# 0.0.2-alpha.78

15 Nov 2026 · Newest · Before you upgrade: No steps listed

- [A page of any type can carry an FAQ section in its structured data, declared by its route or for its whole content type.](https://github.com/nextlyhq/nextly/releases/tag/v0.0.2-alpha.78)
- [The SEO plugin can state the site's X account on every page, and gives a site's own sitemap the date a page's content last changed, which an SEO-only save no longer moves.](https://github.com/nextlyhq/nextly/releases/tag/v0.0.2-alpha.78)
- [On Vercel, the PostgreSQL and MySQL adapters default to a small database pool: at most 3 connections, none kept open, and idle ones closed after 5 seconds.](https://github.com/nextlyhq/nextly/releases/tag/v0.0.2-alpha.78)

Published: 15 Nov 2026
Changes: 20 changes
Notes: [On GitHub](https://github.com/nextlyhq/nextly/releases/tag/v0.0.2-alpha.78)

## Before you upgrade

These notes list no steps.

## Upgrading

<!-- release-notes -->

- Run `nextly migrate` right after upgrading (the scaffolded build script already runs `nextly migrate && next build`).
- Field groups made with an earlier release: their next save in the Schema Builder, or the next `nextly migrate:create`, writes a migration that adds the new nullable columns for the data fields they now store. No existing column changes. Until that migration is applied, the field group reads and writes as before.
- Apps on Vercel with PostgreSQL or MySQL: the adapters now default to a small pool (at most 3 connections, none kept open, idle ones closed after 5 seconds). Install `@vercel/functions` so the adapters can also hand the pool to `attachDatabasePool`; without it they log one warning. To keep the old pool, set `DB_POOL_MAX=10` and `DB_POOL_IDLE_TIMEOUT=30000` (each `DB_POOL_*` variable, or the adapter's `pool` option, replaces only its own default). Nothing changes off Vercel.
- SEO plugin, in production: when Nextly is unavailable (it could not start, or the database is busy, unreachable or timed out), a page's SEO head now fails the render instead of serving the route's own metadata. A revalidation keeps serving the last good render, but a first request with nothing cached answers with an error, and `next build` fails instead of shipping pages without their SEO head. Make sure the database is reachable from the build. `next dev` is unchanged.
- The Single API now refuses a top-level field without a name, with the collection API's existing `REQUIRED` validation error. Before, such a request crashed.

<!-- /release-notes -->

## What's changed

<!-- release-notes -->

Released 21 packages at `0.0.2-alpha.78` in lockstep. Every package below ships at this version.

### Patch Changes

- [#4323](https://github.com/nextlyhq/nextly/pull/4323) [`abec063`](https://github.com/nextlyhq/nextly/commit/abec0634aaf51ddf5415c41fc06336f9fbbcaeb9) @nextly-agent: A collection or Single whose definition file holds a group with no name at the top level no longer stops Schema Builder saves.

  Such a group only lays out the fields inside it, and those fields are stored as columns of the entry's own row, which is where entries already read and write them.
  
  Before, every builder of a table's columns tried to read the group's missing name and crashed with `Cannot read properties of undefined (reading 'replace')`: the table create, the table change, the runtime table, and the desired schema a migration is diffed from. Once such a collection or Single was in the definition file, every later Builder save failed.
  
  All of these builders now read the group's fields as columns of the row. A later save's migration adds or changes only what changed. The rest of the save and of the entry reads and writes take those fields the same way:
  - rename detection and the checks run before a save;
  - relationship keys and field-group row moves;
  - defaults, `select` reads, versions, field-group values, JSON values, dates and relationship values.
  
  **Compatibility:** the Single API now refuses a top-level field without a name, with the same answer the collection API already gives: a validation error, code `REQUIRED`, "Each field requires a name.". This covers creating a Single, saving its schema and applying it from the schema dialog. Before, such a request crashed. A group with no name inside a named group or a repeater stays allowed.

- [#4354](https://github.com/nextlyhq/nextly/pull/4354) [`0e22fd2`](https://github.com/nextlyhq/nextly/commit/0e22fd2201db908dae120253704726737ec2ee51) @mobeenabdullah: The SEO plugin's structured data no longer leaves out `dateModified` when the field a content type maps for it, such as a correction date, is empty: it then uses the date the page's content last changed, the same date as its sitemap `<lastmod>`.

  A date in that field still comes first.
  
  A save that changes only a page's SEO no longer moves its `<lastmod>` and `dateModified` when it sends the page's field groups back unchanged: the ids and dates of the rows a field group is kept in, which such a save rewrites, are no longer read as content. Dates kept by an earlier release stay as they are.

