feat(notes): notekroppe forsegles, og søgningen åbner dem selv

Grunden er den samme som for postkassers kodeord og OAuth-tokens: en
sikkerhedskopi kan hentes gennem fladen, og indtil nu bar hver eneste af dem hver
eneste note i klartekst. Noter er dér, folk skriver kodeord og API-nøgler ned — det
ved vi, fordi der ligger sådan nogle i den installation, det her blev bygget til.

**Både titel og krop.** Titlen *er* første linje, så en forseglet krop ved siden af
en læselig titel ville efterlade begyndelsen af hver eneste note i klartekst.

**Og de to kopier gik med.** Det var derfor det ikke var gjort for længe siden:
`notes_fts` indekserede kroppen, og den genererede `fold`-søjle bar en
translittereret kopi af den. At forsegle kroppen og lade dem stå ville have været
teater — en angriber med filen kunne læse hver note ud af indekset. Begge er væk,
og tabellen er bygget om, fordi SQLite ikke kan fjerne en genereret søjle på plads.

**Søgningen åbner teksten i Go.** Målt frem for gættet: tolv hundrede noter à 3,3 kB
koster 33–43 ms pr. søgning, og fladen venter 250 ms efter sidste tastetryk, før den
spørger — så det ligger inde i pausen. Tiden går ikke med at åbne konvolutterne;
fire megabyte AES-GCM er få millisekunder. Den går med at folde og småskrive hver
krop, og det er lineært: regn med omtrent 350 ms ved ti tusind noter. Dér er det den
forkerte form, og svaret er da et nøglet indeks — hvis pris er, at præfiks- og
delordssøgningen falder bort.

bm25 forsvandt med indekset. Det, der er beholdt, er den halvdel, der betød mest:
titlen vejer ti gange kroppen. Med et loft pr. ord, så en note, der nævner ordet
fyrre gange, ikke slår en, der nævner alle ordene én gang — det er den fejl, en ren
optælling laver, og den er let at lave og svær at få øje på.

**En bagudfyldning ved opstart, ikke efterhånden.** Instansens indstillinger
forsegles, når de skrives, og det er rigtigt for dem. For et arkiv på tolv hundrede
importerede noter, som ingen redigerer igen, betyder "når de skrives" aldrig — og
det er præcis dét arkiv, der er værd at forsegle. Idempotent, så den koster et
gennemløb af id'er hver anden gang, og `updated_at` røres ikke: at lukke en
konvolut om en note er ikke at skrive den, og et stempel med i aften ville kaste
den rækkefølge væk, importen brugte en dag på at hente hjem.

En fejl under bagudfyldningen stopper opstarten. Halvdelen af noterne lukket er
ikke i stykker — `v1:`-præfikset gør, at begge slags kan læses — men en instans,
der starter og lader som om den beskytter noget, den ikke gør, er værre end en, der
siger fra.

Hvad der stadig ligger åbent, står i SECURITY.md og på dokumentationssiden:
opgavetitler, beskrivelser, kommentarer, og `note_links`, som røber hvilke
projekter og opgaver en note peger på, men ikke hvad der står.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

.gitignore

@@ -66,3 +66,6 @@ node_modules.nosync/
 # Kept on disk because it is how a session hands over to the next one.
 HANDOFF.md
 NOTES-PLAN.md
+
+# Agenternes arbejdstræer. Et git-repo inde i et git-repo, som ikke hører til her.
+.claude/worktrees/

NOTES-SPEC.md

@@ -0,0 +1,437 @@
+# Notes, as verdande builds them
+
+A description of one working notes system: the data model, the syntax, the search,
+the rendering contract, the import and export formats, and — more useful than any
+of it — the decisions that look arbitrary until you know what they cost.
+
+It is written to be lifted into another program. Nothing here depends on Go,
+SQLite or Svelte except where it says so; where an answer was forced by the tool,
+that is called out so you can pick a different one on purpose rather than by
+accident.
+
+---
+
+## Read this part first: what to ask before you build
+
+**This document is not a drop-in. It describes a notes system that lives inside a
+task app with projects, roles, a trash and a search box, and about a third of its
+design is inherited from that.** Lifting it into a program with different
+surroundings without deciding which parts to keep produces something that works
+for a week and then fights you.
+
+**If you are an assistant working from this file: ask these questions and get
+answers before writing code.** Do not guess a default. Each one changes what you
+build, not just how you build it.
+
+### About the surroundings
+
+1. **Does the host app already have a place to file things** — projects, folders,
+   notebooks, tags? verdande deliberately gave notes no hierarchy of their own and
+   filed them under the projects that already existed. If your app has nothing,
+   you need an answer for where a note lives, and "a folder tree" is the expensive
+   one.
+2. **Is there more than one person?** Sharing, roles and "who wrote this" run
+   through the whole model below. Single-user changes the schema and removes about
+   half the permission surface.
+3. **Is there already a trash, and does it expire?** Notes lean on one. Without
+   it, deleting is either permanent or you are building a trash.
+4. **Is there an existing full-text index?** The search here is FTS5-shaped. On
+   Postgres it is `tsvector`, on a client-side app it is probably a scan, and the
+   ranking section below changes accordingly.
+
+### About the notes themselves
+
+5. **Markdown, rich text, or blocks?** verdande stores Markdown and renders it.
+   That decision is defended below; it is also the one that most constrains
+   everything else.
+6. **Is the first line the title, or is there a title field?** Both work. Deriving
+   it means one less thing to keep in step and one less thing a person can leave
+   blank, and it forces the rule "the first line is a heading" on the editor.
+7. **What links to what?** `#project`, `[[note]]` and task references exist here
+   because the app has projects and tasks. Yours may have none of them, or others.
+8. **Do notes carry files?** Attachments bring content-type handling, a store, an
+   authorisation path and an XSS surface. Skipping them is a legitimate answer.
+
+### About scale and language
+
+9. **How many notes, realistically?** Several choices below flip at around ten
+   thousand: the list payload, the search fallback, rendering the whole list.
+10. **Which languages?** The folding described under *Search* is Danish. Every
+    language has its own version of that problem and the generic answer (strip
+    diacritics) is wrong for several of them.
+11. **Where does the text end up besides the screen?** Search indexes, AI
+    features, exports and backups each read it. Plaintext makes all of them cheap
+    and makes a stolen backup expensive — see *What is still open*.
+
+### About integration, specifically
+
+12. **Who owns the note's identity — you or the other system?** If notes will sync
+    with something else, ids and timestamps are the whole problem, and the
+    `updated_at` rule below becomes the first thing you have to change.
+13. **Is this a read of somebody else's notes, or the store of record?** A copy
+    needs conflict rules; a store of record needs export.
+14. **What happens to a note whose project, task or link target is deleted?**
+    Answered here as "the link goes dead and the note is untouched". If your host
+    app cascades instead, say so before the first delete.
+
+---
+
+## The shape of a note
+
+```
+id           an opaque string, unique
+project_id   nullable — where it is filed, or loose
+title        derived, never set by a caller
+body         Markdown, the whole note
+created_by   nullable — survives the account being deleted
+pinned       favourite; the list puts these first
+archived_at  nullable timestamp — put away, not thrown away
+deleted_at   nullable timestamp — the trash
+created_at
+updated_at
+```
+
+### Markdown in a column
+
+Not rich text, and not a format of our own.
+
+Export was a requirement before a line was written, and **if the stored form is
+the exported form there is no conversion left to lose anything in.** A note stays
+readable by any editor on the day the program is gone. It is also what makes
+search and anything an AI does over notes cheap: both read text.
+
+The cost is real and you should know it before choosing: everything the editor can
+do has to survive a round trip through Markdown, and anything Markdown cannot say,
+the editor must not offer. There is no place to hide a feature.
+
+### The title is the first line
+
+Always, and re-derived on every save — not "when it is empty".
+
+Derived once and then left alone, a title goes stale the moment somebody rewrites
+the opening of a note, and the list ends up calling a note by a name that is
+nowhere in it. **The column is a cache of the body, not a field beside it**, and no
+caller can set it, so every caller gets the same answer.
+
+This forces one editor rule: a new note opens on its title, and the first line is
+styled as a heading. Otherwise a note whose first line is a sentence is a note
+named by its first sentence.
+
+### Archive and trash are two different bits
+
+The trash says *this was a mistake, and in thirty days it is gone*. The archive
+says *this is finished, and I do not want to read past it.*
+
+A twelve-hundred-note import is exactly the case where most notes are worth
+keeping and almost none are worth looking at. Without an archive, the list is the
+whole archive forever, and finding this week's note means reading past a decade.
+
+Both are nullable timestamps rather than booleans: *when was this put away* answers
+questions a flag cannot, and it costs the same byte.
+
+---
+
+## Filing: no hierarchy of its own
+
+A note belongs to a project, or to nobody.
+
+The reasoning is worth repeating because it is easy to talk yourself out of:
+projects already exist, are already shared, already have roles, already have a
+trash. **A second hierarchy to file things in is one too many**, and a note filed
+under a project inherits every one of those answers instead of needing its own.
+
+- A note with a project: everybody who can read the project can read the note; an
+  editor can change it.
+- A note without one: the author's, and nobody else's.
+
+Sharing a note *is* filing it in a project. There is no access list of its own —
+a second way to give somebody access is a second place to get it wrong, and the
+two would disagree the day one of them changed.
+
+---
+
+## Links: three kinds, parsed out of the text
+
+The body is the truth. Links are an index over it, rebuilt on every save.
+
+| Syntax | Means | Notes |
+|---|---|---|
+| `#project` | a project | Same characters the task quick-add parser uses, on purpose |
+| `[[a note]]` | another note, by title | The spelling every note app has settled on |
+| `/opgave/<uuid>` or `task:<uuid>` | a task | What a link pasted out of the app carries |
+
+Patterns as used here:
+
+```
+project   (?:^|\s)#([\p{L}\p{N}_-]{1,64})
+note      \[\[([^\]\n]{1,200})\]\]
+task      (?:/opgave/|task:)([0-9a-fA-F-]{36})
+```
+
+Three decisions inside that table:
+
+- **`#Firma` in a note means the same project as `#Firma` in the task box.** Not a
+  second kind of tag that happens to look alike. That is the whole of "tag a note
+  and see it in the project".
+- **The link table is torn down and rebuilt on save, not diffed.** A body is small
+  and the set is small; a diff is a second description of the same fact, and it
+  can disagree with the text. That is the one thing this must never do.
+- **The target is not a foreign key.** A link to something deleted is a fact about
+  the note. It should read as a dead link rather than vanish, and it must not
+  block the delete.
+
+The index exists so the question can be asked **backwards**: *what has been written
+about this task?* Answering that by searching every note's prose would be too slow
+to leave a panel open on.
+
+---
+
+## Search
+
+Three layers, in the order they answer.
+
+### 1. Full-text, ranked
+
+FTS5 over `title`, `body` and a folded copy, with `unicode61 remove_diacritics 2`.
+
+Ranking is bm25 with **the title weighted ten times the body**. A word in the title
+is what the note is *about*; the same word once in a long body is a mention.
+
+This matters more than it sounds. Ordering by "last touched" is invisible with ten
+notes and useless with twelve hundred imported the same evening: searching "Tesla"
+answered with a note about nginx first and the note actually called "Tesla Model Y"
+second — which reads as a search that does not work, because from where the person
+is sitting it is one.
+
+### 2. Folding, for the language you are in
+
+`remove_diacritics` does not touch `ø`, `æ` or `å` — Unicode considers them
+letters in their own right, not accented forms. So a Dane cannot find "grøn" by
+typing "gron", which is how people type when the keyboard is not theirs.
+
+The answer here is a generated column carrying a transliterated copy:
+
+```
+ø,Ø → o     æ,Æ → ae     å,Å → aa
+```
+
+indexed alongside the body and **weighted the same as the body** — otherwise a
+folded match outranks an exact one.
+
+> **Ask, do not copy.** This mapping is Danish. German wants `ü → ue`, Swedish
+> disagrees with Danish about `ä`, and Turkish has a dotless `ı` that breaks
+> naive lowercasing entirely. Get the mapping from somebody who types the
+> language.
+
+### 3. A word inside a word
+
+An index is built of tokens, and a token is whatever the tokenizer decided was a
+word. `gamerule keepInventory true` is three of them, and the middle one is
+`keepinventory` — so a search for "keep inventory" asks for two words that are
+nowhere and gets nothing, while the note sits right there.
+
+A prefix query gets half of it: `keep*` matches, `inventory*` cannot, and the two
+are ANDed, so the half that works is thrown away by the half that cannot.
+
+**This is not exotic in a notes app.** A decade of notes is full of commands,
+identifiers and file names, and every one of them is several words the index
+thinks is one.
+
+So: the index answers first; when it has nothing, the text is read, and every term
+must appear somewhere in the note, in any word. A `LIKE '%term%'` over the folded
+column covers title, body and language folding at once.
+
+What it costs: a leading `%` cannot use an index, so it scans. At twelve hundred
+notes that is a few megabytes and a millisecond or two, and it only runs when the
+fast path already came back empty — which is exactly the search that was about to
+disappoint somebody. **At a hundred thousand notes it is the wrong shape**, and the
+answer then is a trigram index, not a bigger scan.
+
+It does not rank. There is no match to rank — that is the whole situation.
+
+---
+
+## The list, and why the body is empty in it
+
+A list showing a title and ninety characters of preview must not fetch whole
+notes. Twelve hundred notes came to 3.9 MB, of which 98 % was body nobody could
+see.
+
+The important half is *how* it is trimmed: **the body is emptied, not shortened.**
+
+A field called `body` holding a *piece* of the body is a trap. The editor opens
+it, the autosave pause writes it back, and the note is now cut down to its own
+excerpt — destroyed by nothing more than having been looked at. Send `preview` as
+its own field and leave `body` empty; the client sees the empty body and fetches
+the whole note.
+
+That costs one more separation, and it is not optional:
+
+> **Which row is selected and which note is loaded are two different things.**
+
+The editor loads its text when the note's id changes. Hand it the excerpt first
+and the whole note second and the id is the same both times — so the second one
+never renders, and somebody sits looking at a note that had a title a moment ago.
+
+Highlight the row immediately from the list; give the editor the note only when it
+is whole.
+
+Rendering cost, separately: with twelve hundred rows, first paint took 2.1 s. CSS
+`content-visibility: auto` with an intrinsic-size guess took it to 0.95 s with no
+virtual list, no scroll maths, and no list that can fall out of step with itself.
+Try that before building windowing.
+
+---
+
+## The editor contract
+
+Rich text on top, Markdown underneath, and a source view that shows the file.
+
+The supported subset — deliberately small, because everything in it has to survive
+the round trip:
+
+| Block | Markdown |
+|---|---|
+| Title | `# ` |
+| Heading | `## ` |
+| Subheading | `### ` |
+| Quote | `> ` |
+| Code block | ```` ``` ```` with an optional language |
+| Bullet and numbered lists | `- `, `1. `, nested |
+
+Inline: `**bold**`, `*italic*`, `~~strikethrough~~`, `` `code` ``, `<u>underline</u>`,
+and images as `![alt](url)`.
+
+Five rules that were each learned the hard way:
+
+1. **Every line becomes exactly one block.** Markdown normally joins consecutive
+   lines into a paragraph. Here it must not: the person typed a line and expects
+   to get it back, and an editor that silently reflows what somebody wrote is an
+   editor they stop trusting.
+2. **Load on id change, never while typing.** Writing to the editor's HTML puts
+   the caret back at the start, which mid-sentence is unusable.
+3. **One definition of what a code block contains.** A copy button inside the
+   `<pre>` and a syntax highlighter reading `textContent` will disagree, and the
+   difference ends up in the file — a click on "Copy" wrote the word *Copied* into
+   the code and the next save put it on disk. Have one function that says what a
+   block holds, used by the highlighter and the Markdown serialiser both.
+4. **A note that cannot be rendered must not take the pane with it.** One note
+   arriving in an unexpected shape threw inside the editor's reactive effect, and
+   from then on *no* note would open — the previous one just stayed on screen. Put
+   a guard around the render, fall back to showing the raw text, and let the rest
+   of the pane keep working.
+5. **Escaping between tags and escaping inside an attribute are different jobs.**
+   An escaper written for text between two tags handles `&`, `<`, `>` — and not
+   quotes, because between tags they are harmless. Inside an attribute they close
+   it. `![" onerror="alert(1)](…)` is the whole exploit.
+
+---
+
+## Attachments
+
+Images pasted or dragged into a note, stored content-addressed by hash, referenced
+from the body as ordinary Markdown images.
+
+The security shape, which is the part worth copying:
+
+- **Served as `application/octet-stream` with `nosniff` by default.** Attachments
+  are the one place a user supplies content another user opens. An uploaded SVG or
+  HTML file rendered inline runs its own script on your origin with the session
+  cookie attached.
+- **One exception, as a named allowlist:** a short list of raster image types
+  served inline as their exact type — `png`, `jpeg`, `gif`, `webp`, `avif`, `heic`.
+  A note carries its pictures in the text, and served as octet-stream every one of
+  them is a broken image.
+- **The list is named one type at a time, never matched on an `image/` prefix.**
+  The prefix lets `image/svg+xml` through, and an SVG is a document that can run
+  script.
+- **The stored path is checked against the files directory again at read time.** A
+  traversal that somehow got into the database must not become one that reads
+  `/etc/passwd`.
+- **Reaching an attachment follows the note, which follows its project.** Same
+  chain as everything else. Do not let a client hand you the parent id and take it
+  on trust.
+
+---
+
+## Import and export
+
+The same format both ways: **a zip of Markdown files**, one per note, with a YAML
+front matter block.
+
+```markdown
+---
+created: 2019-04-02T10:15:00Z
+modified: 2024-11-30T21:04:00Z
+---
+
+# Møde med Anders
+
+Han vil gerne have kaffe hver uge.
+```
+
+- `created` / `created_at` / `date` and `modified` / `updated` / `updated_at` are
+  all accepted on the way in. Obsidian and Bear write this block; so should you.
+- Images travel in the zip beside the notes, and the links in the text are
+  relative — **a note and its pictures must end up in the same archive**, which
+  matters the moment an export is split into parts.
+- File names collide: two notes called "Møde" are two notes, and a zip with one
+  entry twice is a zip that loses one.
+
+Limits that a real import taught, all of them cheap to add and expensive to omit:
+
+- A ceiling on the upload, on the number of entries, and on the **unpacked** size.
+  64 MB of zip expanded to 66 GB is a normal shape for an attack.
+- A per-file limit as well as a whole-archive budget: one enormous photograph and
+  ten thousand small ones are different attacks with the same ending.
+- The entry name is never used as a path. Store by hash; the name is only a key
+  for rewriting links.
+- **Count what you skip, and say so.** A note whose name contained an ellipsis was
+  dropped silently by a filter looking for `..` — one out of 1141, found only by
+  counting at both ends.
+
+### Timestamps on the way in
+
+An importer needs to write `created_at` and `updated_at` from the file, and the
+normal save path must refuse to let it.
+
+`updated_at` means *when did this program last write this note*. Let any caller set
+it and the column means whatever the last writer felt like. Give the importer its
+own narrow entry point instead, and say in the comment why it exists.
+
+Without those dates, a move of twelve hundred notes is a pile that was all written
+the same evening — and the order they were written in is half of what an archive
+is.
+
+---
+
+## What is still open here
+
+Written down because a specification that only lists what works is a specification
+that lies by omission.
+
+- **Note bodies are plaintext at rest.** They hold API keys and passwords, and
+  they are in every backup anybody can download. Encrypting them is bigger than it
+  sounds: the body is in the full-text index and the folded column is generated
+  from it, so both are a second plaintext copy in the same database. Encrypt and
+  both have to go, and search has to be rewritten to decrypt in application code —
+  fine at a thousand notes, a full scan per keystroke at ten thousand, at which
+  point the answer is a keyed index (an HMAC per word).
+- **Import never merges.** Every run creates new notes. Say so out loud in the UI.
+- **There is no sync.** Export and the API are the way out.
+
+---
+
+## A short list of the traps, for the person in a hurry
+
+1. A `body` field holding an excerpt will be saved over the real one.
+2. Selected row and loaded note are two different pieces of state.
+3. The title must be re-derived on every save or it goes stale.
+4. Diacritic stripping is not folding, and folding is language-specific.
+5. A tokenizer cannot see a word inside a word; decide what you do about it.
+6. An attachment served inline is a script on your origin unless you allowlist by
+   exact type and send `nosniff`.
+7. Escaping for text and escaping for attributes are different functions.
+8. One render failure must not be allowed to freeze the whole pane.
+9. An unpacked-size limit is not optional on an archive import.
+10. Let the importer set timestamps; let nobody else.

README.md

@@ -6,7 +6,7 @@
 </p>
 
 <p align="center">
-  <img src="site/screenshots/today.png" alt="The Today view: projects, filters and labels in the sidebar, beside the tasks due today." width="900">
+  <img src="docs/screenshots/today.png" alt="The Today view: projects, filters and labels in the sidebar, beside the tasks due today." width="900">
 </p>
 
 ---
@@ -35,6 +35,11 @@ mail-to-task address, and an MCP endpoint so Claude can work your task list dire
   in Firma, labelled regnskab. Danish and English, mixed freely in one line.
 - **Projects, sections, sub-tasks, labels, saved filters.** List, board and calendar
   views, and foldable groups over the projects in the sidebar.
+- **Notes, beside the work.** Rich text that is Markdown underneath — headings,
+  lists, quotes, fenced code. A note is shared by filing it in a project;
+  `#project` and `[[another note]]` are links, and a task shows the notes that
+  mention it. A zip of Markdown goes in — Obsidian, Bear, or Apple Notes through
+  the exporter in [`tools/`](tools/) — and comes back out the same shape.
 - **Drag where it means something.** Reorder a list or a board, file a project under
   a group, drop a task on another project, or drop it on another day.
 - **Sharing.** Owner, editor and viewer roles. There is no open registration:
@@ -90,21 +95,30 @@ Apple Reminders, forward mail into it, and add it to Claude as a connector.
 | Project templates | Done, tested |
 | Nightly backups with rotation | Done, tested |
 | Comments, attachments, notifications, Web Push | Built |
+| Notes: rich text, Markdown on disk, links to projects, notes and tasks | Done, tested |
+| Notes in and out: a zip of Markdown, both directions | Done, tested |
 | Todoist CSV import and export, account export | Done, tested |
 | MCP server for Claude | Done, tested |
 | CalDAV server (two-way VTODO) | Done, tested |
 | Mail-to-task | Done, tested |
 | AI layer (Anthropic/OpenAI/Google/local) | Built, degrades when unconfigured |
 | Gmail: OAuth2 with PKCE, polling, one-way to tasks | Done, tested |
+| Google Calendar: read-only, laid over the task calendar | Built, never run against Google |
+| Calendar subscriptions: ICS and webcal, several per person | Done, tested |
+| Times on tasks, set by dragging in the week grid | Done, tested |
+| Notes: archive, grouped list, images, source view | Done, tested |
 | Update notice for administrators | Done, tested |
 | Personal API tokens | Done, tested |
 | OpenAPI 3.1 spec, checked against the router | Done, tested |
 | End-to-end smoke tests (Playwright) | Done |
 | Documentation, landing page, licence | Done |
 
-**Not finished:** Web Push, the AI providers and the Gmail API calls are
+**Not finished:** Web Push, the AI providers and the Google Calendar API calls are
 implemented against their specifications and have only ever been exercised there,
-not against live services. Everything else has an interface: opening a task shows
+not against live services. Gmail is no longer in that list — it runs in production,
+including the flagging and de-duplication that took a day to get right. Calendar
+subscriptions over ICS and webcal are proven; the Google half of the calendar is
+not. Everything else has an interface: opening a task shows
 its description, sub-tasks, comments, files and reminders, and the settings pages
 cover the account, notifications, integrations, AI, API tokens, import, export and
 templates.
@@ -116,17 +130,20 @@ accented `o` and a Danish-first app that cannot do this is broken for its own us
 
 | | |
 |---|---|
-| ![Board view](site/screenshots/board.png) | ![Light theme](site/screenshots/today-light.png) |
+| ![Board view](docs/screenshots/board.png) | ![Light theme](docs/screenshots/today-light.png) |
 | Board view — one column per section. | The light theme. |
-| ![A week on Upcoming](site/screenshots/week.png) | ![A task, opened](site/screenshots/detail.png) |
-| A week at a time, so a task can be dragged across a month boundary. | A task, opened. |
-| ![The five themes](site/screenshots/themes.png) | ![Instance history](site/screenshots/history.png) |
-| Five themes, all measured for contrast. | What has been done on the server. |
+| ![A week on Upcoming](docs/screenshots/week.png) | ![A task, opened](docs/screenshots/detail.png) |
+| A week at a time, so a task can be dragged across a month boundary. | A task, opened: description, priority, labels, sub-tasks and its files. |
+| ![Notes](docs/screenshots/notes.png) | ![The same note as Markdown](docs/screenshots/notes-source.png) |
+| Notes, filed in a project to share them. | The same note as its source — which is what the export writes to disk. |
+| ![Themes and looks](docs/screenshots/themes.png) | ![Instance history](docs/screenshots/history.png) |
+| Five themes and four looks: how bright, and how it reads. | What has been done on the server. |
 
 Regenerate these with `node scripts/generate-screenshots.js` from `web/`. It starts
-the real binary, creates its own account, fills it with demo data and photographs
-eleven surfaces — nothing to set up and nothing to remember. (`go run ./tools/shots`
-still works, and is the way to photograph an instance that is already running.)
+the real binary, creates its own account and a second one to share with, fills them
+with demo data and photographs fourteen surfaces — nothing to set up and nothing to
+remember. (`go run ./tools/shots` still works, and is the way to photograph an
+instance that is already running.)
 
