This repo contains deployment configurations and artifact generation machinery for alchemiscale instances hosted under alchemiscale.org.
Instances run on Amazon EKS and are deployed continuously: pull requests build images and stand up throwaway test environments, and publishing a release rolls the modified instances out to production behind a review gate.
| doc | answers |
|---|---|
| docs/quickstart.md | stand it up, try it, tear it down — start here |
| docs/continuous-deployment.md | how PRs and releases deploy; what to configure in the repository |
| docs/infrastructure.md | clusters, AWS account model, setup, upgrades, cost |
| docs/operations.md | identities, neo4j dumps, reaching PR environments, monitoring |
| docs/migration.md | phases and cutover runbooks for the EC2 instances |
| charts/alchemiscale/README.md | the chart, mapped onto the old compose stack |
Each deployment under deployments carries its own conda environments and Dockerfiles — so instances can experiment freely as their needs differ — plus chart values, for the instances this repo deploys.
| path | what it is |
|---|---|
deployments/ |
per-instance conda environments, Dockerfiles, and chart values |
charts/alchemiscale/ |
the Helm chart every instance is deployed from |
infra/opentofu/ |
the EKS clusters and their supporting AWS resources |
scripts/ |
operator tooling: identities, neo4j dump/restore |
.github/workflows/ |
image builds, PR environments, release deployment |
Click a deployment name for its details and client setup instructions.
| Deployment | Deployed Tag | API URI | Description |
|---|---|---|---|
omsf |
— | https://api.omsf.alchemiscale.org | production use, with a combination of HPC, Kubernetes, and Folding@Home compute provisioned |
asap |
2026.04.09-0 | https://api.asap.alchemiscale.org | ASAP Discovery production use, with HPC compute provisioned. Runs on its own host; not deployed by this repo's automation |
openadmet |
— | https://api.openadmet.alchemiscale.org | OpenADMET production use. Runs on its own host; migration to the cluster is not scheduled yet |
root |
2026.02.06-0 | https://api.alchemiscale.org | deprecated — succeeded by omsf; frozen, retired at the end of the parallel-run window |
root (https://api.alchemiscale.org) is succeeded by omsf (https://api.omsf.alchemiscale.org). The two run in parallel for three months from the omsf availability announcement, then root is decommissioned.
omsfstarts with a fresh database and a new S3 prefix. Finish or retrieve work in progress onroot; start new work onomsf.- Identities are not copied over — ask an administrator to register yours on
omsf. - Compute services attached to
root(HPC, Folding@Home) need repointing during the window. rootstays frozen on its EC2 host in the legacy AWS account. Its configuration remains indeployments/rootfor its operators, but has no chart values, so CD cannot deploy it. The directory goes when the host does.
Nothing is deployed by hand. A release with a CalVer tag (YYYY.MM.DD-N) builds images for whichever instances changed and deploys them one at a time, each behind a GitHub Environment with required reviewers.
To exercise a change first, label the pull request test-deploy: it is deployed into a throwaway namespace on the test cluster, smoke tested, and torn down when the label is removed or the PR closes.