- [#4345](https://github.com/nextlyhq/nextly/pull/4345) [`ac0b7fd`](https://github.com/nextlyhq/nextly/commit/ac0b7fd4c8d1eba8acb2d5a0e9fdab8e7ee16ff1) @nextly-agent: An entry list no longer moves its cells when a plugin's column arrives after the rows.

  A column provider whose columns rest on a read of its own can now say so with `ctx.deciding(until)` (experimental), and the entry list keeps its outline a moment longer (250 ms at most, timed from the rows' arrival) while one is still deciding, then draws the rows with the column. The Page Builder's Design column says so while it reads whether the collection has a template, and its cells draw from that answer until their own read lands, so the column no longer changes width either.

- [#4338](https://github.com/nextlyhq/nextly/pull/4338) [`9add3c2`](https://github.com/nextlyhq/nextly/commit/9add3c2537b23ca44d33ba3d4560b18e31224b3b) @nextly-agent: Field groups now store the data fields placed inside a group with no name.

  A group with no name only lays fields out, and everywhere else its fields are stored where the group sits; in a field group that is the instance's own row. Before, the field group's table had no column for such a field, its value was dropped on write, and a read never returned it. Now the creator, the desired table a save or `nextly migrate:create` diffs from, the instance write and read (a relationship's expansion at depth included), and reconcile all read the same walk, so each such field is a column of `comp_<slug>` and its value is stored and read back. A field group placed inside such a group keeps its instances in its own table, as before.
  
  Compatibility: field groups made with an earlier release have no column for these fields. Their next save in the Schema Builder, or the next `nextly migrate:create`, writes a migration that adds each missing column, nullable, and changes no existing column. Where the table was made before Builder saves wrote migrations, the save's review now names each column it adds, for you to confirm. Until that migration is applied, such a field group is read and written as before, without those fields. Reconcile keeps the fields and lists each column still to come (`awaitingColumns`) instead of removing them, and a reconcile of a field group holding a group with no name no longer fails.

- [#4364](https://github.com/nextlyhq/nextly/pull/4364) [`26e751f`](https://github.com/nextlyhq/nextly/commit/26e751f6f1560afeaafccbaf5f7922e146a2036d) @mobeenabdullah: An app that has not installed `@nextlyhq/admin` no longer fails its type check on the generated `plugin-admin-imports.generated.ts`: `nextly generate:types` leaves the admin import and its registration out of the map there, and writes them back once the admin is installed and the command is run again.

- [#4347](https://github.com/nextlyhq/nextly/pull/4347) [`7934af0`](https://github.com/nextlyhq/nextly/commit/7934af0e91060f0033d31b5a7296ea218dec7a87) @nextly-agent: The dashboard's Needs your attention card holds the line that names the stopped or overdue releases under a row while that read is on its way, so the rows below no longer move down when the names arrive.

- [#4363](https://github.com/nextlyhq/nextly/pull/4363) [`3961290`](https://github.com/nextlyhq/nextly/commit/396129093f7b6e1bea3cb9956083a06f8821f33a) @mobeenabdullah: On a phone, each card in an entry list is headed by the entry's title again, not by the Page Builder's Design column, and on a wide screen the Design column sits after Title rather than before it, on a collection with a template.

- [#4346](https://github.com/nextlyhq/nextly/pull/4346) [`aea91fa`](https://github.com/nextlyhq/nextly/commit/aea91fae0a9b822fa9c6bc8cecd47be6022336f6) @nextly-agent: The Releases page no longer moves while it loads when releases exist.

  The tabs and the list wait, as one outline, for the read that says whether the needs-attention strip is drawn, so a strip no longer lands above them and pushes them down; the list's search and filter are drawn with the table rather than above its outline; each tab's count keeps room for two digits, so a count past nine no longer widens its tab once it arrives; and a row's readiness line keeps its height while the checks are on their way.

- [#4353](https://github.com/nextlyhq/nextly/pull/4353) [`0f06edd`](https://github.com/nextlyhq/nextly/commit/0f06edd743cb1874d866f3cc6cb2d413a000e053) @mobeenabdullah: The SEO plugin's Search Console report on **SEO > Settings > Connections** now loads in seconds rather than a minute or more.

  It asks Google for its counts and inspects the most visited pages all at once, where before it made about fourteen requests one after another, and it keeps the report for six hours, reading Google again only after that, so opening the tab again shows it straight away. The tab says when the report was read. Connecting again or disconnecting forgets the kept report. One click now reads "1 click", not "1 clicks".

