Skip to content

notes.md

Latest commit

 

History

History
78 lines (71 loc) · 5.84 KB

notes.md

File metadata and controls

78 lines (71 loc) · 5.84 KB
Raw

Notes - issue #303 doc-lint remainder

Stops / Deferrals

  • packages/auth-better-auth/src/better-auth.ts exposes Better Auth passthrough options whose public fields reference upstream private declaration aliases: NetscriptBetterAuthOptions.plugins -> BetterAuthOptions['plugins'] and betterAuthOptions -> Omit<BetterAuthOptions, 'database' | 'plugins'>. A trivial re-export or BetterAuthPlugin[] replacement did not satisfy deno doc --lint; resolving this cleanly needs a public API redesign for the passthrough surface, so it is deferred per the issue #303 brief.
  • packages/fresh/src/application/vite/vite.ts exposes NetScriptVitePlugin = Plugin from Vite. deno doc --lint reports Vite's Plugin declaration as private. Replacing this with a package-owned structural hook interface would change the public plugin type contract, so it is deferred per the issue #303 brief.
  • packages/contracts still reports 12 private-type-ref diagnostics after local alias/JSDoc cleanup. The remaining refs are tied to the oRPC-bound base contract/CRUD route seam (oc, AnySchema, ContractProcedureBuilderWith*, Schema, and aliases over those types). This is the sanctioned 86eca907 carve-out class; changing it would require redesigning the sound oRPC builder surface.
  • packages/plugin-ai-core is reduced from 23 diagnostics to 2 after JSDoc and public vocabulary re-export cleanup. The remaining refs are the exported aiContract value naming the internal precise contract definition shape and aiContractV1 naming oRPC's Implementer type. Removing those annotations breaks --isolatedDeclarations; replacing them with a package-owned structural type would change the public contract API, so this needs a design decision rather than a local doc-lint edit.
  • packages/plugin-auth-core is reduced from 3 combined diagnostics to 2 after replacing the PluginCapabilities alias with a package-owned AuthCapabilities interface. The remaining refs are the same precise oRPC contract-definition/implementer seam as the other feature-plugin contract packages.
  • packages/plugin-sagas-core is reduced from 11 combined diagnostics to 2 after making store client/delegate/KV aliases public on the store subpath and replacing the PluginCapabilities alias with a package-owned SagasCapabilities interface. The remaining refs are the same precise oRPC contract-definition/implementer seam as the other feature-plugin contract packages.
  • packages/plugin-workers-core is reduced from 16 combined diagnostics to 4 after making KV, registry, runtime enum-schema, and idempotency support aliases public on the relevant subpaths and replacing the PluginCapabilities alias with a package-owned WorkersCapabilities interface. The remaining refs are the same precise oRPC contract-definition/implementer seam.
  • packages/plugin-triggers-core is reduced from 8 combined diagnostics to 2 after making KV and trigger-domain support aliases public on the store subpath and replacing the PluginCapabilities alias with a package-owned TriggersCapabilities interface. The remaining refs are the same precise oRPC contract-definition/implementer seam; exporting that internal shape widens the public surface to oRPC helper types and increases doc-lint noise, so it is deferred for design review.
  • packages/plugin root/protocol wrapper leaks were cleaned up, leaving 13 combined diagnostics in the contract-base and service subpaths. Those refs name the shared oRPC base contract and service builder/binder generics; replacing them would change the plugin service public API shape, so they are deferred as design-seam residue.
  • First-party plugin wrapper entrypoints now re-export CLI/scaffold/E2E support aliases, making plugins/streams doc-lint clean and reducing wrapper noise in plugins/ai, plugins/auth, plugins/sagas, and plugins/workers. Remaining plugin-root refs come from their core contract/runtime/public seams; plugins/triggers remains runtime/core-seam residue.
  • packages/prisma-adapter-mysql reports 6 private-type refs in src/adapter.ts. This is DB layer surface area and the issue brief explicitly routes DB-layer work out of scope (ROUTE-TO-PRISMA), so it is recorded rather than redesigned in this slice.
  • packages/fresh-ui remains a design-heavy interactive surface cluster. A local conversion from exported Preact JSX.* prop aliases to the package-owned FreshUiElementProps vocabulary reduced doc-lint counts but broke hook/component type-checking across runtime call sites. The package needs a planned public-props/internal-props split rather than an opportunistic doc-lint rewrite.

Process Notes

  • Draft PR #483 is open. Requested labels area:packages, priority:high, and epic:road-to-stable were not present in the repository label set; existing equivalents/nearest labels were applied and the mismatch is recorded in run drift.
  • PLAN-EVAL passed via OpenHands run 28758467765; evaluator noted the strict count is 34 publishable roots plus non-publishable @netscript/bench, and the sanctioned slow-types policy covers four packages.
  • Type-soundness residue grep was narrowed to plugin service/router seams plus Fresh/Aspire touchpoints. No unexpected any handler erasure was found in router contexts; remaining hits are accepted contract error-map casts, KV/schema/runtime adapter casts, tests, or Fresh/Fresh UI framework touchpoints already covered by the deferrals above.

Lock Hygiene

  • No committed deno.lock changes.

  • 2026-07-06: deno info npm:better-auth --json added a transient npm:better-auth@* resolution to deno.lock; reverted immediately per the zero-lock-churn rule.

  • 2026-07-06: deno info npm:@orpc/server@1.14.6 / npm:@orpc/contract@1.14.6 touched deno.lock during read-only declaration inspection; reverted immediately per zero-lock-churn rule.