Money: guided first-budget onboarding (assign money step by step) #1141

Open
opened 2026-10-05 12:38:30 +00:00 by kayg · 17 comments
Owner

Owner request (2026-10-05)

"We need to make it incredibly simpler, explaining how our budgeting works step by step and having users click on input fields and assign the money."

A guided first-budget onboarding for Money: teach the budgeting method (give every unit of money a job) step by step with the User acting on real inputs: enter what you have now, then click each category's Assigned field and assign money, watching "Ready to assign" reach zero. Interactive coach marks over the real budget screen (not a slideshow, not sample data), skippable, resumable, reduced-motion safe, plain language (ASD-STE100). Starts after "Create a new budget" in the import wizard (#wizard) and from Money → Help. Needs a grill round before building (steps, copy, which concepts first).

## Owner request (2026-10-05) "We need to make it incredibly simpler, explaining how our budgeting works step by step and having users click on input fields and assign the money." A guided first-budget onboarding for Money: teach the budgeting method (give every unit of money a job) step by step with the User acting on real inputs: enter what you have now, then click each category's Assigned field and assign money, watching "Ready to assign" reach zero. Interactive coach marks over the real budget screen (not a slideshow, not sample data), skippable, resumable, reduced-motion safe, plain language (ASD-STE100). Starts after "Create a new budget" in the import wizard (#wizard) and from Money → Help. Needs a grill round before building (steps, copy, which concepts first).
Author
Owner

The import wizard referred to above is #1140.

The import wizard referred to above is #1140.
Author
Owner

Owner decisions (grill, 2026-10-06) — guided first-budget onboarding. Depends on the Money format + amount input issue (filed alongside).

Format: one short intro sheet (the single idea: give every unit of money a job; wording names the User's chosen currency, never a fixed one), then a very guided walkthrough with coach marks on the REAL budget screen where the User does each action. No slideshow, no sample data.

Steps (teach beats 1–3):

  1. Currency and format: pick currency (searchable, locale default) and confirm the money format from live samples.
  2. Money you have now: add accounts with current balances — one prefilled "Bank account" row (editable name, balance in the shared amount input) + "Add another account". Creates "Ready to assign".
  3. Categories, question-led: about 5 plain yes/no questions ("Do you pay rent or a mortgage?", "Do you own a car?", "Any kids?", "Any pets?", "Any subscriptions?") add matching categories to the base set: Bills (Rent, Utilities, Phone), Everyday (Groceries, Transport, Eating out), Goals (Emergency fund, Savings), Fun (Hobbies, Gifts). Then review the checklist: untick, rename inline, "Add category". Nothing is created until Continue.
  4. Assign: the User types each amount; a big live "Ready to assign" counter (Torph numeric animation) counts down and quietly celebrates at 0; at the end offer "Put the rest in Savings?". No auto-fill.
  5. First spend: "Add something you bought recently" — real amount + category, watch Available drop; "Skip for now" (nudge returns on the first real transaction).

Later beats as one-time just-in-time nudges: first overspend → cover it by moving money; first new month / new income → assign it.

Words: Ready to assign, Assigned, Spent, Available — each with one plain sentence the first time it appears; ASD-STE100.

Control: skippable at every step ("I know how this works"); resumes where left (progress stored server-side per User); replay from Money → Help → "Show me how budgeting works"; never reopens on its own after finish or skip. Starts after "Create a new budget" in the import wizard (#1140) and on an empty Money.

Look and feel: calternal glass, shared spring between steps (reduced motion respected), friendly nudges; glass metric cards follow #1179 (number top, label bottom).

Evidence: e2e like a User from an empty account through all 5 steps (and skip/resume/replay), Money numbers review, screenshots of every step at 390/820/1440 light + dark with 4x alignment crops.

## Owner decisions (grill, 2026-10-06) — guided first-budget onboarding. Depends on the Money format + amount input issue (filed alongside). **Format:** one short intro sheet (the single idea: give every unit of money a job; wording names the User's chosen currency, never a fixed one), then a very guided walkthrough with coach marks on the REAL budget screen where the User does each action. No slideshow, no sample data. **Steps (teach beats 1–3):** 1. **Currency and format:** pick currency (searchable, locale default) and confirm the money format from live samples. 2. **Money you have now:** add accounts with current balances — one prefilled "Bank account" row (editable name, balance in the shared amount input) + "Add another account". Creates "Ready to assign". 3. **Categories, question-led:** about 5 plain yes/no questions ("Do you pay rent or a mortgage?", "Do you own a car?", "Any kids?", "Any pets?", "Any subscriptions?") add matching categories to the base set: Bills (Rent, Utilities, Phone), Everyday (Groceries, Transport, Eating out), Goals (Emergency fund, Savings), Fun (Hobbies, Gifts). Then review the checklist: untick, rename inline, "Add category". Nothing is created until Continue. 4. **Assign:** the User types each amount; a big live "Ready to assign" counter (Torph numeric animation) counts down and quietly celebrates at 0; at the end offer "Put the rest in Savings?". No auto-fill. 5. **First spend:** "Add something you bought recently" — real amount + category, watch Available drop; "Skip for now" (nudge returns on the first real transaction). **Later beats as one-time just-in-time nudges:** first overspend → cover it by moving money; first new month / new income → assign it. **Words:** Ready to assign, Assigned, Spent, Available — each with one plain sentence the first time it appears; ASD-STE100. **Control:** skippable at every step ("I know how this works"); resumes where left (progress stored server-side per User); replay from Money → Help → "Show me how budgeting works"; never reopens on its own after finish or skip. Starts after "Create a new budget" in the import wizard (#1140) and on an empty Money. **Look and feel:** calternal glass, shared spring between steps (reduced motion respected), friendly nudges; glass metric cards follow #1179 (number top, label bottom). **Evidence:** e2e like a User from an empty account through all 5 steps (and skip/resume/replay), Money numbers review, screenshots of every step at 390/820/1440 light + dark with 4x alignment crops.
Author
Owner

Started #1141 on job/onboard-1141, base d0061ec3df127d81c86729d07d899b0bf2b6de91. Read CLAUDE.md, CONTEXT.md, DESIGN §§33, 34, 48 and 59, and the latest owner decisions.

Fetched origin and merged origin/dev: Already up to date.

Dependency finding: job/moneyfmt-1180 currently points at the same base SHA. Its worktree has no changes, and #1180 has no implementation report. The specified finished base, shared amount input, per-Budget money format and Torph counters are not available yet. I will prepare the independent Category question/checklist logic and its tests. Full UI integration must use the finished #1180 branch.

Started #1141 on `job/onboard-1141`, base `d0061ec3df127d81c86729d07d899b0bf2b6de91`. Read CLAUDE.md, CONTEXT.md, DESIGN §§33, 34, 48 and 59, and the latest owner decisions. Fetched origin and merged origin/dev: Already up to date. Dependency finding: `job/moneyfmt-1180` currently points at the same base SHA. Its worktree has no changes, and #1180 has no implementation report. The specified finished base, shared amount input, per-Budget money format and Torph counters are not available yet. I will prepare the independent Category question/checklist logic and its tests. Full UI integration must use the finished #1180 branch.
Author
Owner

Partial implementation — blocked on the specified base

Branch: job/onboard-1141.
Head: daff4e9c48167e9d63e41060197384d724030de0.
Base: d0061ec3df127d81c86729d07d899b0bf2b6de91.
Fetched origin and merged origin/dev once: Already up to date. No push, deploy or merge into dev. Worktree is clean.

Built

Prepared the independent Category question/checklist model for step 3. Five questions select proposed jobs. Checklist drafts support unticks, inline renames and custom rows. Only selected rows enter the write plan. All selected names are checked before returning a plan. Duplicate names remain valid; server block IDs supply identity. Six tests cover answers, edits, independent drafts, invalid names and Unicode name limits. No API writes or visible UI were added.

Files:

  • apps/web/src/lib/money/onboarding-categories.ts
  • apps/web/src/lib/money/onboarding-categories.test.ts

Re-read both files' doc comments before this report. No dependencies added or versions changed. No test expectations changed. Ran cargo clean: Removed 1 file, 356B total. No web build output was created.

Gate output (verbatim)

cd apps/web && bun run check passed:

perf-lint: PASS; 0 violations; 22104 scoped exceptions
svelte-check found 0 errors and 4 warnings in 3 files

cd apps/web && bun run test -- --maxWorkers=2 passed:

 Test Files  267 passed (267)
      Tests  1836 passed (1836)

cd apps/web && bunx vitest run src/lib/money/onboarding-categories.test.ts --maxWorkers=2 passed:

 Test Files  1 passed (1)
      Tests  6 passed (6)

Logs remain under artifacts/. The initial check found seven unchecked array accesses in the new test file; fixed them without changing assertions. Final check passed. Rust gates do not apply: no Rust changed.

