chore(promotion): promote next to 1.1.0 by mfal · Pull Request #3029 · mittwald/flow

mfal · 2026-08-31T10:33:50Z

Release 1.1.0


Source → Target	`next` → `main`
Version	`1.0.16` → `1.1.0`
Bump	minor (18 `feat` commits, no breaking changes)

This PR graduates the line to the stable 1.1.0 (version + changelog already
bumped on the branch). Merging publishes all @mittwald/flow-* packages to npm
under latest, tags 1.1.0, and creates the GitHub release from the curated
notes below.

The branch is next (which carries main through 1.0.16) plus three commits:
the stable changelogs taken from main, the figure below, and the graduation.
It touches no source file of its own.

Upgrade across Flow versions with one command

Highlights

npx @mittwald/flow-codemods@latest upgrade bumps every Flow dependency,
installs, and runs the codemods for the range it crossed.
The new @mittwald/flow-codemods package carries the migration catalogue that
MIGRATION.md is generated from — 23 entries, each carrying a codemod where
the change can be applied mechanically.
flow-codemods list <revision> shows what a target version would touch before
you commit to it.
Rating takes a segment count (maxValue), single-segment filling
(fill="single"), and individual RatingSegment children with their own
icons and labels.
@mittwald/flow-react-components now ships USAGE.md and a JSON component
index, so a coding agent can answer component questions without the docs site.

The upgrade CLI

Upgrading Flow used to mean reading MIGRATION.md, deciding which entries
applied to your jump, and running codemods one at a time. The new
@mittwald/flow-codemods package does that for you.

# Bump every Flow dependency, install, and run the codemods for the range.
# Defaults to "minor", which stays inside your current major.
npx @mittwald/flow-codemods@latest upgrade

# Or name a target: patch | minor | major | a dist-tag | an exact version.
npx @mittwald/flow-codemods@latest upgrade latest

# See what a target version would touch before committing to it.
npx @mittwald/flow-codemods@latest list latest

# Without a revision: the whole catalogue, offline.
npx @mittwald/flow-codemods@latest list

# Run a single migration by its catalogue id.
npx @mittwald/flow-codemods@latest imports-to-package-root src

upgrade detects your package manager, refuses to run on a dirty working tree
(--allow-dirty overrides), and reports every file it declined. --dry writes
nothing, --print shows the transformed output, and --json makes list
machine-readable.

The catalogue is the single source for both the CLI and MIGRATION.md: an entry
describes what changed and how to migrate, and carries a codemod when the change
can be decided mechanically. Entries without one still tell you what to do by
hand — for those, the prose is the migration.

Rating: segment count, single fill, individual segments

Rating gained maxValue (default 5) for the number of segments, and
fill="single" to fill only the selected segment instead of every segment up to
it.

import { Rating, RatingSegment } from "@mittwald/flow-react-components";

<Rating maxValue={10} defaultValue={7}>
  <Label>Cantina rating</Label>
</Rating>;

For a scale whose segments are not interchangeable, pass RatingSegment
children — they define how many segments there are (maxValue no longer
applies) and each one takes its own icons and accessible label.

<Rating fill="single" defaultValue={3}>
  <Label>How was your stay on Tatooine?</Label>
  <RatingSegment
    aria-label="Terrible"
    iconEmpty={<Icon><IconMoodSad /></Icon>}
    iconFilled={<Icon color="danger"><IconMoodSadFilled /></Icon>}
  />
  <RatingSegment
    aria-label="Great"
    iconEmpty={<Icon><IconMoodHappy /></Icon>}
    iconFilled={<Icon color="success"><IconMoodHappyFilled /></Icon>}
  />
</Rating>;

Both the component and RatingSegment work in mStudio extensions.

Documentation a coding agent can read

@mittwald/flow-react-components now ships its own consumer guide, and the docs
site is published in machine-readable form.