 ## Running it
 

SECURITY.md

@@ -10,13 +10,28 @@ reply within a week.
 
 - **Passwords** are Argon2id with the parameters recorded in the hash, so they can
   be raised later without locking anyone out.
-- **Every token** — session, invite, reset, API, calendar feed, mail address — is
-  stored only as a SHA-256 hash. A database dump contains nothing that can be
-  presented.
+- **Session, invite, reset and API tokens** are stored only as a SHA-256 hash. A
+  database dump contains nothing that can be presented.
+- **The calendar feed address and the mail-to-task address are not hashed.** They
+  are looked up by their exact value, and they are shown in the interface more
+  than once, which a hash cannot do. Anybody holding a backup file therefore holds
+  those two URLs: the feed reads that person's tasks without a login, and the mail
+  address files tasks into their inbox. Rotating either from **Settings →
+  Integrations** invalidates the old one. This is written down rather than fixed
+  because fixing it means the address can only ever be shown once, which is a
+  decision about how the feature works and not only about how it is stored.
+- **Mailbox passwords, OAuth tokens, TOTP secrets and AI keys** are sealed with
+  AES-GCM under a key kept outside the database — see `internal/secret`. That is
+  what makes a stray copy of the database less than a stray copy of everything:
+  the backup does not carry the key.
 - **Sessions** are `httpOnly`, `SameSite=Lax`, and `__Host-` prefixed over HTTPS.
-- **Uploads** are stored content-addressed by hash and always served as
-  `application/octet-stream` with `nosniff`, so an uploaded SVG cannot run script
-  on this origin.
+- **Uploads** are stored content-addressed by hash and served as
+  `application/octet-stream` with `nosniff` — with one exception, added so that a
+  note can show the pictures in it: a short allowlist of raster image types is
+  served inline as its own type. The list is named one format at a time rather
+  than matched on an `image/` prefix, because the prefix would let `image/svg+xml`
+  through, and an SVG is a document that can run script. `nosniff` stays, so the
+  browser must read the file as that image or not at all.
 - **The filter language** compiles to parameterised SQL. No typed text is ever
   concatenated into a statement.
 - **Permission checks** are route middleware, and anything you may not see answers
@@ -24,8 +39,28 @@ reply within a week.
 - **The CSP** is computed at startup by hashing the inline scripts in the built
   page, rather than allowing `unsafe-inline`.
 
+- **Note titles and bodies are sealed** under the same key, since 0027. Both, not
+  only the body: a note's title *is* its first line, so a readable title beside a
+  sealed body would leave the beginning of every note in the clear.
+
+  The two plaintext copies went with it. `notes_fts` indexed the body and the
+  generated `fold` column held a transliterated copy of it — sealing the body while
+  leaving those would have been theatre, since a stolen file could be read straight
+  out of the index. Search opens the text in Go instead, which costs 33–43 ms
+  across twelve hundred notes and is the wrong shape somewhere around ten thousand.
+- **Task titles, descriptions and comments are still in the clear**, and so is
+  `note_links` — ids derived from a note's body, so a stolen file still shows which
+  projects and tasks a note points at, though not what it says.
+
 ## What it does not do
 
-verdande has no rate limiting beyond the login and password-reset endpoints, and
-no audit log of reads. It is built to be run behind something — a Cloudflare
-Tunnel, a reverse proxy, a VPN — by somebody who knows what they exposed.
+Rate limiting covers the endpoints a stranger can reach: signing in, resetting a
+password, the second factor, the installation beacon, and inbound mail. **Nothing
+else is limited** — an authenticated caller can ask as often as they like, and a
+signed-in account that wanted to be a nuisance can be one.
+
+There is no audit log of reads. Who *changed* something is recorded per project;
+who looked at it is not.
+
+It is built to be run behind something — a Cloudflare Tunnel, a reverse proxy, a
+VPN — by somebody who knows what they exposed.

cmd/verdande/main.go

@@ -64,6 +64,22 @@ func run() error {
 	}
 	db.UseSecrets(box)
 
+	// Noter, der stadig ligger åbne, lukkes nu.
+	//
+	// Ved opstart og ikke efterhånden som de skrives: et arkiv på tolv hundrede
+	// importerede noter, som ingen redigerer igen, ville ellers aldrig blive
+	// forseglet — og det er præcis dét arkiv, der er værd at forsegle. Idempotent,
+	// så den koster et gennemløb af id'er hver anden gang.
+	//
+	// En fejl her stopper opstarten. Halvdelen af noterne lukket er ikke i stykker,
+	// men en instans, der starter og lader som om den beskytter noget, den ikke gør,
+	// er værre end en, der siger fra.
+	if sealed, err := db.SealNotes(context.Background()); err != nil {
+		return fmt.Errorf("seal notes: %w", err)
+	} else if sealed > 0 {
+		log.Info("notes sealed", "count", sealed)
+	}
+
 	httpapi.Version = version
 	api := httpapi.New(cfg, db, log, webAssets(log))
 
@@ -75,7 +91,9 @@ func run() error {
 	runner := jobs.New(cfg, db, log, api.Mail(), api.Hub())
 	runner.SyncGmail = api.SyncGmail
 	runner.SyncMailbox = api.SyncMailbox
+	runner.SyncCalendars = api.SyncCalendars
 	runner.Push = api.PushToUser
+	runner.SendBeacon = api.SendBeacon
 	runner.Start(jobCtx)
 
 	srv := &http.Server{

docs/caldav.md

@@ -1,13 +1,15 @@
 # Calendar and CalDAV
 
-Two ways to get your tasks into a calendar, and they are not the same thing.
+Two ways to get your tasks into a calendar. Getting somebody else's calendar
+*into* verdande is the other direction and has its own page — see
+[Calendar](calendar.md).
 
-| | [Feed](#subscribing-to-a-feed) | [CalDAV](#connecting-a-caldav-client) |
-|---|---|---|
-| Direction | Read only | Two-way |
-| Works with | Anything that subscribes to a URL | Apple Reminders, Thunderbird, DAVx⁵ |
-| You can tick things off | No | Yes |
-| Setup | Paste a URL | Username and token |
+| | [Feed](#subscribing-to-a-feed) | [CalDAV](#connecting-a-caldav-client) | [Calendars in](calendar.md) |
+|---|---|---|---|
+| Direction | Tasks out, read only | Tasks out, two-way | Events **in**, read only |
+| Works with | Anything that subscribes to a URL | Apple Reminders, Thunderbird, DAVx⁵ | A Google account |
+| You can tick things off | No | Yes | No — an event is not a task |
+| Setup | Paste a URL | Username and token | Connect, then pick calendars |
 
 ## Subscribing to a feed
 
@@ -39,6 +41,13 @@ most people actually keep.
     **Settings → Calendar feed → Rotate** issues a new one and breaks the old
     immediately.
 
+## Showing a Google calendar in verdande
+
+Moved. Connecting a Google account, subscribing to an address, and what the grid
+does with either now live on their own page — see [Calendar](calendar.md).
+
+This page is the other direction: verdande's own tasks, read by something else.
+
 ## Connecting a CalDAV client
 
 This is the one where ticking something off in Apple Reminders ticks it off here.

docs/calendar.md

@@ -0,0 +1,91 @@
+# Calendar
+
+Your dated tasks as a grid, with other people's calendars laid under them.
+
+**Calendar** in the menu shows a month or a week. The week is a day: hours down the
+side, and everything with a time on it in its place. Two things at ten o'clock sit
+beside each other, because that is a clash you should be able to see rather than
+read.
+
+## What is in it
+
+- **Your tasks**, on the day they are due — and at the time, if they have one.
+- **Events from calendars you have connected**, read-only.
+
+Drag a task to another day to move it. Drop it on the hour column and it takes that
+time; drop it in the band across the top and it keeps the day and loses the time.
+Events cannot be dragged: verdande holds a copy of somebody else's calendar, and a
+chip that lets itself be moved is a promise the server cannot keep.
+
+A project's own calendar is under **Calendar** on the project page, with the same
+week and month.
+
+## Subscribing to a calendar
+
+**Settings → Integrations → Subscriptions.** Paste an address and it is fetched at
+once.
+
+This works with anything that publishes an `.ics` file: public holidays, a club's
+fixtures, a shared calendar from work, your own calendar's secret address from
+Google, Apple or Fastmail.
+
+- **`webcal://` addresses work.** They are the same URL over https with a scheme
+  that makes a desktop hand it to the calendar app; it is rewritten for you.
+- **Several at a time.** Unlike the Google connection, which is one account, a
+  subscription is one address and you can have as many as you like.
+- **Read-only, refreshed every fifteen minutes.**
+- The address is fetched once before it is stored, so a typo tells you now rather
+  than by silently never filling.
+
+!!! note "This is the way in for a private account"
+    If the instance's Google registration is *Internal*, Google refuses accounts
+    from outside the organisation with `org_internal` — no setting on your side
+    changes that. Your calendar's own secret ICS address goes through this panel
+    and works regardless.
+
+## Connecting Google Calendar
+
+**Settings → Integrations → Google Calendar.** It shares the OAuth client Gmail
+uses, so the instance needs that registration first — see
+[Configuration](configuration.md#google).
+
+Two things have to be true in Google Cloud, and only the owner of the registration
+can do them:
+
+1. **Google Calendar API** is enabled for the project.
+2. The calendar callback is listed as an authorised redirect URI. The exact string
+   is printed in the panel — copy it from there rather than typing it. Google
+   compares it exactly: scheme, host, path, no trailing slash. `redirect_uri_mismatch`
+   almost always means one of those four differs.
+
+Once connected, tick the calendars you want to see. The list is refreshed on every
+sync, so a calendar added or renamed since last time appears under its new name.
+
+!!! warning "Written to the specification, not yet proven against Google"
+    The Google Calendar half is implemented from Google's documentation and tested
+    against a stand-in server. It has not been exercised against Google itself.
+    The subscription route above and the outgoing feed below have.
+
+## Your own calendar, elsewhere
+
+Two ways out, for two different jobs:
+
+- **[The ICS feed](caldav.md#subscribing-to-a-feed)** — a read-only address for Apple Calendar,
+  Google or Thunderbird. Your tasks appear as dated entries.
+- **[CalDAV](caldav.md)** — two-way. Tick something off in Apple Reminders and it
+  is ticked off here.
+
+The feed address is a credential in itself: anybody holding it can read your tasks
+without logging in, because a calendar client cannot sign in. Rotate it from
+**Settings → Integrations** if it gets out.
+
+## What it does not do
+
+- **It does not write to a connected calendar.** Events are a copy.
+- **It does not expand repeating events** from a subscribed file. A weekly meeting
+  in somebody else's calendar is stored as its first occurrence.
+- **It does not resolve named time zones** from a subscribed file. An event carrying
+  `TZID=Europe/Copenhagen` shows the clock the file wrote, rather than one this
+  server has moved.
+- **Only Google, and only reading**, for account connections. Everything else goes
+  through a subscription.

docs/configuration.md

@@ -39,12 +39,24 @@ Any Go duration: `720h`, `30m`, `45s`.
     yourself, password resets are written to the log, and reminders arrive in the
     app and as Web Push rather than by email.
 
-## Gmail
+## Google
 
-Only needed if anyone will connect a mailbox. Register an OAuth client in Google
-Cloud (**APIs & Services → Credentials → OAuth client ID → Web application**) with
-the redirect URI `https://your-address/oauth/gmail/callback`, enable the Gmail API,
-and add the `gmail.readonly` scope.
+Only needed if anyone will connect a Gmail mailbox or a Google calendar. One
+registration covers both: register an OAuth client in Google Cloud
+(**APIs & Services → Credentials → OAuth client ID → Web application**) and add
+**both** redirect URIs:
+
+```
+https://your-address/oauth/gmail/callback
+https://your-address/oauth/calendar/callback
+```
+
+Then enable the APIs and scopes for whichever you want:
+
+| For | Enable | Scope |
+|---|---|---|
+| Mailboxes | Gmail API | `gmail.readonly` |
+| Calendar | Google Calendar API | `calendar.readonly` |
 
 | Variable | What it does |
 |---|---|
@@ -52,8 +64,15 @@ and add the `gmail.readonly` scope.
 | `VERDANDE_GMAIL_CLIENT_SECRET` | The client secret. |
 
 One registration serves every user on the instance; each person authorises their
-own mailbox from **Settings → Gmail**. The redirect URI is derived from
-`VERDANDE_BASE_URL`, so it cannot drift out of step with what you registered.
+own mailbox from **Settings → Integrations → Gmail** and their own calendar from
+**Settings → Integrations → Google Calendar**. Both redirect URIs are derived from
+`VERDANDE_BASE_URL`, so they cannot drift out of step with what you registered.
+
+!!! note "Two connections, two grants"
+    Connecting a calendar does not disturb a working Gmail connection, and
+    disconnecting one does not take the other with it. Google issues a refresh
+    token per authorisation and keeps up to a hundred of them live per account,
+    each carrying only the scopes it was granted with.
 
 ## Secrets
 

docs/index.md

@@ -22,6 +22,9 @@ forward mail to, and an MCP endpoint so Claude can work your task list directly.
   Firma, labelled regnskab. Danish and English, mixed freely in one line.
 - **Projects, sections, sub-tasks, labels, [saved filters](filters.md).** List,
   board and calendar views, and foldable groups over the projects in the sidebar.
+- **Notes, beside the work.** Rich text to write in, Markdown on disk. A note is
+  shared by filing it in a project; `#project` and `[[another note]]` are links,
+  and a task shows what has been written about it.
 - **Drag where it means something** — reorder a list or a board, file a project
   under a group, drop a task on another project, or drop it on another day.
 - **Sharing** with owner, editor and viewer roles. Anyone outside your instance
@@ -31,6 +34,33 @@ forward mail to, and an MCP endpoint so Claude can work your task list directly.
 - **[Calendar](caldav.md), [mail](mail.md) and [Claude](mcp.md)** connected over
   open standards rather than private formats.
 
+## What it looks like
+
+![Upcoming as a month](screenshots/upcoming.png)
+
+*Upcoming, as a month. A week and a plain list are the other two, and a task can be
+dragged from one day onto another.*
+
+![A project as a list](screenshots/sections.png)
+
+*The same project as a list, where sections read as bands. As a board they are
+columns.*
+
+![Waiting on others](screenshots/delegated.png)
+
+*What you have handed to somebody else, grouped by whoever has it.*
+
+![Settings, integrations](screenshots/settings.png)
+
+*Where [Gmail](mailboxes.md), the [calendar feed](caldav.md) and the
+[mail-to-task address](mail.md) are connected.*
+
+![Settings, people](screenshots/admin.png)
+
+*Inviting somebody to the instance. There is no open registration — everybody
+arrives through a link, and picks their own password. See
+[signing in](signing-in.md).*
+
 ## Getting it running
 
 verdande is built to be deployed as a **Rune** in

docs/mailboxes.md

@@ -16,7 +16,7 @@ administrator who can read yours.
 **Settings → Integrations → Gmail → Connect.**
 
 Gmail needs an OAuth client, which is the instance's registration with Google
-rather than anybody's account — see [Configuration](configuration.md#gmail).
+rather than anybody's account — see [Configuration](configuration.md#google).
 Once it is set up, each person connects their own mailbox through it.
 
 Choose what makes a task:

docs/notes.md

@@ -0,0 +1,163 @@
+# Notes
+
+Markdown in, Markdown out. Nothing in between.
+
+A note is written as rich text — headings, lists, quotations, code — and stored as
+the Markdown it looks like. That is not a detail about the database: it is what
+makes a note yours. Export gives you the same text you were looking at, and on the
+day this program is gone, every note opens in any editor.
+
+## Writing
+
+Start typing. **The first line is the title** — always, and re-derived every time
+you save, so a note you rewrite the top of is called by what it now says rather
+than by what it once did.
+
+| Style | Written as | Shortcut |
+|---|---|---|
+| Title | `# ` at the start of the line | pick it in **Aa** |
+| Heading | `## ` | |
+| Subheading | `### ` | |
+| Quote | `> ` | |
+| Code block | three backticks, with an optional language | |
+| Bullet and numbered lists | `- ` and `1. `, nested with Tab | |
+| Bold | two asterisks around the words | ⌘B |
+| Italic | one asterisk around the words | ⌘I |
+| Underline | a `u` tag around the words | ⌘U |
+| Strikethrough | two tildes around the words | |
+| Code | one backtick around the words | |
+
+Every line becomes exactly one block. Markdown would normally join consecutive
+lines into a paragraph; here it does not, because you typed a line and expect to
+get it back.
+
+Press the source button in the toolbar to see the note as the file it is. Edits there are edits to the
+note.
+
+### Pictures
+
+Paste one, or drag it onto the note. It is stored with the note and travels with
+it in an export. Every raster format works; SVG is stored but not shown inline —
+it is a document that can run script, and a note is not a place to run somebody
+else's.
+
+### Code blocks
+
+A fenced block is coloured for the language it names, and guesses when it does not.
+Each one carries a **Copy** button in its corner.
+
+## Linking
+
+Three kinds of reference, parsed out of the text as you write it:
+
+| You type | It means |
+|---|---|
+| `#Firma` | The project — the same `#Firma` quick add understands, not a second kind of tag |
+| `[[Møde med Anders]]` | Another note, by its title |
+| A pasted task link | That task |
+
+Links are an index over the text, rebuilt every time the note is saved. The text
+stays the truth.
+
+**The interesting direction is backwards.** A task shows what has been written
+about it, and a project shows the notes filed in it — without anybody having to
+remember to attach anything. Write `#Firma` in a note and it appears on that
+project's page.
+
+A link to something that has been deleted reads as a dead link. It is a fact about
+the note, and it should not vanish because something it mentioned did.
+
+## Filing and sharing
+
+A note belongs to a project, or to nobody.
+
+- **In a project:** everybody who can read the project can read the note, and an
+  editor can change it. Choose it under **Share in** at the foot of the note.
+- **In no project:** yours, and nobody else's.
+
+There is no folder tree. Projects, groups and labels already exist, are already
+shared and already have roles — a second hierarchy to file things in is one too
+many, and a note in a project inherits every one of those answers.
+
+## Finding
+
+The list groups itself: favourites first, then **Today**, **Yesterday**, **This
+week**, and then by month. Sort by created, last touched or name, and the grouping
+follows — sorted by name it groups by first letter instead.
+
+Search finds a word wherever it is:
+
+- **`gron` finds `grøn`.** Unicode does not consider `ø` an accented `o`, so
+  stripping diacritics is not enough for Danish. A folded copy of every note is
+  indexed beside the original.
+- **`inventory` finds `keepInventory`.** An index is built of whole words, and a
+  decade of notes is full of commands and identifiers that are several words in
+  one. When the index has nothing, the text itself is read.
+- Results are ranked by how well they match, with the title weighted ten times the
+  body. A word in the title is what the note is about; the same word once in a long
+  body is a mention.
+
+## Putting notes away
+
+Two different things, and the difference matters:
+
+- **Archive** — this is finished and I do not want to read past it. It stays
+  findable and comes back whenever you ask.
+- **Trash** — this was a mistake. It is gone in thirty days.
+
+Select several with ⌘-click and shift-click to archive or delete them together.
+
+## In and out
+
+**Export** gives you a zip of Markdown files, one per note, with the dates in a
+YAML front matter block that Obsidian and Bear also read:
+
+```markdown
+---
+created: 2019-04-02T10:15:00Z
+modified: 2024-11-30T21:04:00Z
+---
+
+# Møde med Anders
+
+Han vil gerne have kaffe hver uge.
+```
+
+**Import** takes the same shape back. Pictures travel in the zip beside the notes,
+and the links in the text are relative — so a note and its pictures must be in the
+same archive, which matters when a large export is split into parts.
+
+Two things to know before importing:
+
+- **Nothing is merged.** Every run creates new notes. Importing the same archive
+  twice gives you two of everything.
+- **The dates are read from the front matter.** Without them, a move of a thousand
+  notes is a pile that was all written the same evening — and the order they were
+  written in is half of what an archive is.
+
+### From Apple Notes
+
+`tools/apple-notes-til-markdown.py` in the repository exports Apple Notes to
+exactly this shape, front matter and pictures included. It is macOS-only, because
+Apple Notes is.
+
+For a large library, `tools/pak-noter-i-portioner.py` splits the export into
+uploadable parts, keeping each note with its own pictures.
+
+## What is encrypted, and what is not
+
+**A note's title and body are sealed** in the database, under a key kept in a file
+beside it rather than inside it. A backup you download does not carry the key, so
+it does not carry your notes.
+
+Both halves, not only the body: a note's title is its first line, so a readable
+title beside a sealed body would leave the beginning of every note in the open.
+
+**What is still in the clear:** task titles, descriptions and comments — and the
+list of what a note links to, so a stolen file shows which projects and tasks a
+note points at, though not what it says.
+
+Sealing the notes cost the search its index — a full-text index over a note's body
+is another copy of that body — so search reads the notes themselves now. At twelve
+hundred notes that is about forty milliseconds, and the box waits a quarter of a
+second after you stop typing, so you will not see it.

docs/openapi.yaml

@@ -1396,6 +1396,204 @@ paths:
                     description: The environment variables that are not set.
         "403": { $ref: "#/components/responses/Forbidden" }
 