Blocker / known gaps

The job requires the finished job/moneyfmt-1180 branch as its base. That branch still points at d0061ec3df127d81c86729d07d899b0bf2b6de91, the same commit as this job's initial dev base. Its worktree is clean and #1180 has no implementation report. The required shared amount input, per-Budget format and Torph counter implementation are unavailable. Asked for the finished SHA. This is not READY FOR MERGE and does not complete #1141.

The intro, five-step walkthrough UI, real screen coach marks, server-side progress, skip/resume/replay, import wizard start, just-in-time nudges, real-account e2e and screenshot matrix remain to build after #1180 is available. No screenshots or performance measurements: no UI or hot path was integrated.

UX gaps closed

No visible UX change yet. The model keeps review edits local, preserves duplicate names and excludes unticked rows from the write plan.

UX gaps left

All requested visible onboarding behavior and input-mode verification remain open.

Decisions

A no housing answer unticks Rent but keeps it visible in review; a yes uses the base Rent row. Other yes answers add one editable Category: Car, Kids, Pets (Everyday), or Subscriptions (Bills). Their labels and groups were not specified by DESIGN. Review can rename or untick them. All base rows except Rent start selected.

For the merge round

Full onboarding acceptance and screenshots cannot run until integration is complete. The follow-up must add a real-account onboarding e2e entry point for all five steps plus skip/resume/replay, and capture every step at 390/820/1440 in light/dark with macOS emulation and 4x alignment crops. Existing Money regression command: cd apps/web && bun e2e/money.mjs; it must retain exact balance, assignment and Available behavior through the guided actions.

## Partial implementation — blocked on the specified base Branch: `job/onboard-1141`. Head: `daff4e9c48167e9d63e41060197384d724030de0`. Base: `d0061ec3df127d81c86729d07d899b0bf2b6de91`. Fetched origin and merged origin/dev once: Already up to date. No push, deploy or merge into dev. Worktree is clean. ### Built Prepared the independent Category question/checklist model for step 3. Five questions select proposed jobs. Checklist drafts support unticks, inline renames and custom rows. Only selected rows enter the write plan. All selected names are checked before returning a plan. Duplicate names remain valid; server block IDs supply identity. Six tests cover answers, edits, independent drafts, invalid names and Unicode name limits. No API writes or visible UI were added. Files: - `apps/web/src/lib/money/onboarding-categories.ts` - `apps/web/src/lib/money/onboarding-categories.test.ts` Re-read both files' doc comments before this report. No dependencies added or versions changed. No test expectations changed. Ran cargo clean: `Removed 1 file, 356B total`. No web build output was created. ### Gate output (verbatim) `cd apps/web && bun run check` passed: ``` perf-lint: PASS; 0 violations; 22104 scoped exceptions svelte-check found 0 errors and 4 warnings in 3 files ``` `cd apps/web && bun run test -- --maxWorkers=2` passed: ``` Test Files 267 passed (267) Tests 1836 passed (1836) ``` `cd apps/web && bunx vitest run src/lib/money/onboarding-categories.test.ts --maxWorkers=2` passed: ``` Test Files 1 passed (1) Tests 6 passed (6) ``` Logs remain under `artifacts/`. The initial check found seven unchecked array accesses in the new test file; fixed them without changing assertions. Final check passed. Rust gates do not apply: no Rust changed. ### Blocker / known gaps The job requires the finished `job/moneyfmt-1180` branch as its base. That branch still points at `d0061ec3df127d81c86729d07d899b0bf2b6de91`, the same commit as this job's initial dev base. Its worktree is clean and #1180 has no implementation report. The required shared amount input, per-Budget format and Torph counter implementation are unavailable. Asked for the finished SHA. This is not READY FOR MERGE and does not complete #1141. The intro, five-step walkthrough UI, real screen coach marks, server-side progress, skip/resume/replay, import wizard start, just-in-time nudges, real-account e2e and screenshot matrix remain to build after #1180 is available. No screenshots or performance measurements: no UI or hot path was integrated. ### UX gaps closed No visible UX change yet. The model keeps review edits local, preserves duplicate names and excludes unticked rows from the write plan. ### UX gaps left All requested visible onboarding behavior and input-mode verification remain open. ### Decisions A no housing answer unticks Rent but keeps it visible in review; a yes uses the base Rent row. Other yes answers add one editable Category: Car, Kids, Pets (Everyday), or Subscriptions (Bills). Their labels and groups were not specified by DESIGN. Review can rename or untick them. All base rows except Rent start selected. ### For the merge round Full onboarding acceptance and screenshots cannot run until integration is complete. The follow-up must add a real-account onboarding e2e entry point for all five steps plus skip/resume/replay, and capture every step at 390/820/1440 in light/dark with macOS emulation and 4x alignment crops. Existing Money regression command: `cd apps/web && bun e2e/money.mjs`; it must retain exact balance, assignment and Available behavior through the guided actions.
Author
Owner

Resumed #1141 on job/onboard-1141, base 7cdc7ad034f4ae00e453e77dcf01f0bc0620c7a5 from #1180. Read CLAUDE.md, CONTEXT.md, DESIGN and owner decisions. Fetched and merged origin/dev: Already up to date.

Finding: #1180 explicitly reports only formatter/parser foundations; persisted Budget format, shared amount input and counters remain absent. I will build the independent onboarding behavior on existing Money actions and preserve that dependency gap in the report. Restoring the prior #1141 category model for reuse. No new dependencies planned.

Resumed #1141 on `job/onboard-1141`, base `7cdc7ad034f4ae00e453e77dcf01f0bc0620c7a5` from #1180. Read CLAUDE.md, CONTEXT.md, DESIGN and owner decisions. Fetched and merged origin/dev: `Already up to date.` Finding: #1180 explicitly reports only formatter/parser foundations; persisted Budget format, shared amount input and counters remain absent. I will build the independent onboarding behavior on existing Money actions and preserve that dependency gap in the report. Restoring the prior #1141 category model for reuse. No new dependencies planned.
Author
Owner

Server slice committed at e4eb818fc. Progress is stored per User in the existing settings section, with checked revisions. The Category checklist is one Budget.md replacement; stable operation IDs make an uncertain retry idempotent. Tests prove no partial publication for an invalid row, distinct identities for duplicate labels, retry identity reuse, User isolation, restart persistence and stale progress rejection.

Money crate test output: test result: ok. 94 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 66.64s and test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 9.23s. Server and final web gates continue. Fetched and merged origin/dev before final gates: Already up to date.

Decisions: progress lives in User settings, not Budget files. One active guide per User resumes the stored Budget/month. Checklist writes use an operation UUID and server-generated stable block IDs. Format controls still depend on unfinished #1180; import-wizard integration depends on #1140.

Server slice committed at `e4eb818fc`. Progress is stored per User in the existing settings section, with checked revisions. The Category checklist is one Budget.md replacement; stable operation IDs make an uncertain retry idempotent. Tests prove no partial publication for an invalid row, distinct identities for duplicate labels, retry identity reuse, User isolation, restart persistence and stale progress rejection. Money crate test output: `test result: ok. 94 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 66.64s` and `test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 9.23s`. Server and final web gates continue. Fetched and merged origin/dev before final gates: `Already up to date.` Decisions: progress lives in User settings, not Budget files. One active guide per User resumes the stored Budget/month. Checklist writes use an operation UUID and server-generated stable block IDs. Format controls still depend on unfinished #1180; import-wizard integration depends on #1140.
Author
Owner

Head: 4c8baa7a3. Money gates passed after the checklist retry fix.

Evidence: cargo test -p calternal-plugin-money reports 94 unit tests and 27 integration tests passed (4 ignored). The production walkthrough reached the first-spend step with Ready to assign = 0, Groceries Assigned = 2000 and Savings Assigned = 8000 minor units. Its purchase-entry selector timed out; the next run uses the Category combobox role.

The server suite found missing action-policy declarations for the three new operations. These are now declared. OpenAPI, action registry and TypeScript generation succeeded. The production screenshot crop exposed the checklist checkbox above the name field; CSS now places the group label on its own row and gives the checkbox a 44 px target. The assignment coach no longer selects its popover, which took focus away from the real amount input.

Small public addition: ConfirmSheet now forwards the existing optional hideSheetTitle property, default false. The introduction opts in to keep one visible heading on a phone. Other callers keep their current behavior.

Decisions: exact guide links use ?guide=accounts/categories/assign/spend. Unknown values cannot reopen a terminal guide. Preference reads share only a pending request for the current User and lifecycle; no preference response is retained.

