split-rich-text-viewer-pages

Objective: Split the generated Evidence Kit rich-text viewer into two pages: a comparison page for apples-to-apples editor rows and a Slate v2 internals page for diagnostic rows that should not pretend to compare against other editors.

Goal plan: docs/plans/2026-05-28-split-rich-text-viewer-pages.md

Completion threshold: The split is complete when benchmarks/render-rich-text-viewer.mjs emits rich-text.html / rich-text-data.json for comparison rows and slate-v2-internals.html / slate-v2-internals-data.json for v2-only rows, generated checks pass, npm run check passes in benchmarks/editor, both served HTML routes return HTTP 200, served JSON proves there are no internal category leaks on the comparison page, and this plan passes the autogoal completion check.

Verification surface:

Generator: /Users/zbeyens/git/plate-2/benchmarks/editor/benchmarks/render-rich-text-viewer.mjs
Comparison output: /Users/zbeyens/git/plate-2/benchmarks/editor/docs/perf/rich-text.html
Comparison data: /Users/zbeyens/git/plate-2/benchmarks/editor/docs/perf/rich-text-data.json
Internals output: /Users/zbeyens/git/plate-2/benchmarks/editor/docs/perf/slate-v2-internals.html
Internals data: /Users/zbeyens/git/plate-2/benchmarks/editor/docs/perf/slate-v2-internals-data.json
Served routes: http://127.0.0.1:8765/rich-text.html and http://127.0.0.1:8765/slate-v2-internals.html

Constraints:

Keep the intentionally ugly js-framework-benchmark-style table.
Do not hide missing adapters on the comparison page.
Move only categories that are truly Slate v2 internal proof rows.
Keep one generator command so docs:perf and docs:perf:check stay simple.

Boundaries:

Source of truth: current Evidence Kit result benchmarks/results/rich-text-editors-latest.json.
Allowed edit scope: viewer generator, generated perf docs/data, and this plan.
External sources: N/A; local generated benchmark data settles the split.
Browser surface: static generated pages served from docs/perf on port 8765.
Tracker sync: N/A; no issue or PR requested.
Non-goals: adding new benchmark artifacts, changing the visual design, or renaming underlying benchmark categories.

Blocked condition: Work would stop only if generated rows could not be deterministically split, the generator could not check both page/data pairs, or the static server could not serve both routes. None of those happened.

Phase / pass table:

Phase	Status	Evidence
Source map	complete	Identified 16 Slate v2-only internal categories and comparison workload fixtures.
Generator split	complete	`render-rich-text-viewer.mjs` now emits comparison and internals page/data pairs.
Generated docs	complete	`docs:rich-text` wrote all four generated files.
Package verification	complete	`npm run check` passed in `benchmarks/editor`.
Served proof	complete	Both HTML routes returned HTTP 200 and both JSON routes had expected row/category counts.
Closure	complete	This file records evidence and passes `check-complete`.

Start Gates:

Gate	Applies	Evidence
Active goal checked or created	yes	Autogoal created for the viewer split.
Source of truth read before edits	yes	Read current viewer generator, package scripts, and generated benchmark data shape.
Edit scope selected	yes	Only viewer generator, generated docs/data, and this plan changed for this task.
Browser route selected	yes	`rich-text.html` and `slate-v2-internals.html` under the existing static server.
Browser tool decision recorded	yes	Browser MCP was not exposed; HTTP served-route proof used against the same local target.

Work Checklist:

Objective includes outcome, completion threshold, verification surface, constraints, boundaries, and blocked condition.
Source data shape is mapped before editing.
Comparison rows and Slate v2 internals rows have explicit filters.
Generator emits both HTML and data JSON files in normal and check modes.
Generated docs are refreshed.
Package checks are green.
Served-route proof covers both pages and both data files.
Browser caveat is recorded.

Completion Gates:

Gate	Applies	Required action	Evidence
Comparison page generated	yes	Generate `rich-text.html` and `rich-text-data.json`	Comparison JSON has 542 rows, 11 groups, and zero internal category leaks.
Internals page generated	yes	Generate `slate-v2-internals.html` and `slate-v2-internals-data.json`	Internals JSON has 326 rows, 16 groups, and all expected internal categories.
Docs check	yes	Run generated docs check	`npm run docs:rich-text:check` checked both page/data pairs.
Package check	yes	Run package gate	`npm run check` passed in `benchmarks/editor`.
Served HTML proof	yes	Hit both served HTML routes	Both returned HTTP 200.
Served JSON proof	yes	Hit both served data routes	JSON smoke proof returned expected split counts.
Goal plan complete	yes	Run autogoal completion check	`check-complete` passed.

Verification evidence:

cd benchmarks/editor && npm run docs:rich-text wrote rich-text.html, rich-text-data.json, slate-v2-internals.html, and slate-v2-internals-data.json.
Local generated-data proof: comparison page has 542 rows, 11 groups, and no internal category leaks.
Local generated-data proof: internals page has 326 rows, 16 groups, and all expected internal categories.
cd benchmarks/editor && npm run docs:rich-text:check passed and checked both page/data pairs.
cd benchmarks/editor && npm run check passed.
Served proof: curl -I --max-time 2 http://127.0.0.1:8765/rich-text.html returned HTTP 200.
Served proof: curl -I --max-time 2 http://127.0.0.1:8765/slate-v2-internals.html returned HTTP 200.
Served JSON proof: comparison data returned 542 rows, 11 groups, and internalLeaks: [].
Served JSON proof: internals data returned 326 rows, 16 groups, and missingExpectedInternalCategories: [].

Reboot status: Complete. rich-text.html is the comparison page; slate-v2-internals.html is the v2-only proof page. Both are generated by npm run docs:rich-text.

Open risks: The category split is manually curated in the generator. If new Slate v2-only categories are added later, they must be added to the internal category set or they will appear on the comparison page.