+  /api/v1/notes/bulk:
+    post:
+      tags: [notes]
+      summary: Archive, unarchive or delete a set of notes in one call.
+      description: |
+        One call rather than one per note, because the thing being asked is one
+        thing: "put these away". Fifty separate requests would be fifty round
+        trips, fifty chances for half of them to fail, and a list that redraws
+        fifty times — and after an import of twelve hundred notes, selecting fifty
+        is the ordinary case rather than the extreme one.
+
+        Each note is still checked on its own. A set is a convenience for the
+        caller and never a way to touch something they could not have touched one
+        at a time; anything they may not write to is counted in `skipped` rather
+        than failing the whole call, so one note somebody else just deleted does
+        not stop the other forty-nine.
+
+        Archiving is not deleting. The trash says "this was a mistake, and in
+        thirty days it is gone"; the archive says "this is finished, and I want to
+        keep it".
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [ids]
+              properties:
+                ids:
+                  type: array
+                  items: { type: string }
+                  description: At most 500.
+                archived:
+                  type: boolean
+                  description: Put away, or bring back out.
+                delete:
+                  type: boolean
+                  description: Into the trash instead. Wins over `archived`.
+      responses:
+        "200":
+          description: How many were changed, and how many were passed over.
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  done: { type: integer }
+                  skipped: { type: integer }
+        "422":
+          description: No ids, too many, or neither archived nor delete was asked for.
+
+  /api/v1/notes/{noteID}/attachments:
+    post:
+      tags: [notes]
+      summary: Put a file into a note.
+      description: |
+        A picture pasted or dropped into a note. Until this existed the import
+        could bring files in and nothing else could, so the most ordinary act in
+        a notes app — paste a screenshot — did nothing at all.
+
+        Editor rights on the note. The same answer for "no such note" and "not
+        yours", so an id cannot be probed.
+
+        The response carries the attachment's URL; the editor writes that into the
+        note's text as a Markdown image. The file is never inlined as a data URL —
+        a photograph as base64 would make the note itself megabytes of text.
+      parameters:
+        - name: noteID
+          in: path
+          required: true
+          schema: { type: string }
+      requestBody:
+        required: true
+        content:
+          multipart/form-data:
+            schema:
+              type: object
+              required: [file]
+              properties:
+                file:
+                  type: string
+                  format: binary
+      responses:
+        "201":
+          description: Stored, and hung on the note.
+        "404":
+          description: No such note, or not yours to write in.
+        "413":
+          description: Larger than 25 MB.
+
+  /api/v1/beacon:
+    post:
+      tags: [auth]
+      summary: Report that an installation exists. Unauthenticated.
+      description: |
+        The receiving end of the beacon. Open on purpose: the caller is a
+        stranger's installation, which is the whole feature.
+
+        Answers 404 unless this instance has been told it is the collector, so an
+        ordinary installation is not a spam sink that answers 200 to anything.
+        Rate limited like the login routes, and the table behind it has a ceiling:
+        the endpoint inserts a row keyed on an id the caller invented, so without
+        one anybody could grow it until the volume is full.
+
+        Nothing about the request is recorded but the two values in the body. No
+        source address is read or written here.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [instance_id]
+              properties:
+                instance_id:
+                  type: string
+                  description: A random id the installation generated for itself.
+                version:
+                  type: string
+      responses:
+        "204":
+          description: Recorded.
+        "400":
+          description: The id or the version is not the expected shape.
+        "404":
+          description: This instance is not a collector.
+
+  /api/v1/beacon/count:
+    get:
+      tags: [auth]
+      summary: How many installations reported in the last thirty days.
+      description: |
+        Unauthenticated and cached for a minute, and only answered when the
+        collector has been told to publish it.
+
+        A window rather than a running total: a total only climbs, so it stops
+        meaning "how many people use this" and starts meaning "how many ever
+        tried it".
+      responses:
+        "200":
+          description: The count.
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  installs: { type: integer }
+                  window_days: { type: integer }
+        "404":
+          description: Not a collector, or not published.
+
+  /api/v1/beacon/settings:
+    get:
+      tags: [auth]
+      summary: What this instance reports, and to whom. Administrators only.
+      description: |
+        Includes the exact payload that will be sent, so the settings page can
+        print it in full rather than describe it.
+      responses:
+        "200":
+          description: The current configuration, and the counts if this is the collector.
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  enabled: { type: boolean }
+                  collector_url: { type: string }
+                  instance_id: { type: string }
+                  version: { type: string }
+                  is_collector: { type: boolean }
+                  publish_count: { type: boolean }
+                  last_ping_at: { type: string, format: date-time }
+    put:
+      tags: [auth]
+      summary: Turn the beacon off, or point it somewhere else. Administrators only.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                enabled: { type: boolean }
+                collector_url:
+                  type: string
+                  description: |
+                    Where reports go. Fetched through a dialer that refuses
+                    loopback and private address space, so this cannot be used to
+                    make the server reach inside its own network.
+                is_collector: { type: boolean }
+                publish_count: { type: boolean }
+      responses:
+        "200":
+          description: The configuration as it now stands.
+        "422":
+          description: The collector address is not an http or https URL.
+
   /api/v1/panel/restart:
     post:
       tags: [auth]
@@ -1851,6 +2049,125 @@ paths:
         "404": { $ref: "#/components/responses/NotFound" }
         "422": { $ref: "#/components/responses/Validation" }
 
+  /api/v1/projects/{projectID}/sections/reorder:
+    parameters:
+      - { name: projectID, in: path, required: true, schema: { type: string } }
+    post:
+      tags: [projects]
+      summary: Set the order the project's sections appear in. Editors and owners.
+      description: |
+        The whole list at once, for the same reasons as `/projects/reorder`: a
+        project has a handful of sections, so sending the order you want is
+        simpler than midpoints and cannot land half-applied.
+
+        Ids belonging to another project are skipped rather than refused.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [ids]
+              properties:
+                ids:
+                  type: array
+                  maxItems: 500
+                  items: { type: string }
+                  description: Section ids, top to bottom.
+      responses:
+        "204": { description: Reordered. }
+        "404": { $ref: "#/components/responses/NotFound" }
+        "422": { $ref: "#/components/responses/Validation" }
+
+  /api/v1/calendar/subscriptions:
+    post:
+      tags: [calendar]
+      summary: Subscribe to a calendar you only have an address to.
+      description: |
+        An ICS or `webcal://` address. Several per person, unlike the Google
+        connection, which is one — the address is the identity here, the way the
+        token is for OAuth.
+
+        The address is fetched once before it is stored: an address that cannot be
+        read is a typo far more often than a server that happens to be down, and
+        saying so while somebody is still looking at the field is worth two
+        seconds. `webcal://` is rewritten to `https://`, and the destination is
+        checked against the private-network wall both when it is written down and
+        again on the resolved address every time it is fetched.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              required: [url]
+              properties:
+                url: { type: string, description: "https:// or webcal://" }
+                name:
+                  type: string
+                  description: Optional. The calendar's own name is used when this is empty.
+      responses:
+        "201":
+          description: Subscribed, and fetched once.
+        "422": { $ref: "#/components/responses/Validation" }
+
+  /api/v1/calendar/subscriptions/{id}:
+    parameters:
+      - { name: id, in: path, required: true, schema: { type: string } }
+    delete:
+      tags: [calendar]
+      summary: Stop subscribing. The cached events go with it.
+      responses:
+        "204": { description: Gone. }
+        "404": { $ref: "#/components/responses/NotFound" }
+
+  /api/v1/push/test:
+    post:
+      tags: [notifications]
+      summary: Send one notification to your own devices, and say what happened.
+      description: |
+        The answer to "no notifications arrive". A push that a service refuses is
+        the ordinary failure — a stale subscription, a browser that revoked
+        permission without telling the server — and every one of them looks
+        exactly like silence from the outside.
+
+        It goes through the same path a real notification takes, so what it proves
+        is the thing that has to work. The title and body come from the caller
+        because the dictionaries live in the interface.
+      requestBody:
+        required: false
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                title: { type: string, maxLength: 200 }
+                body: { type: string, maxLength: 500 }
+      responses:
+        "200":
+          description: What each device did with it.
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  subscriptions:
+                    type: integer
+                    description: Devices registered for this account.
+                  sent:
+                    type: integer
+                  failed:
+                    type: array
+                    items:
+                      type: object
+                      properties:
+                        service:
+                          type: string
+                          description: The push service's host — never the endpoint, which is a credential.
+                        reason:
+                          type: string
+                          description: What the service said.
+
   /api/v1/sections/{sectionID}:
     parameters:
       - { name: sectionID, in: path, required: true, schema: { type: string } }
@@ -2507,6 +2824,126 @@ paths:
                   created: { type: integer }
         "502": { $ref: "#/components/responses/Upstream" }
 
+  /api/v1/calendar:
+    get:
+      tags: [integrations]
+      summary: The Google Calendar connection and the calendars in it.
+      description: |
+        Read-only. verdande shows a Google calendar beside its own dated tasks and
+        never writes one back — writing needs the `calendar.events` scope and an
+        answer to what happens when both sides moved the same meeting, which is a
+        synchronisation model rather than a scope.
+
+        The list of calendars is fetched from Google when it is empty, so an
+        account that connected during a hiccup corrects itself on the next look
+        rather than showing an empty panel for ever.
+      responses:
+        "200": { $ref: "#/components/responses/Calendar" }
+    delete:
+      tags: [integrations]
+      summary: Disconnect the calendar and forget its tokens.
+      description: >
+        The cached events go too. Unlike the tasks a mailbox made, none of this is
+        the person's own work — it is a copy of somebody else's calendar, and it
+        means nothing without the account.
+      responses:
+        "204": { description: Disconnected. }
+
+  /api/v1/calendar/calendars:
+    put:
+      tags: [integrations]
+      summary: Choose which calendars are drawn.
+      description: >
+        The whole set in one write rather than one flip at a time, for the reason
+        the projects' order is written the same way: the question is "these ones",
+        and a list built out of several requests can land half applied. A calendar
+        turned off has its cached events removed; they come back if it is turned
+        on again.
+      requestBody:
+        required: true
+        content:
+          application/json:
+            schema:
+              type: object
+              properties:
+                shown:
+                  type: array
+                  items: { type: string }
+      responses:
+        "204": { description: Saved. }
+        "404": { $ref: "#/components/responses/NotFound" }
+
+  /api/v1/calendar/authorize:
+    post:
+      tags: [integrations]
+      summary: Begin the OAuth flow for a calendar.
+      description: |
+        The same registration Gmail signs in through, asking for
+        `calendar.readonly` and coming back to `/oauth/calendar/callback`. Google
+        issues a refresh token per authorisation, so connecting a calendar does not
+        disturb a working Gmail connection.
+      responses:
+        "200":
+          description: OK
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  url: { type: string }
+        "409":
+          $ref: "#/components/responses/Conflict"
+
+  /api/v1/calendar/sync:
+    post:
+      tags: [integrations]
+      summary: Refresh the chosen calendars now.
+      description: >
+        Refreshes the list of calendars and replaces the cached window of events
+        for the ones that are shown. `partial` means the budget ran out before the
+        last calendar; what was read is kept, and the rest arrive on the next run.
+      responses:
+        "200":
+          description: OK
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  events: { type: integer }
+                  partial: { type: boolean }
+        "502": { $ref: "#/components/responses/Upstream" }
+
+  /api/v1/calendar/events:
+    get:
+      tags: [integrations]
+      summary: The events of the chosen calendars, between two days.
+      description: |
+        Both ends inclusive, and an event that merely overlaps the range comes
+        back — a grid asks which cells an event covers, not where it begins.
+
+        `from` and `to` in the answer are the window verdande holds a copy of,
+        which is not the window that was asked for. A grid paged past it gets no
+        events, and an empty answer with no explanation is a calendar quietly
+        claiming a day is clear.
+      parameters:
+        - { name: from, in: query, required: true, schema: { type: string, format: date } }
+        - { name: to, in: query, required: true, schema: { type: string, format: date } }
+      responses:
+        "200":
+          description: OK
+          content:
+            application/json:
+              schema:
+                type: object
+                properties:
+                  events:
+                    type: array
+                    items: { $ref: "#/components/schemas/CalendarEvent" }
+                  from: { type: string, format: date }
+                  to: { type: string, format: date }
+        "422": { $ref: "#/components/responses/Validation" }
+
   /api/v1/export/notes.zip:
     get:
       tags: [notes]
@@ -3150,6 +3587,26 @@ paths:
       responses:
         "302": { description: "Back into the app, with the outcome as a query parameter." }
 
+  /oauth/calendar/callback:
+    get:
+      tags: [integrations]
+      summary: Where Google sends the browser after consent to a calendar.
+      description: |
+        A URI of its own rather than a reuse of Gmail's. One callback deciding from
+        stored state which of the two flows had come back would put the choice of
+        which token store to write behind a lookup — and when that goes wrong it
+        goes wrong by overwriting a working Gmail connection with calendar tokens.
+
+        Answers with a redirect to `/indstillinger/integrationer`, not with JSON.
+        An `org_internal` error here means the OAuth app is registered as Internal
+        and the account is outside the organisation.
+      parameters:
+        - { name: code, in: query, schema: { type: string } }
+        - { name: state, in: query, schema: { type: string } }
+        - { name: error, in: query, schema: { type: string } }
+      responses:
+        "302": { description: "Back into the app, with the outcome as a query parameter." }
+
   /mcp:
     post:
       tags: [integrations]
@@ -3332,6 +3789,40 @@ components:
               email: { type: string }
               trigger: { type: string, enum: ["", starred, label, both] }
               label: { type: string }
+    Calendar:
+      description: OK
+      content:
+        application/json:
+          schema:
+            type: object
+            properties:
+              connected: { type: boolean }
+              account:
+                type: string
+                description: >
+                  The address of the connected account, taken from the id of its
+                  primary calendar — which is where Google puts it, so no profile
+                  scope has to be asked for to learn it.
+              read_only:
+                type: boolean
+                description: >
+                  Always true today. Stated rather than assumed: an interface that
+                  quietly looks editable is one somebody will try to edit.
+              has_client:
+                type: boolean
+                description: >
+                  Whether an OAuth client is registered on this instance at all, so
+                  the page can explain rather than offer a button that answers 409.
+              redirect_uri:
+                type: string
+                description: >
+                  What has to be registered with Google, spelled out. Derived from
+                  VERDANDE_BASE_URL and matched exactly — the single most likely
+                  thing to be wrong, and the error Google gives names neither value.
+              last_sync_at: { type: string, format: date-time }
+              calendars:
+                type: array
+                items: { $ref: "#/components/schemas/Calendar" }
     ImportResult:
       description: Created
       content:
@@ -3349,6 +3840,62 @@ components:
                 items: { type: string }
 
   schemas:
+    Calendar:
+      type: object
+      description: >
+        One calendar inside a connected account. Kept whether or not it is shown,
+        so turning one off and on again gives back the same row.
+      properties:
+        id: { type: string }
+        remote_id: { type: string, description: "The id Google knows it by; an address for the primary calendar." }
+        name: { type: string }
+        colour:
+          type: string
+          description: >
+            Google's own hex value, not one of verdande's ten named colours. Those
+            are names so the interface can restyle them per theme and they are
+            verdande's palette; a Google calendar is told apart by the colour the
+            person already knows it by.
+        time_zone: { type: string }
+        primary: { type: boolean }
+        writable:
+          type: boolean
+          description: >
+            Whether the connected account could write to it. Nothing does — the
+            scope is read-only — and it is stored so the day something can, the
+            interface knows which calendars could even be offered.
+        shown: { type: boolean }
+
+    CalendarEvent:
+      type: object
+      description: >
+        One occurrence, in the shape a grid asks about. A recurrence is expanded by
+        Google into its occurrences, so what arrives here is a day, not a rule.
+      properties:
+        id: { type: string }
+        calendar_id: { type: string }
+        summary: { type: string }
+        starts_at:
+          type: string
+          description: >
+            RFC 3339 as Google wrote it, offset and all. Empty for an all-day
+            event, which has a day rather than a moment — inventing midnight for it
+            is how an event moves a day when a container moves time zone.
+        ends_at: { type: string }
+        start_day: { type: string, format: date }
+        end_day:
+          type: string
+          format: date
+          description: >
+            The last day the event covers, inclusive. Google's all-day end date is
+            exclusive and is corrected on the way in, once, rather than by every
+            reader.
+        all_day: { type: boolean }
+        location: { type: string }
+        url: { type: string, description: "Opens the event in Google's own interface." }
+        calendar_name: { type: string }
+        colour: { type: string }
+
     Note:
       type: object
       description: A page of Markdown. The body is the stored form and the exported form.

internal/ai/ai.go

@@ -19,6 +19,8 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+
+	"github.com/kristianwind/verdande/internal/safedial"
 	"strings"
 	"time"
 )
@@ -76,7 +78,16 @@ func New(cfg Config) *Client {
 		// Generous: a local model on modest hardware can take a while to produce
 		// its first token, and the alternative is a timeout that only ever fires
 		// for the people running their own.
-		http: &http.Client{Timeout: 120 * time.Second},
+		// safedial, not a plain client. `base_url` is typed into a field in the
+		// interface, so it is a user-supplied address — and a server that fetches a
+		// user-supplied address is a way to ask it to reach things the caller
+		// cannot: the panel next door, a database on the same bridge, the cloud
+		// metadata endpoint. On a homelab that is most of what is worth reaching.
+		//
+		// The check is on the resolved address rather than on the URL, because a
+		// name answers whatever its owner says — and can answer differently the
+		// second time, after a parse-time check has passed.
+		http: safedial.Client(120 * time.Second),
 	}
 }
 
@@ -268,11 +279,35 @@ func (c *Client) do(req *http.Request, out any) error {
 		// The provider's own message is included: for a local model it is usually
 		// the actual problem ("model not found"), and hiding it would leave the
 		// operator with nothing to go on.
-		return fmt.Errorf("ai: %s said %s: %s", req.URL.Host, resp.Status, truncate(string(raw), 300))
+		// The provider's body is no longer passed back to the caller.
+		//
+		// It was, and combined with a free-text base URL that made a working
+		// internal scanner: point it at a host, read the first three hundred bytes
+		// of whatever answered. The operator still needs the message — for a local
+		// model it is usually the actual problem — so it goes to the log, where it
+		// is available to the person running the instance and to nobody else.
+		return &upstreamError{Host: req.URL.Host, Status: resp.Status, Body: truncate(string(raw), 300)}
 	}
 	return json.Unmarshal(raw, out)
 }
 
+// upstreamError keeps the provider's own words for the log and away from the
+// response. Error() is what reaches a caller; Detail() is what is written down.
+type upstreamError struct {
+	Host   string
+	Status string
+	Body   string
+}
+
+func (e *upstreamError) Error() string {
+	return fmt.Sprintf("ai: %s said %s", e.Host, e.Status)
+}
+
+// Detail is the whole of it, for the instance's own error log.
+func (e *upstreamError) Detail() string {
+	return fmt.Sprintf("ai: %s said %s: %s", e.Host, e.Status, e.Body)
+}
+
 func truncate(s string, max int) string {
 	if len(s) <= max {
 		return s

internal/beacon/beacon.go

@@ -0,0 +1,139 @@
+// Package beacon answers one question for the project, and refuses to answer any
+// other: how many installations are out there.
+//
+// It sends two values once a day — an anonymous id the installation made up for
+// itself, and the version it is running — and nothing else. No IP is stored by
+// the collector, no hostname, no domain, no account, no usage. The payload is
+// small enough to be printed in full on the settings page, and it is, because a
+// promise about telemetry is worth exactly as much as the reader's ability to
+// check it.
+//
+// # On by default, and off in one click
+//
+// The alternative that was asked for first was mandatory, and the argument
+// against it is practical rather than principled: it cannot be enforced. The
+// domain can be blocked, the binary can be patched, and a fork removes it in its
+// first commit — so mandatory buys nothing over on-by-default, while costing the
+// trust the rest of this program works for. Secrets are sealed here, the key is
+// kept out of the database, and the backup page warns you what a copy contains.
+// Telemetry you are not allowed to refuse is the one sentence in that story that
+// does not fit.
+//
+// So: on by default, said plainly, one switch to turn it off. That collects
+// essentially the same number and leaves nobody feeling tricked.
+//
+// # Where this disagrees with the rest of the program
+//
+// `config.UpdateCheck` is off unless asked for, and its comment gives the reason:
+// "a self-hosted app that reaches out without being told to has broken the deal
+// its operator made by self-hosting." That is the same situation as this one, and
+// the two defaults do not agree.
+//
+// The disagreement is written down here rather than smoothed over, because it is
+// a real one and the reader deserves to see both halves. The case for treating
+// them differently is that the update check reveals an instance to a third party
+// (GitHub) and returns something the operator can act on, while this returns
+// nothing and goes to the project itself — and that a count nobody sends is a
+// count that does not exist, whereas an update nobody checks for is one you can
+// still go and look up. Whether that is a distinction or an excuse is a judgement
+// call, and it was made deliberately by the person whose instance this is.
+//
+// If it is ever decided the other way, the change is one word: the default in
+// `beaconSettings`. Nothing else in this package assumes which way it points.
+//
+// # What the collector can still see
+//
+// A request has a source address whether or not anybody writes it down. The
+// collector does not store or log it, and the code is arranged so that it never
+// reaches a function that could — but it passes through a reverse proxy and a
+// tunnel on the way, and those keep their own logs. Saying "we send no IP" would
+// be a half-truth; what is true is that nothing here records one.
+package beacon
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"regexp"
+	"time"
+)
+
+// DefaultCollector is where a fresh installation reports, unless it is told
+// otherwise. Somebody running their own fleet points it at their own instance;
+// the field is on the settings page for exactly that.
+const DefaultCollector = "https://verdande.yggdrasilpanel.com/api/v1/beacon"
+
+// Payload is the entire message. Two fields, and adding a third is a change to
+// what this program promises, not a change to a struct.
+type Payload struct {
+	InstanceID string `json:"instance_id"`
+	Version    string `json:"version"`
+}
+
+// InstanceID and Version are checked on the way in as well as written on the way
+// out, because the collector's endpoint is unauthenticated: anything that reaches
+// it is a stranger's bytes, and it should be able to store only the two shapes it
+// expects.
+var (
+	idShape      = regexp.MustCompile(`^[0-9a-fA-F-]{16,64}$`)
+	versionShape = regexp.MustCompile(`^v?[0-9]{1,4}(\.[0-9]{1,5}){0,3}(-[0-9A-Za-z.]{1,20})?$`)
+)
+
+// ValidID reports whether an id is the shape this program writes.
+func ValidID(s string) bool { return idShape.MatchString(s) }
+
+// ValidVersion allows a version number and the word this binary uses when it was
+// built outside a release.
+func ValidVersion(s string) bool { return s == "dev" || versionShape.MatchString(s) }
+
+// Send delivers one ping. It is deliberately dull: a short timeout, no retry, no
+// queue, and a failure that is returned rather than made anybody's problem.
+//
+// No retry, because the next pass is tomorrow and a missed day changes a count of
+// installations by nothing. A telemetry ping that retries is a telemetry ping
+// that can hammer somebody's server on the day it is down.
+func Send(ctx context.Context, client *http.Client, collector string, p Payload) error {
+	if !ValidID(p.InstanceID) {
+		return fmt.Errorf("beacon: refusing to send a malformed id")
+	}
+	// Versionen kontrolleres også her, ikke kun hos modtageren. Kommentaren ved
+	// mønstrene lovede begge veje, og gjorde det ikke — og en installation, der
+	// sender noget mærkeligt, skal opdage det hos sig selv frem for at være noget,
+	// en fremmed collector skal filtrere fra.
+	if !ValidVersion(p.Version) {
+		return fmt.Errorf("beacon: refusing to send a malformed version")
+	}
+	body, err := json.Marshal(p)
+	if err != nil {
+		return err
+	}
+
+	ctx, cancel := context.WithTimeout(ctx, 15*time.Second)
+	defer cancel()
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, collector, bytes.NewReader(body))
+	if err != nil {
+		return err
+	}
+	req.Header.Set("Content-Type", "application/json")
+	// A plain user agent with the version already in the payload. Nothing is
+	// learned from it that the body does not already say, and it makes the request
+	// legible in a log rather than anonymous-looking.
+	req.Header.Set("User-Agent", "verdande-beacon/1")
+
+	resp, err := client.Do(req)
+	if err != nil {
+		return err
+	}
+	defer resp.Body.Close()
+	// Read and discard, so the connection can be reused rather than dropped.
+	_, _ = io.Copy(io.Discard, io.LimitReader(resp.Body, 4<<10))
+
+	if resp.StatusCode >= 400 {
+		return fmt.Errorf("beacon: collector answered %d", resp.StatusCode)
+	}
+	return nil
+}