Head: 4c8baa7a3. Money gates passed after the checklist retry fix. Evidence: `cargo test -p calternal-plugin-money` reports 94 unit tests and 27 integration tests passed (4 ignored). The production walkthrough reached the first-spend step with Ready to assign = 0, Groceries Assigned = 2000 and Savings Assigned = 8000 minor units. Its purchase-entry selector timed out; the next run uses the Category combobox role. The server suite found missing action-policy declarations for the three new operations. These are now declared. OpenAPI, action registry and TypeScript generation succeeded. The production screenshot crop exposed the checklist checkbox above the name field; CSS now places the group label on its own row and gives the checkbox a 44 px target. The assignment coach no longer selects its popover, which took focus away from the real amount input. Small public addition: ConfirmSheet now forwards the existing optional hideSheetTitle property, default false. The introduction opts in to keep one visible heading on a phone. Other callers keep their current behavior. Decisions: exact guide links use ?guide=accounts/categories/assign/spend. Unknown values cannot reopen a terminal guide. Preference reads share only a pending request for the current User and lifecycle; no preference response is retained.
Author
Owner

PARTIAL: The real Budget walkthrough works. #1141 is not complete. The supplied #1180 branch has format/parser foundations only. It does not have the persisted format picker or shared amount input. The #1140 new-Budget wizard entry is also absent from this base.

Head: f1fd658f42bfd74d01a6148e1e5df109be3f6ea9. Branch: job/onboard-1141. Base: 7cdc7ad034f4ae00e453e77dcf01f0bc0620c7a5 from job/moneyfmt-1180. origin/dev was merged before the final gates; Git reported Already up to date. No push, deploy or branch merge was done.

Built

  • One short introduction, locale-default currency search and actual format samples. The existing shared forms create the real Budget and Accounts.
  • Five Category questions, an editable checklist and Add category. No Category is written until Continue. One bounded write publishes all selected rows. Stable operation IDs make ordinary uncertain retries idempotent.
  • A coach on the real Assigned fields. The live Torph counter counts down. Savings receives the remainder only after the User asks for it. A real purchase changes Available.
  • Server-side progress per User, checked revisions, reload/resume, terminal skip/finish, explicit Help replay and links to individual active beats.
  • One-time overspend and new-month/new-income nudges. The cover action moves actual money and offers Undo. State-changing amount and purchase actions use the existing Undo mechanism with User and report-revision guards.
  • Three bounded API operations, action-policy entries, generated OpenAPI/actions/client types, exact live performance pins, focused route and component tests, a production browser flow, and a prepared hot-path profile.

Money numbers and evidence

The User entered EUR 100 in Bank account. They assigned EUR 20 to Groceries and EUR 80 to Savings. Ready to assign became 0. A EUR 15 purchase left Groceries Available at 500 minor units. Another EUR 10 purchase caused a 500-unit overspend. Cover moved 500 units from Savings: Groceries Available became 0 and Savings Available became 7500. EUR 50 income in the next month showed the one-time assignment nudge.

The production flow checks resume, finish, explicit replay and skip. It also checks rejected invalid progress and bounded checklist bodies. All 108 screenshots use macOS platform emulation. Each of the nine states has 390/820/1440 light + dark views and 4x crops. The checklist crop checks cap-height alignment after the CSS fix. Claude must review visual quality.

Gate output, verbatim

cargo fmt --check produced no output and exited 0.

cargo clippy -p calternal-plugin-money --all-targets -- -D warnings:

    Finished `dev` profile [unoptimized + debuginfo] target(s) in 14m 15s

cargo test -p calternal-plugin-money -- --test-threads=4:

test result: ok. 94 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 97.21s
test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 26.44s

cargo clippy -p calternal-server --all-targets -- -D warnings:

    Finished `dev` profile [unoptimized + debuginfo] target(s) in 20m 42s

cargo test -p calternal-server -- --test-threads=4:

test result: FAILED. 255 passed; 1 failed; 10 ignored; 0 measured; 0 filtered out; finished in 340.70s

The new OpenAPI checks pass. The remaining failure is in the existing large-Home startup child, not in a Money call:

search_reconcile_calendar_write_profile files=20000 writes=32 p50_ms=9218 p95_ms=9709 p95_budget_ms=10000
startup Search reconciliation did not finish

This is a SLOW finding: the background completion limit expired. Its Calendar latency stayed within the existing budget. No test expectation or deadline was changed. The merge round must check this child again.

cd apps/web && bun run check:

perf-lint: PASS; 0 violations; 22145 scoped exceptions
svelte-check found 0 errors and 4 warnings in 3 files

cd apps/web && bun run test --maxWorkers=1:

 Test Files  271 passed (271)
      Tests  1851 passed (1851)

The preceding two-worker suite had one 5000 ms timeout in the unchanged dense Calendar test. The one-worker suite passed with the same assertions and timeout.

Focused browser output:

PASS #1141: real guided Budget, exact balances, resume, finish, overspend, new money, replay and skip; macOS screenshot matrix captured.

Generated registry check:

Action registry: 409 operations, 386 generated tools

UX gaps closed

The assignment popover no longer takes focus from the real amount input. Guide reads do not delay an existing Budget. Parallel empty-state actions are hidden while the guide owns that action. Copy link restores the chosen beat. The checklist check is aligned with the name input and has a 44 px target. The phone intro has one visible title. A failed skip keeps the active beat. Old User responses cannot publish after a User change.

UX gaps left / known gaps

  • #1180 must supply persisted per-Budget format confirmation and the shared Numora amount input. This change uses existing exact amount forms and actual locale format samples. Currency search uses the existing common-currency list.
  • #1140 must start this guide after its Create a new budget action. Empty Money already starts it.
  • Help replay starts at Accounts for an existing Budget. It does not replay currency/format confirmation.
  • Progress stores the active beat, Budget and month. It does not store unsaved Account fields or Category question/checklist drafts.
  • Same-month income is detected while this view is mounted. A first income across route reloads needs an event-backed trigger.
  • Account and Category creation have no Undo. Existing Money Undo has no server mutation receipt; the new guards do not provide a full cross-client Undo guarantee.
  • The category batch does not retain a tombstone after every row from that batch is deleted.
  • The separate skip-first-spend/first-transaction nudge path has no production browser evidence. The main flow spends, and marks that teaching beat as seen.
  • Claude visual review remains pending.

Decisions

One active guide is stored in the existing User settings section money_onboarding, with a revision. No balance is stored there. Custom draft Categories start in Goals. Optional questions default to No; Rent stays visible but unticked when housing is No. One UUID identifies each confirmed checklist; at most 100 rows and 32 KiB are accepted. Active guide links use ?guide=accounts|categories|assign|spend. Unknown values cannot reopen a terminal guide. Preference reads share only an in-flight request for the current User/lifecycle; no completed response is retained. These choices extend the issue's owner decisions where they did not specify storage or draft defaults.

Performance / for the merge round

bench/money-onboarding-1141.mjs covers progress-read p50/p95, CPU/RSS, a 100-Category checklist and four concurrent retries. It was not measured: the latest verification policy reserves performance measurements for performance issues on the locked perf VM. There is no onboarding baseline entry in docs/perf/baseline.json.

After the dependency integrations, run the production flow against the combined server:

CALTERNAL_SERVER_BIN="$CARGO_TARGET_DIR/release/calternal-server" node apps/web/e2e/money-onboarding-1141.mjs

It must prove all five beats with the final format and amount controls, the #1140 entry, exact totals and skip/resume/replay. Add the remaining first-spend skip and cross-route income cases.

Check the failed startup child without changing its limits:

cargo test -p calternal-server --bin calternal-server wire::tests::startup_search_reconciliation_keeps_calendar_writes_responsive_on_large_home -- --ignored --exact --test-threads=1

Run the full Money e2e suite (cd apps/web && bun run test:e2e:money) and the existing tests/adversarial/authz_matrix.py / xuser_matrix.py entry points as part of the combined merge round. They must include the new own-data routes. Full adversarial matrices, release/staging checks and real Mac interop were not run in this feature job.

Files

  • apps/web/e2e/money-onboarding-1141.mjs
  • apps/web/src/lib/components/ConfirmSheet.svelte
  • apps/web/src/lib/components/money/AddAccountForm.svelte
  • apps/web/src/lib/components/money/BudgetGuide.svelte
  • apps/web/src/lib/components/money/BudgetGuide.svelte.test.ts
  • apps/web/src/lib/components/money/CategoryChecklist.svelte
  • apps/web/src/lib/components/money/CategoryChecklist.svelte.test.ts
  • apps/web/src/lib/components/money/CreateBudgetForm.svelte
  • apps/web/src/lib/components/money/TransactionForm.svelte
  • apps/web/src/lib/money/api.ts
  • apps/web/src/lib/money/onboarding-api.test.ts
  • apps/web/src/lib/money/onboarding-categories.test.ts
  • apps/web/src/lib/money/onboarding-categories.ts
  • apps/web/src/lib/money/onboarding.test.ts
  • apps/web/src/lib/money/onboarding.ts
  • apps/web/src/routes/money/+page.svelte
  • apps/web/src/routes/money/[budget]/[month]/+page.svelte
  • bench/money-onboarding-1141.mjs
  • contracts/action-policy.json
  • contracts/actions.json
  • contracts/openapi.json
  • contracts/perf/adoption-1058.json
  • contracts/perf/exceptions.json
  • contracts/perf/ratchet.json
  • contracts/perf/registry.json
  • crates/plugins/money/src/lib.rs
  • crates/plugins/money/src/onboarding.rs
  • crates/plugins/money/src/routes.rs
  • crates/plugins/money/src/tests.rs
  • packages/api-client/src/generated.ts
  • tests/adversarial/money_onboarding.mjs

Cleanup

cargo clean output:

     Removed 18316 files, 13.3GiB total

The generated web build and .svelte-kit/output were deleted. Screenshots, archives and logs stay in ignored artifacts/. The worktree is clean. No screenshot or video was committed. No credential was displayed.

