Refine test lanes and expand targeted coverage

.changeset/ai-fix-repeated-placeholders.md

@@ -0,0 +1,5 @@
+---
+'@platejs/ai': patch
+---
+
+- Fix `replacePlaceholders` replacing only the first `{prompt}` and markdown placeholder occurrence in AI prompt templates

.changeset/docx-fix-rtf-image-matching.md

@@ -0,0 +1,5 @@
+---
+'@platejs/docx': patch
+---
+
+- Fix RTF image extraction matching control words like `shp` inside longer tokens such as `shppict`

.changeset/layout-fix-column-group-normalization.md

@@ -0,0 +1,5 @@
+---
+'@platejs/layout': patch
+---
+
+- Fix invalid column group normalization preserving wrapped paragraph content instead of dropping it

.changeset/suggestion-fix-metadata-lookups.md

@@ -0,0 +1,5 @@
+---
+'@platejs/suggestion': patch
+---
+
+- Fix suggestion metadata lookups using the actual per-suggestion keys and IDs for active descriptions, node matching, and line-break detection

.claude/docs/plans/2026-03-22-coverage-priority-map-post-yjs.md

@@ -0,0 +1,224 @@
+---
+title: Coverage Priority Map Post Yjs Merge
+type: testing
+date: 2026-03-22
+status: completed
+---
+
+# Coverage Priority Map Post Yjs Merge
+
+## Goal
+
+Run fresh repo coverage, score every `packages/*/src/**` file for non-React unit-test value, and sync the next recommendations against:
+
+- March 6, 2026 cleanup strategy
+- March 9, 2026 excellence plan
+- March 17, 2026 execution passes
+- the now-merged Yjs suite work
+
+## Coverage Run
+
+- Command:
+  - `bun test --coverage --coverage-reporter=lcov --coverage-dir=/Users/zbeyens/git/plate/.coverage-repo-2026-03-22b --reporter=dots`
+- Result:
+  - `2557 pass`
+  - `0 fail`
+  - `444 files`
+  - `4.25s`
+- Artifact:
+  - [lcov.info](/Users/zbeyens/git/plate/.coverage-repo-2026-03-22b/lcov.info)
+
+## Sync From Completed Work
+
+- March 14, 2026:
+  - `markdown` already got the worthwhile helper and fallback pass.
+- March 15, 2026:
+  - `table` already got the merge, sizing, and selection helper pass.
+- March 17, 2026:
+  - `selection`, `docx-io`, and the focused non-React `core` lane already got follow-up work.
+- March 22, 2026:
+  - `yjs` just got the fast suite, slow collaboration lane, and follow-up upstream-inspired cases.
+
+Strong take: the raw matrix still slightly over-scores `core` and `table` because they are big and still have uncovered leftovers. I would not reopen them next. The actual recommendation order below is the one to follow.
+
+## Scoring Rules
+
+- Scope is every runtime file under `packages/*/src/**`.
+- `/react` files score `0` by design.
+- Barrels, declaration files, and type-only files score `0`.
+- Declarative rule tables under `autoformat/rules/**` score `0`.
+- Recent package passes get penalized so we do not recommend the same work twice.
+- Scores favor deterministic transforms, parsers, serializers, queries, provider logic, plugin contracts, and pure helpers with meaningful uncovered behavior.
+- UI-skewed packages get pushed down even when they are uncovered.
+
+## My Order
+
+### 1. `list-classic`
+
+Best next package.
+
+- Why:
+  - best remaining mix of deterministic query and transform seams
+  - high-value editor behavior, not coverage cosplay
+  - sits exactly where the older plan already wanted the next serious spend
+- Start here:
+  - [getTodoListItemEntry.ts](/Users/zbeyens/git/plate/packages/list-classic/src/lib/queries/getTodoListItemEntry.ts)
+  - [getHighestEmptyList.ts](/Users/zbeyens/git/plate/packages/list-classic/src/lib/queries/getHighestEmptyList.ts)
+  - [withDeleteFragmentList.ts](/Users/zbeyens/git/plate/packages/list-classic/src/lib/withDeleteFragmentList.ts)
+  - [withInsertBreakList.ts](/Users/zbeyens/git/plate/packages/list-classic/src/lib/withInsertBreakList.ts)
+  - [moveListSiblingsAfterCursor.ts](/Users/zbeyens/git/plate/packages/list-classic/src/lib/transforms/moveListSiblingsAfterCursor.ts)
+
+### 2. `suggestion`
+