internal/beacon/beacon_test.go

@@ -0,0 +1,114 @@
+package beacon
+
+import (
+	"context"
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+)
+
+// Hvad der bliver sendt, tegn for tegn.
+//
+// Prøven er skrevet imod *hele* kroppen og ikke imod enkelte felter, fordi det er
+// den, der er løftet: siden siger "præcis det her og ikke mere", og et felt lagt
+// til senere skal vælte noget. En prøve, der kun så efter, at instance_id var med,
+// ville lade en ny nøgle glide igennem uden at nogen opdagede det.
+func TestSendsExactlyTwoFields(t *testing.T) {
+	var got map[string]any
+	var contentType string
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		contentType = r.Header.Get("Content-Type")
+		body, _ := io.ReadAll(r.Body)
+		if err := json.Unmarshal(body, &got); err != nil {
+			t.Errorf("kroppen er ikke JSON: %v", err)
+		}
+		w.WriteHeader(http.StatusNoContent)
+	}))
+	defer srv.Close()
+
+	err := Send(context.Background(), srv.Client(), srv.URL, Payload{
+		InstanceID: "01a01b16-4e54-7000-9eaa-2f8eb0629c3a",
+		Version:    "v0.26.3",
+	})
+	if err != nil {
+		t.Fatalf("Send: %v", err)
+	}
+
+	if contentType != "application/json" {
+		t.Errorf("Content-Type = %q", contentType)
+	}
+	if len(got) != 2 {
+		t.Fatalf("der blev sendt %d felter, ikke to: %v", len(got), got)
+	}
+	if got["instance_id"] != "01a01b16-4e54-7000-9eaa-2f8eb0629c3a" {
+		t.Errorf("instance_id = %v", got["instance_id"])
+	}
+	if got["version"] != "v0.26.3" {
+		t.Errorf("version = %v", got["version"])
+	}
+}
+
+// En collector, der er nede, er ikke denne installations problem.
+func TestSendReportsAFailureRatherThanRetrying(t *testing.T) {
+	calls := 0
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		calls++
+		w.WriteHeader(http.StatusInternalServerError)
+	}))
+	defer srv.Close()
+
+	err := Send(context.Background(), srv.Client(), srv.URL, Payload{
+		InstanceID: "01a01b16-4e54-7000-9eaa-2f8eb0629c3a", Version: "v1.0.0",
+	})
+	if err == nil {
+		t.Fatal("en 500 skal give en fejl")
+	}
+	if calls != 1 {
+		t.Errorf("der blev prøvet %d gange; en telemetri-ping må ikke hamre på en server, der er nede", calls)
+	}
+}
+
+// Et id, der ikke er id-formet, sendes slet ikke — så en fejl her ikke bliver til
+// noget, en fremmed collector skal filtrere fra.
+func TestSendRefusesAMalformedID(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		t.Error("der blev sendt noget, som ikke skulle sendes")
+	}))
+	defer srv.Close()
+
+	if err := Send(context.Background(), srv.Client(), srv.URL, Payload{
+		InstanceID: "; DROP TABLE beacon_installs", Version: "v1.0.0",
+	}); err == nil {
+		t.Fatal("et misformet id skal afvises før det forlader maskinen")
+	}
+}
+
+func TestShapes(t *testing.T) {
+	ok := []string{"01a01b16-4e54-7000-9eaa-2f8eb0629c3a", "89df44d7-3742-4611-9a43-db0b07733c6d"}
+	for _, s := range ok {
+		if !ValidID(s) {
+			t.Errorf("ValidID(%q) = false", s)
+		}
+	}
+	bad := []string{"", "kort", "../../etc/passwd", "01a01b16-4e54-7000-9eaa-2f8eb0629c3a'; --",
+		"<script>alert(1)</script>"}
+	for _, s := range bad {
+		if ValidID(s) {
+			t.Errorf("ValidID(%q) = true", s)
+		}
+	}
+
+	for _, s := range []string{"v0.26.3", "0.26.3", "v1.2.3-rc.1", "dev"} {
+		if !ValidVersion(s) {
+			t.Errorf("ValidVersion(%q) = false", s)
+		}
+	}
+	for _, s := range []string{"", "v<script>", "v" + string(make([]byte, 200))} {
+		if ValidVersion(s) {
+			t.Errorf("ValidVersion(%q) = true", s)
+		}
+	}
+}

internal/config/config.go

@@ -59,10 +59,19 @@ type Config struct {
 	// without spending the real one on every run.
 	GmailSyncBudget time.Duration
 
-	// GmailEndpoints redirects the Gmail calls at a server the tests control. Empty
-	// in production, where it means Google.
-	GmailTokenURL string
-	GmailAPIURL   string
+	// CalendarSyncBudget is the same bound for one calendar refresh, and exists for
+	// the same reason: a slow Google held the request open past what Cloudflare is
+	// willing to wait, and the browser got Cloudflare's HTML instead of an error
+	// this server could explain.
+	CalendarSyncBudget time.Duration
+
+	// Where Google is. Empty in production; a test points them at a server it
+	// controls, which is the only way to exercise a Google that is slow or refuses.
+	// The token endpoint is shared — Gmail and Calendar sign in through the same
+	// registration — and the two APIs are not.
+	GoogleTokenURL string
+	GmailAPIURL    string
+	CalendarAPIURL string
 
 	PanelURL      string
 	PanelToken    string
@@ -98,14 +107,15 @@ func Load() (*Config, error) {
 		TrashRetention: 30 * 24 * time.Hour,
 		Dev:            envBool("VERDANDE_DEV", false),
 
-		GmailClientID:     env("VERDANDE_GMAIL_CLIENT_ID", ""),
-		GmailSyncBudget:   25 * time.Second,
-		PanelURL:          strings.TrimSuffix(env("VERDANDE_PANEL_URL", ""), "/"),
-		PanelToken:        env("VERDANDE_PANEL_TOKEN", ""),
-		PanelServerID:     env("VERDANDE_PANEL_SERVER_ID", ""),
-		GmailClientSecret: env("VERDANDE_GMAIL_CLIENT_SECRET", ""),
-		SecretKey:         env("VERDANDE_SECRET_KEY", ""),
-		UpdateCheck:       envBool("VERDANDE_UPDATE_CHECK", false),
+		GmailClientID:      env("VERDANDE_GMAIL_CLIENT_ID", ""),
+		GmailSyncBudget:    25 * time.Second,
+		CalendarSyncBudget: 25 * time.Second,
+		PanelURL:           strings.TrimSuffix(env("VERDANDE_PANEL_URL", ""), "/"),
+		PanelToken:         env("VERDANDE_PANEL_TOKEN", ""),
+		PanelServerID:      env("VERDANDE_PANEL_SERVER_ID", ""),
+		GmailClientSecret:  env("VERDANDE_GMAIL_CLIENT_SECRET", ""),
+		SecretKey:          env("VERDANDE_SECRET_KEY", ""),
+		UpdateCheck:        envBool("VERDANDE_UPDATE_CHECK", false),
 	}
 
 	var err error
@@ -152,6 +162,15 @@ func (c *Config) BackupsDir() string { return filepath.Join(c.DataDir, "backups"
 // configured separately and left to drift out of step with it.
 func (c *Config) GmailRedirectURL() string { return c.BaseURL + "/oauth/gmail/callback" }
 
+// CalendarRedirectURL is a second registered URI rather than a reuse of Gmail's.
+//
+// One callback that decided from server-side state which of the two flows had come
+// back would save the operator one line in Google Cloud, and would put the choice
+// of *which token store to write* behind a lookup. When that goes wrong it goes
+// wrong by overwriting a working Gmail connection with calendar tokens — a failure
+// nobody would think to look for. Two paths cannot do that.
+func (c *Config) CalendarRedirectURL() string { return c.BaseURL + "/oauth/calendar/callback" }
+
 func env(key, def string) string {
 	if v, ok := os.LookupEnv(key); ok && v != "" {
 		return v

internal/filterquery/filterquery.go

@@ -60,6 +60,20 @@ func Parse(query string, ctx Context) (Compiled, error) {
 	if len(p.tokens) == 0 {
 		return Compiled{}, fmt.Errorf("filteret er tomt")
 	}
+	// Refused before it is parsed, not while.
+	//
+	// `combine` rebuilds the whole accumulated SQL string and re-copies the
+	// argument slice per operator, so the cost is quadratic in the number of
+	// terms: fifty thousand of them — three hundred kilobytes, which fits in a
+	// query string — took seven seconds of CPU and forty-six gigabytes of
+	// allocation in one request. The preview endpoint runs this on whatever is
+	// typed, and there is no write timeout on the server to cut it off.
+	//
+	// Two hundred terms is a filter nobody has ever written and far below where it
+	// costs anything.
+	if len(p.tokens) > maxTokens {
+		return Compiled{}, fmt.Errorf("filteret er for langt")
+	}
 
 	node, err := p.parseExpression()
 	if err != nil {
@@ -140,10 +154,19 @@ func isOperator(r rune) bool {
 
 // --- parser -------------------------------------------------------------------
 
+// The ceilings on one filter. See Parse for why.
+const (
+	maxTokens = 200
+	maxDepth  = 32
+)
+
 type parser struct {
 	tokens []token
 	pos    int
 	ctx    Context
+	// depth guards against a filter that is only parentheses. Each level is a
+	// stack frame, and nothing else stops somebody sending a megabyte of them.
+	depth int
 }
 
 func (p *parser) done() bool  { return p.pos >= len(p.tokens) }
@@ -187,6 +210,13 @@ func (p *parser) parseTerm() (Compiled, error) {
 	if p.done() {
 		return Compiled{}, fmt.Errorf("filteret slutter midt i et udtryk")
 	}
+	// Nesting is stack depth. A filter made of half a million opening parentheses
+	// is a megabyte of text and a stack frame for each.
+	p.depth++
+	defer func() { p.depth-- }()
+	if p.depth > maxDepth {
+		return Compiled{}, fmt.Errorf("filteret har for mange niveauer")
+	}
 
 	switch t := p.peek(); t.kind {
 	case tokNot:

internal/gcal/gcal.go

@@ -0,0 +1,339 @@
+// Package gcal reads a Google Calendar. It never writes one.
+//
+// The scope is calendar.readonly, and that is the whole of the decision: verdande
+// already is a CalDAV server, so a Google event shown here is a second calendar
+// laid over the first, not a copy Google and verdande now both have to agree
+// about. Writing needs `calendar.events`, and it needs an answer to what happens
+// when both sides moved the same meeting — which is a synchronisation model, not a
+// scope. Read first; write later and to one calendar at a time.
+//
+// What the API gives back and what this returns are deliberately different shapes.
+// Google answers with RFC 3339 instants for a timed event and bare dates for an
+// all-day one, and with an *exclusive* end date for the second. A calendar grid
+// asks one question — "which cells does this event cover" — so that is what an
+// Event answers, in days, worked out once here rather than in every caller.
+package gcal
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"net/url"
+	"strings"
+	"time"
+
+	"github.com/kristianwind/verdande/internal/google"
+)
+
+// Scope is read-only, and only the calendar.
+//
+// Google also offers calendar.calendarlist.readonly and calendar.events.readonly,
+// which together are strictly less than this — they leave out access rules and
+// account settings, which verdande has no use for. They are the tighter pair and
+// worth moving to; they are also newer, and this one has been documented and
+// accepted unchanged for a decade. A consent screen that fails with `invalid_scope`
+// is not a place to be adventurous on somebody else's Google account.
+const Scope = "https://www.googleapis.com/auth/calendar.readonly"
+
+const defaultAPIBase = "https://www.googleapis.com/calendar/v3"
+
+// The OAuth flow, which is Google's rather than the calendar's — the same one
+// Gmail signs in through, asking for the scope above. Aliased so a caller reads
+// one package for one feature.
+type (
+	Config = google.Config
+	PKCE   = google.PKCE
+	Token  = google.Token
+)
+
+var NewPKCE = google.NewPKCE
+
+// ErrUnauthorized means the tokens are no longer good — access was revoked, or the
+// password changed. Re-exported so a caller need not know the flow lives elsewhere.
+var ErrUnauthorized = google.ErrUnauthorized
+
+type Client struct {
+	accessToken string
+	// base is where the Calendar API is. A field rather than a constant so a test
+	// can point it at a server it controls; empty is Google. There is no other way
+	// to exercise a calendar that is slow, empty, or refuses.
+	base string
+	http *http.Client
+}
+
+func NewClient(accessToken string) *Client {
+	return &Client{accessToken: accessToken, http: &http.Client{Timeout: 30 * time.Second}}
+}
+
+// At points the client at another server. Tests only; the zero value is Google.
+func (c *Client) At(base string) *Client {
+	c.base = base
+	return c
+}
+
+func (c *Client) api() string {
+	if c.base != "" {
+		return c.base
+	}
+	return defaultAPIBase
+}
+
+// Calendar is one entry in the person's calendar list.
+type Calendar struct {
+	ID string
+	// Name is the summary Google holds. A calendar somebody renamed for themselves
+	// carries summaryOverride instead, and that is the name they will look for.
+	Name string
+	// Colour is Google's own hex value for the calendar, which is what makes two
+	// calendars in one grid tellable apart. Their own colour rather than one picked
+	// here: it is the colour they already know the calendar by.
+	Colour   string
+	TimeZone string
+	Primary  bool
+	// Owner means the connected account can write to it. Nothing writes today —
+	// it is stored so the day something does, the interface knows which calendars
+	// could even be offered.
+	Owner bool
+}
+
+// Calendars lists what the connected account can see.
+func (c *Client) Calendars(ctx context.Context) ([]Calendar, error) {
+	var out []Calendar
+	page := ""
+	// Bounded rather than "until Google stops": a page token that never empties is
+	// a loop that never ends, and the difference between a bug at Google and a
+	// hung sync should not be this server hanging.
+	for range 10 {
+		v := url.Values{}
+		v.Set("minAccessRole", "reader")
+		v.Set("maxResults", "250")
+		if page != "" {
+			v.Set("pageToken", page)
+		}
+
+		var parsed struct {
+			Items []struct {
+				ID              string `json:"id"`
+				Summary         string `json:"summary"`
+				SummaryOverride string `json:"summaryOverride"`
+				BackgroundColor string `json:"backgroundColor"`
+				TimeZone        string `json:"timeZone"`
+				Primary         bool   `json:"primary"`
+				AccessRole      string `json:"accessRole"`
+				Deleted         bool   `json:"deleted"`
+			} `json:"items"`
+			NextPageToken string `json:"nextPageToken"`
+		}
+		if err := c.get(ctx, c.api()+"/users/me/calendarList?"+v.Encode(), &parsed); err != nil {
+			return nil, err
+		}
+
+		for _, item := range parsed.Items {
+			if item.Deleted {
+				continue
+			}
+			name := item.SummaryOverride
+			if name == "" {
+				name = item.Summary
+			}
+			out = append(out, Calendar{
+				ID: item.ID, Name: name, Colour: item.BackgroundColor,
+				TimeZone: item.TimeZone, Primary: item.Primary,
+				Owner: item.AccessRole == "owner" || item.AccessRole == "writer",
+			})
+		}
+		if parsed.NextPageToken == "" {
+			break
+		}
+		page = parsed.NextPageToken
+	}
+	return out, nil
+}
+
+// Event is one occurrence, in the shape a calendar grid asks about.
+type Event struct {
+	ID      string
+	Summary string
+	// StartsAt and EndsAt are RFC 3339 for a timed event and empty for an all-day
+	// one — an all-day event has no instant, only a day, and inventing midnight for
+	// it is how an event moves a day every time a server changes time zone.
+	StartsAt string
+	EndsAt   string
+	// StartDay and EndDay are the first and last day the event covers, inclusive at
+	// both ends. Google's all-day end date is exclusive; that is corrected here,
+	// once, rather than in every reader.
+	StartDay string
+	EndDay   string
+	AllDay   bool
+	Location string
+	// URL opens the event in Google's own interface. Same reasoning as the Gmail
+	// link on a task made from a mail: what verdande shows is a pointer to the
+	// event, not a second copy of it that somebody could try to edit.
+	URL string
+}
+
+// Events lists what falls inside a window, one calendar at a time.
+//
+// `singleEvents` expands a recurrence into its occurrences, which is the only form
+// a grid can draw: an RRULE is an instruction, and a Tuesday cell needs the Tuesday
+// it produced. It also means verdande does not have to grow a second recurrence
+// engine that agrees with Google's about Danish summer time.
+//
+// from and to are dates, inclusive. The window is the caller's business; this
+// turns it into the half-open interval Google wants.
+func (c *Client) Events(ctx context.Context, calendarID, from, to string) ([]Event, error) {
+	start, err := time.Parse(dateLayout, from)
+	if err != nil {
+		return nil, fmt.Errorf("gcal: from: %w", err)
+	}
+	end, err := time.Parse(dateLayout, to)
+	if err != nil {
+		return nil, fmt.Errorf("gcal: to: %w", err)
+	}
+
+	var out []Event
+	page := ""
+	for range 20 {
+		v := url.Values{}
+		v.Set("singleEvents", "true")
+		v.Set("orderBy", "startTime")
+		v.Set("maxResults", "2500")
+		// UTC, and a day of slack at each end. An event at 23:30 on the last day in
+		// Copenhagen is 21:30 UTC that day, but one at 00:30 on the first day is
+		// 22:30 UTC the day *before* — so a window cut exactly at the requested
+		// dates loses the events at both edges. The extra day is thrown away again
+		// by the day fields below, which are computed from the event's own offset.
+		v.Set("timeMin", start.AddDate(0, 0, -1).Format(time.RFC3339))
+		v.Set("timeMax", end.AddDate(0, 0, 2).Format(time.RFC3339))
+		if page != "" {
+			v.Set("pageToken", page)
+		}
+
+		var parsed struct {
+			Items         []rawEvent `json:"items"`
+			NextPageToken string     `json:"nextPageToken"`
+		}
+		url := c.api() + "/calendars/" + url.PathEscape(calendarID) + "/events?" + v.Encode()
+		if err := c.get(ctx, url, &parsed); err != nil {
+			return nil, err
+		}
+
+		for _, item := range parsed.Items {
+			// A cancelled occurrence of a recurring event still comes back, so that
+			// a client keeping its own copy knows to drop it. verdande replaces the
+			// window wholesale, so it only has to not add it.
+			if item.Status == "cancelled" {
+				continue
+			}
+			event, ok := item.event()
+			if !ok {
+				continue
+			}
+			if event.EndDay < from || event.StartDay > to {
+				continue // the slack above, thrown away
+			}
+			out = append(out, event)
+		}
+		if parsed.NextPageToken == "" {
+			break
+		}
+		page = parsed.NextPageToken
+	}
+	return out, nil
+}
+
+const dateLayout = "2006-01-02"
+
+type rawEvent struct {
+	ID       string `json:"id"`
+	Status   string `json:"status"`
+	Summary  string `json:"summary"`
+	Location string `json:"location"`
+	HTMLLink string `json:"htmlLink"`
+	Start    when   `json:"start"`
+	End      when   `json:"end"`
+}
+
+type when struct {
+	Date     string `json:"date"`
+	DateTime string `json:"dateTime"`
+}
+
+// event turns Google's two shapes into the one a grid can read.
+//
+// The second return is false for an event with neither a date nor a dateTime —
+// which Google does return, for an event somebody is still being asked about. It
+// belongs on no day, so there is no cell to draw it in.
+func (r rawEvent) event() (Event, bool) {
+	e := Event{
+		ID: r.ID, Summary: strings.TrimSpace(r.Summary),
+		Location: r.Location, URL: r.HTMLLink,
+	}
+	if e.Summary == "" {
+		// Google's own interface says this for an event with no title, and a blank
+		// chip in a grid reads as a rendering fault rather than as an empty title.
+		e.Summary = "(uden titel)"
+	}
+
+	switch {
+	case r.Start.Date != "":
+		e.AllDay = true
+		e.StartDay = r.Start.Date
+		// Google's all-day end is the day *after* the last one, so a one-day event
+		// ends tomorrow. Subtracting a day is what makes "covers this cell" a
+		// comparison rather than a special case at every call site.
+		e.EndDay = r.Start.Date
+		if day, err := time.Parse(dateLayout, r.End.Date); err == nil {
+			if last := day.AddDate(0, 0, -1); !last.Before(mustDay(r.Start.Date)) {
+				e.EndDay = last.Format(dateLayout)
+			}
+		}
+	case r.Start.DateTime != "":
+		e.StartsAt = r.Start.DateTime
+		e.EndsAt = r.End.DateTime
+		// The day is the date part of the timestamp Google returned, which carries
+		// the event's own offset. Parsing it into a time.Time and formatting it
+		// back would ask *this process* what day it is — and a container running in
+		// UTC would then move a 23:30 meeting in Copenhagen to the day before.
+		e.StartDay = dayOf(r.Start.DateTime)
+		e.EndDay = dayOf(r.End.DateTime)
+		if e.EndDay == "" {
+			e.EndDay = e.StartDay
+		}
+		// An event ending at midnight ends the evening before, as far as a grid is
+		// concerned. Without this, every meeting that runs to 24:00 puts a chip in
+		// the following day as well.
+		if e.EndDay > e.StartDay && strings.Contains(r.End.DateTime, "T00:00:00") {
+			if day, err := time.Parse(dateLayout, e.EndDay); err == nil {
+				e.EndDay = day.AddDate(0, 0, -1).Format(dateLayout)
+			}
+		}
+	default:
+		return Event{}, false
+	}
+	if e.EndDay < e.StartDay {
+		e.EndDay = e.StartDay
+	}
+	return e, true
+}
+
+// dayOf takes the date out of an RFC 3339 timestamp without interpreting it.
+func dayOf(stamp string) string {
+	if len(stamp) < len(dateLayout) {
+		return ""
+	}
+	day := stamp[:len(dateLayout)]
+	if _, err := time.Parse(dateLayout, day); err != nil {
+		return ""
+	}
+	return day
+}
+
+func mustDay(s string) time.Time {
+	day, _ := time.Parse(dateLayout, s)
+	return day
+}
+
+func (c *Client) get(ctx context.Context, url string, out any) error {
+	return google.Get(ctx, c.http, c.accessToken, url, out)
+}