PARTIAL: The real Budget walkthrough works. #1141 is not complete. The supplied #1180 branch has format/parser foundations only. It does not have the persisted format picker or shared amount input. The #1140 new-Budget wizard entry is also absent from this base. Head: `f1fd658f42bfd74d01a6148e1e5df109be3f6ea9`. Branch: `job/onboard-1141`. Base: `7cdc7ad034f4ae00e453e77dcf01f0bc0620c7a5` from `job/moneyfmt-1180`. `origin/dev` was merged before the final gates; Git reported `Already up to date.` No push, deploy or branch merge was done. **Built** - One short introduction, locale-default currency search and actual format samples. The existing shared forms create the real Budget and Accounts. - Five Category questions, an editable checklist and Add category. No Category is written until Continue. One bounded write publishes all selected rows. Stable operation IDs make ordinary uncertain retries idempotent. - A coach on the real Assigned fields. The live Torph counter counts down. Savings receives the remainder only after the User asks for it. A real purchase changes Available. - Server-side progress per User, checked revisions, reload/resume, terminal skip/finish, explicit Help replay and links to individual active beats. - One-time overspend and new-month/new-income nudges. The cover action moves actual money and offers Undo. State-changing amount and purchase actions use the existing Undo mechanism with User and report-revision guards. - Three bounded API operations, action-policy entries, generated OpenAPI/actions/client types, exact live performance pins, focused route and component tests, a production browser flow, and a prepared hot-path profile. **Money numbers and evidence** The User entered EUR 100 in Bank account. They assigned EUR 20 to Groceries and EUR 80 to Savings. Ready to assign became 0. A EUR 15 purchase left Groceries Available at 500 minor units. Another EUR 10 purchase caused a 500-unit overspend. Cover moved 500 units from Savings: Groceries Available became 0 and Savings Available became 7500. EUR 50 income in the next month showed the one-time assignment nudge. The production flow checks resume, finish, explicit replay and skip. It also checks rejected invalid progress and bounded checklist bodies. All 108 screenshots use macOS platform emulation. Each of the nine states has 390/820/1440 light + dark views and 4x crops. The checklist crop checks cap-height alignment after the CSS fix. Claude must review visual quality. - [intro](https://git.kayg.org/attachments/403fc8f6-984b-48eb-87bd-77ae0ff70cc3) - [currency](https://git.kayg.org/attachments/aff95c1a-e397-4361-9265-54b77192e753) - [accounts](https://git.kayg.org/attachments/99f58b40-81ed-49dd-a70e-563d783e1fd8) - [questions](https://git.kayg.org/attachments/885d6f87-2fe3-44e4-8c92-bc029a1ee56c) - [categories](https://git.kayg.org/attachments/6d0d9b84-8ae4-41a6-8d52-db336135d445) - [assign](https://git.kayg.org/attachments/06d0ace6-58bb-4135-8173-f21e5ef54dac) - [spend](https://git.kayg.org/attachments/71efb0a0-8dd6-486e-90ae-9843c2dd0e9a) - [overspend](https://git.kayg.org/attachments/41c27c84-8d0d-4421-aef7-aa973c38f85b) - [new-money](https://git.kayg.org/attachments/9659c159-fdfa-4d53-89ac-ddcab3aa08b9) **Gate output, verbatim** `cargo fmt --check` produced no output and exited 0. `cargo clippy -p calternal-plugin-money --all-targets -- -D warnings`: ``` Finished `dev` profile [unoptimized + debuginfo] target(s) in 14m 15s ``` `cargo test -p calternal-plugin-money -- --test-threads=4`: ``` test result: ok. 94 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 97.21s test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 26.44s ``` `cargo clippy -p calternal-server --all-targets -- -D warnings`: ``` Finished `dev` profile [unoptimized + debuginfo] target(s) in 20m 42s ``` `cargo test -p calternal-server -- --test-threads=4`: ``` test result: FAILED. 255 passed; 1 failed; 10 ignored; 0 measured; 0 filtered out; finished in 340.70s ``` The new OpenAPI checks pass. The remaining failure is in the existing large-Home startup child, not in a Money call: ``` search_reconcile_calendar_write_profile files=20000 writes=32 p50_ms=9218 p95_ms=9709 p95_budget_ms=10000 startup Search reconciliation did not finish ``` This is a SLOW finding: the background completion limit expired. Its Calendar latency stayed within the existing budget. No test expectation or deadline was changed. The merge round must check this child again. `cd apps/web && bun run check`: ``` perf-lint: PASS; 0 violations; 22145 scoped exceptions svelte-check found 0 errors and 4 warnings in 3 files ``` `cd apps/web && bun run test --maxWorkers=1`: ``` Test Files 271 passed (271) Tests 1851 passed (1851) ``` The preceding two-worker suite had one 5000 ms timeout in the unchanged dense Calendar test. The one-worker suite passed with the same assertions and timeout. Focused browser output: ``` PASS #1141: real guided Budget, exact balances, resume, finish, overspend, new money, replay and skip; macOS screenshot matrix captured. ``` Generated registry check: ``` Action registry: 409 operations, 386 generated tools ``` **UX gaps closed** The assignment popover no longer takes focus from the real amount input. Guide reads do not delay an existing Budget. Parallel empty-state actions are hidden while the guide owns that action. Copy link restores the chosen beat. The checklist check is aligned with the name input and has a 44 px target. The phone intro has one visible title. A failed skip keeps the active beat. Old User responses cannot publish after a User change. **UX gaps left / known gaps** - #1180 must supply persisted per-Budget format confirmation and the shared Numora amount input. This change uses existing exact amount forms and actual locale format samples. Currency search uses the existing common-currency list. - #1140 must start this guide after its Create a new budget action. Empty Money already starts it. - Help replay starts at Accounts for an existing Budget. It does not replay currency/format confirmation. - Progress stores the active beat, Budget and month. It does not store unsaved Account fields or Category question/checklist drafts. - Same-month income is detected while this view is mounted. A first income across route reloads needs an event-backed trigger. - Account and Category creation have no Undo. Existing Money Undo has no server mutation receipt; the new guards do not provide a full cross-client Undo guarantee. - The category batch does not retain a tombstone after every row from that batch is deleted. - The separate skip-first-spend/first-transaction nudge path has no production browser evidence. The main flow spends, and marks that teaching beat as seen. - Claude visual review remains pending. **Decisions** One active guide is stored in the existing User settings section `money_onboarding`, with a revision. No balance is stored there. Custom draft Categories start in Goals. Optional questions default to No; Rent stays visible but unticked when housing is No. One UUID identifies each confirmed checklist; at most 100 rows and 32 KiB are accepted. Active guide links use `?guide=accounts|categories|assign|spend`. Unknown values cannot reopen a terminal guide. Preference reads share only an in-flight request for the current User/lifecycle; no completed response is retained. These choices extend the issue's owner decisions where they did not specify storage or draft defaults. **Performance / for the merge round** `bench/money-onboarding-1141.mjs` covers progress-read p50/p95, CPU/RSS, a 100-Category checklist and four concurrent retries. It was not measured: the latest verification policy reserves performance measurements for performance issues on the locked perf VM. There is no onboarding baseline entry in `docs/perf/baseline.json`. After the dependency integrations, run the production flow against the combined server: ``` CALTERNAL_SERVER_BIN="$CARGO_TARGET_DIR/release/calternal-server" node apps/web/e2e/money-onboarding-1141.mjs ``` It must prove all five beats with the final format and amount controls, the #1140 entry, exact totals and skip/resume/replay. Add the remaining first-spend skip and cross-route income cases. Check the failed startup child without changing its limits: ``` cargo test -p calternal-server --bin calternal-server wire::tests::startup_search_reconciliation_keeps_calendar_writes_responsive_on_large_home -- --ignored --exact --test-threads=1 ``` Run the full Money e2e suite (`cd apps/web && bun run test:e2e:money`) and the existing `tests/adversarial/authz_matrix.py` / `xuser_matrix.py` entry points as part of the combined merge round. They must include the new own-data routes. Full adversarial matrices, release/staging checks and real Mac interop were not run in this feature job. **Files** - `apps/web/e2e/money-onboarding-1141.mjs` - `apps/web/src/lib/components/ConfirmSheet.svelte` - `apps/web/src/lib/components/money/AddAccountForm.svelte` - `apps/web/src/lib/components/money/BudgetGuide.svelte` - `apps/web/src/lib/components/money/BudgetGuide.svelte.test.ts` - `apps/web/src/lib/components/money/CategoryChecklist.svelte` - `apps/web/src/lib/components/money/CategoryChecklist.svelte.test.ts` - `apps/web/src/lib/components/money/CreateBudgetForm.svelte` - `apps/web/src/lib/components/money/TransactionForm.svelte` - `apps/web/src/lib/money/api.ts` - `apps/web/src/lib/money/onboarding-api.test.ts` - `apps/web/src/lib/money/onboarding-categories.test.ts` - `apps/web/src/lib/money/onboarding-categories.ts` - `apps/web/src/lib/money/onboarding.test.ts` - `apps/web/src/lib/money/onboarding.ts` - `apps/web/src/routes/money/+page.svelte` - `apps/web/src/routes/money/[budget]/[month]/+page.svelte` - `bench/money-onboarding-1141.mjs` - `contracts/action-policy.json` - `contracts/actions.json` - `contracts/openapi.json` - `contracts/perf/adoption-1058.json` - `contracts/perf/exceptions.json` - `contracts/perf/ratchet.json` - `contracts/perf/registry.json` - `crates/plugins/money/src/lib.rs` - `crates/plugins/money/src/onboarding.rs` - `crates/plugins/money/src/routes.rs` - `crates/plugins/money/src/tests.rs` - `packages/api-client/src/generated.ts` - `tests/adversarial/money_onboarding.mjs` **Cleanup** `cargo clean` output: ``` Removed 18316 files, 13.3GiB total ``` The generated web build and `.svelte-kit/output` were deleted. Screenshots, archives and logs stay in ignored `artifacts/`. The worktree is clean. No screenshot or video was committed. No credential was displayed.
Author
Owner

