The markdown had no writing section at all, so posts, blocks, drafts and
deletes were only visible in the admin. Adds them with every field, plus
date and number rules and the media delete in both places.
Rows carry a checkbox, the bar above publishes, withdraws, resets to draft,
sets the audience or deletes drafts for the whole selection and reports
what went through.
- the admin uses the full window width, table cells stop wrapping
- columns sort by number, title, project, status, date and author
- filters sit in one row with search first and a page size next to them
- entries carry a real date: publish_at on create and change, publishing
keeps that date, a date in the past publishes backdated
- the status cell offers every transition the entry allows, scheduling
asks for the date right there
- POST /api/v1/posts takes cover_media_id, no second call needed
- the row grid follows its content so long type names stop overlapping
The list showed the status but only the editor could change it. Each row
now carries the next sensible step and names the blockers when publishing
is not possible. Also real umlauts in the API texts.
- the plate carries project colour on the lead image, in the cards and in
every row, not only when an entry lacks an image
- entries no longer fade in one after another, the page is readable at once
- detail pages lead on: previous, next and more from the project
- the month band shows from the first month, the nav item hides without one
- side cards and row teasers get room instead of ellipsis
- admin opens on drafts, scheduled and published plus what was edited last,
and its tables use the full width
Tiptap replaces the plain textarea: bold, italic, subheading, lists and
links. Stored as Markdown so API clients keep reading and writing the
same field, rendered with react-markdown.
- drafts stay unnumbered, the number is handed out when an entry goes live
- DELETE /api/v1/media/:id and a delete control in the media library
- cover_media_id in the draft read, coverMediaId and coverUrl in the archive read
- build the runtime image so libvips lands next to the sharp binary
- GET /api/v1/posts/:slug finds drafts by slug for write clients
- media upload documented with project slug instead of a project id
- own select control instead of the browser default
- quieter sidebar footer, project name leads to its entries
- ship libvips with the runtime image so POST /api/v1/media stops failing
- admin downloads openapi.json and the style guide via session, no token
- copy rows for both API addresses
- anchor mark in the wordmark and a favicon
Standalone Next output, migration script run on start, storage volume
at /app/storage, port 4700. Database connection is created lazily so
the image builds without a database.
Public archive with sidebar navigation, project pages, month archive,
search and entry pages with image blocks. Admin area for brands,
projects, post types, users, api clients, media and the entry editor
with drag and drop images, preview per audience, scheduling and
publish checks. Read and write API with bearer tokens, audience
scoping, idempotent creation, OpenAPI document and editorial guide.
Magic link login with configurable allowed domains, whole app behind
the session gate. 456 tests including design rule checks.
Matthias G