A tribe project to build a Model Platform
- Build a project as a team
- Test, identify, consolidate convictions on model management at scale
- Be as portable as possible
- Use opensource technologies
Vision
For models developers, so that they can focus on building the best model, we offer a model platform that can version, deploy, host, govern models with as few configuration as possible
For application developers, so that they can integrate models seamlessly, we offer a model platform that can provide a simple API to call.
A working minikube is needed for dev purposes
brew install minikube kubectl
Install helm
brew install helm
Start minikube cluster with recommended specs
#recommended configuration to avoid freezing/timeouts
minikube start --cpus 2 --memory 7800 --disk-size 50g(You may need to start colima with custom config)
colima start --cpu 4 --memory 8Activate Ingress controller add-on
minikube addons enable ingressActivate ssh tunnel to you minikube cluster
minikube tunnelAdd the following line to your /etc/hosts
# Mac
127.0.0.1 model-platform.com
# Linux
minikube ip
IP.RESULT model-platform.commake k8s-inframake create-backend-secret POSTGRES_PWD=your_postgres_password JWT_SECRET="ask for the JWT secret" ADMIN_EMAIL=alice@example.com ADMIN_PWD=pass!You should now have a working k8s environment
make k8s-modelplatformVia fronted
http://model-platform.com
or
mp login --username alice@example.com --password pass!
You will need to install nektos act https://nektosact.com/installation/
#mac ARM
make run-ci-armor
#Intel processors
make run-ci-amdAccess the pgsql via local db client
Run
kubectl port-forward svc/modelplatform-pgsql 5432:5432 -n pgsqlAdd to you db client
host: localhost
port: 5432
user: postgres
password: your_postgres_password
db: model_platform_db
you can now access the model_platform_db database