USAGE.md, shipped inside the package: how to build an application with
Flow — component selection, layout and spacing, what is safe to depend on, and
the mistakes that come up most. Available offline, which is what makes it
usable from an agent's tool loop.
@mittwald/flow-react-components/component-index — a JSON index of every
public component with its lifecycle status and props.
The docs site at https://flow.mittwald.de serves /llms.txt, /llms.json,
/llms-full.txt, and every page as Markdown at /raw/<path>.md.

Point your agent at USAGE.md and the component index; it stops guessing
component names and props.

Raw commits (148 total, 94 non-merge) — reference only, not published

feat (18)

feat(Rating): individual segments, segment count and single fill (feat(Rating): individual segments, segment count and single fill #2863)
feat(codemods): add detect/verify commands and the checks module shape
feat(codemods): add the list command
feat(codemods): add the migration catalogue and its generated module
feat(codemods): add the single-codemod command
feat(codemods): add the upgrade command
feat(codemods): adopt the transform suite and test harness from fix(codemods): make the composite codemod runnable from a URL #2942
feat(codemods): detect the package manager and guard a dirty tree
feat(codemods): find and rewrite Flow dependencies in a consumer manifest
feat(codemods): generate MIGRATION.md from the catalogue
feat(codemods): make list output readable in a terminal
feat(codemods): publish the package with a CLI entry point
feat(codemods): rename transforms to catalogue ids and retire the URL path
feat(codemods): resolve a revision to a published target version
feat(codemods): run a codemod by its catalogue id
feat(codemods): select catalogue entries for a version range
feat(codemods): wire the upgrade command into the CLI
feat: ship a consumer agent guide and a compact component index (feat: ship a consumer agent guide and a compact component index #2897)

fix (24)

fix(codemods): correct the since version of the package-root import migration
fix(codemods): cover a prerelease current and name every registry failure
fix(codemods): drive jscodeshift programmatically instead of scraping its output
fix(codemods): drop the lower bound for codemod catch-up, keep it for manual migrations
fix(codemods): drop the manual-entry lower bound, mark catch-up everywhere
fix(codemods): explain why a revision could not be resolved, not just that it couldn't
fix(codemods): make the catalogue generator deterministic and its detect runnable
fix(codemods): make the catalogue's detect, apply and verify fields true
fix(codemods): make the gate's exhaustiveness check and its prerelease test real
fix(codemods): make the interactive gate an input, not an environment read
fix(codemods): move list's summary below the entries, make colour deterministic
fix(codemods): pin the list command's tool asymmetry and lower its sentinel
fix(codemods): resolve the target from what every Flow dependency published
fix(codemods): rewrite value positions, catalogue the password-tools subpath
fix(codemods): satisfy the repo's lint rules in three modules
fix(codemods): set exitCode instead of calling process.exit
fix(codemods): shorten list header to three lines
fix(codemods): stop a guard test racing the remote-component generator
fix(codemods): stop a transform that declines every file reading as success
fix(codemods): stop imports-to-package-root corrupting current code
fix(codemods): stop the install and the dirty-tree guard failing silently
fix(codemods): stop upgrade damaging a project and reporting success
fix(codemods): tidy list/upgrade output — zero-width range, legend, and duplicate header
fix(codemods): wrap the catch-up legend like every other line

refactor (5)

refactor(codemods): colocate each migration's entry, codemod and test
refactor(codemods): cut detect and verify from the migration catalogue
refactor(codemods): drop detect/verify prose and list's title line
refactor(codemods): drop to-remote-package from the migration catalogue
refactor(codemods): make list take a revision instead of --from/--to

revert (1)

revert(codemods): take the detect and verify commands back out

test (2)

test(codemods): pin the manifest cases that could hide a defect
test(docs): keep the Vitest and node:test globs disjoint

docs (27)

