@qoretechnologies/reqraft 0.10.65 → 0.10.66-pr.137.ga1983bf

package/.tasks/FORM_ENGINE_EXPRESSIONS.md

@@ -1,197 +0,0 @@
-# Form engine — expression support
-
-Design: [`design/FORM_ENGINE_EXPRESSIONS.md`](../design/FORM_ENGINE_EXPRESSIONS.md).
-Adds expression values to the form engine — visual builder (ported) +
-DPQL text editor (existing), one AST, plus an Explain button.
-
-**Status:** all 5 phases DONE pending user verify — implemented 2026-06-09/10,
-uncommitted. Foundation → TemplateField toggle + shell → DPQL text mode →
-visual ExpressionBuilder → expression validation. jest 267/267, ~14 new
-expression play tests pass, prod typecheck + eslint clean. One open item:
-the Explain transport (server `render-expression` exposure) — held behind
-`useRenderExpression` (interim client-side render works today). Builds on the
-also-uncommitted field-migration batch.
-
-**Server contract:** verified live (see design doc table + memory
-"Qorus expression/DPQL contract"). Explain transport is the one open
-server decision — held behind `useRenderExpression`, does not block.
-
-## Surface area
-
-| Path | Change |
-|---|---|
-| `src/components/form/expressions/types.ts` | **new** — `IExpression`, `IExpressionSchema`, arg types |
-| `src/components/form/expressions/useExpressions.ts` | **new** — catalogue fetch (`/system?action=expressions` + `expressions_url`) |
-| `src/components/form/expressions/useRenderExpression.ts` | **new** — Explain seam |
-| `src/components/form/expressions/ExpressionField.tsx` | **new** — Visual⇄Text shell + Explain |
-| `src/components/form/expressions/ExpressionBuilder.tsx` (+ argument/group/modal files) | **new** — ported visual tree |
-| `src/components/form/expressions/*.stories.tsx` | **new** — mock catalogue + mock LSP |
-| `src/components/form/fields/template/TemplateField.tsx` | edit — "Use Expression" toggle → render `ExpressionField` |
-| `src/components/form/engine/FormEngine.tsx` / `fields/Field.tsx` | edit — thread `is_expression` value plumbing |
-| `src/components/form/index.tsx` | edit — exports |
-
-## Phases (each ends at a STOP — user verifies before commit)
-
-### Phase 0 — Foundation (unblocked) — DONE pending verify (2026-06-09)
-- [x] `types.ts` (`IExpression`, `IExpressionValue`, `IExpressionSchema`,
-      `IExpressionSchemaArg`, `ExpressionDefaultValue`)
-- [x] `useExpressions` — `useFetch('system?action=expressions&context=ui')`,
-      `expressionsUrl` merge (marks `from_server`/`from_both`), `override`/
-      `errorOverride` seams, cached
-- [x] `useRenderExpression` seam — `render(value, expressions) => Promise<string>`.
-      **Interim impl is client-side** (`renderExpressionToText` walks the AST
-      using the catalogue: infix for operators, join for `&&`/`||`, function
-      form otherwise) — decoupled, no session needed; swaps to server
-      `render-expression` in one place. (Design said `dpql/serialize`; chose
-      client-side as it's session-free and gives a real string today.)
-- [x] `mockExpressions.ts` fixture (real `==` / `&&` / `contains` entries)
-- [x] `Foundation.stories.tsx` — catalogue count/names + rendered AST;
-      play test passes; verified visually
-- [x] prod typecheck + base typecheck clean
-
-### Phase 1 — TemplateField expression mode + `ExpressionField` shell — DONE pending verify
-- [x] `TemplateField`: "Use Expression" / "Use value" controls in the More
-      menu, gated on `supports_expressions`; seeds `{ args: [] }`
-- [x] When `isExpression`, render `ExpressionField` (main + disabled branches)
-- [x] `ExpressionField` shell: holds the AST, **Visual ⇄ Text toggle**
-      (placeholders for now), Explain button (`useRenderExpression`), live
-      preview of the current expression
-- [x] Value plumbing: `FormEngine.handleValueChange` detects an
-      `is_expression` value and stores the canonical sibling shape
-      `{ type, value:<AST>, is_expression:true }`; passes `is_expression`
-      down to `TemplateField`
-- [x] Stories: `Default`, `Empty`, `ViaFormEngine` (round-trip),
-      `ToggleInFormEngine` (verified live: bool field → Use Expression →
-      editor; → Use value → back). Barrel exports added.
-- [x] Regression: jest 267/267, form play tests pass (only pre-existing
-      `OnValidityChange` fails)
-- Note: expression-aware **validation** deferred to Phase 4 (an empty
-  expression currently shows the base-type validation error).
-
-### Phase 2 — Text mode (DPQL) — DONE pending verify
-- [x] `ExpressionField` Text mode renders `DpqlEditor` (via its ref for
-      parse/serialize)
-- [x] Boundary glue: parse-on-edit (debounced 300ms) `text→AST`;
-      serialize-on-enter `AST→text`; flush-parse on switch-to-Visual
-- [x] DPQL-context: optional `provider`/`recordType` passed through when
-      supplied; parse/serialize work without a provider (only `@field`
-      completions are reduced) — matches the design's worst-case note
-- [x] `dpqlMockLsp.ts` — compact mock-socket LSP (initialize / parse /
-      serialize / validate / completion); `dpql/parse` echoes the typed
-      text into an `==` AST for deterministic verification
-- [x] Story `TextMode` — types DPQL → parse → the Parsed preview reflects
-      it (verified in real chromium). 5/5 ExpressionField play tests pass.