- [#4342](https://github.com/nextlyhq/nextly/pull/4342) [`73a524b`](https://github.com/nextlyhq/nextly/commit/73a524b68de73c0894498ead5d30a83307e7ef29) @mobeenabdullah: SEO: a page with no stored words of its own, a page built in the site's code or a Page Builder layout, now has a plain-text copy made from the page as it renders: its main content as Markdown, asked for as the public asks for it at its listed address, only where the site names its own address (a production address or `baseUrl`), and kept for five minutes. llms.txt and the page's head link these copies too.

- [#4320](https://github.com/nextlyhq/nextly/pull/4320) [`1e74c1e`](https://github.com/nextlyhq/nextly/commit/1e74c1e5a31c8aaf941f410b452d4074b31000d1) @mobeenabdullah: SEO: every brand fact the plugin's options can give (description, email, telephone, contact points, founding date and the parent organization) can now be saved on Settings › Organization, and each saved fact wins over the options' same fact one at a time, so moving a fact from the code to the admin changes nothing else.

  The brand is named by its own name before the site's. The Organization tab shows each value the site's code gives, marked as set in the code. Run `nextly migrate` for the settings' new field.

- [#4350](https://github.com/nextlyhq/nextly/pull/4350) [`d24b003`](https://github.com/nextlyhq/nextly/commit/d24b0037803bea8a5e1fc2641c6c18e272f242c8) @mobeenabdullah: A page of any type can now carry an FAQ section in its structured data.

  Declare it for one page with `faqSection: true` in a route's `structuredDataHints`, or for a whole content type with the new "Questions on its pages are an FAQ section" switch in SEO › Settings › Content types (or `structuredData.types.<type>.faqSection` in code). The page keeps its own type, and its questions (from the route, or its FAQ field) are described as an `FAQPage` that is part of the page. Pages that declare nothing are unchanged.

- [#4356](https://github.com/nextlyhq/nextly/pull/4356) [`4048b19`](https://github.com/nextlyhq/nextly/commit/4048b19f35235db87132aabc2d454c36dbda5409) @mobeenabdullah: In production, a page's SEO head now fails the render when Nextly is unavailable (it could not be started, the database is busy, unreachable or timed out), instead of serving the route's own metadata.

  A revalidation keeps serving the page's last good render, a first request with nothing cached answers with an error, and `next build` fails rather than shipping pages without their SEO head. `next dev` still serves the route's own metadata, and any other failure (a bad answer, a plugin bug) still falls back everywhere. This covers the route helpers and the SEO plugin's `seoMetadata` and `seoStructuredData`. Plugins can apply the same rule with `failRenderWhenUnavailable` and `isNextlyUnavailable` from `@nextlyhq/plugin-sdk` (experimental).

- [#4358](https://github.com/nextlyhq/nextly/pull/4358) [`37fecdf`](https://github.com/nextlyhq/nextly/commit/37fecdf9ad2f62f9a76a58be981f90783072dee8) @mobeenabdullah: The SEO plugin's install and uninstall no longer print "The SEO address record could not be brought up to date" while `nextly plugins install` runs: the hooks wait for the address record catch-up the command's start began, so it no longer runs on after the command has stopped, and the first server start after an install still brings the record up to date.

- [#4351](https://github.com/nextlyhq/nextly/pull/4351) [`9301f12`](https://github.com/nextlyhq/nextly/commit/9301f12b4f3ab43843b025005127147f407121fc) @mobeenabdullah: The SEO Pages list shows how many other pages link to each page and how many it links to, and the problem list names a listed page no other page links to ("Page no other page links to"), both worked out from the links each page's check already reads, so the site is not read twice.

  The page links are a new field of the plugin's own list of checks: a site that applies its schema with migrations runs `nextly migrate:create` and then `nextly migrate` after upgrading.

- [#4341](https://github.com/nextlyhq/nextly/pull/4341) [`8c3abf9`](https://github.com/nextlyhq/nextly/commit/8c3abf96c799b69673a73b90469a037164bddf01) @mobeenabdullah: SEO: llms.txt can carry the site's own words for AI tools under its summary (Settings › Search engines and AI › What AI tools should know).

  `/llms-full.txt`, its own switch and off by default, serves every page llms.txt lists with the words of its plain-text copy, in one file capped at 5 MB, and llms.txt links it. Core: `withNextly` rewrites `/llms-full.txt` to the plugins' root routes, as it does `/llms.txt`.

- [#4355](https://github.com/nextlyhq/nextly/pull/4355) [`07d8172`](https://github.com/nextlyhq/nextly/commit/07d817209b7088b356836c26c08e67c7c13f1c13) @mobeenabdullah: SEO › Pages now says when the server cannot read the build's prerender manifest, instead of silently listing only the routes with no variable part.