docs(codemods): document the catalogue and the upgrade CLI
docs: adopt PR fix(codemods): make the composite codemod runnable from a URL #2942 into the plan and add the deprecation follow-up
docs: correct the ADR to the CLI that was actually built
docs: correct the docs-page layout and the cross-version CI split
docs: correct the prerelease target-resolution expectation
docs: do not hardcode the package version in the plan
docs: drive jscodeshift programmatically and respect the consumer's registry
docs: drop the execution plan
docs: fix the install and dirty-tree modules in the plan
docs: fix the since value and the tool filter in the plan
docs: format the plan document
docs: handle the string private flag in the package-list generator
docs: make the gate snippet exhaustive and its prerelease test discriminating
docs: pin the list command's tool asymmetry in the plan
docs: pin the manifest cases that could hide a defect
docs: plan detect and verify as executable modules
docs: plan the codemods upgrade CLI and migration catalogue
docs: point every migration pointer at the catalogue and the CLI
docs: reconcile task 14's AGENTS.md step with what task 4 already did
docs: record the ESM extension and tsc findings from task 1
docs: record the accepted red commit in task 3
docs: record the task 6 findings in the plan
docs: report declined files in the upgrade loop too
docs: restore the remoteScope invariant as a task 4 step
docs: stop the docs implying upgrade has a lower bound
docs: stop the plan from hardcoding the package version
docs: tell the installation page how to stay current

chore (17)

chore(codemods): wire the generators into nx
chore(release): bump version to 1.1.0-next.0
chore(release): bump version to 1.1.0-next.1
chore(release): bump version to 1.1.0-next.10
chore(release): bump version to 1.1.0-next.11
chore(release): bump version to 1.1.0-next.12
chore(release): bump version to 1.1.0-next.2
chore(release): bump version to 1.1.0-next.3
chore(release): bump version to 1.1.0-next.4
chore(release): bump version to 1.1.0-next.5
chore(release): bump version to 1.1.0-next.6
chore(release): bump version to 1.1.0-next.7
chore(release): bump version to 1.1.0-next.8
chore(release): bump version to 1.1.0-next.9
chore(sync): fix what the main → next merge left red
chore(sync): resolve forward-merge conflict (main into next) (chore(sync): resolve forward-merge conflict (main into next) #2969)
chore: ignore the superpowers scratch directory

Co-authored-by: mathis <37186532+maaaathis@users.noreply.github.com> Co-authored-by: maaaathis <maaaathis@users.noreply.github.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com> Co-authored-by: Lisa Stockmann <84317589+Lisa18289@users.noreply.github.com> Co-authored-by: Lisa18289 <Lisa18289@users.noreply.github.com>

@deprecated

Three conflicts, plus two cross-side interactions the merge creates. `loadProperties.ts` — `next` moved the properties table onto the generated component index (#2897); `main` taught it to hide `@deprecatedValues` and to tell that tag apart from a prop-level `@deprecated` (#2960, #2961). Kept `next`'s data source and ported `main`'s value hiding onto the new prop shape (key-based name, `type` as a string, optional `description`). Top-level `undefined`/`null` are not filtered here any more — the index generator already strips them, so the docs side only removes deprecated values. `buildComponentIndex.ts` — `main`'s new `@deprecatedValues` tag contains the substring `next`'s generator matched to set `deprecated: true`, so `Button`'s `color` came out of the merge flagged deprecated as a whole. Uses `main`'s `/@deprecated(?!\w)/` now; regression test added. `RemoteRatingSegmentElement.ts` — `main` put react-aria's `render` on the generator's global ignore list (#2966) after `next` had generated `RatingSegment` (#2863). Regenerated, which drops the `render` property that would have broken the whole mutation batch in remote. ADR 0004 §3 — took `main`'s passage: it already carries both the `publish-next.yml` → `publish.yml` rename (#2968) and the version-consistency guard paragraph (#2958). `pnpm-lock.yaml` — kept `next`'s. It is a strict superset of `main`'s (zero entries only on `main`; `next` already had the Dependabot bumps), and `pnpm install` reports it up to date against every merged manifest. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