Resumed #1141 on job/onboard-1141 at f1fd658f42. Read CLAUDE.md, CONTEXT.md, DESIGN §§33,34,48,59, previous report and owner spec. Worktree contains the pending merge of completed #1180 head 2f005fc516. Resolving both guide and format behavior, then finishing walkthrough UX and evidence. No push or deploy.

Resumed #1141 on job/onboard-1141 at f1fd658f42bfd74d01a6148e1e5df109be3f6ea9. Read CLAUDE.md, CONTEXT.md, DESIGN §§33,34,48,59, previous report and owner spec. Worktree contains the pending merge of completed #1180 head 2f005fc5168d767d050dd8a0cecd033b9c9e72a7. Resolving both guide and format behavior, then finishing walkthrough UX and evidence. No push or deploy.
Author
Owner

#1180 integration now uses the persisted Budget convention in walkthrough Account and Transaction forms and overspend amounts. Shared Torph counters remain in the real Money screen. Money clippy passed: Finished dev profile [unoptimized + debuginfo] target(s) in 4m 40s.

Focused guide tests: Test Files 5 passed (5) / Tests 20 passed (20). Added recovery for raw Account and checklist drafts using existing User-owned device storage, full format replay on an existing Budget, a zero Ready to assign requirement before first spend, and same-month new-money detection after route reload. Partial currency text previously reached Intl.NumberFormat and could throw; preview now waits for a complete canonical code. Production evidence is being extended for those paths.

Dependency finding: origin/dev still has the old import surface; the finished #1140 wizard is on job/wizard-1140, outside this job's explicitly authorized dependency merge. The combined merge round must connect its new-Budget callback to the start-guide link. I am preserving that boundary while making the guide entry reusable.

#1180 integration now uses the persisted Budget convention in walkthrough Account and Transaction forms and overspend amounts. Shared Torph counters remain in the real Money screen. Money clippy passed: `Finished dev profile [unoptimized + debuginfo] target(s) in 4m 40s`. Focused guide tests: `Test Files 5 passed (5)` / `Tests 20 passed (20)`. Added recovery for raw Account and checklist drafts using existing User-owned device storage, full format replay on an existing Budget, a zero Ready to assign requirement before first spend, and same-month new-money detection after route reload. Partial currency text previously reached Intl.NumberFormat and could throw; preview now waits for a complete canonical code. Production evidence is being extended for those paths. Dependency finding: origin/dev still has the old import surface; the finished #1140 wizard is on job/wizard-1140, outside this job's explicitly authorized dependency merge. The combined merge round must connect its new-Budget callback to the start-guide link. I am preserving that boundary while making the guide entry reusable.
Author
Owner

Integration committed at 364d5d240. Fetched origin/dev and merged once before final gates: Already up to date. Refreshed exact live pins without changing rule logic or increasing the ratchet: perf-lint: PASS; 0 violations; 22396 scoped exceptions.

Money gates (verbatim):

    Finished `dev` profile [unoptimized + debuginfo] target(s) in 4m 40s
    Finished `test` profile [unoptimized + debuginfo] target(s) in 11m 12s
test result: ok. 96 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 167.99s
test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 15.96s

The calternal-money core suite also passed. No Rust expectation changed. Production build passed; final web check/test and the local server build are in progress. The existing shared-server CURRENT file is absent, so the screenshot server must be built in this worktree.

Integration committed at 364d5d240. Fetched origin/dev and merged once before final gates: `Already up to date.` Refreshed exact live pins without changing rule logic or increasing the ratchet: `perf-lint: PASS; 0 violations; 22396 scoped exceptions`. Money gates (verbatim): ``` Finished `dev` profile [unoptimized + debuginfo] target(s) in 4m 40s Finished `test` profile [unoptimized + debuginfo] target(s) in 11m 12s test result: ok. 96 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 167.99s test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 15.96s ``` The calternal-money core suite also passed. No Rust expectation changed. Production build passed; final web check/test and the local server build are in progress. The existing shared-server CURRENT file is absent, so the screenshot server must be built in this worktree.
Author
Owner

UX review found that the checklist's accessible name included its internal draft key. Names now use the visible group, and test selectors use a non-visible test ID; the existing assertions are unchanged. The checklist's 44 px target now belongs to the shared Checkbox via an optional hitFloor prop, with tests and docs; feature CSS only positions it. The new prop defaults off.

Correction to the earlier wizard-link note: creating another Budget must NOT reopen a finished or skipped guide. The owner explicitly requires terminal progress to remain closed. The newly added regression now enforces this rule. Only Help or an explicit guide replay link overrides terminal progress. A pending first guide still starts at Accounts after the new-Budget form confirms currency and format.

The complete two-worker web run reports:

 Test Files  9 failed | 276 passed | 1 skipped (286)
      Tests  9 failed | 1942 passed | 1 skipped (1952)

Its failures are timeouts in unchanged files; no assertion or deadline was weakened. Web check passed with zero errors. A one-worker final suite will include the native target and accessible-name fixes.

UX review found that the checklist's accessible name included its internal draft key. Names now use the visible group, and test selectors use a non-visible test ID; the existing assertions are unchanged. The checklist's 44 px target now belongs to the shared Checkbox via an optional hitFloor prop, with tests and docs; feature CSS only positions it. The new prop defaults off. Correction to the earlier wizard-link note: creating another Budget must NOT reopen a finished or skipped guide. The owner explicitly requires terminal progress to remain closed. The newly added regression now enforces this rule. Only Help or an explicit guide replay link overrides terminal progress. A pending first guide still starts at Accounts after the new-Budget form confirms currency and format. The complete two-worker web run reports: ``` Test Files 9 failed | 276 passed | 1 skipped (286) Tests 9 failed | 1942 passed | 1 skipped (1952) ``` Its failures are timeouts in unchanged files; no assertion or deadline was weakened. Web check passed with zero errors. A one-worker final suite will include the native target and accessible-name fixes.
Author
Owner

Continuation finding: the cover action hid its one-time lesson after a transfer even when the donor could cover only part of the shortfall. Commit cd85613e9 now checks the returned Category balance and keeps the lesson open until the shortfall is funded. Its focused regression passed. Commit 77957ce27 uses #1180 Torph settings for the guide step and transfer counters as well as the existing Money report.

Focused gate output:

 Test Files  1 passed (1)
      Tests  10 passed (10)

Exact live perf pins were refreshed in dfad597fc; the exception count did not change. The final check, production build and full serial test retry are running. The local server build is still required because the shared prebuilt CURRENT file was absent.

Continuation finding: the cover action hid its one-time lesson after a transfer even when the donor could cover only part of the shortfall. Commit cd85613e9 now checks the returned Category balance and keeps the lesson open until the shortfall is funded. Its focused regression passed. Commit 77957ce27 uses #1180 Torph settings for the guide step and transfer counters as well as the existing Money report. Focused gate output: ``` Test Files 1 passed (1) Tests 10 passed (10) ``` Exact live perf pins were refreshed in dfad597fc; the exception count did not change. The final check, production build and full serial test retry are running. The local server build is still required because the shared prebuilt CURRENT file was absent.
Author
Owner

Production touch finding: Playwright tapped the native Rent checkbox, but the decorative SVG intercepted the hit. The tap timed out after 30 seconds. Commit e9581b51d makes the native-form drawing ignore pointer events, so pointer and touch reach the input. The focused Checkbox and guide files passed:

 Test Files  2 passed (2)
      Tests  12 passed (12)

The browser flow now checks pointer, Space and touch activation without force. It also reloads each draft screen and checks the real stored fields. The active guide hides the duplicate Account prompt.

