Skip to content

[Wave 5d] @netscript/fresh — umbrella: target architecture + 6 sub-gates (5d1…5d6)#40

Merged
rickylabs merged 285 commits into
feat/package-quality-wave5-appsfrom
feat/package-quality-wave5-apps-5d-fresh
Jun 14, 2026
Merged

[Wave 5d] @netscript/fresh — umbrella: target architecture + 6 sub-gates (5d1…5d6)#40
rickylabs merged 285 commits into
feat/package-quality-wave5-appsfrom
feat/package-quality-wave5-apps-5d-fresh

Conversation

@rickylabs

@rickylabs rickylabs commented Jun 12, 2026

Copy link
Copy Markdown
Owner

What this is

Umbrella for Wave 5d — @netscript/fresh, the most important unit of Wave 5: it ties pretty much everything NetScript is together in one DSL and API set usable server-side and client-side, borrowing best-in-class features from the market (TanStack Start, Next.js App Router, Remix data APIs) and targeting the absolute best DX possible — fully E2E typesafe and E2E telemetry-instrumented.

This branch forked from the 5c head (c64cb16). PR #30 (5c) is now merged (da14cff) and this branch is synced with its base (fee1e23) — the diff here is pure 5d content.

Supervisor artifacts (.llm/tmp/run/feat-package-quality-wave5-apps--5d-fresh/)

  • plan.md — BINDING target architecture. One coherent final-structure goal all six sub-plans derive from: mission + DX benchmark; archetype (A3 + SCOPE-frontend, per-cluster nuances); final public surface (the 12 existing subpaths + ./testing as the only sanctioned growth, F-16 locked); final folder shape (builders decomposition, components/ and hooks/ dissolutions, unified telemetry convention, docs/ + tests/ scaffolds); the whole-package quality bar (combined doc-lint 276→0, 13 over-cap files→0, dry-run PASS incl. slow types, doctested README, abort/cleanup tests for every streaming surface, RFC 14 seams protected not implemented); accepted drift tolerance (±2 slices per sub-plan; internal folder moves free; surface/dep changes need umbrella drift + supervisor review; 5d2 may split in two plans); sequencing + evaluation protocol.
  • handover-5d1-plan.mdhandover-5d6-plan.md — full PLAN-phase prompts (start use harness; skills; deep-dive directives; MEASURE-FIRST; concept of done; per-slice commit/push/PR routine). These docs are merged into every sub-branch so cloud checkouts can read them.
  • README.md (sub-gate map), worklog.md, commits.md, drift.md.

Sub-gate structure (branch · worktree · run dir · Draft PR — all created & seeded)

Gate Scope PR
5d1 support spine: error · utils · ./vite · interactive · curated root · doctrine spine (docs/, ./testing, single config). Sets conventions binding on all later units (error taxonomy, telemetry vocabulary). #34
5d2 ./builders — definePage DSL; heaviest cluster (41.5K/38.6K/22.6K/20.7K/18.6K files); may split into two plans #35
5d3 ./route — manifest + contract runtime; the typesafety spine (contract → handler → sdk client → island props) #36
5d4 ./defer + ./streams + server streaming internals — RFC 13 PSR + RFC 16 e2e streams over Wave 4 plugin-streams; abort/cleanup is the spine #37
5d5 ./form — RFC 15, consuming the 5c fresh-ui seams (form-field, control-props); progressive enhancement; Standard Schema adapter evaluation #38
5d6 ./query (RFC 17 bridge over 5b sdk) + defineFreshApp + final surface pass (F-16/F-18) + RFC 14 seam audit + wave closeout (lift fresh into root gates, package-wide dry-run) #39

All six branches fork from c64cb16. PLAN phases run in parallel — dispatched 2026-06-12 to OpenHands agents (openrouter/moonshotai/kimi-k2.7-code, output=pr-comment) via trigger comments on PRs #34#39; implementation is chained 5d1→5d6, each unit merging prior landings first (recorded as drift D-5d-0 vs split-strategy's literal fork-off-prior-merge).