# Conflicts: # apps/docs/package.json

Review found that dropping the gate's lower bound turned an era-specific transform into a corrupter of correct code, and I reproduced all of it. `imports-to-package-root` (since alpha.28) collapsed 94 subpath exports onto the package root, and it did that with a catch-all `else`. With no lower bound it reaches every 1.x project, where five of today's nine subpaths were collateral: `all-layered.css` became a JS import with the stylesheet silently gone, and `Rule` from `mittwald-password-tools-js` moved onto a root that does not export it. A default import became a named specifier along the way. It also cancelled another migration. Sorted by `since` it runs first, so it flattened `…/password-tools` before `password-tools-subpath-renamed` — the entry whose whole job is that path — could see it. Both then reported success while the file no longer compiled. Both have one cause, so one fix: an explicit leave-alone set. It holds every subpath the package still exports, plus `password-tools`, which another entry owns. Era code still migrates — verified both directions. Two independent tests pin it, one comparing the set against the real `exports` map so a new subpath cannot quietly become collateral again; both fail if a single entry is removed. Idempotency could never have caught this, and a comment in manifest.ts leaned on it as if it could. It proves a second pass over a transform's own output changes nothing — not that an era-specific transform is harmless against later code. Also from the review: - The dirty-tree guard treated every git exit 128 as "not a repository", so `detected dubious ownership` — the minimal CI container the guard exists for — read as a clean tree. It now matches the message, with `LC_ALL=C` pinning it. - `peerDependencies` were rewritten. A peer range states what a package supports, not what it installs; narrowing `^1.0.0` to `^1.0.14` is a breaking change to someone else's package made by a command they ran on their own. They are now reported and left, through one reporter `--dry` shares. - `engines.node` was `>=24.0.0` on a CLI consumers reach through `npx`. The only thing needing 24 was an import attribute for reading the version; that is a `readFileSync` now, and the floor is `>=22.0.0`. - `components-src` included the `MIGRATION.md` that `codemods:build` writes, with no edge between the projects, so its hash depended on task order. Excluded. - Ctrl+C at the codemod prompt rejected out of `runUpgrade` after the bump and install, showing an inquirer trace instead of what had happened. It now reports that and runs nothing. `@inquirer/prompts` also loads lazily, so `list` no longer pays for it. - `list --json` was a bare array that could not say which range it described — on the 1.x line `list` and `list minor` select the same entries. It now carries `range` and per-entry `catchUp`. This replaces a test that pinned the bare array; that test recorded the old shape, it did not defend it. - `kind`'s docblock still described the selection it once drove. It is descriptive now, and says so. Not fixed here, because it is not in the repo: `@mittwald/flow-codemods` does not exist on npm, and npm binds one Trusted Publisher per package. The first publish needs an npm-side bootstrap or `lerna publish from-package` aborts mid-release with the packages ahead of it already out. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…pgrade-cli-f87427

… read The `unit` job failed on the cancel case I added with the review fixes. It passed here and failed on the runner for one reason: `createChoose` read `process.env.CI` inside itself, underneath the parameters the caller passes. On a runner `CI` is set, so the prompt never ran, so it never rejected, so the cancel path the test asserts was unreachable. That is the second time a hidden environment read under an explicit parameter has produced a green local suite and a red CI one — picocolors self-disabling beneath the `color` argument was the first. So `isCI` is an input now, stated by `cli.ts` where reading the environment is the job, and TypeScript makes every call site name it. Two things fell out. The behaviour that read gave — a TTY on an unattended runner must not prompt — was untested, because no test could reach it; it has a case now. And the one existing test that wanted the prompt shown had to save, delete and restore `process.env.CI` around itself, with a comment explaining the workaround. That scaffolding is gone, and no test in this package touches `process.env` any more. Verified both ways: 251 tests with `CI=true` and with `CI` unset. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

…pgrade-cli-f87427

github-actions · 2026-08-31T10:34:57Z

Coverage Report for ./packages/components/

Status	Category	Percentage	Covered / Total
🔵	Lines	78.69%	746 / 948
🔵	Statements	78.57%	763 / 971
🔵	Functions	80.09%	165 / 206
🔵	Branches	70.33%	377 / 536