+- Why:
+  - deterministic diff and suggestion-transform behavior
+  - strong value without any `/react` detour
+  - better next seam quality than reopening `core` leftovers
+- Start here:
+  - [diffToSuggestions.ts](/Users/zbeyens/git/plate/packages/suggestion/src/lib/diffToSuggestions.ts)
+  - [withSuggestion.ts](/Users/zbeyens/git/plate/packages/suggestion/src/lib/withSuggestion.ts)
+  - [removeNodesSuggestion.ts](/Users/zbeyens/git/plate/packages/suggestion/src/lib/transforms/removeNodesSuggestion.ts)
+  - [getSuggestionNodeEntries.ts](/Users/zbeyens/git/plate/packages/suggestion/src/lib/utils/getSuggestionNodeEntries.ts)
+  - [findSuggestionNode.ts](/Users/zbeyens/git/plate/packages/suggestion/src/lib/queries/findSuggestionNode.ts)
+
+### 3. `docx`
+
+- Why:
+  - very unit-testable cleaner and import seams
+  - big user-facing value without dragging app wiring into package tests
+  - cleaner next move than `docx-io`, which already got its pure-helper pass
+- Start here:
+  - [cleanDocxImageElements.ts](/Users/zbeyens/git/plate/packages/docx/src/lib/docx-cleaner/utils/cleanDocxImageElements.ts)
+  - [getRtfImageHex.ts](/Users/zbeyens/git/plate/packages/docx/src/lib/docx-cleaner/utils/getRtfImageHex.ts)
+  - [getRtfImagesMap.ts](/Users/zbeyens/git/plate/packages/docx/src/lib/docx-cleaner/utils/getRtfImagesMap.ts)
+  - [getRtfImagesByType.ts](/Users/zbeyens/git/plate/packages/docx/src/lib/docx-cleaner/utils/getRtfImagesByType.ts)
+  - [getRtfImageMimeType.ts](/Users/zbeyens/git/plate/packages/docx/src/lib/docx-cleaner/utils/getRtfImageMimeType.ts)
+
+### 4. `ai`
+
+- Why:
+  - the value here is in pure transforms and prompt helpers, not model fakery
+  - coverage is still brutally low in the exact deterministic seams that matter
+  - worth doing after `suggestion` and `docx`, not before them
+- Start here:
+  - [withAIBatch.ts](/Users/zbeyens/git/plate/packages/ai/src/lib/transforms/withAIBatch.ts)
+  - [getEditorPrompt.ts](/Users/zbeyens/git/plate/packages/ai/src/lib/utils/getEditorPrompt.ts)
+  - [insertAINodes.ts](/Users/zbeyens/git/plate/packages/ai/src/lib/transforms/insertAINodes.ts)
+  - [replacePlaceholders.ts](/Users/zbeyens/git/plate/packages/ai/src/lib/utils/replacePlaceholders.ts)
+  - [undoAI.ts](/Users/zbeyens/git/plate/packages/ai/src/lib/transforms/undoAI.ts)
+
+### 5. `layout`
+
+- Why:
+  - still mostly deterministic column operations
+  - real regression surface
+  - no `/react` requirement
+- Start here:
+  - [resizeColumn.ts](/Users/zbeyens/git/plate/packages/layout/src/lib/transforms/resizeColumn.ts)
+  - [moveMiddleColumn.ts](/Users/zbeyens/git/plate/packages/layout/src/lib/transforms/moveMiddleColumn.ts)
+  - [insertColumnGroup.ts](/Users/zbeyens/git/plate/packages/layout/src/lib/transforms/insertColumnGroup.ts)
+  - [withColumn.ts](/Users/zbeyens/git/plate/packages/layout/src/lib/withColumn.ts)
+  - [insertColumn.ts](/Users/zbeyens/git/plate/packages/layout/src/lib/transforms/insertColumn.ts)
+
+### 6. `list`
+
+- Why:
+  - still has clean query and toggle seams left
+  - smaller and cheaper than the packages above
+  - good follow-up once `list-classic` is moving
+- Start here:
+  - [getSiblingListStyleType.ts](/Users/zbeyens/git/plate/packages/list/src/lib/queries/getSiblingListStyleType.ts)
+  - [toggleListByPath.ts](/Users/zbeyens/git/plate/packages/list/src/lib/transforms/toggleListByPath.ts)
+  - [someList.ts](/Users/zbeyens/git/plate/packages/list/src/lib/queries/someList.ts)
+  - [someTodoList.ts](/Users/zbeyens/git/plate/packages/list/src/lib/queries/someTodoList.ts)
+  - [setListNode.ts](/Users/zbeyens/git/plate/packages/list/src/lib/transforms/setListNode.ts)
+
+### 7. `code-block`
+
+- Why:
+  - remaining non-React logic is still real
+  - cleaner than reopening `core` or `table`
+- Start here:
+  - [withCodeBlock.ts](/Users/zbeyens/git/plate/packages/code-block/src/lib/withCodeBlock.ts)