internal/gcal/gcal_test.go

@@ -0,0 +1,255 @@
+package gcal
+
+import (
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+)
+
+// fakeGoogle answers the two calls this package makes, out of whatever the test
+// hands it. There is no other way to exercise this: the real Calendar needs an
+// account, a registered client and somebody's consent, and none of the three can be
+// had inside a test — so the shapes Google returns are written down here and the
+// parsing is checked against them.
+func fakeGoogle(t *testing.T, calendarList any, events map[string]any) *Client {
+	t.Helper()
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		switch {
+		case strings.HasPrefix(r.URL.Path, "/users/me/calendarList"):
+			_ = json.NewEncoder(w).Encode(calendarList)
+		case strings.HasPrefix(r.URL.Path, "/calendars/"):
+			id := strings.TrimSuffix(strings.TrimPrefix(r.URL.Path, "/calendars/"), "/events")
+			body, ok := events[id]
+			if !ok {
+				http.Error(w, `{"error":"not found"}`, http.StatusNotFound)
+				return
+			}
+			_ = json.NewEncoder(w).Encode(body)
+		default:
+			http.Error(w, "unexpected path "+r.URL.Path, http.StatusNotFound)
+		}
+	}))
+	t.Cleanup(srv.Close)
+	return NewClient("token").At(srv.URL)
+}
+
+func TestCalendarsPrefersTheNameThePersonGaveIt(t *testing.T) {
+	client := fakeGoogle(t, map[string]any{
+		"items": []map[string]any{
+			{"id": "primary", "summary": "kw@nolimit.dk", "backgroundColor": "#3f51b5",
+				"timeZone": "Europe/Copenhagen", "primary": true, "accessRole": "owner"},
+			{"id": "fam", "summary": "Family", "summaryOverride": "Familien",
+				"backgroundColor": "#0b8043", "accessRole": "reader"},
+			// A calendar removed from the list still arrives, flagged.
+			{"id": "gone", "summary": "Væk", "deleted": true, "accessRole": "reader"},
+		},
+	}, nil)
+
+	got, err := client.Calendars(t.Context())
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 2 {
+		t.Fatalf("got %d calendars, want 2 — a deleted one is not one you can show", len(got))
+	}
+	if got[1].Name != "Familien" {
+		t.Errorf("name = %q; summaryOverride is the name the person will look for", got[1].Name)
+	}
+	if !got[0].Primary || !got[0].Owner {
+		t.Errorf("the primary calendar came back as %+v", got[0])
+	}
+	if got[1].Owner {
+		t.Error("a reader was reported as an owner; nothing may offer to write to it")
+	}
+	if got[0].Colour != "#3f51b5" {
+		t.Errorf("colour = %q; two calendars in one grid are told apart by it", got[0].Colour)
+	}
+}
+
+// The three shapes an event arrives in, and the one shape a grid can read.
+func TestEventsBecomeDaysAGridCanAskAbout(t *testing.T) {
+	client := fakeGoogle(t, nil, map[string]any{
+		"primary": map[string]any{
+			"items": []map[string]any{
+				{
+					"id": "timed", "summary": "Bestyrelsesmøde", "htmlLink": "https://cal/1",
+					"start": map[string]string{"dateTime": "2026-08-20T14:00:00+02:00"},
+					"end":   map[string]string{"dateTime": "2026-08-20T15:30:00+02:00"},
+				},
+				{
+					// All-day, one day. Google's end is the day after.
+					"id": "allday", "summary": "Fridag",
+					"start": map[string]string{"date": "2026-08-21"},
+					"end":   map[string]string{"date": "2026-08-22"},
+				},
+				{
+					// All-day, three days.
+					"id": "trip", "summary": "Aarhus",
+					"start": map[string]string{"date": "2026-08-24"},
+					"end":   map[string]string{"date": "2026-08-27"},
+				},
+				{
+					// Runs to midnight, which is the evening before as far as a
+					// grid is concerned.
+					"id": "late", "summary": "Nattevagt",
+					"start": map[string]string{"dateTime": "2026-08-22T20:00:00+02:00"},
+					"end":   map[string]string{"dateTime": "2026-08-23T00:00:00+02:00"},
+				},
+				{
+					// A cancelled occurrence of a recurring event.
+					"id": "off", "status": "cancelled", "summary": "Aflyst",
+					"start": map[string]string{"date": "2026-08-25"},
+					"end":   map[string]string{"date": "2026-08-26"},
+				},
+				{
+					// Neither a date nor a dateTime: belongs on no day at all.
+					"id": "nowhere", "summary": "Uden tid",
+				},
+			},
+		},
+	})
+
+	got, err := client.Events(t.Context(), "primary", "2026-08-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	byID := map[string]Event{}
+	for _, e := range got {
+		byID[e.ID] = e
+	}
+	if len(got) != 4 {
+		t.Fatalf("got %d events, want 4: %+v", len(got), got)
+	}
+	if _, ok := byID["off"]; ok {
+		t.Error("a cancelled occurrence was kept; it is an event that is not happening")
+	}
+	if _, ok := byID["nowhere"]; ok {
+		t.Error("an event with no time was kept; there is no cell to draw it in")
+	}
+
+	timed := byID["timed"]
+	if timed.AllDay || timed.StartDay != "2026-08-20" || timed.EndDay != "2026-08-20" {
+		t.Errorf("timed = %+v", timed)
+	}
+	if timed.StartsAt != "2026-08-20T14:00:00+02:00" {
+		t.Errorf("the instant was rewritten: %q", timed.StartsAt)
+	}
+
+	allday := byID["allday"]
+	if !allday.AllDay || allday.StartDay != "2026-08-21" || allday.EndDay != "2026-08-21" {
+		t.Errorf("allday = %+v; Google's end date is exclusive and has to be corrected here",
+			allday)
+	}
+	if allday.StartsAt != "" {
+		t.Errorf("an all-day event was given an instant: %q — it has a day, not a moment",
+			allday.StartsAt)
+	}
+
+	if trip := byID["trip"]; trip.StartDay != "2026-08-24" || trip.EndDay != "2026-08-26" {
+		t.Errorf("trip = %+v, want 24th to 26th inclusive", trip)
+	}
+	if late := byID["late"]; late.EndDay != "2026-08-22" {
+		t.Errorf("late ends on %q; an event running to midnight ends the evening before, "+
+			"or every one of them puts a chip in the following day", late.EndDay)
+	}
+}
+
+// An event with no title has to say so. A blank chip in a grid reads as a
+// rendering fault rather than as an empty title.
+func TestAnEventWithNoTitleStillSaysSomething(t *testing.T) {
+	client := fakeGoogle(t, nil, map[string]any{
+		"c": map[string]any{"items": []map[string]any{{
+			"id":    "x",
+			"start": map[string]string{"date": "2026-08-20"},
+			"end":   map[string]string{"date": "2026-08-21"},
+		}}},
+	})
+	got, err := client.Events(t.Context(), "c", "2026-08-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 1 || got[0].Summary == "" {
+		t.Fatalf("got %+v", got)
+	}
+}
+
+// The window is asked for with a day of slack at each end, because an event late
+// in the evening in Copenhagen is the next day in UTC. The slack must not leak
+// out: an event on the day before the window is not in the window.
+func TestTheSlackAtTheEdgesIsThrownAwayAgain(t *testing.T) {
+	client := fakeGoogle(t, nil, map[string]any{
+		"c": map[string]any{"items": []map[string]any{
+			{"id": "before", "summary": "For tidligt",
+				"start": map[string]string{"date": "2026-08-09"},
+				"end":   map[string]string{"date": "2026-08-10"}},
+			{"id": "inside", "summary": "Med",
+				"start": map[string]string{"date": "2026-08-10"},
+				"end":   map[string]string{"date": "2026-08-11"}},
+			{"id": "after", "summary": "For sent",
+				"start": map[string]string{"date": "2026-08-21"},
+				"end":   map[string]string{"date": "2026-08-22"}},
+		}},
+	})
+	got, err := client.Events(t.Context(), "c", "2026-08-10", "2026-08-20")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 1 || got[0].ID != "inside" {
+		t.Fatalf("got %+v, want only the one inside the window", got)
+	}
+}
+
+// A page token has to be followed, or a busy calendar silently shows its first
+// page and nothing else — which is a calendar that lies about a day being clear.
+func TestEveryPageIsRead(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		body := map[string]any{
+			"items": []map[string]any{{"id": "a", "summary": "Første",
+				"start": map[string]string{"date": "2026-08-11"},
+				"end":   map[string]string{"date": "2026-08-12"}}},
+			"nextPageToken": "more",
+		}
+		if r.URL.Query().Get("pageToken") == "more" {
+			body = map[string]any{"items": []map[string]any{{"id": "b", "summary": "Anden",
+				"start": map[string]string{"date": "2026-08-12"},
+				"end":   map[string]string{"date": "2026-08-13"}}}}
+		}
+		_ = json.NewEncoder(w).Encode(body)
+	}))
+	defer srv.Close()
+
+	got, err := NewClient("t").At(srv.URL).Events(t.Context(), "c", "2026-08-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 2 {
+		t.Fatalf("got %d events, want both pages: %+v", len(got), got)
+	}
+}
+
+// A 401 is not a failure to retry; it is a connection that has ended. Telling the
+// two apart is what stops a revoked account being polled every ten minutes for ever.
+func TestARefusedTokenIsItsOwnError(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		http.Error(w, `{"error":"invalid_credentials"}`, http.StatusUnauthorized)
+	}))
+	defer srv.Close()
+
+	_, err := NewClient("spent").At(srv.URL).Calendars(t.Context())
+	if err != ErrUnauthorized {
+		t.Fatalf("err = %v, want ErrUnauthorized", err)
+	}
+}
+
+// Read-only, and only the calendar. The consent screen is what somebody agrees to,
+// and a scope that quietly grew is a promise quietly broken.
+func TestTheScopeIsReadOnly(t *testing.T) {
+	if !strings.HasSuffix(Scope, "calendar.readonly") {
+		t.Errorf("scope = %q; nothing here writes a calendar", Scope)
+	}
+}

internal/gmail/gmail.go

@@ -1,5 +1,5 @@
-// Package gmail implements the OAuth2 authorisation-code flow with PKCE against
-// Google, and the small slice of the Gmail API verdande needs.
+// Package gmail is the small slice of the Gmail API verdande needs, and the scope
+// it asks for.
 //
 // The connection is one-way on purpose: a starred or labelled message becomes a
 // task, and nothing verdande does afterwards touches the mailbox. Unstarring the
@@ -9,200 +9,51 @@
 //
 // Scope is gmail.readonly. verdande never needs to send, modify or delete, and
 // asking for less is what makes the consent screen something a person can agree to.
+//
+// The OAuth flow itself moved to internal/google when Calendar needed the same one.
+// The names below are aliases so that every call site — and this package's own
+// tests — still read as Gmail's, which is what they are: one registration, one
+// flow, one scope each.
 package gmail
 
 import (
 	"context"
-	"crypto/rand"
-	"crypto/sha256"
-	"encoding/base64"
-	"encoding/json"
 	"fmt"
 	"net/http"
 	"net/url"
 	"strings"
 	"time"
+
+	"github.com/kristianwind/verdande/internal/google"
 )
 
-const (
-	authEndpoint         = "https://accounts.google.com/o/oauth2/v2/auth"
-	defaultTokenEndpoint = "https://oauth2.googleapis.com/token"
-	defaultAPIBase       = "https://gmail.googleapis.com/gmail/v1"
+// Read-only, and only mail. Not profile, not contacts, not send.
+const Scope = "https://www.googleapis.com/auth/gmail.readonly"
 
-	// Read-only, and only mail. Not profile, not contacts, not send.
-	Scope = "https://www.googleapis.com/auth/gmail.readonly"
+const defaultAPIBase = "https://gmail.googleapis.com/gmail/v1"
+
+// The OAuth flow, which is Google's rather than Gmail's.
+type (
+	Config = google.Config
+	PKCE   = google.PKCE
+	Token  = google.Token
 )
 
-// Endpoints is where Google is.
-//
-// A field rather than a constant so a test can point it at a server it controls.
-// There is no other way to exercise what happens when Gmail is slow or refuses —
-// and "the sync comes back before whatever sits in front of this server gives up
-// on it" is exactly the kind of promise worth a test and impossible to make
-// against the real thing.
-//
-// The zero value is Google, so nothing outside a test has to know it exists.
-type Endpoints struct {
-	Token string
-	API   string
-}
-
-func (e Endpoints) token() string {
-	if e.Token != "" {
-		return e.Token
-	}
-	return defaultTokenEndpoint
-}
-
-func (e Endpoints) api() string {
-	if e.API != "" {
-		return e.API
-	}
-	return defaultAPIBase
-}
-
-// Config is the OAuth client the operator registered in Google Cloud. It belongs to
-// the instance rather than to a person: one registration, any number of users.
-type Config struct {
-	ClientID     string
-	ClientSecret string
-	RedirectURL  string
-	// Endpoints is Google unless a test says otherwise.
-	Endpoints Endpoints
-}
-
-func (c Config) Configured() bool {
-	return c.ClientID != "" && c.ClientSecret != "" && c.RedirectURL != ""
-}
-
-// PKCE is the proof-key pair for one authorisation attempt.
-//
-// PKCE on a server-side client is not strictly required — the client secret already
-// authenticates the token exchange — but it costs nothing and closes the window in
-// which an authorisation code intercepted from the redirect could be replayed
-// before verdande gets to it.
-type PKCE struct {
-	Verifier  string
-	Challenge string
-	State     string
-}
-
-func NewPKCE() (PKCE, error) {
-	verifier, err := randomString(64)
-	if err != nil {
-		return PKCE{}, err
-	}
-	state, err := randomString(32)
-	if err != nil {
-		return PKCE{}, err
-	}
-	sum := sha256.Sum256([]byte(verifier))
-	return PKCE{
-		Verifier:  verifier,
-		Challenge: base64.RawURLEncoding.EncodeToString(sum[:]),
-		State:     state,
-	}, nil
-}
-
-// AuthURL is where the browser is sent to ask the person for consent.
-func (c Config) AuthURL(p PKCE) string {
-	v := url.Values{}
-	v.Set("client_id", c.ClientID)
-	v.Set("redirect_uri", c.RedirectURL)
-	v.Set("response_type", "code")
-	v.Set("scope", Scope)
-	v.Set("state", p.State)
-	v.Set("code_challenge", p.Challenge)
-	v.Set("code_challenge_method", "S256")
-	// Without both of these Google returns no refresh token on a re-authorisation,
-	// and the connection silently stops working an hour later.
-	v.Set("access_type", "offline")
-	v.Set("prompt", "consent")
-	return authEndpoint + "?" + v.Encode()
-}
-
-type Token struct {
-	AccessToken  string
-	RefreshToken string
-	ExpiresAt    time.Time
-}
-
-// Exchange trades the authorisation code for tokens.
-func (c Config) Exchange(ctx context.Context, code string, p PKCE) (Token, error) {
-	form := url.Values{}
-	form.Set("code", code)
-	form.Set("client_id", c.ClientID)
-	form.Set("client_secret", c.ClientSecret)
-	form.Set("redirect_uri", c.RedirectURL)
-	form.Set("grant_type", "authorization_code")
-	form.Set("code_verifier", p.Verifier)
-
-	return c.token(ctx, form)
-}
-
-// Refresh gets a fresh access token. Google does not return a new refresh token
-// here, so the caller keeps the one it has.
-func (c Config) Refresh(ctx context.Context, refreshToken string) (Token, error) {
-	form := url.Values{}
-	form.Set("refresh_token", refreshToken)
-	form.Set("client_id", c.ClientID)
-	form.Set("client_secret", c.ClientSecret)
-	form.Set("grant_type", "refresh_token")
-
-	token, err := c.token(ctx, form)
-	if err != nil {
-		return Token{}, err
-	}
-	token.RefreshToken = refreshToken
-	return token, nil
-}
-
-func (c Config) token(ctx context.Context, form url.Values) (Token, error) {
-	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.Endpoints.token(),
-		strings.NewReader(form.Encode()))
-	if err != nil {
-		return Token{}, err
-	}
-	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
-
-	resp, err := (&http.Client{Timeout: 30 * time.Second}).Do(req)
-	if err != nil {
-		return Token{}, err
-	}
-	defer resp.Body.Close()
-
-	var parsed struct {
-		AccessToken      string `json:"access_token"`
-		RefreshToken     string `json:"refresh_token"`
-		ExpiresIn        int    `json:"expires_in"`
-		Error            string `json:"error"`
-		ErrorDescription string `json:"error_description"`
-	}
-	if err := json.NewDecoder(resp.Body).Decode(&parsed); err != nil {
-		return Token{}, fmt.Errorf("gmail: reading token response: %w", err)
-	}
-	if parsed.Error != "" {
-		return Token{}, fmt.Errorf("gmail: %s: %s", parsed.Error, parsed.ErrorDescription)
-	}
-	if resp.StatusCode != http.StatusOK || parsed.AccessToken == "" {
-		return Token{}, fmt.Errorf("gmail: token exchange failed with %s", resp.Status)
-	}
-
-	return Token{
-		AccessToken:  parsed.AccessToken,
-		RefreshToken: parsed.RefreshToken,
-		// A minute of slack, so a token that is about to expire is refreshed
-		// rather than used and rejected.
-		ExpiresAt: time.Now().Add(time.Duration(parsed.ExpiresIn-60) * time.Second),
-	}, nil
-}
+var (
+	NewPKCE         = google.NewPKCE
+	ErrUnauthorized = google.ErrUnauthorized
+)
 
 // --- the API ---------------------------------------------------------------------
 
 type Client struct {
 	accessToken string
-	endpoints   Endpoints
-	http        *http.Client
+	// base is where the Gmail API is. A field rather than a constant so a test can
+	// point it at a server it controls — and "the sync comes back before whatever
+	// sits in front of this server gives up on it" is exactly the kind of promise
+	// worth a test and impossible to make against the real thing. Empty is Google.
+	base string
+	http *http.Client
 }
 
 func NewClient(accessToken string) *Client {
@@ -210,11 +61,18 @@ func NewClient(accessToken string) *Client {
 }
 
 // At points the client at another server. Tests only; the zero value is Google.
-func (c *Client) At(e Endpoints) *Client {
-	c.endpoints = e
+func (c *Client) At(base string) *Client {
+	c.base = base
 	return c
 }
 
+func (c *Client) api() string {
+	if c.base != "" {
+		return c.base
+	}
+	return defaultAPIBase
+}
+
 // Message is the part of a Gmail message a task is made from.
 type Message struct {
 	ID       string
@@ -233,7 +91,7 @@ func (c *Client) Profile(ctx context.Context) (string, error) {
 	var parsed struct {
 		EmailAddress string `json:"emailAddress"`
 	}
-	if err := c.get(ctx, c.endpoints.api()+"/users/me/profile", &parsed); err != nil {
+	if err := c.get(ctx, c.api()+"/users/me/profile", &parsed); err != nil {
 		return "", err
 	}
 	return parsed.EmailAddress, nil
@@ -257,7 +115,7 @@ func (c *Client) List(ctx context.Context, query string, max int) ([]string, err
 			ID string `json:"id"`
 		} `json:"messages"`
 	}
-	if err := c.get(ctx, c.endpoints.api()+"/users/me/messages?"+v.Encode(), &parsed); err != nil {
+	if err := c.get(ctx, c.api()+"/users/me/messages?"+v.Encode(), &parsed); err != nil {
 		return nil, err
 	}
 
@@ -288,7 +146,7 @@ func (c *Client) Get(ctx context.Context, id string) (Message, error) {
 			} `json:"headers"`
 		} `json:"payload"`
 	}
-	if err := c.get(ctx, c.endpoints.api()+"/users/me/messages/"+url.PathEscape(id)+"?"+v.Encode(), &parsed); err != nil {
+	if err := c.get(ctx, c.api()+"/users/me/messages/"+url.PathEscape(id)+"?"+v.Encode(), &parsed); err != nil {
 		return Message{}, err
 	}
 
@@ -308,31 +166,9 @@ func (c *Client) Get(ctx context.Context, id string) (Message, error) {
 }
 
 func (c *Client) get(ctx context.Context, url string, out any) error {
-	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
-	if err != nil {
-		return err
-	}
-	req.Header.Set("Authorization", "Bearer "+c.accessToken)
-
-	resp, err := c.http.Do(req)
-	if err != nil {
-		return err
-	}
-	defer resp.Body.Close()
-
-	if resp.StatusCode == http.StatusUnauthorized {
-		return ErrUnauthorized
-	}
-	if resp.StatusCode != http.StatusOK {
-		return fmt.Errorf("gmail: %s", resp.Status)
-	}
-	return json.NewDecoder(resp.Body).Decode(out)
+	return google.Get(ctx, c.http, c.accessToken, url, out)
 }
 
-// ErrUnauthorized means the tokens are no longer good — the person revoked access,
-// or changed their password. The caller disconnects rather than retrying.
-var ErrUnauthorized = fmt.Errorf("gmail: authorisation was refused")
-
 // SenderName pulls "Anders Jensen" out of "Anders Jensen <anders@example.dk>",
 // falling back to the address when there is no display name.
 func SenderName(from string) string {
@@ -375,11 +211,3 @@ func Query(trigger, label string) string {
 	terms = append(terms, "newer_than:30d")
 	return strings.Join(terms, " ")
 }
-
-func randomString(n int) (string, error) {
-	b := make([]byte, n)
-	if _, err := rand.Read(b); err != nil {
-		return "", err
-	}
-	return base64.RawURLEncoding.EncodeToString(b)[:n], nil
-}

internal/gmail/gmail_test.go

@@ -1,113 +1,19 @@
 package gmail
 
 import (
-	"crypto/sha256"
-	"encoding/base64"
-	"net/url"
 	"strings"
 	"testing"
 )
 
-// The challenge has to be the SHA-256 of the verifier, base64url without padding.
-// Getting this wrong means Google refuses every exchange with a message that does
-// not say which end is at fault.
-func TestPKCEChallengeMatchesItsVerifier(t *testing.T) {
-	p, err := NewPKCE()
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	sum := sha256.Sum256([]byte(p.Verifier))
-	want := base64.RawURLEncoding.EncodeToString(sum[:])
-	if p.Challenge != want {
-		t.Errorf("challenge = %q, want %q", p.Challenge, want)
-	}
-
-	// RFC 7636 requires 43–128 characters, and no padding.
-	if n := len(p.Verifier); n < 43 || n > 128 {
-		t.Errorf("verifier is %d characters, outside the 43–128 the spec allows", n)
-	}
-	if strings.ContainsAny(p.Verifier, "=+/") {
-		t.Errorf("verifier %q contains characters that are not URL-safe", p.Verifier)
-	}
-	if p.State == "" {
-		t.Error("no state was generated")
-	}
-
-	// Two attempts must not share anything.
-	other, err := NewPKCE()
-	if err != nil {
-		t.Fatal(err)
-	}
-	if other.Verifier == p.Verifier || other.State == p.State {
-		t.Error("two authorisation attempts produced the same secrets")
-	}
-}
-
-func TestAuthURL(t *testing.T) {
-	cfg := Config{
-		ClientID:     "client-123",
-		ClientSecret: "secret",
-		RedirectURL:  "https://todo.example.dk/oauth/gmail/callback",
-	}
-	p, err := NewPKCE()
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	raw := cfg.AuthURL(p)
-	parsed, err := url.Parse(raw)
-	if err != nil {
-		t.Fatalf("the URL does not parse: %v", err)
-	}
-	q := parsed.Query()
-
-	checks := map[string]string{
-		"client_id":             cfg.ClientID,
-		"redirect_uri":          cfg.RedirectURL,
-		"response_type":         "code",
-		"scope":                 Scope,
-		"code_challenge":        p.Challenge,
-		"code_challenge_method": "S256",
-		"state":                 p.State,
-		// Without both of these Google returns no refresh token on a
-		// re-authorisation, and the connection dies silently an hour later.
-		"access_type": "offline",
-		"prompt":      "consent",
-	}
-	for key, want := range checks {
-		if got := q.Get(key); got != want {
-			t.Errorf("%s = %q, want %q", key, got, want)
-		}
-	}
-
-	// The verifier itself must never leave the server.
-	if strings.Contains(raw, p.Verifier) {
-		t.Error("the code verifier was put in the authorisation URL")
-	}
-	// Read-only, and only mail.
+// The scope is the one thing verdande chooses about the connection, and it is the
+// one the person reads on the consent screen. The flow itself is exercised in
+// internal/google; what belongs here is that Gmail still asks for read-only mail.
+func TestTheScopeIsReadOnlyMail(t *testing.T) {
 	if !strings.HasSuffix(Scope, "gmail.readonly") {
 		t.Errorf("scope = %q; verdande never needs to send or modify", Scope)
 	}
 }
 
