Where and how your declarative runtime runs.

A typical Pwas project consists of a few containerised components: the interpreted backend, the client that generates interfaces, the orchestrator when automation is needed, one or more microservices when dedicated processing is needed, and a standard relational database.

There's no cluster of products to install and no ecosystem of proprietary dependencies. The minimum requirement is an environment that can run containers: a virtual machine with Docker is enough to start, a container orchestrator is the natural choice when you need high availability and horizontal scaling.

That's also why installation in your own server room isn't a reduced version of the product: it's exactly the same software, running elsewhere.

Where you decide, and the decision is a deployment configuration rather than a product choice. The three modes that cover almost every case are: European cloud managed by us or by you, on-premise in your own infrastructure, and hybrid — typically with the outward-facing part on cloud and the application core on site.

The choice stays reversible: because the project is containerised and the database is standard relational, moving it means moving containers and data, not redesigning the application. It's worth verifying explicitly during evaluation, because it's a real difference from platforms whose operation depends on one provider's managed services.

Where a project involves language models, the model's location is a separate decision that must be taken explicitly, including options that keep processing within a perimeter you control.

It depends how you want to set it up, and all three configurations are workable.

• Managed by us, on European cloud: we handle it, under agreed service levels. The choice for those without, or unwilling to commit, an infrastructure team.
• Managed by you, with our support: your team operates the infrastructure, we work on the application. Frequent in organisations that already have a capability and perimeter constraints.
• Managed by a partner: your usual infrastructure provider operates the stack. Because there's no exotic technology, a competent infrastructure team works with what it already knows.

In all three cases, application monitoring, log collection and backups integrate with the tools you already use instead of requiring a dedicated console.