+  - [formatter.ts](/Users/zbeyens/git/plate/packages/code-block/src/lib/formatter/formatter.ts)
+  - [withInsertDataCodeBlock.ts](/Users/zbeyens/git/plate/packages/code-block/src/lib/withInsertDataCodeBlock.ts)
+  - [setCodeBlockToDecorations.ts](/Users/zbeyens/git/plate/packages/code-block/src/lib/setCodeBlockToDecorations.ts)
+
+### 8. `autoformat`
+
+Surgical only.
+
+- Worth testing:
+  - [AutoformatPlugin.ts](/Users/zbeyens/git/plate/packages/autoformat/src/lib/AutoformatPlugin.ts)
+  - [autoformatBlock.ts](/Users/zbeyens/git/plate/packages/autoformat/src/lib/transforms/autoformatBlock.ts)
+  - [autoformatMark.ts](/Users/zbeyens/git/plate/packages/autoformat/src/lib/transforms/autoformatMark.ts)
+  - [autoformatText.ts](/Users/zbeyens/git/plate/packages/autoformat/src/lib/transforms/autoformatText.ts)
+- Not worth the time:
+  - rule-table files under `packages/autoformat/src/lib/rules/**`
+
+### 9. `csv`
+
+Tiny but sharp.
+
+- Start here:
+  - [deserializeCsv.ts](/Users/zbeyens/git/plate/packages/csv/src/lib/deserializer/utils/deserializeCsv.ts)
+  - [CsvPlugin.ts](/Users/zbeyens/git/plate/packages/csv/src/lib/CsvPlugin.ts)
+
+### 10. `comment`
+
+Good file-level value, but not a bigger campaign than the packages above.
+
+- Start here:
+  - [BaseCommentPlugin.ts](/Users/zbeyens/git/plate/packages/comment/src/lib/BaseCommentPlugin.ts)
+  - [withComments.ts](/Users/zbeyens/git/plate/packages/comment/src/lib/withComments.ts)
+  - [getCommentKeys.ts](/Users/zbeyens/git/plate/packages/comment/src/lib/utils/getCommentKeys.ts)
+  - [getCommentCount.ts](/Users/zbeyens/git/plate/packages/comment/src/lib/utils/getCommentCount.ts)
+
+## Not Next
+
+- `yjs`
+  - just landed a serious fast plus slow pass
+  - current matrix score is `0`
+- `media`
+  - current non-`/react` matrix score is `0`
+  - the old March 17 recommendation is stale here
+- `markdown`
+  - down to one score-1 leftover on [defaultRules.ts](/Users/zbeyens/git/plate/packages/markdown/src/lib/rules/defaultRules.ts)
+  - not worth the next cycle
+- `selection`
+  - only one real leftover file still registers
+  - March 17 already consumed the good internal seams
+- `docx-io`
+  - pure-helper pass already happened on March 17
+  - remaining debt is lower-value export/app crossover
+- `table`
+  - still has leftovers, especially [withSetFragmentDataTable.ts](/Users/zbeyens/git/plate/packages/table/src/lib/withSetFragmentDataTable.ts)
+  - still not worth reopening right now
+- `core`
+  - still has low-coverage leftovers
+  - most of them are not as clean or as urgent as the packages above
+
+## Ranking Snapshot
+
+This is my actual recommendation order, not just the raw heuristic sort.
+
+| Rank | Package | Score | Take |
+| --- | --- | ---: | --- |
+| 1 | `list-classic` | 49 | Best next deterministic editor-behavior lane |
+| 2 | `suggestion` | 46 | Rich transform/diff value without `/react` |
+| 3 | `docx` | 45 | Very unit-testable cleaner/import seams |
+| 4 | `ai` | 47 | Good pure transform/helper lane, avoid network theater |
+| 5 | `layout` | 44 | Strong column transform seams |
+| 6 | `list` | 41 | Cheap query/toggle follow-up |
+| 7 | `code-block` | 36 | Real non-React behavior still left |
+| 8 | `autoformat` | 25 | Surgical pass only |
+| 9 | `csv` | 19 | Two-file slice, clean ROI |
+| 10 | `comment` | 26 | Good file-level value, smaller overall win |
+
+## Artifacts
+
+- [lcov.info](/Users/zbeyens/git/plate/.coverage-repo-2026-03-22b/lcov.info)
+- [2026-03-22-coverage-priority-packages-post-yjs.tsv](/Users/zbeyens/git/plate/.claude/docs/plans/2026-03-22-coverage-priority-packages-post-yjs.tsv)
+- [2026-03-22-coverage-priority-files-post-yjs.tsv](/Users/zbeyens/git/plate/.claude/docs/plans/2026-03-22-coverage-priority-files-post-yjs.tsv)
+
+The TSVs are the exhaustive matrix for every package and file under `packages/*/src/**`. The markdown is the actual call on what to do next.