Evaluation protocol

Per sub-gate: PLAN (OpenHands generator) → plan review by Fable 5 on the sub-PR (validating the handover directives were respected + probing areas the planner missed) → PLAN-EVAL (separate session) → implementation → IMPL-EVAL (separate session) → --no-ff merge into this branch. This umbrella merges into the Wave 5 umbrella (#17) once, at 5d completeness.

Next steps

  1. ✅ PR Wave 5c — fresh-ui end product: foundation (Run 1) + design system (Run 2) DELIVERED · scaffold revamp deferred to post-5d · READY TO MERGE #30 (5c) merged; this branch synced with base — diff is 5d-only.
  2. ✅ Six PLAN runs dispatched (OpenHands × Kimi K2.7 trigger comments on [5d1] fresh support spine — error · utils · vite config · interactive · mod skeleton (PLAN pending) #34[5d6] fresh query + server + final surface — RFC 17 bridge, defineFreshApp, RFC 14 seam audit (PLAN pending) #39).
  3. ▶️ As each plan lands with READY FOR PLAN-EVAL: Fable 5 plan review on the sub-PR, then PLAN-EVAL.
  4. After 5d completes: CLI scaffold revamp (public-facing TanStack-starter-level showcase app) + fresh-ui doctrine cleanup (C-1…C-12) compile everything together — see PR Wave 5c — fresh-ui end product: foundation (Run 1) + design system (Run 2) DELIVERED · scaffold revamp deferred to post-5d · READY TO MERGE #30 status note.

🤖 Generated with Claude Code

…rst findings

- Verify 113 combined deno doc --lint baseline per entrypoint.
- Confirm private-type-ref symbols and per-file breakdown.
- Resolve 3-vs-27 plugin-streams coupling divergence.
- Fill abort/cleanup/backpressure audit table.
- Add telemetry baseline and market comparison.

Co-Authored-By: openhands <openhands@all-hands.dev>
- MEASURE-FIRST table: deno doc --lint (39 errors), deno check no-op,
  14 private-type-refs, over-cap inventory, publish dry-run 62 problems.
- Inventory of error/, utils/, config/vite.ts, interactive/, mod.ts.
- Telemetry fork raw material from defer/telemetry.ts + form/telemetry.ts.
- Drift entries D-5d1-003 through D-5d1-007.

Co-Authored-By: openhands <openhands@all-hands.dev>
Add plan.md, design.md, context-pack.md for 5d4 streaming wave.
Update drift.md with D-5d4-6 and D-5d4-7.

Scope: public-surface type fixes, abort/cleanup hardening,
telemetry port polish, and tests for server/stream, server/sse,
and streams/ client lifecycle. PLAN phase only; no implementation.

Co-authored-by: openhands <openhands@all-hands.dev>
Independent evaluator session (OpenHands 27461651497).
Three blocking findings: incomplete Arch 3 gate set selection,
doc-lint/over-cap budgets not retired by slices, jsr-audit rubric
not applied pre-slice.

Co-authored-by: openhands <openhands@all-hands.dev>
Independent evaluation identified 4 critical blockers:
- No commit slices enumerated
- 'One plan or two plans' decision unresolved
- Fitness gate matrix not selected from archetype-gate-matrix.md
- Required tail section (Review map/Assumptions/Questions/Dependencies/Side-effect ledger) missing

Plus 7 significant gaps including incomplete design document (6/7 sections TODO),
non-operationalized budgets, and missing form-package leak strategy.

Second PLAN-EVAL cycle permitted per protocol.

Co-authored-by: openhands <openhands@all-hands.dev>
- Lock L-5d4-7: local fake-timer/clock test helper in packages/fresh;
  promote to shared ./testing only if 5d5/5d6 needs it.
- Reword Open-Decision Sweep clock-port item to RESOLVED.
- Close Question for supervisor #1 (answered).
- Mark drift D-5d4-7 RESOLVED.

Plan-only revision; zero implementation.

Co-authored-by: openhands <openhands@all-hands.dev>
@rickylabs
rickylabs merged commit 2c4efbe into feat/package-quality-wave5-apps Jun 14, 2026
@rickylabs
rickylabs deleted the feat/package-quality-wave5-apps-5d-fresh branch June 22, 2026 21:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants