Auto-Provisioning

in-progress

Detect bindings your code needs (KV, R2, Hyperdrive, Queues) from build output, prompt once, then create them in your Cloudflare account. Code → infrastructure, no wrangler.toml hand-editing. Killer feature in development.

Source reference: shared/nextcore/ (DeriveBindings stub), CLOUDFLARE_PARITY.md, NEXTCOMPILE_ROADMAP.md

The pitch

TODO most CF projects spend the first hour wiring wrangler.toml — auto-provisioning kills that step.

Detection sources

TODO env var patterns, fetch URLs, import statements — what NextCompile scans.

Prompt flow

TODO CLI shows: "I see KV reads of NAMESPACE_X — create it in account Y? [Y/n]" with binding name suggestion.

Idempotent reconcile

TODO re-runs only create what's missing, never deletes without explicit destroy.

What can be auto-provisioned (target)

TODO KV namespace, R2 bucket, Queues, Vectorize index, AI Gateway, Hyperdrive (with secret URL).

What cannot be auto-provisioned

TODO anything needing external creds (e.g. Postgres URL for Hyperdrive must be supplied).

Status today

TODO detection stub exists; full prompt flow + apply on roadmap (Phase 1 of NextCompile, ~2 weeks).

Comparison: AWS path

TODO AWS already provisions Lambda + S3 + CloudFront + ACM today — auto-provisioning brings parity to CF.