WARNING: We develop on GitLab. The GitHub repository is a read-only mirror — please open issues and merge requests on GitLab instead. To see the whole application code, see https://gitlab.com/ryax-tech/ryax/
⭐ About the project
Ryax is the new open-source hybrid workflow orchestrator to optimize your AI workflows and applications.
Create AI workflows and orchestrate them over hybrid infrastructure (multi-HPC and multi-cloud).
⚙ The source code
Ryax is composed of multiple micro-services and tools. To learn more about the internal architecture, see the documentation. The main components are:
User interfaces
Micro-services
- studio: Handles the editing of workflows.
- runner: A trigger or an action run? It handles it.
- authorization: Answers the question: "do you have the rights to do so?".
- repository: Scans git repositories to find actions.
- action-builder: Builds actions and triggers.
Ryax Actions
Others
- action-wrappers: Nix-based building tool for Ryax actions that injects the Ryax wrapper.
Main technologies used by Ryax
The code: Python, dependency-injector, SQLalchemy, GRPC.
Around the code: Nix, uv, ruff, mypy.
Deploying Ryax: Kubernetes, Helm, Slurm.
⚡ Getting started with Ryax
Requirements
All you need to install Ryax is a Kubernetes cluster. Supported versions are:
Hardware:
- At least 2 CPU cores
- 4GB of memory
- 40GB of disk available
helm install ryax oci://registry.ryax.org/release-charts/ryax-engine -n ryaxns --create-namespace
Note that depending on the actions that you run on your cluster, you might need more resources.
On an existing Kubernetes cluster
This is the standard and recommended approach.
It works on most managed Kubernetes services, like AWS EKS, Azure AKS, GCP GKE, and Scaleway Kapsule.
For more details on the configuration, see our documentation.
Test locally with Kubernetes in Docker
We recommend this option if you wish to test our product with a minimal amount of configuration steps, and if you have enough RAM (~3GB) and disk (20GB) available.
/!\ Warning: Do not use this setup in production!
/!\ Warning: To make it easier for you to access the cluster from your
browser, we expose the ports 80 (HTTP) and 443 (HTTPS) on your local machine.
Make sure these aren't already used!
Copy the docker-compose.yml and the minimal.yaml files from this repository and run:
docker-compose up -d
Check that your local k3s is started with:
export KUBECONFIG=$PWD/kubeconfig.yaml
kubectl get pods -A
Once all pods are running, run:
helm install ryax oci://registry.ryax.org/release-charts/ryax-engine -n ryaxns --create-namespace -f ./minimal.yaml
Be patient, this may take some minutes depending on your internet connection.
Once it's done, you can access your cluster at:
http://localhost/app/login
The admin password is generated for your installation. The helm install output
above tells you how to read it back; here are the same two commands:
RYAX_USER=$(kubectl -n ryaxns get secret ryax-admin-credentials -o jsonpath='{.data.admin-user}' | base64 -d)
RYAX_PASSWORD=$(kubectl -n ryaxns get secret ryax-admin-credentials -o jsonpath='{.data.admin-password}' | base64 -d)
echo "$RYAX_USER / $RYAX_PASSWORD"
Change it from Users once you are logged in; the secret is not updated when
you do.
Go to the Infrastructure > New Site and create a Kubernetes site called "Local".
Now click on Add node pool and add a "k3s" node pool with the quantity of resources you want to give to Ryax actions, for example: 1000 mCPU and 2GB of memory.
Copy the Site ID and the Node pool ID into variables or directly in the command and run:
SITE_ID="GET ME FROM UI"
NODE_POOL_ID="GET ME FROM UI"
helm install ryax-worker-k8s oci://registry.ryax.org/release-charts/ryax-worker-k8s -n ryaxns \
--set config.site.id=$SITE_ID \
--set 'config.site.spec.nodePools[0].id'=$NODE_POOL_ID \
--set 'config.site.spec.nodePools[0].selector.node\.kubernetes\.io/instance-type'=k3s
The same two objects can be created over the REST API, which is handy for
scripting the whole install. It needs jq:
JWT=$(curl -s -X POST http://localhost/api/authorization/login \
-H 'Content-Type: application/json' \
-d "{\"username\":\"$RYAX_USER\",\"password\":\"$RYAX_PASSWORD\"}" | jq -r .jwt)
SITE_ID=$(curl -s -X POST http://localhost/api/runner/sites \
-H "Authorization: Bearer $JWT" -H 'Content-Type: application/json' \
-d '{"name":"Local","type":"KUBERNETES"}' | jq -r .site_id)
# cpu is in millicores and memory in bytes: 1000 mCPU and 2GiB below.
# Use "arm64" as architecture on an ARM machine.
NODE_POOL_ID=$(curl -s -X POST http://localhost/api/runner/node-pools \
-H "Authorization: Bearer $JWT" -H 'Content-Type: application/json' \
-d "{\"site_id\":\"$SITE_ID\",\"name\":\"k3s\",\"architecture\":\"x86_64\",
\"cpu\":1000,\"memory\":2147483648,
\"energy_score\":50,\"performance_score\":50,\"cost_score\":50,
\"filter_no_gpu_action\":false}" | jq -r .node_pool_id)
echo "SITE_ID=$SITE_ID NODE_POOL_ID=$NODE_POOL_ID"
Then run the helm install above — $SITE_ID and $NODE_POOL_ID are already set.
Now you can add actions to your Library by adding the default-action repository: https://gitlab.com/ryax-tech/workflows/default-actions.git
Scan it, build useful triggers like, for example, Emit Every, HTTP API JSON, Run once, and HTTP POST.
Also, build some example actions like Echo and Cat content of a file.
Then, you can create your own workflows from the Dashboard and deploy them on your cluster!
For more detailed use cases, see the Quick Start Guide.
To uninstall your cluster, stop it with:
docker-compose down -v
🛹 Roadmap
A more complete roadmap will be published soon.
🤗 Contributing
If you want to say thank you and/or support the active development of Ryax:
- Add a GitHub Star to the project.
- Talk about the project on your favorite social network.
- Write a review or tutorial on Medium, Dev.to, or your personal blog.
- Share some triggers and actions with the community.
- Fix bugs and implement features in our code.
Discord,
Reddit,
Slack