-- Note: serialize-on-enter is best-effort if the LSP session isn't ready
-      yet; matters mainly when coming from a Visual-built expression
-      (Phase 3).
-
-### Phase 3 — Visual ExpressionBuilder — DONE pending verify (core)
-- [x] `ExpressionBuilder.tsx` — recursive tree: function `Select` (filtered
-      by return type), operands, AND/OR groups, add/remove condition,
-      recursion via `level`/`nested`
-- [x] **Operands render through `TemplateField`** (→ literal / template /
-      nested expression via the Phase 1 toggle) and `AutoFormField` for
-      `any`-typed operands. No re-port of qorus-ide's operand stack.
-- [x] `returnType` filters the function list; nested-expression ↔ value
-      conversion (the operand "fx"/Use-Expression toggle + "use value")
-- [x] Stories: `Default` (`==` + operands), `Empty`, `Group` (`&&` tree +
-      Add condition), `ReadOnly`. 4/4 play tests pass; verified visually
-      (nested AND tree with per-operand type pickers).
-- [x] Regression: jest 267/267, form play tests pass (only pre-existing
-      `OnValidityChange`)
-- **Deferred to a polish pass** (documented): wrap/unwrap, the
-  `types_mismatch` confirm modal + full type-compat filtering, varargs
-  "add argument" for non-group expressions, fullscreen/focused editing,
-  AI-assist button. The core build/edit/group/recurse flow is complete.
-
-### Phase 3b — visual polish to match qorus-ide — DONE pending verify (2026-06-10)
-- [x] Bumped ReQore **0.68.3 → 0.68.5** (matches the IDE; devDep + peer
-      range). Full suite still green (188/189, jest 267).
-- [x] Rebuilt `ExpressionBuilder` to mirror qorus-ide's layout: each
-      expression is a dashed-border **card** titled by the function picker
-      (`Σ` chip); operands laid **horizontally** (label on top, field below;
-      `fluid` cards + 220px leaf width so they sit in a row and wrap
-      responsively); tiny-uppercase `LABEL * [type]` labels; **AND/OR
-      convert** buttons (`.expression-and`/`.expression-or`); `&&`/`||` as
-      group cards (`IF group`/`AND group`/`OR group`, `.expression-group`)
-      with AND/OR separators + Add condition + per-condition remove; per-
-      level `main:darken:${level}` tint. Picker always keeps the selected
-      expression even when its return type doesn't match the filter.
-- [x] Stories rewritten to mirror foxhoundn's IDE stories using the
-      **`.expression` count** convention + the `.expression-and/or`
-      interaction (1→2→3). 11/11 play tests pass; verified visually vs `:6009`.
-- [x] **Live story** (`ExpressionField › Live`, `tags: ['!test']`,
-      `parameters: { live: true }`) — fetches the REAL catalogue (no
-      `expressions` override), mirroring the DpqlEditor live stories.
-      `.storybook/preview.tsx` instance is now env-configurable
-      (`REACT_APP_QORUS_INSTANCE` || hq default). Full suite **189/189**
-      (Live excluded from the offline runner; defaults to hq when env unset).
-- Verification gap (honest): could NOT render-verify the Live story against
-      real data — `localhost:8012` returns the catalogue to curl but sends no
-      CORS header for the storybook origin (browser blocks the fetch), and I
-      have no `hq:8092` token. The data path is proven via curl (121
-      expressions) and the wiring mirrors the working DpqlEditor live pattern;
-      the user verifies by running storybook with their token/instance.
-
-### Phase 3c — faithful mock + offline parity with the IDE — DONE (2026-06-10)
-- [x] **Rewrote `mockExpressions`** as a faithful subset of the LIVE catalogue
-      (captured via curl): `==`, `<`, `>=`, `&&`, `||`, `contains`,
-      `starts-with`, `ends-with` with the REAL arg types (`richtext` strings,
-      `any` comparators), `label_after` operator words ("contains", "starts
-      with", "=="), and the Ignore Case bool. This is why the offline stories
-      were looking worse — the old 3-entry mock used `any` for `==` (→ bulky
-      type-pickers) and omitted operator words.
-- [x] **`WithComplexValue` now uses qorus-ide's EXACT value** (`&&[ ||[contains,
-      &&[starts-with, starts-with], >=], ends-with, < ]`) → 6 expression cards
-      across 3 groups; was a flat 2-card `&&`.
-- [x] **Delete is now a card action** (`onRemove` prop → `expression-group-remove`)
-      instead of a sibling button. The sibling stole flex width and forced
-      operands to wrap vertically; as a card action the cards fill width and
-      operands lay out **horizontally** — matching the IDE.
-- [x] Verified side-by-side vs `:6009` at the same viewport: nesting, function
-      chips, horizontal operands, operator words, group containers all match.
-      12/12 expression play tests; full suite 188/189 (pre-existing flaky only).
