Documentation/Start here
Start here · v0.0.1

Installation overview

Select a supported topology, prepare dependencies, and use the release-owned preflight and verification surfaces.

AudiencePlatform administrators, SRE, infrastructureReading time14 minCommands in context5 guided stepsReviewed sourcefcc5871 · 30 July 2026
RC

Exact release scope. This page documents the reviewed v0.0.1 source at fcc5871. Check release status before enabling a gated capability.

Choose a deployment profile

ProfileUse whenCritical requirements
profile-single-nodeEvaluation or a controlled small installation where one host is acceptable.One API/worker placement; durable PostgreSQL and artifacts still required.
profile-splitAPI, worker, database, artifact storage, and agent need separate placement.Documented network paths, time sync, TLS trust, and explicit secret delivery.
profile-ha-readyThe control plane must tolerate instance loss.External HA PostgreSQL, shared filesystem or S3-compatible artifacts, leader election, multiple replicas.

Required dependencies

  • Signed release bundle with checksums, compatibility manifest, release metadata, and digest-pinned API and agent images.
  • Durable PostgreSQL reachable from the API and workers. The Helm chart intentionally does not install PostgreSQL.
  • Artifact storage using a durable filesystem or S3-compatible service. More than one replica requires shared storage.
  • Secrets protection key containing exactly 32 random bytes, base64 encoded, with a managed non-development key ID.
  • Trusted time source for every API and agent host. Default agent rejection begins when reported skew exceeds 30 seconds.
  • TLS and name resolution for the user-facing endpoint and every API-to-agent, provider, registry, database, and object-store path.

Baseline capacity and storage decisions

The Helm chart starts with conservative container requests of 100m CPU and 256 MiB memory, and limits of 1 CPU and 1 GiB. These are scheduling defaults—not a production sizing guarantee. Size from your expected concurrent OperationRuns, evidence volume, retention, provider latency, and add-on/conformance workload.

Data classPlacementBackup requirement
Control-plane stateExternal PostgreSQLConsistent database backup and restore test
Run logs and artifactsDurable filesystem or S3-compatible storageBack up or replicate with checksums and retention metadata
Secrets master keyExternal secret manager or protected Kubernetes SecretSeparate encrypted escrow; never store only beside the database dump
Release evidenceImmutable artifact store or release bundleRetain exact checksums, signatures, SBOM, and provenance used by the environment

Required installation flow

  1. 01

    Verify the release

    Validate checksums, signed compatibility metadata, image signatures, and the exact digests you will deploy.

  2. 02

    Prepare configuration and secret references

    Keep credentials outside values files, command history, logs, and support bundles. Use Kubernetes Secret references or the supported external delivery boundary.

  3. 03

    Run preflight-only

    Do not begin with installation. Resolve every hard failure in storage, database, trust, clock, ports, or release integrity first.

  4. 04

    Run dry-run or render review

    Review the final Compose/Helm configuration and confirm replica, storage, ingress, and feature-gate values.

  5. 05

    Install and verify

    Require both /health and /ready, inspect build information, and create the bootstrap administrator only through the supported onboarding path.

  6. 06

    Capture a baseline backup

    Verify a restorable backup before registering production credentials or starting lifecycle work.

Helm or Compose?

DecisionHelmCompose
Best fitExisting Kubernetes platform and enterprise scheduling controlsSingle Linux host or first controlled evaluation
ScaleSupports multi-replica API/worker topology with external dependenciesSingle-node profile
SecretsKubernetes Secret referencesProtected environment file and/or file references
IngressOptional chart-managed IngressOperator-managed reverse proxy or local bind
UpgradeChart values, signed target release, staged rolloutRelease-owned installer entrypoint and Compose manifest