Skip to content

Add heavy-task dual completion status - #128

Merged
likun666661 merged 1 commit into
apache:mainfrom
likun666661:issue102-heavy-task-p0d
Jun 23, 2026
Merged

Add heavy-task dual completion status#128
likun666661 merged 1 commit into
apache:mainfrom
likun666661:issue102-heavy-task-p0d

Conversation

@likun666661

Copy link
Copy Markdown
Member

Summary

Implements the P0-d export-only dual status slice for issue #102.

  • Add typed HeavyTaskCompletionStatus evaluation for heavy-task runs.
  • Project heavyTaskCompletion after task-run replay and export it as stable heavyTask.completion in full and compact result exports.
  • Keep official taskRun.status, scorer/verifier taxonomy, score, verifier, and legacy result fields authoritative.
  • Gate semantic completion on heavy-task mode enabled, accepted public P0-c self-check with status: "pass", latest P0-b todos present/non-empty, and all todos resolved or evidence-backed nonblocking.
  • Expose finalization.eligible as advisory metadata only; no bounded finalization model turn is implemented in this PR (boundedTurnImplemented: false).

Verification

  • npm --workspace @maka/headless run typecheck
  • npm --workspace @maka/headless run build
  • node --test dist/__tests__/heavy-task-finalization.test.js dist/__tests__/task-run-store.test.js dist/__tests__/result-export.test.js dist/__tests__/scorer.test.js dist/__tests__/heavy-task-self-check.test.js (36 tests)
  • npm --workspace @maka/headless test (274 tests, 37 suites)
  • git diff --cached --check

Rive workflow

Used a 3-node Rive workflow as requested: design reader, code writer/tester, reviewer. Reviewer returned ACCEPT; final steward reran verification and committed the final diff.

@likun666661
likun666661 merged commit 2698d20 into apache:main Jun 23, 2026
jackwener added a commit that referenced this pull request Jun 25, 2026
… one primitive + dead selector sweep (#254)

@kenji styles inventory task #128
(`notes/maka-styles-css-inventory-2026-06-25-task-128.md`) flagged that
`styles.css` (14066 lines) had become the product's UI database, with
the Settings select and Plan Reminder select drifted into two
near-identical wrappers + two CSS recipes. The Plan Reminder copy
got a real bug fix (selected-trigger icon) in PR #250; the Settings
copy never did.

@WAWQAQ msg `dea869c5` "做久一点 / 一次多做点" + msg `61862591`
authorizing this round; coordination ack `a137b9d7`. This bundles
Round A (dead selector sweep) and Round B (shared Rich Select
primitive + call-site migration) into one PR.

**Round B — shared SettingsSelect primitive.**
- New `packages/ui/src/primitives/settings-select.tsx`. Option type
  `[value, label, icon?]`; `width` variant (`compact` | `select` |
  `full`); the trigger uses `SelectValue`'s function-child so the
  collapsed picked state renders the same icon + label row as the
  popup items.
- `PlanReminderSelect` in `components.tsx` collapsed to a thin
  `<SettingsSelect width="full" {...props} />` specialization. Both
  surfaces now share one source of truth.
- `SettingsModal.tsx` drops its local `SettingsSelect` definition; the
  5 existing call sites (Daily Review 分析模型, Proxy 代理协议, Gateway
  监听地址, Bot 域名, Usage 状态筛选) automatically pick up the imported
  primitive — same `[value, label]` props still work.
- CSS: rename `.maka-plan-select*` / `.settingsBaseSelectTrigger*` →
  `.settingsSelect*` family on the primitive side; drop the
  `.settingsBaseSelectPopup` block that no longer has a consumer.
  `.settingsRow[data-control-width="select"]` and
  `.settingsField[data-orient="horizontal"]` keep the same width
  contract, now keyed off `.settingsSelectTrigger`.

**Round A — dead selector sweep.**
Per @kenji inventory the following had zero JSX/TSX consumers; I
re-verified each with `grep --include='*.tsx'`:
  - `.maka-nav-primary` (3 rules under `.maka-session-panel[data-
    collapsed="true"]`)
  - `.maka-skill-workbench-rail`

Promoted both into `renderer-style-pruning-contract.test.ts`'s
retired-hook list alongside the 12 already locked there. The contract
test (`✔ does not keep CSS for retired renderer hooks`) passes after
the deletion.