-
-### Phase 3d — operand layout: flex-fill (the real refinement) — DONE (2026-06-10)
-- [x] Root cause of "old / crooked / unaligned": operands sat in **fixed
-      220px boxes**, left-bunched with dead gaps, fields `fluid={false}`.
-      Compared at full width (`:6007`) the IDE spreads operands across the
-      whole row.
-- [x] Reworked: flattened the operand row (label_before / column / label_after
-      as flat siblings), each operand column `flex: 1 1 200px` (nested
-      expressions `1 1 100%`), field now `fluid` (fills its column). Result:
-      operands fill the row evenly with roomy fields, operator words centered
-      in the gap — matches the IDE's refinement. eslint clean, 12/12 play
-      tests, full suite 188/189.
-
-### Phase 4 — Explain + validation + polish — DONE pending verify
-- [x] Explain panel renders `useRenderExpression` output (done since Phase 1;
-      works in Visual + Text modes)
-- [x] **Expression-aware validation** — `_validateField` short-circuits on
-      `field.isFunction`: an expression with a chosen `exp` is valid; an
-      empty one reports "Expression operation is required". FormEngine passes
-      `isFunction` from the field value's `is_expression`. Verified live (the
-      FormEngine invalid banner clears for a valid expression) + a story guard.
-- [x] Full pass: prod typecheck, eslint, jest 267/267, form play tests pass
-      (only pre-existing `OnValidityChange`)
-- [ ] When the server transport lands, swap `useRenderExpression` impl
-      (one file) — awaiting the meeting decision
-- Deferred (optional): fullscreen/focused-editing affordance
-
-### Checks (final)
-- [x] Typecheck, lint, jest, storybook play tests pass
-- [ ] **STOP — user verifies in browser** before commit
-
-## Open decisions
-
-- **Explain transport** — REST `?action=render-expression` vs LSP
-  `dpql/explain` vs interim approximation. Server ask in
-  `/tmp/render-expression-exposure.md`; user raising it at the
-  2026-06-10 meeting. Until then: interim seam impl.
-- **DPQL text-mode context** (Phase 2) — how to bind a generic field
-  expression's "fields" when there's no data provider.
-
-## Out of scope
-
-- `server_expression_handling` filtering nuances; AI-assist button;
-  wiring concrete consumers (alert-rule editor is qorus-ide-side).

package/.tasks/FORM_ENGINE_OPTIONS_SYNC.md

@@ -1,95 +0,0 @@
-# TASK — FormEngine ↔ IDE Options sync (fetch layer)
-
-**Status:** Phases 1+2+3 DONE pending user verify (uncommitted) — 6/6 new play tests. Follows the verbatim-port method
-(`design/EXPRESSION_BUILDER_REPORT_STRATEGY.md`); audit findings below.
-
-## Research summary (2026-06-10)
-
-FormEngine was extracted from the IDE's `Options` (`systemOptions.tsx`) on
-**2026-03-18**; both evolved independently since (8 IDE commits). Full diff
-audit found the drift is small — loop fixes, help text, dependencies,
-operators UI, required groups all present on both sides; helpers materially
-identical. The REAL gaps:
-
-- **IDE-only:** the remote-fetch layer — `url`/`customUrl` (options schema
-  from `/options/{url}` or a custom endpoint), `onOptionsLoaded` (declared in
-  FormEngine but never called), `operatorsUrl` (operator schema; without it
-  FormEngine's operators UI is dead code — `operators` state is hardcoded
-  `undefined` with no setter). Plus a two-line `stretch` layout for
-  `tool-catalog` options (commit 8e6b7781 — the "external data support" in
-  that commit message was ApiKeys/OAuth2 views, NOT Options).
-- **reqraft-only (IDE behind us):** detailed `onValidityChange` payload,
-  any/auto type preservation, expression value unwrapping.
-- **Tests:** the IDE has **zero** coverage for the fetch features (all 30
-  Options stories pass fixture schemas as props) — so there is nothing to
-  copy; we write new coverage with a `window.fetch` mock (the dpqlMockLsp
-  pattern).
-
-## Decisions
-
-- Port the fetch trio verbatim-adapted: IDE `fetchData` → reqraft `query()`
-  (same `{data, ok, error}` response shape; reqraft prepends
-  `${instance}api/latest/` like the IDE's `${apiHost}api/latest/`).
-  Normalise the IDE's leading-slash URLs (avoid `api/latest//options/...`).
-- Port the IDE's loading-skeleton branch + the `stretch`/`STRECHABLE_TYPES`
-  two-liner.
-- **Skip `allowAi`** — IDE-only AI infra; seam pattern exists (expression
-  builder's `extraActions`) if ever needed.
-- **Skip `interfaceContext`** — only meaningful through the IDE's template
-  fetching, which reqraft's `useTemplates` deliberately stubs.