.claude/docs/plans/2026-03-23-ai-deterministic-coverage-pass.md

@@ -0,0 +1,33 @@
+# AI Deterministic Coverage Pass
+
+## Goal
+
+Add a narrow, high-value non-React coverage pass for `@platejs/ai` focused on the deterministic lib seam:
+
+- `withAIBatch.ts`
+- `getEditorPrompt.ts`
+- `insertAINodes.ts`
+- `replacePlaceholders.ts`
+- `undoAI.ts`
+- `removeAINodes.ts`
+- `removeAIMarks.ts`
+
+## Constraints
+
+- Fast-lane only.
+- No `/react`.
+- No model mocks, network tests, or streaming coverage.
+- Prefer tiny transform and helper contracts over wrapper coverage.
+
+## Slice
+
+1. Add pure stub-based specs for `withAIBatch` and `undoAI`.
+2. Add helper specs for prompt resolution and placeholder replacement.
+3. Add real-editor specs for insert/remove AI node transforms.
+4. Allow one small runtime fix if direct tests expose an actual bug.
+
+## Notes
+
+- Existing non-React `ai` coverage is almost empty outside `getMarkdown.spec.tsx`.
+- `replacePlaceholders.ts` currently uses `String.prototype.replace`, so repeated-placeholder coverage is likely to expose a real one-occurrence bug.
+- Execution confirmed that repeated placeholders were in fact broken; the helper only replaced the first occurrence of `{prompt}` and the same bug pattern applied to other placeholder tokens.

