining these technologies, we create a streamlined, highly efficient solution that allows businesses to focus on innovation and driving real results without the complexity of managing AI platforms.