-
-## Phases
-
-### Phase 1 — fetch layer in FormEngine — DONE (2026-06-10)
-- [x] Props: `url`, `customUrl`, `operatorsUrl`; `onOptionsLoaded` now fires
-      (was declared-but-dead since extraction)
-- [x] Mount fetch (seeds `fixOptions(value, schema)`), refetch-on-change
-      (clears value, seeds `fixOptions({}, schema)`) — IDE semantics verbatim
-      (`query()` in place of the IDE's `fetchData`; same response shape)
-- [x] Operators fetch (own state — `setOperators` finally exists; the loading
-      gate waits for BOTH schemas when both urls are set, per the IDE)
-- [x] Skeleton branch extended with the remote gates; `stretch` +
-      `STRECHABLE_TYPES` (tool-catalog) ported
-- [x] GATE: prod typecheck + eslint clean, jest 267/267
-
-### Phase 2 — tests (new coverage; the IDE has none) — DONE (2026-06-10)
-- [x] `mockFetchRoutes` helper in `FormEngineRemote.stories.tsx` — patches
-      `window.fetch`, serves matching routes, passes the rest through;
-      distinct URLs per story (the `query()` GET cache is 5 min)
-- [x] Stories (4): `SchemaFromUrl` (3 options render, preselected default
-      seeded, `onOptionsLoaded` called with the schema), `OperatorsFromUrl`
-      (operator select resolves `Like` + the WHERE/IS tags — the previously
-      DEAD operators UI working end-to-end, verified by screenshot),
-      `FetchFailure` (500 → "No options available", no crash),
-      `UrlChangeResetsValue` (url switch clears the old value, new schema
-      seeds)
-- [x] GATE: **4/4 play tests** (repeatedly, incl. after a flake-hardening
-      timeout bump); jest 267/267; FormEngine.stories 24/24 isolated;
-      useWebSocket 15/15, Object 11/11 isolated.
-### Phase 3 — seam parity ("item 5") — DONE (2026-06-10)
-- [x] `SelectFormField` now **rest-spreads** caller props into all four render
-      branches (IDE parity; computed `flat`/`intent`/`size`/handlers stay
-      authoritative by coming after the spread) — closes the
-      "every pass-through prop needs hand-forwarding" seam for good.
-- [x] **`optionActions`** on FormEngine — per-option injected hover actions
-      (array or factory receiving `{name, schema, value}`), prepended where
-      the IDE renders its `allowAi` AiAssistanceAction. Same pattern as the
-      builder's `extraActions`; reqraft stays AI-free.
-- [x] **`useTemplates` opt-in fetching** — ported the IDE's
-      `system/getContextData` fetch + `buildTemplates` (FSM `states` param
-      seamed out; structural payload types), gated on a NEW explicit
-      `interfaceContext` (FormEngine prop, threaded through). Without a
-      context: the original stub behavior, byte-for-byte — no existing
-      consumer starts fetching.
-- [x] Coverage: `InjectedOptionActions` (hover → injected action) and
-      `TemplatesFromContext` (context fetch hit + loading gate) stories.
-      **6/6** FormEngineRemote, **17/17** expressions, 23/24 FormEngine
-      (the 1 = pre-existing `OnValidityChange` flake), jest 267/267,
-      typecheck + eslint clean.
-
-- Verification caveat (environment, not code): full-suite runs tonight flap
-      on LIVE-hq stories (CurrentUser, Log/WS) — same suites pass and fail
-      across runs with zero code delta (CurrentUser failed 2/2 then passed
-      2/2 a minute later), and the same full suite was green 194/194 earlier
-      today. hq intermittency + dev-server wear; re-run the full suite on a
-      fresh server when hq is stable for the final pre-commit pass.

package/.tasks/FORM_FIELD_EXTRAS.md

@@ -1,72 +0,0 @@
-# Form field extras — bucket-1 portable fields
-
-Follow-on to the field migration (see
-[`design/FORM_ENGINE_FIELD_MIGRATION.md`](../design/FORM_ENGINE_FIELD_MIGRATION.md)).
-Picks off the **cleanly portable** bucket-1 fields one at a time, and
-records an evidence-based triage of the ones that are *not* a quick port
-(so "deferred" stays honest, not vague).
-
-**Status:** `byte-size`, `url`, `multi-select`/`select-array` done pending
-user verify — uncommitted, 2026-06-09. 3 new Field stories pass
-(`test-storybook`), jest 267/267, prod typecheck + eslint clean. The rest
-remain triaged below.
-
-## Done this pass
-
-| Type | What | How |
-|---|---|---|
-| `byte-size` | **new** `byte-size/ByteSize.tsx` | amount `NumberFormField` + unit `SelectFormField` (KiB/MiB/GiB/TiB); value is the combined `"512MiB"` string. `splitByteSize` reimplemented (3 lines). No app deps. |
-| `url` | **new** `url/Url.tsx` | thin `ReqoreInput type="url"` + `new URL()` validity hint. Built fresh — qorus-ide's `URLField` imports `context/init` (app-coupled); a URL field doesn't need it. |
-| `multi-select` / `select-array` | wired existing `MultiSelectFormField` | reqraft already had the component; added a `renderField` branch **before** the single-value `allowed_values` short-circuit (multi-select owns its options + is always an array), and a `renderAllowedValues` guard so the picker isn't doubled. |
-
-> **Superseded 2026-06-10 (FIELD_STACK_REPORT phase 3):** the `url` and
-> `byte-size` rows above describe the original from-scratch versions — both
-> were replaced by verbatim IDE ports (url: protocol select + `://` + address;
-> byte-size: KiB/MiB only). Descriptions kept for history; see
-> FIELD_STACK_REPORT for the current implementations.
