À propos
Cloudflare D1's official engineering article explains the database infrastructure behind AI-generated apps and vibe coding platforms. It covers scaling a control plane for many small databases per user, application, preview or tenant, keeping routing and usage accounting off the query hot path. For Generative UI builders, it is a backend architecture reference for persistent generated applications.
Résumé
Operating Many Small Application Databases:
1. Provisioning: Separate database lifecycle operations from query execution
2. Routing: Cache read-heavy metadata with Workers KV and use Hyperdrive for central Postgres access
3. Usage accounting: Queue and aggregate usage updates rather than scan every database during a request
4. Scale evidence: Cloudflare reports 15× growth in live databases and 18× in daily active databases over one year, measured on October 8, 2026
Implementation context: Check per-database capacity limits before adopting this pattern. Tenant authorization and application access control still belong in the platform design.
1. Provisioning: Separate database lifecycle operations from query execution
2. Routing: Cache read-heavy metadata with Workers KV and use Hyperdrive for central Postgres access
3. Usage accounting: Queue and aggregate usage updates rather than scan every database during a request
4. Scale evidence: Cloudflare reports 15× growth in live databases and 18× in daily active databases over one year, measured on October 8, 2026
Implementation context: Check per-database capacity limits before adopting this pattern. Tenant authorization and application access control still belong in the platform design.
Tags
cloudflare-d1ai-generated-appsvibe-codingdatabase-per-tenantapplication-provisioningcontrol-plane