Net: +162 / -148 lines. One source of truth for the Settings/Plan
Reminder Select chrome; one less wrapper and ~24 lines of dead CSS to
maintain. Round C (ChoiceCard/RadioCard) and Round D (CSS ownership
split) still on @kenji's plan.
jackwener added a commit that referenced this pull request Jun 25, 2026
…onnectionStatus) (#255)

* refactor(ui,settings): unify SettingsSelect + PlanReminderSelect into one primitive + dead selector sweep

@kenji styles inventory task #128
(`notes/maka-styles-css-inventory-2026-06-25-task-128.md`) flagged that
`styles.css` (14066 lines) had become the product's UI database, with
the Settings select and Plan Reminder select drifted into two
near-identical wrappers + two CSS recipes. The Plan Reminder copy
got a real bug fix (selected-trigger icon) in PR #250; the Settings
copy never did.

@WAWQAQ msg `dea869c5` "做久一点 / 一次多做点" + msg `61862591`
authorizing this round; coordination ack `a137b9d7`. This bundles
Round A (dead selector sweep) and Round B (shared Rich Select
primitive + call-site migration) into one PR.

**Round B — shared SettingsSelect primitive.**
- New `packages/ui/src/primitives/settings-select.tsx`. Option type
  `[value, label, icon?]`; `width` variant (`compact` | `select` |
  `full`); the trigger uses `SelectValue`'s function-child so the
  collapsed picked state renders the same icon + label row as the
  popup items.
- `PlanReminderSelect` in `components.tsx` collapsed to a thin
  `<SettingsSelect width="full" {...props} />` specialization. Both
  surfaces now share one source of truth.
- `SettingsModal.tsx` drops its local `SettingsSelect` definition; the
  5 existing call sites (Daily Review 分析模型, Proxy 代理协议, Gateway
  监听地址, Bot 域名, Usage 状态筛选) automatically pick up the imported
  primitive — same `[value, label]` props still work.
- CSS: rename `.maka-plan-select*` / `.settingsBaseSelectTrigger*` →
  `.settingsSelect*` family on the primitive side; drop the
  `.settingsBaseSelectPopup` block that no longer has a consumer.
  `.settingsRow[data-control-width="select"]` and
  `.settingsField[data-orient="horizontal"]` keep the same width
  contract, now keyed off `.settingsSelectTrigger`.

**Round A — dead selector sweep.**
Per @kenji inventory the following had zero JSX/TSX consumers; I
re-verified each with `grep --include='*.tsx'`:
  - `.maka-nav-primary` (3 rules under `.maka-session-panel[data-
    collapsed="true"]`)
  - `.maka-skill-workbench-rail`

Promoted both into `renderer-style-pruning-contract.test.ts`'s
retired-hook list alongside the 12 already locked there. The contract
test (`✔ does not keep CSS for retired renderer hooks`) passes after
the deletion.

Net: +162 / -148 lines. One source of truth for the Settings/Plan
Reminder Select chrome; one less wrapper and ~24 lines of dead CSS to
maintain. Round C (ChoiceCard/RadioCard) and Round D (CSS ownership
split) still on @kenji's plan.

* chore(css): drop 2 more verified-dead selectors + lock in retired-hook contract

Round A-extended (yuejing 2026-06-25). @WAWQAQ 「继续」 (msg
`76e36d3b`) after #254 landed. Sticking to zero-visual-risk cleanup
while @kenji is offline and the ChoiceCard / RadioCard primitive
(Round C) still needs a visual-smoke harness to be safe.

Found two more selectors with zero JSX/TSX consumers via the standard
sweep against `apps/desktop/src/renderer/styles.css`:

- `.providerCatalog` — bare base class. `.providerCatalogRow` /
  `.providerCatalogTitle` / `.providerCatalogDesc` etc. are live in
  `ProvidersPanel.tsx`, but nothing renders the bare class itself.
- `.connectionStatus[data-ok="false"]` — combinator. No element ever
  carries the `connectionStatus` class in the renderer; the live
  status arm in the codebase is `.settingsConnectionBadge[data-tone]`,
  and `.providerError` (the other arm of the same rule) is unrelated
  and stays.

Both promoted into `renderer-style-pruning-contract.test.ts`'s
retired-hook list (now locks 16). `\.${hook}\b` boundary keeps the
live `.providerCatalogRow` / `.settingsConnectionBadge` siblings
clear.
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.

1 participant