-
-Surface area: `src/components/form/fields/byte-size/ByteSize.tsx` (new),
-`src/components/form/fields/url/Url.tsx` (new),
-`src/components/form/fields/Field.tsx` (3 cases + multi-select branch +
-allowed-values guard + 2 imports), `src/components/form/index.tsx`
-(exports), `src/components/form/fields/Field.stories.tsx` (`ByteSize`,
-`Url`, `MultiSelect` stories).
-
-## Triage of the remaining bucket-1 / bucket-2 fields
-
-Evidence is the qorus-ide source's size + its outside-`Field/` imports
-(the portability tell — app context can't come to the toolkit).
-
-| Type | qorus-ide source | Blocker / verdict |
-|---|---|---|
-| `option_hash` | `optionHash.tsx` (129) | imports `../../App`, `context/init`, `withTextContext` — **app-coupled**; needs a rewrite, not a port. |
-| `array-of-pairs` | `multiPair.tsx` (207) | imports `../../App`, `withTextContext` — app-coupled; moderate rewrite. |
-| `enum`-as-radio | `radioField.tsx` (127) | imports `../../App`, two HOCs, **image assets**. Reqraft already has `RadioGroup`; `enum` currently routes to `Select`. Switching is a UX decision, low value. |
-| `code-editor` | `CodeEditor/` | **Monaco** — a heavy new runtime dependency reqraft doesn't carry. A dependency call for the maintainers, not something to add unilaterally. |
-| `data-provider` | `connectors/` | connection-management + **OAuth2** + instance context — squarely bucket-3 (app integration), excluded by the design. |
-| `options` | `systemOptions.tsx` (1500+) | this *is* `FormEngine`, which reqraft already has. Exposing a `type: 'options'` wrapper is possible but needs option-set value-shape decisions; defer until a consumer needs it. |
-| `tool-catalog` | `toolCatalog/` (434) | server-driven feature editor (All/Some/None per source). Schema-editor-sized; own task. |
-| `active-windows` | `activeWindows/` (290) | server-driven cron-window editor. Own task. |
-| `collection-documents` | `collectionDocuments.tsx` (347) | managed uploads tied to AI-collection APIs. Own task. |
-| `test-cases` | `testCases/` (**1469**) | large server-driven editor (`/tests/step-kinds`, `/tests/assertions`). Its own multi-day task. |
-| `processor-mappings` | `ProcessorMappingPanel` | processor field-mapping editor; feature-specific. Own task. |
-
-**Principle:** small + dependency-free → done now; app-coupled or
-server-driven feature editors → their own task, pulled when a real
-consumer needs them so they can be verified against that use (the lesson
-from the schema editor's `ui_type` path).
-
-## Tasks
-
-- [x] `ByteSizeFormField` + `case 'byte-size'` + export + story
-- [x] `UrlFormField` + `case 'url'` + export + story
-- [x] `multi-select`/`select-array` branch + allowed-values guard + story
-- [x] Prod typecheck, eslint, jest (267), form play tests (only the
-      pre-existing `OnValidityChange` fails)
-- [ ] **STOP — user verifies in browser** before commit
-
-## Out of scope
-
-- Everything in the triage table marked "own task" / "app-coupled" /
-  "dependency call" — see the per-row verdicts above.
-- Bucket-3 interface selectors (excluded by the design doc).

package/.tasks/INDEX.md

@@ -1,111 +0,0 @@
-# SmartEditor task index
-
-One-glance status across every task file in `.tasks/`. **Update this
-file whenever a task transitions between statuses** (per the
-convention in `.claude/CLAUDE.md` § "Task & design docs workflow").
-
-## Release strategy
-
-**Revised 2026-05-26.** All `.tasks/` items below ship together as a
-single `0.10.0` release — no incremental `-beta` / `0.10.1` / `0.10.2`
-tags between. Rationale recorded in
-[`design/SMART_EDITOR_UX.md` § "Release strategy"](../design/SMART_EDITOR_UX.md):
-shipping incremental releases would mean cutting tags with no real
-consumer (qorus-ide alert-rule editor needs `alertPayloadContext`/`fsmContext`
-which are in the CONTEXT_AND_POLISH task, not in what was originally
-scoped for `0.10.0`). Better to land everything in one coherent
-release that an actual consumer can adopt.
-
-| Release | Status | Tag / sha |
-|---|---|---|
-| `0.9.0` | shipped | (current published) |
-| `0.10.0` | in progress | held until every task row below is `committed` |
-| `0.11.0` (tentative) | planned | Form engine field migration (separate workstream, see below) |
-
-## Tasks
-
-All tasks ship as part of `0.10.0`. Sequence below is the suggested
-implementation order; release happens after the last row commits.
