Skip to main content

Use Cases

The two use cases discussed in the framework's introduction share every mechanism in this reference architecture but tune it differently.

Neon for Agents​

An AI-powered app builder gives each generated app its own project.

  • Lifecycle: very high create/delete churn and short-lived instances. Make provisioning and teardown fully idempotent.
  • Compute: short suspend_timeout_seconds so idle projects cost nothing; scale-to-zero is doing most of the economic work.
  • Quotas: set tight per-project quotas as a guardrail against a runaway agent; expect and handle quota-driven suspension gracefully.
  • Branching: use branches for agent checkpoints, per-attempt sandboxes, and instant rollback.
  • Connectivity: the serverless driver (HTTP/WebSocket) suits agent runtimes and edge functions; pool connections when many agents run concurrently.

Neon as Your Platform Database​

A platform offers Lakebase as part of its own product, using Neon to provision and manage a dedicated Lakebase for each customer, application, or workload.

  • Lifecycle: tenants persist; provisioning is steady rather than bursty, and deletion follows customer offboarding with appropriate retention.
  • Compute: size min_cu/max_cu for steady traffic with headroom for peaks; use a longer suspend timeout for consistently active tenants to avoid cold starts.
  • Quotas: map quota profiles to plan tiers (e.g. trial vs. paid) and PATCH them on upgrade/downgrade.
  • Migrations: design a fleet-wide schema-migration strategy so every tenant project stays on a consistent schema version.
  • Connectivity: standard drivers/ORMs through the pooled endpoint for web traffic; direct connections for migrations.
  • Telemetry: per-project consumption underpins usage-based billing and margin analysis per customer.

What's Next​