This reverts commit 0fc470a. The feature was not wanted. Migration
00013 stays on disk because databases have already run it (goose refuses
a missing applied migration); 00014 drops the column and table it added.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

A page's kind says what it shows between its intro and outro: posts,
as before, or something special. The first special kind is a gallery —
every image in the files library as a masonry wall (CSS columns, no
script), newest first, each linked to its full-size file. The page form
lists the library's images with a Hide tick per picture; exclusions are
stored (page_hidden_files) rather than inclusions so a new upload shows
up without editing the page, and the design's own pictures — logo,
favicon, header and background image — are never shown. Special pages
keep the page chrome (menu, announcements, home, ordering) but hold no
posts: the post forms skip them and a forced page id is refused.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

Announcements showed on every public page. Now each page has a "Show
announcements on this page" checkbox (pages.show_notices, on by default)
that covers the page and its posts, and the Design tab's Content section
has "Show announcements on post pages" (theme.post_notices, on by default)
for bloggers who want notices only on listings, not while reading a post.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

Recreating gizmolab.net as a blog showed three things the design system
could not express, all of them ordinary blog features:

- Listings can show a summary and a "Read more" link instead of the whole
  post (theme.list_style). The summary is the source up to <!--more-->;
  without the marker a Markdown post is cut at a block boundary after 70
  words (never inside a code fence) and an HTML post is shown whole, since
  cutting hand-written markup blind would leave tags open. Computed at
  render time (Post.Excerpt/HasMore), nothing new is stored.
- A post can carry a featured image (posts.image, a library file that is
  unset when the file is deleted): a thumbnail floated beside the entry on
  listings, and on the post page above or below the text, or not at all,
  as theme.post_image says. The picker is the design page's imagepick; its
  library panel moves out of design.html into partials/imagelib.html so
  both forms share it — which also removes a stale pickAction call the
  design form's click handler would have hit.
- A page can have a closing text after its posts (pages.outro_md/html), in
  the intro's format: the per-page counterpart of the blog-wide "after the
  posts" module, for a home page that ends in a site map or a widget. The
  editor partial takes formatof to share another editor's Format row, and
  its upload field is now <name>_file so two editors fit in one form.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

Markdown is the default and unchanged; the editor's new Format switch
stores the text as raw HTML instead, put on the blog exactly as written.
Like the custom HTML module it is unsanitised on purpose: the escape
hatch for embeds, scripts and inline styles that Markdown cannot express.

The source column keeps holding the text in both modes and a `format`
column says how to read it, so the public templates and the feed still
print the stored `*_html`. The dashboard preview of HTML goes into a
sandboxed iframe rather than the page, because a superadmin edits other
people's blogs and their markup must never run on the dashboard origin.
Search snippets of HTML posts are cut from a tag-stripped copy.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

Posts can now carry tags, set on the post form as a checklist of the
blog's existing tags plus a comma-separated box for new ones (no JS).
A tag is a name and a unique slug, so "Go" and "go" are one tag and
Greek tags get readable URLs; tags no post uses any more are deleted.

On the blog, tags appear under the post date and link to /tag/<slug>,
which lists the published posts from every page, paginated like a page.
Two new layout modules show them: a Tags list (with counts, by use) and
a Tag cloud (alphabetical, sized by use). Both are only fetched when a
visible module needs them.

The article loop and pager move to a shared postlist partial; while
there, the pager stops adding a trailing slash — /news/?p=2 was a 404
because a {page} wildcard never matches one.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

A blog is now a database of its own (blog_<sub>) on the same server: one
pg_dump is a complete backup of a blog, one psql restores it, and nothing a
blog's queries do can reach another blog's rows. The control database
(DATABASE_URL) keeps only users and the blog registry (id, owner, subdomain,
db_name); title, tagline and theme move into a one-row settings table next
to the content so the dump really is everything.

db.Cluster holds the control pool plus small, lazily opened per-blog pools.
store.Store (control) hands out a store.BlogStore per blog; every blog_id
parameter and column is gone, the database is the scope. Handlers reach it
through blogStore(r), which resolveBlog puts in the context next to the blog.

Existing data is moved in place by control migration 00006, a Go migration
that runs inside the control transaction: it creates and migrates each blog
database, copies the rows preserving ids, and marks the registry; 00007 then
drops the old tables. Either every blog is moved or the control database is
untouched.

/media/{id} now serves the host's blog only, so dashboard previews on the
root domain use /b/{sub}/media/{id}. Subdomains are capped at 58 chars so
"blog_" + name fits a Postgres identifier. Deleting a user drops their
database. Store.Open resets a blog's pool and retries once so a database
restored under a running app (dropdb --force, createdb, psql) just works.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

A blog page is now header, left column, main content, right column and
footer, each holding an ordered list of modules (blog title, logo, menu,
archive by year/month, recent posts, custom HTML, footer text, RSS link,
site map). Side columns have percentage widths, can be hidden, and can
keep their space when empty; on phones they stack under the posts. The
menu becomes its own ordered list mixing pages and outside links.
Announcements are placed per column at its top or bottom.

Modules and menu items get tables; the migration derives them from each
blog's old theme (nav position, footer text, show-in-nav pages) so
existing blogs look the same. Custom HTML is deliberately stored and
served as-is (owner's decision; see AGENTS.md for why the blast radius
is the blogger's own origin).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A

Go + Postgres application serving a management dashboard on the base
domain and one public blog per subdomain. Markdown posts organised in
pages, form-based theme customisation, image uploads stored in Postgres,
JWT cookie sessions with CSRF, superadmin user management, RSS feeds.
Docker/compose deployment and a Makefile-driven dev environment with
seed data.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Sd8UPWrvyYCLj97JexNw3A