Verification findings: running check and build together produced mismatched SvelteKit bootstrap identifiers in HTML and JavaScript, so the app did not boot. A sequential build has matching identifiers. The server then refused a theme save with status 507 because the shared disk fell below its 5% reserve. I retained the server binary and cleaned this job's Cargo output:

     Removed 12758 files, 9.1GiB total

The current production capture run uses the sequential build and the retained binary. No test assertion, deadline or server reserve was weakened.

Production touch finding: Playwright tapped the native Rent checkbox, but the decorative SVG intercepted the hit. The tap timed out after 30 seconds. Commit e9581b51d makes the native-form drawing ignore pointer events, so pointer and touch reach the input. The focused Checkbox and guide files passed: ```text Test Files 2 passed (2) Tests 12 passed (12) ``` The browser flow now checks pointer, Space and touch activation without force. It also reloads each draft screen and checks the real stored fields. The active guide hides the duplicate Account prompt. Verification findings: running check and build together produced mismatched SvelteKit bootstrap identifiers in HTML and JavaScript, so the app did not boot. A sequential build has matching identifiers. The server then refused a theme save with status 507 because the shared disk fell below its 5% reserve. I retained the server binary and cleaned this job's Cargo output: ```text Removed 12758 files, 9.1GiB total ``` The current production capture run uses the sequential build and the retained binary. No test assertion, deadline or server reserve was weakened.
Author
Owner

Production primary-action finding: after the User typed a balance and clicked Add account, the Account API still returned zero Accounts. No alert was shown, and every native field was valid. The focused amount preview disappeared on blur and moved the button between pointer press and release.

Commit 786036c98 keeps the preview slot in flow. Inactive text is hidden from sight and speech. Amount input names remain fixed while the preview changes. The shared Assigned variant keeps its screen-reader-only preview. This fixes the control itself, with no forced click or layout measurement.

Focused output:

 Test Files  3 passed (3)
      Tests  6 passed (6)

The production test requires the real Add another account confirmation, then Continue. It focuses the filled shared amount fields before capturing and clicking Save. The new Slot regression also checks the accessible description and hidden state across blur. A full test run that overlapped this code change cached the old component while loading the new regression. It is not final evidence; final gates will run with the source held fixed.

Root-disk reserve remains unchanged. This job's required target/tmp path now points to a private job-owned fixture directory on /mnt/hdd, which has enough free space for real API writes.

Production primary-action finding: after the User typed a balance and clicked Add account, the Account API still returned zero Accounts. No alert was shown, and every native field was valid. The focused amount preview disappeared on blur and moved the button between pointer press and release. Commit 786036c98 keeps the preview slot in flow. Inactive text is hidden from sight and speech. Amount input names remain fixed while the preview changes. The shared Assigned variant keeps its screen-reader-only preview. This fixes the control itself, with no forced click or layout measurement. Focused output: ```text Test Files 3 passed (3) Tests 6 passed (6) ``` The production test requires the real Add another account confirmation, then Continue. It focuses the filled shared amount fields before capturing and clicking Save. The new Slot regression also checks the accessible description and hidden state across blur. A full test run that overlapped this code change cached the old component while loading the new regression. It is not final evidence; final gates will run with the source held fixed. Root-disk reserve remains unchanged. This job's required `target/tmp` path now points to a private job-owned fixture directory on `/mnt/hdd`, which has enough free space for real API writes.
Author
Owner

#1141 continuation report

Branch: job/onboard-1141. Base: f1fd658f42bfd74d01a6148e1e5df109be3f6ea9. Head: 13a955864e.

Status: continuation committed and verified; #1141 still has the gaps below. The job time box has been reached, so no further feature work is done in this run. No push, deployment or merge into dev was done. The pending #1180 merge was completed, origin was fetched, and git merge origin/dev reported Already up to date.

Built

The real Budget walkthrough now uses #1180's persisted Budget money format in its Account form, first purchase, report and overspend action. It uses the shared AmountInput and FormatPicker. Its step and transfer counters use Torph, with the same motion durations and reduced-motion preference as the Money report. Help replays the format beat on the existing Budget.

Raw Budget, Account, checklist and purchase drafts survive reload on the same device. Drafts are owned by the current User, bounded, optional and cleared after a confirmed write or explicit terminal replay. The checklist retains its operation ID for safe retries. Changing a purchase date no longer clears the other fields.

The guide teaches new money after returning from another route in the same month. Assignment Continue requires Ready to Assign to be zero. A partial cover keeps the overspend lesson open. Skipped and finished guides stay closed until explicit Help replay.

Files and shared controls

Continuation files: AmountInput.svelte and its test, BudgetGuide.svelte, CategoryChecklist.svelte, CreateBudgetForm.svelte, AddAccountForm.svelte, TransactionForm.svelte, their focused tests, money/guide-drafts.ts and its tests, money/onboarding.ts, the Money month route, packages/ui/src/components/Checkbox.svelte, its focused test and UI README. money/api.ts, generated API/action contracts and exact perf pins were integrated with #1180. The production browser regression is apps/web/e2e/money-onboarding-1141.mjs. The complete file list, including incoming branch changes, is in artifacts/onboard2-files.txt.

The AmountInput preview now keeps its slot on blur, so native Save buttons stay under the pointer. Its amount names stay fixed while the parsed preview changes. Inactive text is hidden from sight and speech.

Visible controls reuse Pill for actions, Card for guide surfaces, Select through FormatPicker, AmountInput for editable amounts, SegmentedControl for Account type, Checkbox for selection, and PopoverSurface for assignment teaching. Inputs keep the existing Money form style. Native Checkbox has an optional shared hitFloor; its default geometry is unchanged.

UX gaps closed

  • Incomplete currency typing does not crash the preview. Continue waits for a complete code; the server validates it.
  • Budget format remains consistent in every guide amount. Help includes format confirmation.
  • Unsubmitted fields survive route changes and reload. Oversized drafts cannot grow retained state.
  • Category names have readable screen-reader names; stable row keys are not spoken.
  • Checklist checkboxes have shared touch targets and native keyboard activation. Their decorative SVG no longer intercepts native pointer or touch input.
  • Amount previews keep their space on blur. Add account and Add purchase stay under the pointer between press and release. Screen-reader input names remain fixed as previews change.
  • The active guide hides the duplicate Account prompt.
  • Add another account appears after an account exists. The purchase form has one Skip for now action.
  • New-money teaching works across route reloads. A partial cover does not dismiss its lesson.
  • New-budget entry respects skipped and finished progress.

Gates, verbatim

cargo fmt --check: no output; exit 0.

cargo clippy -p calternal-plugin-money -p calternal-money --all-targets -- -D warnings:

    Finished `dev` profile [unoptimized + debuginfo] target(s) in 4m 40s

cargo test -p calternal-money -- --test-threads=4:

test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.12s
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
test result: ok. 13 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 23.99s
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.14s
test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.60s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
test result: ok. 4 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.53s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.11s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

cargo test -p calternal-plugin-money -- --test-threads=4:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 96 filtered out; finished in 148.48s
test result: ok. 96 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 167.99s
test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 15.96s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s

bun run check:

perf-lint: PASS; 0 violations; 22396 scoped exceptions
svelte-check found 0 errors and 2 warnings in 2 files

bun run test --maxWorkers=2 (initial run):

 Test Files  9 failed | 276 passed | 1 skipped (286)
      Tests  9 failed | 1942 passed | 1 skipped (1952)

One timeout-diagnosis retry, bun run test --maxWorkers=1 --no-file-parallelism:

 Test Files  2 failed | 284 passed | 1 skipped (287)
      Tests  2 failed | 1952 passed | 1 skipped (1955)

The timeout-diagnosis failures were five-second timeouts in unchanged RangeBar.svelte.test.ts and CalendarFeedsGroup.svelte.test.ts. Both files then passed together, with no edits to assertions or deadlines:

 Test Files  2 passed (2)
      Tests  11 passed (11)

Final focused BudgetGuide regression:

 Test Files  1 passed (1)
      Tests  10 passed (10)

Final amount/transaction/currency checks:

 Test Files  3 passed (3)
      Tests  6 passed (6)

bun run test --maxWorkers=2 on the completed, fixed source:

 Test Files  286 passed | 1 skipped (287)
      Tests  1955 passed | 1 skipped (1956)

This final run passed with the original deadlines and assertions. A prior run overlapped the preview fix and cached its old component while loading the new regression; that run is not final evidence.