.claude/docs/plans/2026-03-23-autoformat-coverage-pass.md

@@ -0,0 +1,28 @@
+# Autoformat Coverage Pass
+
+## Goal
+
+Add a surgical, fast-lane-only coverage pass for `@platejs/autoformat` focused on the last worthwhile non-rule seams:
+
+- `AutoformatPlugin.ts`
+- `autoformatBlock.ts`
+- `isPreviousCharacterEmpty.ts`
+
+## Constraints
+
+- No `/react`.
+- No rule-table coverage.
+- Keep the slice in normal `*.spec.ts[x]`.
+- Prefer real editor behavior for plugin contracts and tiny stubs for pure transform helpers.
+
+## Slice
+
+1. Add a direct `AutoformatPlugin` spec for query veto, `insertTrigger`, and undo-on-delete array-match restore.
+2. Deepen `autoformatBlock.spec.ts` for `triggerAtBlockStart: false`, `allowSameTypeAbove`, and single-character match safety.
+3. Add one tiny `isPreviousCharacterEmpty` edge for empty-string ranges.
+4. Stop unless a direct test exposes a real bug.
+
+## Notes
+
+- Existing `withAutoformat` coverage already exercises most end-to-end mark and text flows.
+- The remaining value is in the plugin override seam and a few pure `autoformatBlock` branches.

.claude/docs/plans/2026-03-23-basic-styles-and-math-coverage-pass.md

@@ -0,0 +1,61 @@
+---
+title: Basic Styles And Math Coverage Pass
+type: testing
+date: 2026-03-23
+status: completed
+---
+
+# Goal
+
+Do two tiny non-React passes:
+
+- `@platejs/basic-styles`
+- `@platejs/math`
+
+# Scope
+
+## `basic-styles`
+
+- `setLineHeight`
+- `toUnitLess`
+- `BaseLineHeightPlugin`
+- `BaseFontColorPlugin`
+
+## `math`
+
+- `insertEquation`
+- `insertInlineEquation`
+- `BaseEquationPlugin`
+- `BaseInlineEquationPlugin`
+- `getEquationHtml`
+
+# Explicit Deferrals
+
+- `/react`
+- broad font plugin sweep
+- extra KaTeX behavior matrices
+
+# Verification Plan
+
+- targeted `bun test` on touched files
+- `bun test packages/basic-styles/src/lib packages/math/src/lib`
+- `pnpm test:profile -- --top 20 packages/basic-styles/src packages/math/src`
+- `pnpm test:slowest -- --top 20 packages/basic-styles/src packages/math/src`
+- `pnpm install`
+- `pnpm turbo build --filter=./packages/basic-styles --filter=./packages/math`
+- `pnpm turbo typecheck --filter=./packages/basic-styles --filter=./packages/math`
+- `pnpm lint:fix`
+
+# Result
+
+- added focused `basic-styles` specs for `setLineHeight`, `toUnitLess`, `BaseLineHeightPlugin`, and `BaseFontColorPlugin`
+- added focused `math` specs for `insertEquation`, `insertInlineEquation`, `BaseEquationPlugin`, `BaseInlineEquationPlugin`, and `getEquationHtml`
+- verification passed:
+  - targeted touched-file `bun test`
+  - `bun test packages/basic-styles/src/lib packages/math/src/lib`
+  - `pnpm test:profile -- --top 20 packages/basic-styles/src packages/math/src`
+  - `pnpm test:slowest -- --top 20 packages/basic-styles/src packages/math/src`
+  - `pnpm install`
+  - `pnpm turbo build --filter=./packages/basic-styles --filter=./packages/math`
+  - `pnpm turbo typecheck --filter=./packages/basic-styles --filter=./packages/math`
+  - `pnpm lint:fix`