-func TestConfigured(t *testing.T) {
-	full := Config{ClientID: "a", ClientSecret: "b", RedirectURL: "c"}
-	if !full.Configured() {
-		t.Error("a complete config is not reported as configured")
-	}
-	for _, partial := range []Config{
-		{},
-		{ClientID: "a"},
-		{ClientID: "a", ClientSecret: "b"},
-		{ClientSecret: "b", RedirectURL: "c"},
-	} {
-		if partial.Configured() {
-			t.Errorf("%+v was reported as configured", partial)
-		}
-	}
-}
-
 func TestQuery(t *testing.T) {
 	cases := []struct {
 		trigger, label string

internal/google/google.go

@@ -0,0 +1,224 @@
+// Package google implements the OAuth2 authorisation-code flow with PKCE against
+// Google.
+//
+// It was written inside internal/gmail, because Gmail was the only thing that
+// needed it. Calendar needs the same flow — the same endpoints, the same PKCE, the
+// same refresh — and the only thing that differs between them is one string: the
+// scope. A second copy would be a second place for "access_type=offline" to be
+// forgotten, which is the mistake that makes a connection die silently an hour
+// after it was made.
+//
+// What is *not* here is the API calls. Those differ completely, so gmail and gcal
+// keep their own clients and share only the way in.
+package google
+
+import (
+	"context"
+	"crypto/rand"
+	"crypto/sha256"
+	"encoding/base64"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"net/url"
+	"strings"
+	"time"
+)
+
+const (
+	authEndpoint         = "https://accounts.google.com/o/oauth2/v2/auth"
+	defaultTokenEndpoint = "https://oauth2.googleapis.com/token"
+)
+
+// Config is the OAuth client the operator registered in Google Cloud, plus the
+// scope one feature is asking for.
+//
+// The client belongs to the instance rather than to a person: one registration,
+// any number of users. The scope belongs to the feature: Gmail and Calendar sign
+// in through the same registration and ask for different things, and Google issues
+// a refresh token per authorisation — so connecting one does not disturb the other.
+type Config struct {
+	ClientID     string
+	ClientSecret string
+	RedirectURL  string
+	Scope        string
+
+	// TokenURL is where Google is. A field rather than a constant so a test can
+	// point it at a server it controls: there is no other way to exercise what
+	// happens when the token endpoint is slow or refuses. Empty is Google, so
+	// nothing outside a test has to know it exists.
+	TokenURL string
+}
+
+func (c Config) Configured() bool {
+	return c.ClientID != "" && c.ClientSecret != "" && c.RedirectURL != ""
+}
+
+func (c Config) tokenURL() string {
+	if c.TokenURL != "" {
+		return c.TokenURL
+	}
+	return defaultTokenEndpoint
+}
+
+// PKCE is the proof-key pair for one authorisation attempt.
+//
+// PKCE on a server-side client is not strictly required — the client secret already
+// authenticates the token exchange — but it costs nothing and closes the window in
+// which an authorisation code intercepted from the redirect could be replayed
+// before verdande gets to it.
+type PKCE struct {
+	Verifier  string
+	Challenge string
+	State     string
+}
+
+func NewPKCE() (PKCE, error) {
+	verifier, err := randomString(64)
+	if err != nil {
+		return PKCE{}, err
+	}
+	state, err := randomString(32)
+	if err != nil {
+		return PKCE{}, err
+	}
+	sum := sha256.Sum256([]byte(verifier))
+	return PKCE{
+		Verifier:  verifier,
+		Challenge: base64.RawURLEncoding.EncodeToString(sum[:]),
+		State:     state,
+	}, nil
+}
+
+// AuthURL is where the browser is sent to ask the person for consent.
+func (c Config) AuthURL(p PKCE) string {
+	v := url.Values{}
+	v.Set("client_id", c.ClientID)
+	v.Set("redirect_uri", c.RedirectURL)
+	v.Set("response_type", "code")
+	v.Set("scope", c.Scope)
+	v.Set("state", p.State)
+	v.Set("code_challenge", p.Challenge)
+	v.Set("code_challenge_method", "S256")
+	// Without both of these Google returns no refresh token on a re-authorisation,
+	// and the connection silently stops working an hour later.
+	v.Set("access_type", "offline")
+	v.Set("prompt", "consent")
+	return authEndpoint + "?" + v.Encode()
+}
+
+type Token struct {
+	AccessToken  string
+	RefreshToken string
+	ExpiresAt    time.Time
+}
+
+// Exchange trades the authorisation code for tokens.
+func (c Config) Exchange(ctx context.Context, code string, p PKCE) (Token, error) {
+	form := url.Values{}
+	form.Set("code", code)
+	form.Set("client_id", c.ClientID)
+	form.Set("client_secret", c.ClientSecret)
+	form.Set("redirect_uri", c.RedirectURL)
+	form.Set("grant_type", "authorization_code")
+	form.Set("code_verifier", p.Verifier)
+
+	return c.token(ctx, form)
+}
+
+// Refresh gets a fresh access token. Google does not return a new refresh token
+// here, so the caller keeps the one it has.
+func (c Config) Refresh(ctx context.Context, refreshToken string) (Token, error) {
+	form := url.Values{}
+	form.Set("refresh_token", refreshToken)
+	form.Set("client_id", c.ClientID)
+	form.Set("client_secret", c.ClientSecret)
+	form.Set("grant_type", "refresh_token")
+
+	token, err := c.token(ctx, form)
+	if err != nil {
+		return Token{}, err
+	}
+	token.RefreshToken = refreshToken
+	return token, nil
+}
+
+func (c Config) token(ctx context.Context, form url.Values) (Token, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.tokenURL(),
+		strings.NewReader(form.Encode()))
+	if err != nil {
+		return Token{}, err
+	}
+	req.Header.Set("Content-Type", "application/x-www-form-urlencoded")
+
+	resp, err := (&http.Client{Timeout: 30 * time.Second}).Do(req)
+	if err != nil {
+		return Token{}, err
+	}
+	defer resp.Body.Close()
+
+	var parsed struct {
+		AccessToken      string `json:"access_token"`
+		RefreshToken     string `json:"refresh_token"`
+		ExpiresIn        int    `json:"expires_in"`
+		Error            string `json:"error"`
+		ErrorDescription string `json:"error_description"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&parsed); err != nil {
+		return Token{}, fmt.Errorf("google: reading token response: %w", err)
+	}
+	if parsed.Error != "" {
+		return Token{}, fmt.Errorf("google: %s: %s", parsed.Error, parsed.ErrorDescription)
+	}
+	if resp.StatusCode != http.StatusOK || parsed.AccessToken == "" {
+		return Token{}, fmt.Errorf("google: token exchange failed with %s", resp.Status)
+	}
+
+	return Token{
+		AccessToken:  parsed.AccessToken,
+		RefreshToken: parsed.RefreshToken,
+		// A minute of slack, so a token that is about to expire is refreshed
+		// rather than used and rejected.
+		ExpiresAt: time.Now().Add(time.Duration(parsed.ExpiresIn-60) * time.Second),
+	}, nil
+}
+
+// ErrUnauthorized means the tokens are no longer good — the person revoked access,
+// or changed their password. The caller disconnects rather than retrying.
+var ErrUnauthorized = fmt.Errorf("google: authorisation was refused")
+
+// Get performs an authorised GET and decodes the JSON body into out.
+//
+// Shared by the two API clients because the interesting part is the same in both:
+// a 401 is not a failure to retry, it is a connection that has ended, and telling
+// those apart is what stops a revoked account being polled every ten minutes for
+// ever.
+func Get(ctx context.Context, client *http.Client, accessToken, url string, out any) error {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return err
+	}
+	req.Header.Set("Authorization", "Bearer "+accessToken)
+
+	resp, err := client.Do(req)
+	if err != nil {
+		return err
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode == http.StatusUnauthorized {
+		return ErrUnauthorized
+	}
+	if resp.StatusCode != http.StatusOK {
+		return fmt.Errorf("google: %s", resp.Status)
+	}
+	return json.NewDecoder(resp.Body).Decode(out)
+}
+
+func randomString(n int) (string, error) {
+	b := make([]byte, n)
+	if _, err := rand.Read(b); err != nil {
+		return "", err
+	}
+	return base64.RawURLEncoding.EncodeToString(b)[:n], nil
+}

internal/google/google_test.go

@@ -0,0 +1,127 @@
+package google
+
+import (
+	"crypto/sha256"
+	"encoding/base64"
+	"net/url"
+	"strings"
+	"testing"
+)
+
+// The challenge has to be the SHA-256 of the verifier, base64url without padding.
+// Getting this wrong means Google refuses every exchange with a message that does
+// not say which end is at fault.
+func TestPKCEChallengeMatchesItsVerifier(t *testing.T) {
+	p, err := NewPKCE()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	sum := sha256.Sum256([]byte(p.Verifier))
+	want := base64.RawURLEncoding.EncodeToString(sum[:])
+	if p.Challenge != want {
+		t.Errorf("challenge = %q, want %q", p.Challenge, want)
+	}
+
+	// RFC 7636 requires 43–128 characters, and no padding.
+	if n := len(p.Verifier); n < 43 || n > 128 {
+		t.Errorf("verifier is %d characters, outside the 43–128 the spec allows", n)
+	}
+	if strings.ContainsAny(p.Verifier, "=+/") {
+		t.Errorf("verifier %q contains characters that are not URL-safe", p.Verifier)
+	}
+	if p.State == "" {
+		t.Error("no state was generated")
+	}
+
+	// Two attempts must not share anything.
+	other, err := NewPKCE()
+	if err != nil {
+		t.Fatal(err)
+	}
+	if other.Verifier == p.Verifier || other.State == p.State {
+		t.Error("two authorisation attempts produced the same secrets")
+	}
+}
+
+func TestAuthURL(t *testing.T) {
+	cfg := Config{
+		ClientID:     "client-123",
+		ClientSecret: "secret",
+		RedirectURL:  "https://todo.example.dk/oauth/gmail/callback",
+		Scope:        "https://www.googleapis.com/auth/gmail.readonly",
+	}
+	p, err := NewPKCE()
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	raw := cfg.AuthURL(p)
+	parsed, err := url.Parse(raw)
+	if err != nil {
+		t.Fatalf("the URL does not parse: %v", err)
+	}
+	q := parsed.Query()
+
+	checks := map[string]string{
+		"client_id":             cfg.ClientID,
+		"redirect_uri":          cfg.RedirectURL,
+		"response_type":         "code",
+		"scope":                 cfg.Scope,
+		"code_challenge":        p.Challenge,
+		"code_challenge_method": "S256",
+		"state":                 p.State,
+		// Without both of these Google returns no refresh token on a
+		// re-authorisation, and the connection dies silently an hour later.
+		"access_type": "offline",
+		"prompt":      "consent",
+	}
+	for key, want := range checks {
+		if got := q.Get(key); got != want {
+			t.Errorf("%s = %q, want %q", key, got, want)
+		}
+	}
+
+	// The verifier itself must never leave the server.
+	if strings.Contains(raw, p.Verifier) {
+		t.Error("the code verifier was put in the authorisation URL")
+	}
+}
+
+// The scope is the one thing that differs between the two features signing in
+// through this, so it has to come out of the config rather than out of a constant
+// somebody could forget to change.
+func TestTheScopeComesFromTheConfig(t *testing.T) {
+	p, err := NewPKCE()
+	if err != nil {
+		t.Fatal(err)
+	}
+	calendar := "https://www.googleapis.com/auth/calendar.readonly"
+	raw := Config{ClientID: "a", Scope: calendar}.AuthURL(p)
+
+	parsed, err := url.Parse(raw)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got := parsed.Query().Get("scope"); got != calendar {
+		t.Errorf("scope = %q, want %q — asking for the wrong one is a consent screen "+
+			"that grants the wrong thing", got, calendar)
+	}
+}
+
+func TestConfigured(t *testing.T) {
+	full := Config{ClientID: "a", ClientSecret: "b", RedirectURL: "c"}
+	if !full.Configured() {
+		t.Error("a complete config is not reported as configured")
+	}
+	for _, partial := range []Config{
+		{},
+		{ClientID: "a"},
+		{ClientID: "a", ClientSecret: "b"},
+		{ClientSecret: "b", RedirectURL: "c"},
+	} {
+		if partial.Configured() {
+			t.Errorf("%+v was reported as configured", partial)
+		}
+	}
+}

internal/httpapi/auth_handlers.go

@@ -154,7 +154,21 @@ func (s *Server) handleLogin(w http.ResponseWriter, r *http.Request) {
 		}
 	}
 
-	s.loginLimiter.reset(clientIP(r))
+	// The bucket is cleared only when the login is *finished*.
+	//
+	// It used to be cleared here, on the strength of the password alone — before
+	// the second factor had been asked for at all. That handed somebody who
+	// already had the password an unlimited supply of ten-guess batches: log in,
+	// get a fresh bucket and a fresh pending session, spend ten guesses on the
+	// code, log in again. At three valid codes in any moment and twenty requests a
+	// second, that is even odds inside an afternoon — against the one control that
+	// exists precisely for when the password is already gone.
+	//
+	// So an account with a second factor clears nothing yet. handleLoginTOTP does
+	// it, once the code is right.
+	if !user.TOTPEnabled {
+		s.loginLimiter.reset(clientIP(r))
+	}
 
 	// A session that still needs a code gets a short life of its own: long enough
 	// to fetch a phone, not long enough to be worth stealing.
@@ -191,6 +205,19 @@ func (s *Server) handleLoginTOTP(w http.ResponseWriter, r *http.Request) {
 	user := userFrom(r.Context())
 	session := sessionFrom(r.Context())
 
+	// Counted per account as well as per address.
+	//
+	// The address bucket alone is defeated by anybody with more than one, which is
+	// the ordinary case — a phone and a laptop is already two. A code is six
+	// digits: the only thing that makes it safe is that guesses are scarce, and
+	// scarcity has to be counted against the thing being attacked.
+	if ok, retry := s.totpLimiter.allow(user.ID); !ok {
+		w.Header().Set("Retry-After", retryAfterSeconds(retry))
+		writeError(w, http.StatusTooManyRequests, CodeRateLimited,
+			"too many verification codes; wait a moment and try again")
+		return
+	}
+
 	err := auth.VerifyTOTP(user.TOTPSecret, req.Code, time.Now())
 	if err != nil {
 		// A recovery code is accepted in the same field. Making somebody find a
@@ -198,6 +225,15 @@ func (s *Server) handleLoginTOTP(w http.ResponseWriter, r *http.Request) {
 		// introduce a second concept.
 		used, rerr := s.db.UseRecoveryCode(r.Context(), user.ID, req.Code)
 		if rerr != nil || !used {
+			// The pending session is left alive on purpose.
+			//
+			// Ending it on the first wrong digit was considered and rejected: it
+			// would make somebody who mistyped one number type their password again,
+			// and it buys almost nothing. What makes a six-digit code safe is that
+			// guesses are scarce, and the per-account bucket above is what makes
+			// them scarce — ten an hour, however many times the password is
+			// re-entered. The attack this closes was never "one session, many
+			// guesses"; it was "a fresh bucket for every password check".
 			writeError(w, http.StatusUnauthorized, CodeUnauthorized,
 				"that verification code is not valid")
 			return
@@ -205,6 +241,11 @@ func (s *Server) handleLoginTOTP(w http.ResponseWriter, r *http.Request) {
 		s.log.Info("recovery code used", "user", user.ID)
 	}
 
+	// Finished. Now the buckets may be cleared — both of them, so somebody who
+	// fumbled the code twice is not still waiting once they get it right.
+	s.loginLimiter.reset(clientIP(r))
+	s.totpLimiter.reset(user.ID)
+
 	if err := s.db.PromoteSession(r.Context(), session.ID, s.cfg.SessionTTL); err != nil {
 		s.internal(w, r, "promote session", err)
 		return

internal/httpapi/auth_test.go

@@ -27,6 +27,9 @@ type testServer struct {
 	*httptest.Server
 	db     *store.DB
 	client *http.Client
+	// api er den samme server, routeren hænger på. Et par prøver har brug for at
+	// nå ind i den — se `icsFetch` — og uden den her er der kun håndtaget udefra.
+	api *Server
 }
 
 func newTestServer(t *testing.T) *testServer {
@@ -62,7 +65,8 @@ func newTestServerWith(t *testing.T, configure func(*config.Config)) *testServer
 	// Discard: a passing test should print nothing, and these handlers log freely.
 	log := slog.New(slog.NewTextHandler(io.Discard, nil))
 
-	srv := httptest.NewServer(New(cfg, db, log, nil))
+	api := New(cfg, db, log, nil)
+	srv := httptest.NewServer(api)
 	t.Cleanup(srv.Close)
 
 	jar, err := cookiejar.New(nil)
@@ -75,6 +79,7 @@ func newTestServerWith(t *testing.T, configure func(*config.Config)) *testServer
 		// The cookie jar is what makes this a session test rather than a
 		// sequence of unrelated requests.
 		client: &http.Client{Jar: jar, Timeout: 10 * time.Second},
+		api:    api,
 	}
 }
 
@@ -301,7 +306,11 @@ func TestTOTPEnrolmentAndTwoStepLogin(t *testing.T) {
 
 	// Begin: a secret exists, but 2FA is not on yet — closing the tab here must
 	// not lock the account.
-	resp, body := ts.do(t, "POST", "/api/v1/auth/totp/setup", nil)
+	// Med adgangskoden: at slå andenfaktoren til er mindst lige så indgribende som
+	// at slå den fra, og den har altid spurgt.
+	resp, body := ts.do(t, "POST", "/api/v1/auth/totp/setup", map[string]string{
+		"password": "et langt kodeord",
+	})
 	if resp.StatusCode != http.StatusOK {
 		t.Fatalf("totp setup: %d %v", resp.StatusCode, body)
 	}
@@ -368,7 +377,9 @@ func TestRecoveryCodeCompletesLoginOnce(t *testing.T) {
 	ts := newTestServer(t)
 	ts.bootstrap(t)
 
-	_, body := ts.do(t, "POST", "/api/v1/auth/totp/setup", nil)
+	_, body := ts.do(t, "POST", "/api/v1/auth/totp/setup", map[string]string{
+		"password": "et langt kodeord",
+	})
 	secret := body["secret"].(string)
 	_, body = ts.do(t, "POST", "/api/v1/auth/totp/confirm", map[string]string{
 		"code": codeFor(t, secret),
@@ -897,3 +908,65 @@ func TestUserAgentsAreSummarisedForAPerson(t *testing.T) {
 		}
 	}
 }
+
+// Andenfaktoren kan ikke slås til uden adgangskoden.
+//
+// Døren åbnede før med et token og lukkede kun med et kodeord: `disable` har
+// altid spurgt, `setup` spurgte om ingenting. Et stjålet token kunne derfor hente
+// hemmeligheden, lægge den i sin egen godkender, bekræfte den og beholde den
+// eneste kopi af gendannelseskoderne — hvorefter ejerens næste login stoppede ved
+// en kode, de ikke havde, og de ikke kunne slå den fra uden at logge ind først.
+func TestTOTPCannotBeSwitchedOnWithoutThePassword(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	resp, _ := ts.do(t, "POST", "/api/v1/auth/totp/setup", map[string]string{
+		"password": "det forkerte kodeord",
+	})
+	if resp.StatusCode != http.StatusUnprocessableEntity {
+		t.Fatalf("status %d, ventede at en forkert adgangskode blev afvist", resp.StatusCode)
+	}
+
+	resp, _ = ts.do(t, "POST", "/api/v1/auth/totp/setup", map[string]string{})
+	if resp.StatusCode == http.StatusOK {
+		t.Fatal("en tom adgangskode blev accepteret")
+	}
+}
+
+// Ratebegrænseren må ikke nulstilles af adgangskoden alene.
+//
+// Den blev det, og det gav én, der havde kodeordet, ubegrænsede portioner på ti
+// gæt: log ind, få en frisk spand og en frisk ventende session, brug ti gæt på
+// koden, log ind igen. Andenfaktoren findes præcis til det tilfælde, hvor
+// kodeordet allerede er tabt, så den portion skal betales for hver gang.
+func TestAPasswordAloneDoesNotRefillTheCodeAttempts(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	_, body := ts.do(t, "POST", "/api/v1/auth/totp/setup", map[string]string{
+		"password": "et langt kodeord",
+	})
+	secret := body["secret"].(string)
+	ts.do(t, "POST", "/api/v1/auth/totp/confirm", map[string]string{"code": codeFor(t, secret)})
+	ts.do(t, "POST", "/api/v1/auth/logout", nil)
+
+	// Kodeordet er rigtigt hver gang; kun koden er forkert. Uden loftet ville det
+	// her kunne blive ved i det uendelige.
+	refused := 0
+	for i := 0; i < 14; i++ {
+		resp, _ := ts.do(t, "POST", "/api/v1/auth/login", map[string]string{
+			"email": "kristian@example.dk", "password": "et langt kodeord",
+		})
+		if resp.StatusCode == http.StatusTooManyRequests {
+			refused++
+			continue
+		}
+		resp, _ = ts.do(t, "POST", "/api/v1/auth/login/totp", map[string]string{"code": "000000"})
+		if resp.StatusCode == http.StatusTooManyRequests {
+			refused++
+		}
+	}
+	if refused == 0 {
+		t.Fatal("fjorten runder med forkerte koder blev aldrig afvist; spanden fyldes stadig af adgangskoden")
+	}
+}