Shared Checkbox/checklist/guide checks passed. Draft, currency and purchase focused regressions passed. Full logs remain in artifacts/. The production build passed:

Compressed 916 static variants; saved 21206280 bytes.

Production evidence

144 production images are attached to #1141: 12 states × 390/820/1440 px × light/dark × native/4× crop. Every browser context emulates macOS. The run used a fresh User and real API writes. It passed exact balances, Budget format, draft recovery, pointer/Space/touch selection, confirmed saves, finish, overspend, next-month and same-month income, Help format replay, skip-first-spend teaching and terminal skip.

Verbatim browser result:

PASS #1141: real guided Budget, exact balances, resume, finish, overspend, new money, replay and skip; macOS screenshot matrix captured.

Verification uses a sequential production build. Check and build both write .svelte-kit; a concurrent run produced mismatched HTML/client bootstrap identifiers. The final build checks that these identifiers match. The main and touch browser contexts both use the current production asset override.

The root disk crossed its 5% reserve during fixture writes. This job cleaned its Cargo output and moved the private fixture directory to /mnt/hdd, through the required worktree target/tmp path. No server reserve, test deadline or existing expectation was weakened.

UX gaps left and known gaps

  • #1140's first-run wizard is not on origin/dev. The pending guide supports ?start-guide=1, but the wizard's new-budget callback still needs to open it when that branch is combined. A skipped or finished guide must remain closed. No unrelated wizard branch was merged here.
  • Creation Undo and cross-client receipt-backed Undo remain shared Money work. Existing inverse actions are User-fenced and revision-guarded where available; they are not durable mutation receipts.
  • Category batch deduplication has no retained tombstone after all batch rows are deleted.
  • Drafts are per device. Disabled storage or a draft beyond the cap keeps edits in memory only.
  • First-spend teaching uses the selected month's aggregate activity. A spend outside that month or a net refund can hide this signal.
  • Final web check and full test pass. Wizard integration, receipt work, #1211 and visual review still remain; this report does not mark all of #1141 complete.
  • #1211: the overspend source select looks blank although the action uses the first funded Category. Its transfer label also loses spaces around the Torph amount. The exact-balance assertions passed; these presentation gaps remain for the follow-up.
  • Claude's visual review is still required. Native and 4× crops were inspected for alignment; this is not a visual-quality approval.

Decisions

Use the existing User-owned device draft store for unsaved fields; server progress remains authoritative. Limit each stored draft to 32,768 UTF-16 code units and keep oversized edits in memory. Retain one operation ID with the checklist draft.

On Help replay, confirm the current Budget currency and allow its number convention to change. Currency remains the existing immutable Budget value; replay does not create another Budget. A new-budget entry can continue pending progress, but explicit Help is required after skip or finish.

Reuse the existing common-currency datalist instead of adding another currency service. Extend Checkbox with a default-off touch target instead of local visual styling. Keep the amount preview slot in flow and hide inactive text, so blur cannot move an action target. Explicit amount names keep the native accessible name stable. A cover lesson completes only when the returned target balance is nonnegative.

Expected cost and performance

A progress read/write uses one bounded User preference section. A draft update encodes one bounded value, with no draft history or network request. The checklist prepares at most 100 rows, then sends one atomic checked batch. Account and purchase submissions use the existing Money actions. Guide derivations scan the currently loaded Categories, not the Home; no runtime layout loop was added.

bench/money-onboarding-1141.mjs covers progress p50/p95, CPU/RSS, the largest allowed 100-Category batch and a four-request retry burst. No matching baseline exists in docs/perf/baseline.json. No measurement was run: the later verification policy reserves perf VM measurements for performance issues. No latency number is claimed.

For the merge round

  • Connect #1140's confirmed Budget creation to /money/<id>/<month>?start-guide=1 and run CALTERNAL_E2E_ASSET_OVERRIDE=1 node apps/web/e2e/money-onboarding-1141.mjs on the combined branch. It must prove the actual wizard entry, exact totals, reload drafts, pointer/touch/keyboard, format replay and terminal skip.
  • cd apps/web && bun run test --maxWorkers=2: the complete suite must pass.
  • cargo clippy -p calternal-server --all-targets -- -D warnings and cargo test -p calternal-server -- --test-threads=4: verify the combined server and route contracts. The previous job found a large-Home startup timeout.
  • tests/adversarial/run-split.sh: verify hostile input, robustness, XUser and authorization scopes on the combined API, with separate data directories. The focused Money guide validation is already part of the browser regression.
  • Claude reviews the attached native-size and 4× images. Full e2e, release, staging and Apple-client checks remain merge-round work.

Cleanup and handoff

Cargo output was cleaned:

     Removed 12758 files, 9.1GiB total

Web build output and the temporary retained binary/runtime are removed. Review PNGs, ZIPs and logs stay under gitignored artifacts/. The worktree is clean. No push, deployment or merge into dev was done.

