Self-host Glossia

Install Glossia on your own Kubernetes cluster with the bundled Helm chart, so your team runs the language OS on its own infrastructure.

Glossia is open source under the O'Saasy License. You can self-host it, modify it, and run it for your organization's internal use. The one thing the license does not permit is offering it to third parties as a hosted or SaaS product that competes with the hosted service at glossia.ai.

This guide gets you from an empty Kubernetes cluster to a running Glossia instance.

Before you start

You'll need:

A Kubernetes cluster you can install Helm charts on (v1.28 or newer)
helm and kubectl locally
A domain you can point at the cluster ingress
An OpenID Connect provider or SMTP relay for authentication (Glossia supports both)

The Helm chart bundles Postgres (via CloudNativePG) and ClickHouse (via the official operator) so you don't need external databases. If you'd rather bring your own, both can be disabled in values.yaml.

Install the operators

Install whichever operators match the components you plan to enable. At minimum:

CloudNativePG operator for the app database
ClickHouse Kubernetes operator for the analytics database
An ingress controller (for example ingress-nginx)
cert-manager if you want automatic TLS

Install the Glossia chart

Clone the repository, then install the chart:

git clone https://github.com/glossia/glossia.git
cd glossia

helm install glossia ./deploy/helm/glossia \
  --namespace glossia --create-namespace \
  --set image.tag=main \
  --set ingress.enabled=true \
  --set ingress.hosts[0].host=glossia.example.com

Provide the app secrets

Create a Kubernetes Secret named glossia-app-env with at least these keys:

Key	Purpose
`GLOSSIA_SECRET_KEY_BASE`	Phoenix session signing key
`GLOSSIA_METRICS_BEARER_TOKEN`	Bearer token guarding `/metrics`
`GLOSSIA_OPS_AUTH_PASSWORD`	Basic-auth for `/ops` dashboards
`RELEASE_COOKIE`	Erlang distribution cookie shared by every pod
`GLOSSIA_SMTP_*`	Outbound email settings

You can provision these directly, use Sealed Secrets, or wire the chart to your secret manager via the External Secrets Operator integration described in the chart README.

What lives in the chart

The Glossia web application
Postgres for application data (optional, on by default)
ClickHouse for analytics (optional, on by default)
Background workers for translation jobs, which run as Kubernetes Jobs so they outlive rolling deploys

Where to go next

The Helm chart README has the full reference for every value, plus notes on backups, object storage, and observability.
Configure a model provider once the instance is running so translations can call an LLM.
Report issues or suggest improvements at the GitHub repository.