-
-| Task | Status | Sequence rationale |
-|---|---|---|
-| [SMART_EDITOR_UX_POLISH](./SMART_EDITOR_UX_POLISH.md) | Phase 1 committed `6fb9bf3` · Phases 2–7 + 6 committed `a9b7f1b` and **verified in browser 2026-05-26** · Phase 8 (release prep) deferred to end of batch | The original UX batch. All design-doc items (1–7) shipped and verified; Phase 8's "bump version + push" step now happens after every follow-up below also lands. |
-| [SMART_EDITOR_CONTEXT_AND_POLISH](./SMART_EDITOR_CONTEXT_AND_POLISH.md) | items 1/2/4 committed (`5e94fd7`, `8e6fe0f`, `2a53a1a`); item 3 (README) deferred to release prep — **awaiting browser verification** | **1st** done |
-| [SMART_EDITOR_LSP_FEATURES](./SMART_EDITOR_LSP_FEATURES.md) | committed (`d54597e`) — **awaiting browser verification** | **2nd** done |
-| [QONSOLE_ASSIST_FEATURES](./QONSOLE_ASSIST_FEATURES.md) | committed (`189fe52`) — **awaiting browser verification** | **3rd** done |
-| [SMART_EDITOR_VISUAL_POLISH](./SMART_EDITOR_VISUAL_POLISH.md) | committed (`c006a79`) — **awaiting browser verification** | **4th** done |
-| [SMART_EDITOR_REVIEW_FOLLOWUPS](./SMART_EDITOR_REVIEW_FOLLOWUPS.md) | all 4 items done pending user verify (uncommitted) | **5th** — reviewer feedback on PR #62 (the `0.10.0` editor batch). |
-| `0.10.0` release prep (Phase 8 of UX_POLISH) | ready to start once all above complete | **6th (final)** — bump `package.json` to `0.10.0`; final README pass against the final API; tag; push. |
-
-## Form engine field migration (separate workstream)
-
-Closing the gap between Reqraft's form engine and qorus-ide's `Field`
-system. Design locked in
-[`design/FORM_ENGINE_FIELD_MIGRATION.md`](../design/FORM_ENGINE_FIELD_MIGRATION.md).
-Independent of the `0.10.0` SmartEditor batch above; targets a later
-minor (`0.11.0` tentative). Sequence is a hard dependency — phase 2
-needs phase 1.
-
-| Task | Status | Sequence rationale |
-|---|---|---|
-| [REQRAFT_AUTO_FIELD](./REQRAFT_AUTO_FIELD.md) | done pending user verify (uncommitted) — 8/8 play tests pass | **1st (keystone).** Type-picker + polymorphic dispatch. The schema editor and most of the remaining field gap render their fields through it. |
-| [SCHEMA_DEFINITION_FIELD](./SCHEMA_DEFINITION_FIELD.md) | done pending user verify (uncommitted, stacked on phase 1) — 8/8 play tests pass (2 added by FIELD_STACK phase 3, ReadOnly play restored in remediation) | **2nd.** DataSchema editor — catalogue-driven nested form on `FormEngine`. Its leaf adapter emits `auto`, so phase 1 must land first. |
-| [FORM_FIELD_EXTRAS](./FORM_FIELD_EXTRAS.md) | done pending user verify (uncommitted) — `byte-size` / `url` / `multi-select` landed; rest triaged | **3rd (ongoing).** Portable bucket-1 fields, picked off one at a time. Contains the evidence-based triage of what's not a quick port. |
-| [FORM_ENGINE_EXPRESSIONS](./FORM_ENGINE_EXPRESSIONS.md) | plumbing kept; **visual builder superseded by [EXPRESSION_BUILDER_REPORT](./EXPRESSION_BUILDER_REPORT.md)** | **4th (new feature, large).** Expression values in fields. The catalogue hook / FormEngine `is_expression` / DPQL bridge / validation are kept; the from-scratch visual builder is being replaced by a verbatim re-port. |
-| [EXPRESSION_BUILDER_REPORT](./EXPRESSION_BUILDER_REPORT.md) | Phases A+B+C done & verified (uncommitted) | **5th.** Verbatim re-port of the IDE ExpressionBuilder + field-stack parity + IDE tests (Filip's method: copy → 1:1 → improve), replacing the from-scratch builder. Design: `design/EXPRESSION_BUILDER_REPORT_STRATEGY.md`. |
-| [FORM_ENGINE_OPTIONS_SYNC](./FORM_ENGINE_OPTIONS_SYNC.md) | Phases 1+2+3 done pending user verify (uncommitted) — 6/6 new play tests | **6th.** Sync FormEngine with the IDE's Options drift since the 2026-03-18 extraction: the remote-fetch layer (`url`/`customUrl`/`onOptionsLoaded`/`operatorsUrl`) + `stretch`; new test coverage (the IDE has none for these). |
-| [LSP_SHARED_CONNECTION](./LSP_SHARED_CONNECTION.md) | done pending user verify (uncommitted) — multiplex proof test + 268/268 jest | **7th.** N editors = 1 LSP socket: shared per-endpoint connection multiplexed by document URI (matches the server's `uriToDpqlSession` design); answers Filip's 20-builders question. Client class renamed `LspClient` → `ReqraftLspClient` (the `Reqraft*` convention, 2026-06-11); API otherwise unchanged. |
-| [RENDER_EXPRESSION_LSP](./RENDER_EXPRESSION_LSP.md) | done pending user verify, branch `feature/render-expression-lsp` — server deployed + live-verified end-to-end 2026-06-10 (final gate 201/201 storybook, 268/268 jest, browser proof of server rendering) | **8th.** Server "Explain" over the LSP: new `dpql/renderExpression` in `QorusLspWebSocketHandler.qc` + `useRenderExpression` swap. Closes EXPRESSION_BUILDER_REPORT Phase D. Design: `design/RENDER_EXPRESSION_TRANSPORT.md`. |