File Coverage

File	Stmts	Branches	Functions	Lines	Uncovered Lines
Changed Files
packages/components/dev/component-index/buildComponentIndex.ts	93.65%	83.67%	100%	93.54%	50, 107, 113, 129
packages/components/dev/component-index/filterProps.ts	100%	93.54%	100%	100%
packages/components/dev/component-index/remoteAvailability.ts	100%	100%	100%	100%
packages/components/src/components/propTypes/index.ts	100%	100%	100%	100%

Generated in workflow #6457 for commit 66aeb52 by the Vitest Coverage Report Action

github-actions · 2026-08-31T10:41:17Z

🚀 Preview Deployment

Preview environments are ready:

Type	URL
docs	pr-3029.docs.review.flow-components.de
storybook	pr-3029.storybook.review.flow-components.de

Images:

docs: ghcr.io/mittwald/flow/docs:pr-3029
storybook: ghcr.io/mittwald/flow/storybook:pr-3029

# Conflicts: # packages/components/MIGRATION.md

`guide.test.ts` pins the guide's newest anchor by id. The merge added a newer entry (`use-design-tokens-build-metadata-removed`, `since: 1.0.16`), so the expectation moves with it. `eslint.config.js` arrives from `main` carrying the `no-restricted-imports` ban on `@mittwald/flow-design-tokens/json/*` under `src/` — and `next` has a call site `main` never saw. `collectTokensInPath` reads only `path` and `value`, so `json-runtime/all-light.json` is a straight swap. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

chore(sync): resolve forward-merge conflict (main into next)

The promotion direction inverts `**/CHANGELOG.md merge=ours` (ADR 0004 §3), which is written for main → next. next's changelogs carry the prerelease line; promoting them would overwrite the stable history publish.yml extracts GitHub release bodies from. The graduation commit prepends the 1.1.0 entry. Its own commit because the merge above is a no-op in the healthy steady state — next already carries main — so there is no merge commit to absorb it (see #3030).

Captured from the Rating stories on this branch: maxValue={10}, fill="single" and RatingSegment children, composed into one figure at 2x. Committed rather than uploaded as a GitHub attachment: user-attachments has no API, so that path cannot be automated. The release notes reference it by this commit's SHA, which resolves during review and keeps resolving after the branch is deleted — see #3030.

The 1.1.0 promotion (#3029) brought feature work from `next` written against the old, numbered docs paths. Two prose conflicts (Rating's scale guideline, Align's codemod pointer) keep main's wording with this branch's paths, and the two new Rating examples land in the renamed directory. It also brought seven references to paths this branch removed, in files that did not exist when the rename ran: - `packages/codemods/project.json` pointed an nx input at `01-get-started/versioning/index.mdx`. A glob matching nothing contributes nothing to the task hash, so `test:unit` would have served stale cache when that page changed — silently. - `documented.test.ts` and `mdxToMarkdown.test.ts` read or assert old paths. - `USAGE.md` lists `/raw/**` paths for consumers. - The CodeBlock migration entry links the old path, and while it was being touched, the legacy `/overview` tab redirect with it. `MIGRATION.md` is an output of `nx build codemods`, so the fix went into `entry.md` and the artifacts were regenerated. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

mfal and others added 30 commits August 26, 2026 09:35

chore(sync): forward-merge main into next (e59cd05)

82db678

chore(sync): forward-merge main into next (e2dab1f)

fe33a64

chore(sync): forward-merge main into next (636df51)

7fdf505

chore(sync): forward-merge main into next (8758466)

d20682e

chore(sync): forward-merge main into next (3430cfe)

34dcf32

chore(sync): forward-merge main into next (ccaf11a)

fd8cd5e

chore(sync): forward-merge main into next (b346fec)

c2b733e

chore(sync): forward-merge main into next (2beea5e)

933f27a

chore(sync): forward-merge main into next (c24c5ef)

50fa0c0

chore(sync): forward-merge main into next (d44d61c)

59e23b6

chore(sync): forward-merge main into next (984e080)

56dc66a

chore(sync): forward-merge main into next (9bbd236)