  A banner names the file, the server logs one warning, and the pages the build recorded for routes with a variable part are listed as Unconfirmed. The sitemap, llms.txt and plain-text copies still list confirmed pages only.

- [#4349](https://github.com/nextlyhq/nextly/pull/4349) [`69282f8`](https://github.com/nextlyhq/nextly/commit/69282f8653a823c8b4a8c2aa44eb87fa88ccc170) @mobeenabdullah: The SEO plugin judges a title or description by how wide it draws on Google, not by its number of characters: Google shows about 600 px of a title and 920 px of a description, so a title of narrow letters past 60 characters now fits and one of wide capitals short of 60 is named as cut.

  The SEO tab shows a bar under the title and description with the width and the characters, and the search preview, Edit mode, Settings and the problem list measure the same way. `hardLengthLimits: true` now refuses a save wider than Google shows rather than one past 60 or 160 characters; a site that had it on runs `nextly migrate:create` and `nextly migrate` once, since the two fields lose their character cap.

- [#4357](https://github.com/nextlyhq/nextly/pull/4357) [`d1f42cd`](https://github.com/nextlyhq/nextly/commit/d1f42cd73a3ecf687345e32b786bebe94e7591cd) @mobeenabdullah: **The SEO plugin states the site's X account, and gives site code its content dates.** `socialCard.xAccount` (experimental) names the site's X (Twitter) account, as a handle with or without the `@` or a profile address such as `https://x.com/example`, and every page's head then carries `<meta name="twitter:site" content="@example">`; anything else stops the plugin at start with an error that says so.

  The page-metadata answer gains `social.xAccount` for it. `seoLastChanged` (experimental, from `@nextlyhq/plugin-seo`) gives a site's own sitemap or `og:article:modified_time` the date the plugin's sitemap uses, when the page's content last changed, instead of `updatedAt`, which every save moves, an SEO-only one included: `seoLastChanged({ collection: "posts", entries: posts })` dates a whole list in one query per hundred entries, and an entry with no date kept falls back to its own `updatedAt`.

- [#4359](https://github.com/nextlyhq/nextly/pull/4359) [`b967a5f`](https://github.com/nextlyhq/nextly/commit/b967a5f2b29fe67399503d0014b79153d46061d5) @mobeenabdullah: On Vercel (where `VERCEL` is set), the PostgreSQL and MySQL adapters now default to a small database pool: at most 3 connections, none kept open, and idle connections closed after 5 seconds, instead of up to 10 kept for 30 seconds.

  Suspended Vercel function instances keep their open connections, which could exhaust small Postgres plans. Install `@vercel/functions` in the app and the adapters also hand their pool to its `attachDatabasePool`; without it they log one warning. To opt out, set `DB_POOL_MAX`, `DB_POOL_MIN` or `DB_POOL_IDLE_TIMEOUT` (or pass `pool` to the adapter): each replaces only its own default, so `DB_POOL_MAX=10` and `DB_POOL_IDLE_TIMEOUT=30000` give back a long-running server's pool. Off Vercel the pool is unchanged. Errors for a database out of connections (PostgreSQL `53300`, MySQL `1040`) or a session closed for idling (PostgreSQL `57P05`) are now reported as connection errors.

### Packages

- `@nextlyhq/adapter-drizzle`
- `@nextlyhq/adapter-mysql`
- `@nextlyhq/adapter-postgres`
- `@nextlyhq/adapter-sqlite`
- `@nextlyhq/admin`
- `@nextlyhq/admin-css`
- `@nextlyhq/blocks-engine`
- `@nextlyhq/blocks-react`
- `@nextlyhq/builder`
- `@nextlyhq/eslint-plugin`
- `@nextlyhq/plugin-form-builder`
- `@nextlyhq/plugin-mcp`
- `@nextlyhq/plugin-page-builder`
- `@nextlyhq/plugin-sdk`
- `@nextlyhq/plugin-seo`
- `@nextlyhq/storage-s3`
- `@nextlyhq/storage-uploadthing`
- `@nextlyhq/storage-vercel-blob`
- `@nextlyhq/ui`
- `create-nextly-app`
- `nextly`

<!-- /release-notes -->

Previous release: [0.0.2-alpha.77](https://nextlyhq.com/changelog/0.0.2-alpha.77) · [All releases](https://nextlyhq.com/changelog)

## Earlier releases

Releases before 0.0.2-alpha.66 are listed in [Nextly's changelog on GitHub](https://github.com/nextlyhq/nextly/blob/main/packages/nextly/CHANGELOG.md).