-| [FIELD_STACK_REPORT](./FIELD_STACK_REPORT.md) | done pending user verify (uncommitted) — final gate 234/234 storybook (incl. 37 ported Template/Auto plays), 268/268 jest | **9th.** Verbatim re-port of the IDE's `Field/template.tsx` + `Field/auto.tsx` (the operand layer under the ported builder) + their IDE stories, replacing the from-scratch TemplateField/AutoFormField; closes the `isFunction` operand-flow gap (behavioral difference #4 in EXPRESSION_BUILDER_REPORT). Supersedes the from-scratch implementation from REQRAFT_AUTO_FIELD (its stories/assertions are kept and reconciled). |
-| [BRANCH_REMEDIATION](./BRANCH_REMEDIATION.md) | done pending user verify (uncommitted) — jest 268→356, plays 241→255, lint/tsc clean | **10th.** Burn-down of the 2026-06-10 full-branch audit: tests/stories/API/comment-voice remediation + behavior fixes (Select modal-click collapse, edge-triggered Url, ReqraftLspClient rename, CI auth wiring). Ledger: gitignored `SWEEP.local.md`. |
-| [STORY_NETWORK_ISOLATION](./STORY_NETWORK_ISOLATION.md) | done pending user verify, folded into `fix/per-field-templates` (PR #115), not pushed | **Standalone.** ExpressionField stories failed off-network on `develop` (passed in CI only because CI reaches the live instance): LSP connections leaking across stories, an unhandled socket give-up in `LspSharedConnection`, unmocked background requests, and mock URLs hard-coded to the hq host. |
-| [TEMPLATE_TOKEN_GRAMMAR](./TEMPLATE_TOKEN_GRAMMAR.md) | committed `0ef167c`, in PR (branch `bugfix/template-token-grammar`) — user-verified 2026-08-27; 13/13 unit, 21/21 template plays, qlip PNGs read | **11th (bugfix).** Braced `$data:{…}` state-output refs rehydrate as raw text: canonical token grammar + strict whole-token check + TemplateField renders whole-token values as the picker chip. Twin fix applied in qorus-ide the same day; ships as a normal per-PR patch release (`0.10.37`). |
-| [EXPRESSION_FIELD_SEAMS](./EXPRESSION_FIELD_SEAMS.md) | committed `f5fb2ee`, pushed (branch `bugfix/116_expression-field-seams`) — user-verified 2026-09-07; PR not yet opened | **12th (bugfix).** The `extraActions` seam stopped one hop short of the `ExpressionField` shell hosts mount, and the builder's operand `TemplateField` mounts dropped it again — qorus-ide lost its AI-assist button on every expression card when it adopted the shell ([#116](https://github.com/qoretechnologies/toolkit-react/issues/116)). Forward it through both hops; pin with a nested-operand story; ships as a per-PR patch (`0.10.51`; re-based onto `4860e6e` after #114 published 0.10.50). |
-| [EXPORT_EXPRESSION_BUILDER](./EXPORT_EXPRESSION_BUILDER.md) | committed `0eea9e7`, pushed, PR open (branch `feature/119_export-expression-builder`) | **13th (surface fix).** The barrel exported the `ExpressionField` shell but not the bare `ExpressionBuilder`, so a host mirroring the library's root/nested rule had to deep-import `dist/…/builder` ([#119](https://github.com/qoretechnologies/toolkit-react/issues/119)). Export + a barrel test; ships as `0.10.53`. |
-
-Not yet scoped (acknowledged in the design doc): the remaining
-`ui_type` editors (`tool-catalog`, `processor-mappings`, `test-cases`,
-`active-windows`) and the rest of bucket-1 fields (`data-provider`,
-`options`, `byte-size`, `url`, `code-editor`, …). They get their own
-task files when picked up.
-
-## Status vocabulary
-
-- **ready to start** — task file exists, design locked, work hasn't begun
-- **in progress, branch `<name>`** — branch created, code being written
-- **done pending user verify** — code complete, tests pass, sitting uncommitted or on a branch awaiting human in-browser check
-- **committed `<sha>`** — landed on the working branch
-- **shipped `<tag>`** — released (tag pushed, npm published)
-- **paused** — intentionally on hold (note the reason in the task file's "Status" header)
-- **superseded by `<file>`** — work absorbed into another task; the file remains as history
-
-## How to use this dashboard
-
-1. **Picking what to work on next** — read top to bottom. Pick the
-   highest-up task whose status is `ready to start` or
-   `in progress` and whose sequence rationale doesn't put you out of
-   order.
-2. **Picking what to verify before commit** — when a task is "done
-   pending user verify", check the **task file** (not this dashboard)
-   for the verification checklist, plus the per-batch
-   `VERIFY.local.md` (untracked) if one exists.
-3. **Adding a new task** — write the planning doc first
-   (`.tasks/<NAME>.md`), then add a row here.
-4. **Closing a task** — flip its status here AND in the task file's
-   `**Status:**` header. Don't delete completed task files;
-   they become history.