internal/httpapi/beacon_handlers.go

@@ -0,0 +1,297 @@
+package httpapi
+
+import (
+	"context"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/kristianwind/verdande/internal/beacon"
+	"github.com/kristianwind/verdande/internal/safedial"
+	"github.com/kristianwind/verdande/internal/store"
+)
+
+// The beacon, from the instance's side.
+//
+// Two halves that happen to live in the same binary: the part that reports in,
+// which every installation runs, and the part that receives, which only the one
+// acting as collector turns on. See internal/beacon for what is sent and why it
+// is on by default rather than compulsory.
+
+const beaconScope = "beacon"
+
+// beaconConfig is what this instance has been told to do. The zero value is what
+// a fresh installation gets, which is why Enabled defaults to true elsewhere and
+// not here — a zero struct must never mean "silently on".
+type beaconConfig struct {
+	Enabled      bool   `json:"enabled"`
+	Collector    string `json:"collector_url"`
+	InstanceID   string `json:"instance_id"`
+	IsCollector  bool   `json:"is_collector"`
+	PublishCount bool   `json:"publish_count"`
+	LastPingAt   int64  `json:"last_ping_at,omitempty"`
+}
+
+// beaconSettings reads the configuration, filling in the defaults and minting the
+// anonymous id on first use.
+//
+// The id is generated here rather than derived from anything about the machine.
+// A hash of the hostname or the base URL would be stable and convenient and would
+// also be a value the collector could compare against a guess — which is the
+// difference between anonymous and pseudonymous, and it is the whole point.
+func (s *Server) beaconSettings(ctx context.Context) (beaconConfig, error) {
+	cfg := beaconConfig{Enabled: true, Collector: beacon.DefaultCollector}
+
+	stored, err := s.db.InstanceSettings(ctx, beaconScope)
+	if err != nil {
+		return cfg, err
+	}
+
+	if v, ok := stored["enabled"].(bool); ok {
+		cfg.Enabled = v
+	}
+	if v, ok := stored["collector_url"].(string); ok && strings.TrimSpace(v) != "" {
+		cfg.Collector = strings.TrimSpace(v)
+	}
+	if v, ok := stored["instance_id"].(string); ok && beacon.ValidID(v) {
+		cfg.InstanceID = v
+	}
+	if v, ok := stored["is_collector"].(bool); ok {
+		cfg.IsCollector = v
+	}
+	if v, ok := stored["publish_count"].(bool); ok {
+		cfg.PublishCount = v
+	}
+	if v, ok := stored["last_ping_at"].(float64); ok {
+		cfg.LastPingAt = int64(v)
+	}
+
+	if cfg.InstanceID == "" {
+		cfg.InstanceID = store.NewID()
+		if err := s.saveBeaconSettings(ctx, cfg); err != nil {
+			return cfg, err
+		}
+	}
+	return cfg, nil
+}
+
+func (s *Server) saveBeaconSettings(ctx context.Context, cfg beaconConfig) error {
+	return s.db.SetInstanceSettings(ctx, beaconScope, map[string]any{
+		"enabled":       cfg.Enabled,
+		"collector_url": cfg.Collector,
+		"instance_id":   cfg.InstanceID,
+		"is_collector":  cfg.IsCollector,
+		"publish_count": cfg.PublishCount,
+		"last_ping_at":  cfg.LastPingAt,
+	})
+}
+
+// --- what the settings page shows ------------------------------------------------
+
+type beaconStatusJSON struct {
+	Enabled      bool               `json:"enabled"`
+	Collector    string             `json:"collector_url"`
+	InstanceID   string             `json:"instance_id"`
+	Version      string             `json:"version"`
+	IsCollector  bool               `json:"is_collector"`
+	PublishCount bool               `json:"publish_count"`
+	LastPingAt   string             `json:"last_ping_at,omitempty"`
+	Stats        *store.BeaconStats `json:"stats,omitempty"`
+}
+
+func (s *Server) handleBeaconStatus(w http.ResponseWriter, r *http.Request) {
+	cfg, err := s.beaconSettings(r.Context())
+	if err != nil {
+		s.internal(w, r, "beacon settings", err)
+		return
+	}
+
+	out := beaconStatusJSON{
+		Enabled: cfg.Enabled, Collector: cfg.Collector, InstanceID: cfg.InstanceID,
+		Version: Version, IsCollector: cfg.IsCollector, PublishCount: cfg.PublishCount,
+	}
+	if cfg.LastPingAt > 0 {
+		out.LastPingAt = time.Unix(cfg.LastPingAt, 0).Format(time.RFC3339)
+	}
+	if cfg.IsCollector {
+		stats, err := s.db.BeaconSummary(r.Context())
+		if err != nil {
+			s.internal(w, r, "beacon summary", err)
+			return
+		}
+		out.Stats = &stats
+	}
+	writeJSON(w, http.StatusOK, out)
+}
+
+type beaconRequest struct {
+	Enabled      *bool   `json:"enabled"`
+	Collector    *string `json:"collector_url"`
+	IsCollector  *bool   `json:"is_collector"`
+	PublishCount *bool   `json:"publish_count"`
+}
+
+func (s *Server) handleSetBeacon(w http.ResponseWriter, r *http.Request) {
+	var req beaconRequest
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+	cfg, err := s.beaconSettings(r.Context())
+	if err != nil {
+		s.internal(w, r, "beacon settings", err)
+		return
+	}
+
+	if req.Enabled != nil {
+		cfg.Enabled = *req.Enabled
+	}
+	if req.IsCollector != nil {
+		cfg.IsCollector = *req.IsCollector
+	}
+	if req.PublishCount != nil {
+		cfg.PublishCount = *req.PublishCount
+	}
+	if req.Collector != nil {
+		url := strings.TrimSpace(*req.Collector)
+		if url == "" {
+			url = beacon.DefaultCollector
+		}
+		// http and https only, and said as a field error rather than a 500: an
+		// operator pointing this at their own collector will mistype it, and the
+		// page should say which field is wrong.
+		if !strings.HasPrefix(url, "http://") && !strings.HasPrefix(url, "https://") {
+			writeFieldErrors(w, map[string]string{"collector_url": "must start with http:// or https://"})
+			return
+		}
+		cfg.Collector = url
+	}
+
+	if err := s.saveBeaconSettings(r.Context(), cfg); err != nil {
+		s.internal(w, r, "save beacon settings", err)
+		return
+	}
+	s.handleBeaconStatus(w, r)
+}
+
+// --- the collector's side ---------------------------------------------------------
+
+// handleBeaconPing receives one report. Unauthenticated, because the whole point
+// is that a stranger's installation can reach it.
+//
+// Nothing about the request is recorded except the two values in the body. The
+// source address is not read here, and there is no code path from this function
+// to anything that could write one down — which is the only version of that
+// promise worth making.
+func (s *Server) handleBeaconPing(w http.ResponseWriter, r *http.Request) {
+	cfg, err := s.beaconSettings(r.Context())
+	if err != nil {
+		s.internal(w, r, "beacon settings", err)
+		return
+	}
+	// An instance that is not a collector does not quietly accept and drop pings —
+	// it says there is nothing here. Otherwise every installation in the world is a
+	// spam sink that answers 200 to anything.
+	if !cfg.IsCollector {
+		writeError(w, http.StatusNotFound, CodeNotFound, "this instance is not a beacon collector")
+		return
+	}
+
+	r.Body = http.MaxBytesReader(w, r.Body, 4<<10)
+	var p beacon.Payload
+	if err := decodeJSON(w, r, &p); err != nil {
+		return
+	}
+
+	// Checked, not trusted. This is the one endpoint anybody can post to, and the
+	// two columns behind it should only ever hold the two shapes this program
+	// writes — an id that is not id-shaped is somebody probing, not an install.
+	if !beacon.ValidID(p.InstanceID) {
+		writeFieldErrors(w, map[string]string{"instance_id": "malformed"})
+		return
+	}
+	if p.Version != "" && !beacon.ValidVersion(p.Version) {
+		writeFieldErrors(w, map[string]string{"version": "malformed"})
+		return
+	}
+
+	if err := s.db.RecordBeacon(r.Context(), p.InstanceID, p.Version); err != nil {
+		s.internal(w, r, "record beacon", err)
+		return
+	}
+	w.WriteHeader(http.StatusNoContent)
+}
+
+// handleBeaconCount publishes the number, when the collector has been told to.
+//
+// Thirty days rather than a running total: a total only climbs, so it stops
+// meaning "how many people use this" and starts meaning "how many ever tried it".
+func (s *Server) handleBeaconCount(w http.ResponseWriter, r *http.Request) {
+	cfg, err := s.beaconSettings(r.Context())
+	if err != nil {
+		s.internal(w, r, "beacon settings", err)
+		return
+	}
+	if !cfg.IsCollector || !cfg.PublishCount {
+		writeError(w, http.StatusNotFound, CodeNotFound, "not published")
+		return
+	}
+
+	n, err := s.db.BeaconCount(r.Context(), 30*24*time.Hour)
+	if err != nil {
+		s.internal(w, r, "beacon count", err)
+		return
+	}
+	w.Header().Set("Cache-Control", "public, max-age=60")
+	writeJSON(w, http.StatusOK, map[string]any{"installs": n, "window_days": 30})
+}
+
+// SendBeacon reports this installation, at most once a day.
+//
+// Exported and handed to the jobs runner the same way the mailbox syncs are: the
+// settings, the anonymous id and the version all live on this side, and the
+// runner should not have to know about any of them.
+//
+// The day is counted from the last successful send, not from a wall-clock hour.
+// A fleet of installations that all pinged at 03:00 would arrive in one spike;
+// counted this way they spread themselves out by when each was started.
+func (s *Server) SendBeacon(ctx context.Context) error {
+	cfg, err := s.beaconSettings(ctx)
+	if err != nil {
+		return err
+	}
+	if !cfg.Enabled {
+		return nil
+	}
+	// The collector does not report to itself over the network. It would work, and
+	// it would also mean the one instance whose count matters most depends on its
+	// own tunnel being up to be counted.
+	if cfg.IsCollector {
+		if err := s.db.RecordBeacon(ctx, cfg.InstanceID, Version); err != nil {
+			return err
+		}
+		cfg.LastPingAt = time.Now().Unix()
+		return s.saveBeaconSettings(ctx, cfg)
+	}
+
+	if time.Since(time.Unix(cfg.LastPingAt, 0)) < 24*time.Hour {
+		return nil
+	}
+
+	// safedial, ikke http.DefaultClient. Adressen kan skrives i fladen, og en
+	// administrator, der peger den på 127.0.0.1 eller på 169.254.169.254, ville
+	// ellers have en maskine, der henter indefra på kommando — og svaret på om den
+	// svarede, er i sig selv det, sådan et forsøg leder efter.
+	if err := beacon.Send(ctx, safedial.Client(15*time.Second), cfg.Collector, beacon.Payload{
+		InstanceID: cfg.InstanceID,
+		Version:    Version,
+	}); err != nil {
+		// Logged at info, not error. The collector being unreachable is not this
+		// installation's problem and must never look like one — it is somebody
+		// else's server, and the only consequence is a number being one lower.
+		s.log.Info("beacon not delivered", "err", err)
+		return nil
+	}
+
+	cfg.LastPingAt = time.Now().Unix()
+	return s.saveBeaconSettings(ctx, cfg)
+}

internal/httpapi/caldav_handlers.go

@@ -231,6 +231,33 @@ func (s *Server) handleCalDAVPut(w http.ResponseWriter, r *http.Request) {
 
 	existing, getErr := s.db.GetTask(r.Context(), taskID, user.ID)
 	if getErr == nil {
+		// The task decides, not the URL.
+		//
+		// The check above is about the project the *client named*, and a client
+		// names both halves of this address. Asking the project for permission and
+		// then taking the task id from the same string ties nothing together: a
+		// viewer on a shared project could name a project of their own — the Inbox,
+		// which everybody owns and may edit — and put the shared task's id in the
+		// filename. The project check passed on the Inbox and the write landed on a
+		// task they may only read.
+		//
+		// GetTask does not close it either: it answers "may this person see it",
+		// and a viewer may. `CanEdit` is the question, and it is the same one
+		// handleUpdateTask asks — the API door has always asked it.
+		role, err := store.TaskRole(r.Context(), s.db, taskID, user.ID)
+		if err != nil || !role.CanEdit() {
+			http.Error(w, "forbidden", http.StatusForbidden)
+			return
+		}
+		// And it has to live where the address says it does. Without this an editor
+		// on two projects can reach either through the other's URL — nothing is
+		// escalated by that on its own, but a URL allowed to lie about where a
+		// resource lives is how the case above got in.
+		if existing.ProjectID != projectID {
+			http.NotFound(w, r)
+			return
+		}
+
 		// An update. Only the fields a task client can express are touched;
 		// everything else the task carries stays as it is.
 		content := parsed.Summary

