Kubernetes (Helm)
The repository includes a starter Helm chart at
infrastructure/helm/pragence. It has not been installed against a live
cluster, so validate it in a test environment before production use. The chart
bundles single-replica Postgres and Redis by default, plus the API, optional
console, migration startup, and Prefect-backed job lane. It does not include
the managed SaaS control plane.
Install from the repository
Build or obtain compatible application and database images, then install the local chart. Supply real secret values through your approved secret-management process; the chart defaults are placeholders.
helm lint infrastructure/helm/pragence
helm template pragence infrastructure/helm/pragence \
--set images.app.repository=<registry>/pragence-app \
--set images.app.tag=<version> \
--set images.postgres.repository=<registry>/pragence-db \
--set images.postgres.tag=<version>Create a deployment-specific values file for provider keys, database
credentials, JWT and encryption secrets, storage, and ingress. The chart's
values and starter chart README
document the available settings. Use an external PostgreSQL service only if it
has both required extensions; pg_textsearch must be available through
shared_preload_libraries.
Scaling and persistence
The bundled Postgres, Redis, and Prefect services are single-replica starter
installations. For high availability, configure external managed services and
review database, cache, and job-lane failure behavior. The chart's
app.replicaCount can be changed, but the migration init container runs in each
app pod; follow the chart README's upgrade guidance before scaling replicas.
Uploads are durable across app and worker pods only when shared upload persistence is enabled and the storage class supports the required access mode. Connector syncs and text ingestion use the job lane; the chart can fall back to in-process execution when Prefect is unavailable.
The repository does not ship the previously described remote Helm repository, Prometheus/Grafana dashboards, HPA defaults, or fixed CPU and memory sizing. Set monitoring, ingress, resource requests, and production availability controls in your own cluster configuration.