-
-## Out-of-band trackers
-
-These don't live in `.tasks/`:
-
-- **Reqore-side bugs** — spawn-task chips (see chat history) for
-  upstream PRs needed by SmartEditor (Textarea null-check,
-  renderElement deps fix). Track via your normal PR workflow on the
-  Reqore repo, not here.
-- **qorus-ide integration** — `qorus-frontend/qorus-ide/.tasks/ALERT_RULES_AND_SILENCES.md`
-  is the consumer-side task that depends on
-  `SMART_EDITOR_CONTEXT_AND_POLISH` item 4. Update the qorus-ide
-  task when item 4 ships so the qorus-ide consumer can unblock.
-- **Per-batch verification** — `VERIFY.local.md` (untracked). One
-  per work batch — overwrite the previous when starting a new batch.

package/.tasks/LSP_SHARED_CONNECTION.md

@@ -1,75 +0,0 @@
-# TASK — Shared LSP connection (N editors = 1 socket)
-
-**Status:** DONE pending user verify (uncommitted, 2026-06-10).
-
-## Why
-
-Filip's call question (2026-06-10): with 20 expression builders open, "would
-it have different connections or how would you associate the message?"
-Research findings:
-- The server is BUILT for multiplexing: `QorusLspWebSocketHandler.qc:531`
-  keeps `uriToDpqlSession` — a per-connection map of URI → DPQL session
-  (created on didOpen :1004, removed on didClose :1213); `dpql/setContext`
-  takes a `uri` param. One connection serves N documents.
-- reqraft's old `LspClient` opened one isolated socket PER editor
-  (`pooled: false`) — 20 Text-mode editors = 20 sockets + 20 handshakes. The
-  old comment's rationale ("sharing conflates diagnostics and correlation")
-  was wrong: JSON-RPC ids correlate responses; `publishDiagnostics` carries
-  the uri.
-- qorus-ide's own Creator WS already uses the shared-socket + `request_id`
-  pattern (`helpers/functions.tsx:405`).
-
-## What changed (`src/utils/lspClient.ts`)
-
-- New internal `LspSharedConnection`: one WebSocket + one `initialize` per
-  endpoint URL (registry; first acquirer's reconnect config wins). Owns the
-  id counter + pending map; routes responses by id, `publishDiagnostics` by
-  `params.uri`, other notifications to every registered client. Closes when
-  the last client releases it.
-- `LspClient` is now a per-document façade with an UNCHANGED public API
-  (`useDpqlSession`/`useLspSession`/`DpqlEditor` untouched);
-  `capabilities`/`semanticTokensLegend` became getters proxying the
-  connection. Reconnect: the connection re-initializes once, then each client
-  re-sends its own `didOpen` (`_resyncAfterReconnect`).
-- Editor URIs were already unique per instance
-  (`${languageId}://session/${counter}` in `useLspSession`) — collision-safe.
-- Test hook `_resetSharedLspConnectionsForTests()` (mirrors the websocket
-  manager reset in jest).
-- Answer to Filip, in one line: **one connection per endpoint, sessions
-  multiplexed by URI; requests correlate by JSON-RPC id, diagnostics by uri —
-  matching the server's design and the IDE's request_id pattern.**
-
-## Consumer audit (what the refactor touches)
-
-- **Exactly one `LspClient` creation site** in the library:
-  `useLspSession.ts:110` — SmartEditor, DpqlEditor, Qonsole and the
-  expressions Text mode ALL flow through it; its API was preserved verbatim.
-- `capabilities`/`semanticTokensLegend` readers (`useLspSession`,
-  `useLspSemanticTokens`, `useLspSignatureHelp`) only READ them → the
-  field→getter change is API-identical.
-- NEW behavior introduced: different `languageId`s (dpql + qonsole) can now
-  share one connection (same `'lsp'` URL). Verified safe — the server
-  resolves language PER DOCUMENT (`detectLanguage(textDoc.languageId, uri)`
-  at didOpen, `uriToLanguage{uri}`/`uriToSource{uri}` maps), and verified
-  LIVE (below).
-
-## Verification
-
-- New jest proof: two clients → `connections === 1`, one `initialize`,
-  diagnostics delivered only to the matching uri, releasing one client keeps
-  the socket for the other. lspClient suite **15/15**; full jest **268/268**.
-- **LIVE e2e against the real server** (`lsp-multiplex-e2e.local.mjs`,
-  gitignored — one socket to `wss://localhost:8012/lsp`): initialize once;
-  two DPQL docs + one Qonsole doc coexist; interleaved per-document
-  `semanticTokens` requests return DISTINCT results; diagnostics route per
-  uri (invalid doc got 3, valid got 0, qonsole got 0); closing doc A leaves
-  doc B fully functional. `E2E OK`.
-- **Browser-level proof**: new `DpqlEditor › TwoEditorsOneConnection` story —
-  two mounted editors, the mock server accepts exactly ONE connection, one
-  `initialize`, two `didOpen`s with distinct URIs. dpqlEditor suite **17/17**.
-- Storybook (fresh server): smartEditor+qonsole **8/8**, expressions
-  **17/17**. Typecheck + eslint clean.
-- Known semantics note: with >1 client on a connection, a disconnecting
-  client's in-flight requests are left to their timeouts (results ignored
-  post-unmount); the sole-client case still rejects them via the connection
-  close — covered by the existing test.