84ba91e

chore(sync): forward-merge main into next (c826ea4)

b4622eb

chore(sync): forward-merge main into next (ac3ef26)

e1c83bf

chore(sync): forward-merge main into next (2f6210d)

ea5861c

chore(sync): forward-merge main into next (6102d2e)

0fbe956

chore(sync): forward-merge main into next (8f98f2f)

79fb4fc

chore(sync): forward-merge main into next (419e17f)

b93b529

chore(sync): forward-merge main into next (2a2eb3a)

59edace

chore(sync): forward-merge main into next (6bf57c7)

c93ad39

chore(sync): forward-merge main into next (cb1fd12)

c70874f

feat: ship a consumer agent guide and a compact component index (#2897)

797f512

chore(sync): forward-merge main into next (4809aa3)

43f3cac

chore(sync): forward-merge main into next (47748ba)

5de4f8e

feat(Rating): individual segments, segment count and single fill (#2863)

aea4e59

chore(release): bump version to 1.1.0-next.0

b171f12

chore(release): bump version to 1.1.0-next.1

3d1fcf6

chore(sync): forward-merge main into next (ce5141d)

0bc2694

# Conflicts: # apps/docs/package.json

mittwald-machine and others added 10 commits August 31, 2026 08:54

chore(release): bump version to 1.1.0-next.9

5d2639c

chore(sync): forward-merge main into next (f7040d6)

ebd772a

Merge remote-tracking branch 'origin/next' into claude/flow-codemod-u…

8f66c28

…pgrade-cli-f87427

chore(release): bump version to 1.1.0-next.10

479b9c4

Merge remote-tracking branch 'origin/next' into claude/flow-codemod-u…

fe103ab

…pgrade-cli-f87427

chore(sync): forward-merge main into next (85bbc3b)

7efb5cf

feat(codemods): add an upgrade CLI backed by a migration catalogue

0334cef

chore(release): bump version to 1.1.0-next.11

bc0ae71

mfal self-assigned this Aug 31, 2026

mfal marked this pull request as ready for review August 31, 2026 10:35

mfal requested a review from a team August 31, 2026 10:35

mfal mentioned this pull request Aug 31, 2026

prepare-release: capture the release-notes screenshots automatically #3030

Open

mfal enabled auto-merge (squash) August 31, 2026 11:08

mfal disabled auto-merge August 31, 2026 11:17

mfal enabled auto-merge (squash) August 31, 2026 11:17

mfal and others added 6 commits August 31, 2026 13:42

chore(sync): forward-merge main into next (a552b4a)

e86c584

# Conflicts: # packages/components/MIGRATION.md

Merge pull request #3033 from mittwald/sync/main-to-next

96e2c3e

chore(sync): resolve forward-merge conflict (main into next)

chore(release): bump version to 1.1.0

66aeb52

mfal force-pushed the release/1.1.0 branch from e7796eb to 66aeb52 Compare August 31, 2026 12:03

ins0 approved these changes Aug 31, 2026

View reviewed changes

mfal merged commit 3a17e79 into main Aug 31, 2026
26 of 27 checks passed

mfal deleted the release/1.1.0 branch August 31, 2026 12:09

Navigation Menu

Uh oh!

chore(promotion): promote next to 1.1.0 - #3029

chore(promotion): promote next to 1.1.0#3029
mfal merged 150 commits into
mainfrom
release/1.1.0

mfal commented Aug 31, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 31, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 31, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Uh oh!

Conversation

mfal commented Aug 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Release 1.1.0

Upgrade across Flow versions with one command

Highlights

The upgrade CLI

Rating: segment count, single fill, individual segments

Documentation a coding agent can read

Uh oh!

github-actions Bot commented Aug 31, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Coverage Report for ./packages/components/

Uh oh!

github-actions Bot commented Aug 31, 2026

🚀 Preview Deployment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

mfal commented Aug 31, 2026 •

edited

Loading

github-actions Bot commented Aug 31, 2026 •

edited

Loading