# #1141 continuation report Branch: `job/onboard-1141`. Base: `f1fd658f42bfd74d01a6148e1e5df109be3f6ea9`. Head: 13a955864eb6739d5c6b2fe249a929429c9c98e2. Status: continuation committed and verified; #1141 still has the gaps below. The job time box has been reached, so no further feature work is done in this run. No push, deployment or merge into dev was done. The pending #1180 merge was completed, origin was fetched, and `git merge origin/dev` reported `Already up to date.` ## Built The real Budget walkthrough now uses #1180's persisted Budget money format in its Account form, first purchase, report and overspend action. It uses the shared AmountInput and FormatPicker. Its step and transfer counters use Torph, with the same motion durations and reduced-motion preference as the Money report. Help replays the format beat on the existing Budget. Raw Budget, Account, checklist and purchase drafts survive reload on the same device. Drafts are owned by the current User, bounded, optional and cleared after a confirmed write or explicit terminal replay. The checklist retains its operation ID for safe retries. Changing a purchase date no longer clears the other fields. The guide teaches new money after returning from another route in the same month. Assignment Continue requires Ready to Assign to be zero. A partial cover keeps the overspend lesson open. Skipped and finished guides stay closed until explicit Help replay. ## Files and shared controls Continuation files: `AmountInput.svelte` and its test, `BudgetGuide.svelte`, `CategoryChecklist.svelte`, `CreateBudgetForm.svelte`, `AddAccountForm.svelte`, `TransactionForm.svelte`, their focused tests, `money/guide-drafts.ts` and its tests, `money/onboarding.ts`, the Money month route, `packages/ui/src/components/Checkbox.svelte`, its focused test and UI README. `money/api.ts`, generated API/action contracts and exact perf pins were integrated with #1180. The production browser regression is `apps/web/e2e/money-onboarding-1141.mjs`. The complete file list, including incoming branch changes, is in `artifacts/onboard2-files.txt`. The AmountInput preview now keeps its slot on blur, so native Save buttons stay under the pointer. Its amount names stay fixed while the parsed preview changes. Inactive text is hidden from sight and speech. Visible controls reuse Pill for actions, Card for guide surfaces, Select through FormatPicker, AmountInput for editable amounts, SegmentedControl for Account type, Checkbox for selection, and PopoverSurface for assignment teaching. Inputs keep the existing Money form style. Native Checkbox has an optional shared `hitFloor`; its default geometry is unchanged. ## UX gaps closed - Incomplete currency typing does not crash the preview. Continue waits for a complete code; the server validates it. - Budget format remains consistent in every guide amount. Help includes format confirmation. - Unsubmitted fields survive route changes and reload. Oversized drafts cannot grow retained state. - Category names have readable screen-reader names; stable row keys are not spoken. - Checklist checkboxes have shared touch targets and native keyboard activation. Their decorative SVG no longer intercepts native pointer or touch input. - Amount previews keep their space on blur. Add account and Add purchase stay under the pointer between press and release. Screen-reader input names remain fixed as previews change. - The active guide hides the duplicate Account prompt. - Add another account appears after an account exists. The purchase form has one Skip for now action. - New-money teaching works across route reloads. A partial cover does not dismiss its lesson. - New-budget entry respects skipped and finished progress. ## Gates, verbatim `cargo fmt --check`: no output; exit 0. `cargo clippy -p calternal-plugin-money -p calternal-money --all-targets -- -D warnings`: ```text Finished `dev` profile [unoptimized + debuginfo] target(s) in 4m 40s ``` `cargo test -p calternal-money -- --test-threads=4`: ```text test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.12s test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s test result: ok. 13 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 23.99s test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.14s test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.60s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s test result: ok. 4 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.53s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.11s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` `cargo test -p calternal-plugin-money -- --test-threads=4`: ```text test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 96 filtered out; finished in 148.48s test result: ok. 96 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 167.99s test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 15.96s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s ``` `bun run check`: ```text perf-lint: PASS; 0 violations; 22396 scoped exceptions svelte-check found 0 errors and 2 warnings in 2 files ``` `bun run test --maxWorkers=2` (initial run): ```text Test Files 9 failed | 276 passed | 1 skipped (286) Tests 9 failed | 1942 passed | 1 skipped (1952) ``` One timeout-diagnosis retry, `bun run test --maxWorkers=1 --no-file-parallelism`: ```text Test Files 2 failed | 284 passed | 1 skipped (287) Tests 2 failed | 1952 passed | 1 skipped (1955) ``` The timeout-diagnosis failures were five-second timeouts in unchanged `RangeBar.svelte.test.ts` and `CalendarFeedsGroup.svelte.test.ts`. Both files then passed together, with no edits to assertions or deadlines: ```text Test Files 2 passed (2) Tests 11 passed (11) ``` Final focused BudgetGuide regression: ```text Test Files 1 passed (1) Tests 10 passed (10) ``` Final amount/transaction/currency checks: ```text Test Files 3 passed (3) Tests 6 passed (6) ``` `bun run test --maxWorkers=2` on the completed, fixed source: ```text Test Files 286 passed | 1 skipped (287) Tests 1955 passed | 1 skipped (1956) ``` This final run passed with the original deadlines and assertions. A prior run overlapped the preview fix and cached its old component while loading the new regression; that run is not final evidence. Shared Checkbox/checklist/guide checks passed. Draft, currency and purchase focused regressions passed. Full logs remain in `artifacts/`. The production build passed: ```text Compressed 916 static variants; saved 21206280 bytes. ``` ## Production evidence 144 production images are attached to #1141: 12 states × 390/820/1440 px × light/dark × native/4× crop. Every browser context emulates macOS. The run used a fresh User and real API writes. It passed exact balances, Budget format, draft recovery, pointer/Space/touch selection, confirmed saves, finish, overspend, next-month and same-month income, Help format replay, skip-first-spend teaching and terminal skip. Verbatim browser result: ```text PASS #1141: real guided Budget, exact balances, resume, finish, overspend, new money, replay and skip; macOS screenshot matrix captured. ``` - [intro](https://git.kayg.org/attachments/e6e9f93c-6b21-46f4-90fd-b3d9e4774aed) - [currency](https://git.kayg.org/attachments/62a95900-6496-4de1-b2e1-16525647d202) - [accounts](https://git.kayg.org/attachments/78afd38b-f4f7-4234-922f-da49296ad38f) - [questions](https://git.kayg.org/attachments/0528e0d7-4cc8-4194-8c65-18a73bfaef2d) - [categories](https://git.kayg.org/attachments/8d3913a6-3df5-43ff-ad02-d6711cb32e5b) - [assign](https://git.kayg.org/attachments/9e0f3881-6a43-49b7-bfb5-f93519c992e9) - [spend](https://git.kayg.org/attachments/6be308a5-ef80-492a-ba83-790f8956674a) - [finished](https://git.kayg.org/attachments/e50b0ec4-a795-4ef5-841b-d2b8642ad685) - [overspend](https://git.kayg.org/attachments/1dd22343-8f99-42ac-85f6-471dab688984) - [new-money](https://git.kayg.org/attachments/5c291900-b532-4b55-a986-bf399467e4c7) - [replay-format](https://git.kayg.org/attachments/12f81136-1fe2-4d53-9ff1-02c5028423be) - [first-spend-nudge](https://git.kayg.org/attachments/baf4aa96-d737-4e6e-9fde-01770886f1fb) Verification uses a sequential production build. Check and build both write `.svelte-kit`; a concurrent run produced mismatched HTML/client bootstrap identifiers. The final build checks that these identifiers match. The main and touch browser contexts both use the current production asset override. The root disk crossed its 5% reserve during fixture writes. This job cleaned its Cargo output and moved the private fixture directory to `/mnt/hdd`, through the required worktree `target/tmp` path. No server reserve, test deadline or existing expectation was weakened. ## UX gaps left and known gaps - #1140's first-run wizard is not on origin/dev. The pending guide supports `?start-guide=1`, but the wizard's new-budget callback still needs to open it when that branch is combined. A skipped or finished guide must remain closed. No unrelated wizard branch was merged here. - Creation Undo and cross-client receipt-backed Undo remain shared Money work. Existing inverse actions are User-fenced and revision-guarded where available; they are not durable mutation receipts. - Category batch deduplication has no retained tombstone after all batch rows are deleted. - Drafts are per device. Disabled storage or a draft beyond the cap keeps edits in memory only. - First-spend teaching uses the selected month's aggregate activity. A spend outside that month or a net refund can hide this signal. - Final web check and full test pass. Wizard integration, receipt work, #1211 and visual review still remain; this report does not mark all of #1141 complete. - [#1211](https://git.kayg.org/kayg/calternal/issues/1211): the overspend source select looks blank although the action uses the first funded Category. Its transfer label also loses spaces around the Torph amount. The exact-balance assertions passed; these presentation gaps remain for the follow-up. - Claude's visual review is still required. Native and 4× crops were inspected for alignment; this is not a visual-quality approval. ## Decisions Use the existing User-owned device draft store for unsaved fields; server progress remains authoritative. Limit each stored draft to 32,768 UTF-16 code units and keep oversized edits in memory. Retain one operation ID with the checklist draft. On Help replay, confirm the current Budget currency and allow its number convention to change. Currency remains the existing immutable Budget value; replay does not create another Budget. A new-budget entry can continue pending progress, but explicit Help is required after skip or finish. Reuse the existing common-currency datalist instead of adding another currency service. Extend Checkbox with a default-off touch target instead of local visual styling. Keep the amount preview slot in flow and hide inactive text, so blur cannot move an action target. Explicit amount names keep the native accessible name stable. A cover lesson completes only when the returned target balance is nonnegative. ## Expected cost and performance A progress read/write uses one bounded User preference section. A draft update encodes one bounded value, with no draft history or network request. The checklist prepares at most 100 rows, then sends one atomic checked batch. Account and purchase submissions use the existing Money actions. Guide derivations scan the currently loaded Categories, not the Home; no runtime layout loop was added. `bench/money-onboarding-1141.mjs` covers progress p50/p95, CPU/RSS, the largest allowed 100-Category batch and a four-request retry burst. No matching baseline exists in `docs/perf/baseline.json`. No measurement was run: the later verification policy reserves perf VM measurements for performance issues. No latency number is claimed. ## For the merge round - Connect #1140's confirmed Budget creation to `/money/<id>/<month>?start-guide=1` and run `CALTERNAL_E2E_ASSET_OVERRIDE=1 node apps/web/e2e/money-onboarding-1141.mjs` on the combined branch. It must prove the actual wizard entry, exact totals, reload drafts, pointer/touch/keyboard, format replay and terminal skip. - `cd apps/web && bun run test --maxWorkers=2`: the complete suite must pass. - `cargo clippy -p calternal-server --all-targets -- -D warnings` and `cargo test -p calternal-server -- --test-threads=4`: verify the combined server and route contracts. The previous job found a large-Home startup timeout. - `tests/adversarial/run-split.sh`: verify hostile input, robustness, XUser and authorization scopes on the combined API, with separate data directories. The focused Money guide validation is already part of the browser regression. - Claude reviews the attached native-size and 4× images. Full e2e, release, staging and Apple-client checks remain merge-round work. ## Cleanup and handoff Cargo output was cleaned: ```text Removed 12758 files, 9.1GiB total ``` Web build output and the temporary retained binary/runtime are removed. Review PNGs, ZIPs and logs stay under gitignored `artifacts/`. The worktree is clean. No push, deployment or merge into dev was done.
Author
Owner

Final handoff: head 13a955864eb6739d5c6b2fe249a929429c9c98e2; worktree clean. All 12 screenshot archives uploaded with status 201. The final full web suite passed (1,955 tests), the web check passed, and Money's Rust gates passed. No push or deploy was done. Build output and private fixtures were removed.

Expected render cost detail: AmountInput retains one preview span per field, including Assigned fields. Its work grows with the field count and the normal integer parse/format work. No new latency number is claimed in this non-performance job. The existing hot-path profile remains in bench/money-onboarding-1141.mjs.

Presentation follow-up: #1211. Remaining dependency work: #1140 wizard entry and shared receipt-backed/creation Undo. Claude's visual review remains required. The complete final report and gate output are in the previous comment.

Final handoff: head `13a955864eb6739d5c6b2fe249a929429c9c98e2`; worktree clean. All 12 screenshot archives uploaded with status 201. The final full web suite passed (1,955 tests), the web check passed, and Money's Rust gates passed. No push or deploy was done. Build output and private fixtures were removed. Expected render cost detail: AmountInput retains one preview span per field, including Assigned fields. Its work grows with the field count and the normal integer parse/format work. No new latency number is claimed in this non-performance job. The existing hot-path profile remains in `bench/money-onboarding-1141.mjs`. Presentation follow-up: #1211. Remaining dependency work: #1140 wizard entry and shared receipt-backed/creation Undo. Claude's visual review remains required. The complete final report and gate output are in the previous comment.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/calternal#1141
No description provided.