rustybuns deploy: adopt existing Cloudflare resources by the names in the config #140

rustybuns plan/deploy doesn't pass the config's names to Alchemy. The Worker gets no name (the config's name), and D1 gets no databaseName, nor R2 a bucketName (packages/cli/src/gen/alchemy.ts). So Alchemy makes <app>-<id>-<stage>-<random> names, and a stack whose local state is gone plans to create everything again.

What happened: codesplitters is live as Worker codesplitters with D1 codesplitters-db, made with wrangler. rustybuns plan wanted to create a new Worker and a new, empty database beside them, so we deployed with wrangler deploy instead. The same thing already left four orphaned druids-curse-...-DB-live-ryanquinn-* databases (see the cleanup issue).

Fix:

• Emit the explicit names: Worker name: cfg.name, D1 name: databaseName, R2 name: bucketName. Alchemy's D1 already adopts a database by name when creating it hits "already exists".
• Pass --adopt when nothing's in state.
• Careful with live Alchemy-managed apps (hippo, druids): a name change replaces resources there, so do it per app or behind a flag.
• Alchemy keeps state in .alchemy/ inside the worktree (Alchemy.localState()), so state is lost when a worktree is deleted. Consider a state store that outlives the worktree.

🤖 Generated with Claude Code