internal/httpapi/calendar_handlers.go

@@ -0,0 +1,808 @@
+package httpapi
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"net/http"
+	"strings"
+	"time"
+
+	"github.com/go-chi/chi/v5"
+
+	"github.com/kristianwind/verdande/internal/gcal"
+	"github.com/kristianwind/verdande/internal/ics"
+	"github.com/kristianwind/verdande/internal/safedial"
+	"github.com/kristianwind/verdande/internal/store"
+)
+
+// defaultCalendarSyncBudget bounds one "fetch now" when nothing else says
+// otherwise. Well under the hundred seconds a proxy is willing to wait, so this
+// server is the one that answers a slow Google rather than Cloudflare's own HTML
+// page — which is how a failure arrives with no code, no message and nothing in
+// the error log.
+const defaultCalendarSyncBudget = 25 * time.Second
+
+// The window verdande keeps a copy of.
+//
+// A month grid pages, so asking Google on every arrow press would mean a round trip
+// for a question answered ninety seconds ago — a calendar slower than the calendar
+// it is showing. A cached window is the alternative, and a window has to end
+// somewhere. Ninety days back is roughly "when was that", a year forward is past
+// anything anybody has booked; between them they cover every screen the grid can
+// reach without a person noticing an edge.
+//
+// The edge is still real, and the events endpoint says where it is rather than
+// answering an empty list for a month outside it. A calendar that quietly shows no
+// events is a calendar that lies about a day being clear.
+const (
+	calendarWindowBack   = 90
+	calendarWindowAhead  = 365
+	calendarDateLayout   = "2006-01-02"
+	calendarProviderName = "google"
+)
+
+func (s *Server) calendarClient(accessToken string) *gcal.Client {
+	return gcal.NewClient(accessToken).At(s.cfg.CalendarAPIURL)
+}
+
+// calendarConfig is the same registration Gmail signs in through, asking for a
+// different scope and coming back to a different URI.
+//
+// One OAuth client, two features. Google issues a refresh token per authorisation
+// and keeps up to a hundred of them live per account, each carrying the scopes it
+// was granted with — so connecting the calendar does not disturb a working Gmail,
+// and disconnecting one does not take the other with it.
+func (s *Server) calendarConfig(ctx context.Context) gcal.Config {
+	cfg := s.gmailConfig(ctx)
+	cfg.Scope = gcal.Scope
+	cfg.RedirectURL = s.cfg.CalendarRedirectURL()
+	return cfg
+}
+
+// --- connecting ------------------------------------------------------------------
+
+// handleCalendarAuthorize starts the flow. Same shape as Gmail's, and the same
+// reason for keeping the verifier server-side: the callback arrives as a top-level
+// navigation from Google, and a SameSite=Lax cookie is the kind of thing that works
+// until a browser tightens its defaults.
+func (s *Server) handleCalendarAuthorize(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+	cfg := s.calendarConfig(r.Context())
+
+	if !cfg.Configured() {
+		writeError(w, http.StatusConflict, CodeGmailNotConfigured,
+			"no Google OAuth client is registered on this server")
+		return
+	}
+
+	pkce, err := gcal.NewPKCE()
+	if err != nil {
+		s.internal(w, r, "calendar pkce", err)
+		return
+	}
+
+	// Its own settings scope, not Gmail's. Two handshakes in one row would mean a
+	// person who started both flows had one verifier, and whichever came back
+	// second would be refused for a reason nothing on the screen could explain.
+	settings, err := s.db.UserSettings(r.Context(), user.ID, "calendar")
+	if err != nil {
+		s.internal(w, r, "calendar settings", err)
+		return
+	}
+	settings["pkce_verifier"] = pkce.Verifier
+	settings["pkce_state"] = pkce.State
+	// The attempt expires: a stored verifier that lives forever is a replay window
+	// that lives forever.
+	settings["pkce_expires"] = time.Now().Add(10 * time.Minute).Unix()
+
+	if err := s.db.SetUserSettings(r.Context(), user.ID, "calendar", settings); err != nil {
+		s.internal(w, r, "calendar settings", err)
+		return
+	}
+	writeJSON(w, http.StatusOK, map[string]string{"url": cfg.AuthURL(pkce)})
+}
+
+// handleCalendarCallback is where Google sends the browser back.
+//
+// It answers with a redirect into the app rather than with JSON: the person is
+// looking at a browser tab, not at a fetch response.
+func (s *Server) handleCalendarCallback(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+	settingsURL := s.cfg.BaseURL + "/indstillinger/integrationer"
+
+	if errParam := r.URL.Query().Get("error"); errParam != "" {
+		// Somebody clicked "cancel" on the consent screen — or, on an Internal
+		// OAuth app, Google refused an account outside the organisation with
+		// `org_internal`. Both arrive here, and the page names what it was given.
+		http.Redirect(w, r, settingsURL+"?calendar="+errParam, http.StatusFound)
+		return
+	}
+
+	code := r.URL.Query().Get("code")
+	state := r.URL.Query().Get("state")
+	if code == "" || state == "" {
+		http.Redirect(w, r, settingsURL+"?calendar=invalid", http.StatusFound)
+		return
+	}
+
+	settings, err := s.db.UserSettings(r.Context(), user.ID, "calendar")
+	if err != nil {
+		s.internal(w, r, "calendar settings", err)
+		return
+	}
+	verifier, _ := settings["pkce_verifier"].(string)
+	expected, _ := settings["pkce_state"].(string)
+	expires, _ := settings["pkce_expires"].(float64)
+
+	// The state has to match, or this endpoint accepts a code from anywhere.
+	if verifier == "" || expected == "" || state != expected {
+		http.Redirect(w, r, settingsURL+"?calendar=state", http.StatusFound)
+		return
+	}
+	if int64(expires) < time.Now().Unix() {
+		http.Redirect(w, r, settingsURL+"?calendar=expired", http.StatusFound)
+		return
+	}
+
+	token, err := s.calendarConfig(r.Context()).Exchange(r.Context(), code,
+		gcal.PKCE{Verifier: verifier})
+	if err != nil {
+		s.log.Error("calendar exchange", "err", err, "user", user.ID)
+		http.Redirect(w, r, settingsURL+"?calendar=failed", http.StatusFound)
+		return
+	}
+	if token.RefreshToken == "" {
+		// Without one the connection dies in an hour. Google only withholds it when
+		// the consent prompt was skipped, which access_type and prompt are meant to
+		// prevent — so this means something is wrong with the client registration
+		// rather than with this request.
+		s.log.Error("calendar returned no refresh token", "user", user.ID)
+		http.Redirect(w, r, settingsURL+"?calendar=norefresh", http.StatusFound)
+		return
+	}
+
+	// The one-time values are cleared: keeping a spent verifier serves nothing and
+	// is one more secret at rest.
+	delete(settings, "pkce_verifier")
+	delete(settings, "pkce_state")
+	delete(settings, "pkce_expires")
+	if err := s.db.SetUserSettings(r.Context(), user.ID, "calendar", settings); err != nil {
+		s.internal(w, r, "clear calendar handshake", err)
+		return
+	}
+
+	account := &store.CalendarAccount{
+		UserID: user.ID, Provider: calendarProviderName,
+		RefreshToken: token.RefreshToken, AccessToken: token.AccessToken,
+		ExpiresAt: token.ExpiresAt,
+	}
+	if err := s.db.SaveCalendarAccount(r.Context(), account); err != nil {
+		s.internal(w, r, "save calendar tokens", err)
+		return
+	}
+
+	// The list of calendars, fetched now so the page that opens next has something
+	// to tick rather than an empty panel and a button. Best effort: an account that
+	// connects but cannot list itself is still connected, and the status endpoint
+	// asks again when it finds nothing.
+	if _, err := s.refreshCalendarList(r.Context(), account); err != nil {
+		s.log.Warn("calendar list", "err", err, "user", user.ID)
+	}
+
+	s.log.Info("calendar connected", "user", user.ID, "account", account.Account)
+	http.Redirect(w, r, settingsURL+"?calendar=connected", http.StatusFound)
+}
+
+// --- what the settings page reads --------------------------------------------------
+
+type calendarStatusJSON struct {
+	Connected bool             `json:"connected"`
+	Account   string           `json:"account,omitempty"`
+	Calendars []store.Calendar `json:"calendars"`
+	// RedirectURI is what has to be registered with Google, spelled out. Derived
+	// from VERDANDE_BASE_URL and matched exactly — the single most likely thing to
+	// be wrong, and the error Google gives for it names neither value.
+	RedirectURI string `json:"redirect_uri"`
+	// HasClient says whether there is an OAuth registration to connect through at
+	// all, so the page can explain rather than offer a button that answers 409.
+	HasClient  bool   `json:"has_client"`
+	LastSyncAt string `json:"last_sync_at,omitempty"`
+	// ReadOnly is stated rather than assumed. Nothing writes a Google calendar, and
+	// an interface that quietly looks editable is one somebody will try to edit.
+	ReadOnly bool `json:"read_only"`
+
+	// Abonnementer står ved siden af Google-kontoen og ikke inde i den: man kan
+	// have det ene uden det andet, og det er hele pointen med dem — en privat
+	// konto, der ikke kan bruge en Internal OAuth-klient, kan abonnere.
+	Subscriptions []store.CalendarAccount `json:"subscriptions"`
+}
+
+func (s *Server) handleGetCalendar(w http.ResponseWriter, r *http.Request) {
+	// Læst først, så de også kommer med, når der ingen Google-konto er.
+	subscriptions, err := s.db.Subscriptions(r.Context(), userFrom(r.Context()).ID)
+	if err != nil {
+		s.internal(w, r, "subscriptions", err)
+		return
+	}
+
+	user := userFrom(r.Context())
+
+	out := calendarStatusJSON{
+		Calendars:   []store.Calendar{},
+		RedirectURI: s.cfg.CalendarRedirectURL(),
+		HasClient:   s.calendarConfig(r.Context()).Configured(),
+		ReadOnly:    true,
+	}
+
+	account, err := s.db.CalendarAccountFor(r.Context(), user.ID, calendarProviderName)
+	if err != nil {
+		s.internal(w, r, "calendar account", err)
+		return
+	}
+	if account == nil || account.RefreshToken == "" {
+		out.Subscriptions = subscriptions
+		writeJSON(w, http.StatusOK, out)
+		return
+	}
+
+	out.Connected = true
+	out.Account = account.Account
+	if !account.LastSyncAt.IsZero() {
+		out.LastSyncAt = account.LastSyncAt.Format(time.RFC3339)
+	}
+
+	calendars, err := s.db.Calendars(r.Context(), account.ID)
+	if err != nil {
+		s.internal(w, r, "calendars", err)
+		return
+	}
+	// An account connected during a Google hiccup has no calendars and no way back:
+	// the person sees an empty list and nothing to press. Asking again here costs
+	// one request on a page nobody opens often, and means the wrong answer lasts
+	// until the next look rather than for ever. Same reasoning as the Gmail address.
+	if len(calendars) == 0 {
+		if fetched, err := s.refreshCalendarList(r.Context(), account); err == nil {
+			calendars = fetched
+			out.Account = account.Account
+		} else {
+			s.log.Warn("calendar list", "err", err, "user", user.ID)
+		}
+	}
+	if calendars != nil {
+		out.Calendars = calendars
+	}
+	out.Subscriptions = subscriptions
+	writeJSON(w, http.StatusOK, out)
+}
+
+type showCalendarsRequest struct {
+	// Shown is the whole set, not one flip. Same reasoning as the projects' order:
+	// a person has a handful of calendars, the question is "these ones", and a list
+	// that can only be built by several requests can land half applied.
+	Shown []string `json:"shown"`
+}
+
+func (s *Server) handleSetCalendars(w http.ResponseWriter, r *http.Request) {
+	var req showCalendarsRequest
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+	user := userFrom(r.Context())
+
+	account, err := s.db.CalendarAccountFor(r.Context(), user.ID, calendarProviderName)
+	if err != nil {
+		s.internal(w, r, "calendar account", err)
+		return
+	}
+	if account == nil {
+		// A 404 rather than a 409: everything the caller may not see answers 404,
+		// and "there is no such connection" is the true answer either way.
+		writeError(w, http.StatusNotFound, CodeNotFound, "no calendar is connected")
+		return
+	}
+	if err := s.db.ShowCalendars(r.Context(), account.ID, req.Shown); err != nil {
+		s.internal(w, r, "show calendars", err)
+		return
+	}
+	w.WriteHeader(http.StatusNoContent)
+}
+
+// handleDisconnectCalendar forgets the connection, and the copy of the calendars
+// with it. Unlike the tasks a mailbox made, none of this is the person's own work.
+func (s *Server) handleDisconnectCalendar(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+	account, err := s.db.CalendarAccountFor(r.Context(), user.ID, calendarProviderName)
+	if err != nil {
+		s.internal(w, r, "calendar account", err)
+		return
+	}
+	if account != nil {
+		if err := s.db.DeleteCalendarAccount(r.Context(), user.ID, account.ID); err != nil {
+			s.internal(w, r, "disconnect calendar", err)
+			return
+		}
+	}
+	w.WriteHeader(http.StatusNoContent)
+}
+
+// --- what the grid reads -----------------------------------------------------------
+
+type calendarEventsJSON struct {
+	Events []store.CalendarEvent `json:"events"`
+	// From and To are the window verdande holds a copy of, not the window that was
+	// asked for. A grid that has paged past it gets no events, and an empty answer
+	// with no explanation is a calendar quietly claiming a day is clear.
+	From string `json:"from"`
+	To   string `json:"to"`
+}
+
+func (s *Server) handleCalendarEvents(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+	from := r.URL.Query().Get("from")
+	to := r.URL.Query().Get("to")
+
+	fields := map[string]string{}
+	if !validDay(from) {
+		fields["from"] = "must be a date, as 2026-08-20"
+	}
+	if !validDay(to) {
+		fields["to"] = "must be a date, as 2026-08-20"
+	}
+	if len(fields) > 0 {
+		writeFieldErrors(w, fields)
+		return
+	}
+
+	events, err := s.db.CalendarEvents(r.Context(), user.ID, from, to)
+	if err != nil {
+		s.internal(w, r, "calendar events", err)
+		return
+	}
+	if events == nil {
+		events = []store.CalendarEvent{}
+	}
+	windowFrom, windowTo := calendarWindow(time.Now())
+	writeJSON(w, http.StatusOK, calendarEventsJSON{
+		Events: events, From: windowFrom, To: windowTo,
+	})
+}
+
+// handleCalendarSyncNow refreshes immediately, so somebody who has just connected
+// can see it work rather than wondering whether it did.
+func (s *Server) handleCalendarSyncNow(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+
+	// A budget of its own, well under what sits in front of this server. An unset
+	// budget is not a budget of nothing: a zero here would expire the context
+	// before the first call and make every sync a silent no-op.
+	budget := s.cfg.CalendarSyncBudget
+	if budget <= 0 {
+		budget = defaultCalendarSyncBudget
+	}
+	ctx, cancel := context.WithTimeout(r.Context(), budget)
+	defer cancel()
+
+	count, err := s.SyncCalendars(ctx, user)
+	if err != nil {
+		if errors.Is(err, context.DeadlineExceeded) {
+			// Not a failure: Google is slow or busy, some calendars were read, and
+			// the rest arrive on the next run. Every calendar is committed as it is
+			// read, so a short budget is a shorter run rather than a lost one.
+			writeJSON(w, http.StatusOK, map[string]any{"events": count, "partial": true})
+			return
+		}
+		s.upstream(w, r, CodeCalendarFailed, "sync calendar", err)
+		return
+	}
+	writeJSON(w, http.StatusOK, map[string]int{"events": count})
+}
+
+// SyncCalendars refreshes every calendar this person has chosen to see.
+//
+// Returns how many events were written. Also used by the background job, which is
+// why it takes a context and a user rather than a request.
+func (s *Server) SyncCalendars(ctx context.Context, user *store.User) (int, error) {
+	// Abonnementer først, og deres fejl tæller ikke mod Google.
+	//
+	// De to har intet med hinanden at gøre ud over at ende samme sted: et
+	// abonnement er en adresse, der hentes, hvor Google er et token, der udløber.
+	// En kalender, hvis vært er nede, må ikke få det til at se ud som om
+	// Google-forbindelsen er væk — og omvendt.
+	written, err := s.syncSubscriptions(ctx, user)
+	if err != nil {
+		s.log.Warn("calendar subscriptions", "err", err, "user", user.ID)
+	}
+
+	account, err := s.db.CalendarAccountFor(ctx, user.ID, calendarProviderName)
+	if err != nil {
+		return written, err
+	}
+	if account == nil || account.RefreshToken == "" {
+		return written, nil // not connected; nothing to do and nothing to complain about
+	}
+
+	// The same guard the mailboxes have, for the same reason: a press of "fetch
+	// now" during the sweep would otherwise have two runs replacing the same window
+	// at once, and the loser's delete would land after the winner's insert.
+	unlock := s.lockSync("calendar:" + account.ID)
+	defer unlock()
+	if fresh, err := s.db.CalendarAccountFor(ctx, user.ID, calendarProviderName); err == nil && fresh != nil {
+		account = fresh
+	}
+
+	access, err := s.calendarAccessToken(ctx, account)
+	if err != nil {
+		return 0, err
+	}
+	client := s.calendarClient(access)
+
+	// The list first, so a calendar added or renamed since the last run is drawn
+	// under its new name and one that has gone stops being drawn at all.
+	calendars, err := s.calendarList(ctx, client, account)
+	if err != nil {
+		return 0, s.forgetIfRevoked(ctx, user, account, err)
+	}
+
+	from, to := calendarWindow(time.Now())
+	for _, calendar := range calendars {
+		if !calendar.Shown {
+			continue
+		}
+		// A spent budget ends the run rather than turning into a failed fetch per
+		// calendar, logged one at a time.
+		if ctx.Err() != nil {
+			return written, ctx.Err()
+		}
+		events, err := client.Events(ctx, calendar.RemoteID, from, to)
+		if err != nil {
+			if errors.Is(err, gcal.ErrUnauthorized) {
+				return written, s.forgetIfRevoked(ctx, user, account, err)
+			}
+			// One calendar somebody was unshared from must not stop the others.
+			s.log.Warn("calendar events", "err", err, "calendar", calendar.Name)
+			continue
+		}
+
+		rows := make([]store.CalendarEvent, 0, len(events))
+		for _, e := range events {
+			rows = append(rows, store.CalendarEvent{
+				RemoteID: e.ID, Summary: e.Summary,
+				StartsAt: e.StartsAt, EndsAt: e.EndsAt,
+				StartDay: e.StartDay, EndDay: e.EndDay,
+				AllDay: e.AllDay, Location: e.Location, URL: e.URL,
+			})
+		}
+		if err := s.db.ReplaceCalendarEvents(ctx, calendar.ID, from, to, rows); err != nil {
+			return written, err
+		}
+		written += len(rows)
+	}
+
+	if err := s.db.MarkCalendarSynced(ctx, account.ID, time.Now()); err != nil {
+		return written, err
+	}
+	return written, nil
+}
+
+// calendarAccessToken returns a usable access token, refreshing only when the one
+// held has actually expired.
+func (s *Server) calendarAccessToken(ctx context.Context, account *store.CalendarAccount) (string, error) {
+	if account.AccessToken != "" && account.ExpiresAt.After(time.Now()) {
+		return account.AccessToken, nil
+	}
+	token, err := s.calendarConfig(ctx).Refresh(ctx, account.RefreshToken)
+	if err != nil {
+		return "", err
+	}
+	account.AccessToken = token.AccessToken
+	account.ExpiresAt = token.ExpiresAt
+	if err := s.db.SaveCalendarAccount(ctx, account); err != nil {
+		return "", err
+	}
+	return account.AccessToken, nil
+}
+
+// refreshCalendarList fetches the calendar list with a token it obtains itself.
+// Used by the paths that have an account but no client yet — connecting, and the
+// status endpoint correcting an empty list.
+func (s *Server) refreshCalendarList(ctx context.Context, account *store.CalendarAccount) ([]store.Calendar, error) {
+	access, err := s.calendarAccessToken(ctx, account)
+	if err != nil {
+		return nil, err
+	}
+	return s.calendarList(ctx, s.calendarClient(access), account)
+}
+
+// calendarList writes what the provider says and returns what is stored, which is
+// not the same list: the provider says what exists, and only verdande knows which
+// of them somebody chose to look at.
+func (s *Server) calendarList(ctx context.Context, client *gcal.Client, account *store.CalendarAccount) ([]store.Calendar, error) {
+	remote, err := client.Calendars(ctx)
+	if err != nil {
+		return nil, err
+	}
+
+	list := make([]store.Calendar, 0, len(remote))
+	for _, c := range remote {
+		list = append(list, store.Calendar{
+			RemoteID: c.ID, Name: c.Name, Colour: c.Colour,
+			TimeZone: c.TimeZone, Primary: c.Primary, Writable: c.Owner,
+		})
+		// The primary calendar is the account, and its id is the address. It is the
+		// only place the connected account names itself without a second scope —
+		// asking for a profile scope to learn an address already in the answer would
+		// be one more thing on the consent screen for nothing.
+		if c.Primary && account.Account == "" {
+			account.Account = c.ID
+			if err := s.db.SaveCalendarAccount(ctx, account); err != nil {
+				return nil, err
+			}
+		}
+	}
+	if err := s.db.ReplaceCalendars(ctx, account.ID, list); err != nil {
+		return nil, err
+	}
+	return s.db.Calendars(ctx, account.ID)
+}
+
+// forgetIfRevoked disconnects when Google says the grant is gone, and passes every
+// other failure through untouched.
+//
+// The distinction is the point. A revoked account polled every ten minutes for ever
+// is a log line a week long and a connection the person believes they removed; a
+// Google that is merely down is one that will answer next time.
+func (s *Server) forgetIfRevoked(ctx context.Context, user *store.User, account *store.CalendarAccount, err error) error {
+	if errors.Is(err, gcal.ErrUnauthorized) {
+		s.log.Info("calendar access was revoked; disconnecting", "user", user.ID)
+		_ = s.db.DeleteCalendarAccount(ctx, user.ID, account.ID)
+	}
+	return err
+}
+
+// calendarWindow is the span verdande keeps a copy of, measured from today.
+//
+// Built from the calendar date rather than by truncating the instant: Truncate
+// rounds against the Unix epoch, which is midnight UTC, and in Copenhagen that is
+// two in the morning — so for two hours of every summer day the window would start
+// and end a day out.
+func calendarWindow(now time.Time) (string, string) {
+	day := time.Date(now.Year(), now.Month(), now.Day(), 0, 0, 0, 0, now.Location())
+	return day.AddDate(0, 0, -calendarWindowBack).Format(calendarDateLayout),
+		day.AddDate(0, 0, calendarWindowAhead).Format(calendarDateLayout)
+}
+
+func validDay(s string) bool {
+	_, err := time.Parse(calendarDateLayout, s)
+	return err == nil
+}
+
+// --- abonnementer -----------------------------------------------------------------
+
+// maxICSBytes bounds one fetch. A published calendar of a decade of holidays is a
+// few hundred kilobytes; ten megabytes is not a calendar, and reading it into
+// memory on a schedule with nobody watching is how a homelab instance falls over
+// at three in the morning.
+const maxICSBytes = 10 << 20
+
+// normaliseCalendarURL turns what somebody pastes into something that can be
+// fetched.
+//
+// `webcal://` is not a scheme anything speaks. It is a convention — the same URL
+// over https, with a scheme that makes a desktop hand it to the calendar app
+// instead of the browser — and every publisher writes their links that way. A
+// person pasting one has pasted the right address; refusing it would be refusing
+// them on a technicality.
+func normaliseCalendarURL(raw string) (string, error) {
+	raw = strings.TrimSpace(raw)
+	if after, ok := strings.CutPrefix(raw, "webcal://"); ok {
+		raw = "https://" + after
+	}
+	if !strings.HasPrefix(raw, "https://") && !strings.HasPrefix(raw, "http://") {
+		return "", errors.New("must be an https:// or webcal:// address")
+	}
+	// The same wall the AI provider's base_url goes through, and for the same
+	// reason: this is an address a user typed, and the server is about to fetch it
+	// on a schedule of their choosing.
+	if reason := safedial.CheckURL(raw); reason != "" {
+		return "", errors.New(reason)
+	}
+	return raw, nil
+}
+
+// icsClient is the client a subscribed calendar is fetched with.
+//
+// safedial, checked on the *resolved* address at connect time and on every
+// redirect hop — because a name that passed the check when it was pasted can
+// answer 127.0.0.1 when the sweep runs at three in the morning.
+//
+// The field exists so a test can put an ordinary client here: httptest listens on
+// loopback, which is exactly what this refuses, so a test of the *format* cannot
+// otherwise reach its own server. It is never set outside a test, and the wall it
+// bypasses is tested where it belongs — `normaliseCalendarURL` refuses a private
+// address where it is written down, and `internal/push` proves the connect-time
+// half.
+func (s *Server) icsClient() *http.Client {
+	if s.icsFetch != nil {
+		return s.icsFetch
+	}
+	return safedial.Client(30 * time.Second)
+}
+
+// fetchICS reads a subscribed calendar.
+//
+// safedial, not a plain client — checked on the resolved address at connect time
+// and on every redirect hop, because a name that passed the check when it was
+// pasted can answer 127.0.0.1 when the sweep runs at three in the morning.
+func (s *Server) fetchICS(ctx context.Context, url string) (ics.Subscribed, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
+	if err != nil {
+		return ics.Subscribed{}, err
+	}
+	req.Header.Set("Accept", "text/calendar, text/plain;q=0.9, */*;q=0.5")
+	req.Header.Set("User-Agent", "verdande/"+Version)
+
+	resp, err := s.icsClient().Do(req)
+	if err != nil {
+		return ics.Subscribed{}, err
+	}
+	defer resp.Body.Close()
+	if resp.StatusCode >= 300 {
+		return ics.Subscribed{}, fmt.Errorf("%s said %s", resp.Request.URL.Host, resp.Status)
+	}
+
+	body, err := io.ReadAll(io.LimitReader(resp.Body, maxICSBytes+1))
+	if err != nil {
+		return ics.Subscribed{}, err
+	}
+	if len(body) > maxICSBytes {
+		return ics.Subscribed{}, errors.New("that calendar is too large")
+	}
+	return ics.ParseCalendar(string(body)), nil
+}
+
+// syncSubscriptions refreshes every calendar this person subscribes to by address.
+//
+// One failing address does not stop the others. A subscription is somebody else's
+// server, and the ordinary state of somebody else's server is that it is fine
+// until one morning it is not — a holiday feed that 404s must not take the
+// football club's fixtures down with it.
+func (s *Server) syncSubscriptions(ctx context.Context, user *store.User) (int, error) {
+	subs, err := s.db.Subscriptions(ctx, user.ID)
+	if err != nil || len(subs) == 0 {
+		return 0, err
+	}
+
+	from, to := calendarWindow(time.Now())
+	written := 0
+	var failures []string
+
+	for _, sub := range subs {
+		if ctx.Err() != nil {
+			return written, ctx.Err()
+		}
+		unlock := s.lockSync("calendar:" + sub.ID)
+
+		calendarID, err := s.db.SubscribedCalendarID(ctx, sub.ID)
+		if err != nil {
+			unlock()
+			failures = append(failures, sub.URL)
+			continue
+		}
+
+		cal, err := s.fetchICS(ctx, sub.URL)
+		if err != nil {
+			unlock()
+			s.log.Warn("subscribed calendar", "err", err, "url", sub.URL, "user", user.ID)
+			failures = append(failures, sub.URL)
+			continue
+		}
+
+		// The name the file calls itself, once. A subscription is added under its
+		// address because that is all there is to go on before it has been fetched.
+		if err := s.db.RenameSubscribedCalendar(ctx, sub.ID, cal.Name); err != nil {
+			s.log.Warn("name subscribed calendar", "err", err)
+		}
+
+		rows := make([]store.CalendarEvent, 0, len(cal.Events))
+		for _, e := range cal.Events {
+			// Outside the window is not stored. The grid asks about a window, the
+			// sweep replaces a window, and a decade of holidays in one file would
+			// otherwise be a decade of rows rewritten every fifteen minutes.
+			if e.EndDay < from || e.StartDay > to {
+				continue
+			}
+			rows = append(rows, store.CalendarEvent{
+				RemoteID: e.UID, Summary: e.Summary, Location: e.Location, URL: e.URL,
+				StartsAt: e.StartsAt, EndsAt: e.EndsAt,
+				StartDay: e.StartDay, EndDay: e.EndDay, AllDay: e.AllDay,
+			})
+		}
+		if err := s.db.ReplaceCalendarEvents(ctx, calendarID, from, to, rows); err != nil {
+			unlock()
+			return written, err
+		}
+		if err := s.db.MarkCalendarSynced(ctx, sub.ID, time.Now()); err != nil {
+			s.log.Warn("mark subscription synced", "err", err)
+		}
+		written += len(rows)
+		unlock()
+	}
+
+	if len(failures) == len(subs) {
+		return written, fmt.Errorf("none of the %d subscribed calendars could be read", len(subs))
+	}
+	return written, nil
+}
+
+// handleSubscribeCalendar adds a calendar somebody only has an address to.
+//
+// Fetched once, here, before it is stored. An address that cannot be read is a
+// typo far more often than it is a server that happens to be down, and telling
+// somebody now — while they are looking at the field they just pasted into — is
+// worth the two seconds. A subscription that silently never fills is the same
+// silence the notifications had.
+func (s *Server) handleSubscribeCalendar(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		URL  string `json:"url"`
+		Name string `json:"name"`
+	}
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+
+	url, err := normaliseCalendarURL(req.URL)
+	if err != nil {
+		writeFieldErrors(w, map[string]string{"url": err.Error()})
+		return
+	}
+
+	user := userFrom(r.Context())
+	ctx, cancel := context.WithTimeout(r.Context(), 20*time.Second)
+	defer cancel()
+
+	cal, err := s.fetchICS(ctx, url)
+	if err != nil {
+		s.log.Warn("subscribe calendar", "err", err, "user", user.ID)
+		writeFieldErrors(w, map[string]string{"url": "that address could not be read"})
+		return
+	}
+
+	name := strings.TrimSpace(req.Name)
+	if name == "" {
+		name = strings.TrimSpace(cal.Name)
+	}
+	if name == "" {
+		name = url
+	}
+
+	account, err := s.db.SubscribeCalendar(r.Context(), user.ID, url, name)
+	if err != nil {
+		// The unique index is the one refusal worth naming: subscribing twice to
+		// the same address is the same calendar fetched twice, with no way to say
+		// which one counts.
+		if errors.Is(err, store.ErrDuplicate) {
+			writeFieldErrors(w, map[string]string{"url": "that calendar is already subscribed"})
+			return
+		}
+		s.storeError(w, r, "subscribe calendar", err)
+		return
+	}
+
+	// The events it already fetched, rather than making the person wait fifteen
+	// minutes for the sweep to prove it worked.
+	if _, err := s.syncSubscriptions(r.Context(), user); err != nil {
+		s.log.Warn("first subscription sync", "err", err)
+	}
+	writeJSON(w, http.StatusCreated, account)
+}
+
+func (s *Server) handleUnsubscribeCalendar(w http.ResponseWriter, r *http.Request) {
+	user := userFrom(r.Context())
+	if err := s.db.DeleteCalendarAccount(r.Context(), user.ID, chi.URLParam(r, "id")); err != nil {
+		s.storeError(w, r, "unsubscribe calendar", err)
+		return
+	}
+	w.WriteHeader(http.StatusNoContent)
+}

internal/httpapi/calendar_test.go

@@ -0,0 +1,471 @@
+package httpapi
+
+import (
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/kristianwind/verdande/internal/config"
+	"github.com/kristianwind/verdande/internal/store"
+)
+
+// fakeCalendarGoogle stands in for the parts of Google this feature touches: the
+// token endpoint, the calendar list and one calendar's events.
+//
+// There is no other way to exercise any of it. The real thing needs a registered
+// OAuth client, an account and somebody's consent, and none of the three can be had
+// inside a test — so the shapes Google returns are written down and the behaviour
+// is checked against them, the way internal/imap is checked against go-imap's own
+// in-process server.
+type fakeCalendarGoogle struct {
+	*httptest.Server
+	// refreshes counts token refreshes, so a test can prove one is not made for a
+	// token that has not expired.
+	refreshes int
+	// refuse makes every API call answer 401, which is what a revoked grant does.
+	refuse bool
+	events map[string][]map[string]any
+}
+
+func newFakeCalendarGoogle(t *testing.T) *fakeCalendarGoogle {
+	t.Helper()
+	g := &fakeCalendarGoogle{events: map[string][]map[string]any{}}
+	g.Server = httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		switch {
+		case r.URL.Path == "/token":
+			g.refreshes++
+			_ = json.NewEncoder(w).Encode(map[string]any{
+				"access_token": "ya29.fresh", "refresh_token": "1//0-refresh",
+				"expires_in": 3600,
+			})
+		case g.refuse:
+			http.Error(w, `{"error":"invalid_credentials"}`, http.StatusUnauthorized)
+		case strings.HasPrefix(r.URL.Path, "/users/me/calendarList"):
+			_ = json.NewEncoder(w).Encode(map[string]any{"items": []map[string]any{
+				{"id": "kw@nolimit.dk", "summary": "kw@nolimit.dk", "primary": true,
+					"accessRole": "owner", "backgroundColor": "#3f51b5",
+					"timeZone": "Europe/Copenhagen"},
+				{"id": "fam", "summary": "Familien", "accessRole": "reader",
+					"backgroundColor": "#0b8043"},
+			}})
+		case strings.HasPrefix(r.URL.Path, "/calendars/"):
+			id := strings.TrimSuffix(strings.TrimPrefix(r.URL.Path, "/calendars/"), "/events")
+			items := g.events[id]
+			if items == nil {
+				items = []map[string]any{}
+			}
+			_ = json.NewEncoder(w).Encode(map[string]any{"items": items})
+		default:
+			http.Error(w, "unexpected "+r.URL.Path, http.StatusNotFound)
+		}
+	}))
+	t.Cleanup(g.Close)
+	return g
+}
+
+// connectedCalendar returns a server with a calendar account already connected —
+// the state the OAuth callback leaves behind, without the consent screen nothing
+// can automate.
+func connectedCalendar(t *testing.T, g *fakeCalendarGoogle, tune func(*config.Config)) (*testServer, *store.User) {
+	t.Helper()
+	ts := newTestServerWith(t, func(cfg *config.Config) {
+		cfg.GmailClientID = "test-client"
+		cfg.GmailClientSecret = "test-secret"
+		cfg.GoogleTokenURL = g.URL + "/token"
+		cfg.CalendarAPIURL = g.URL
+		if tune != nil {
+			tune(cfg)
+		}
+	})
+	ts.bootstrap(t)
+
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	account := &store.CalendarAccount{
+		UserID: user.ID, Provider: "google", RefreshToken: "1//0-refresh",
+	}
+	if err := ts.db.SaveCalendarAccount(t.Context(), account); err != nil {
+		t.Fatal(err)
+	}
+	return ts, user
+}
+
+// The scope on the consent screen is what somebody agrees to. It has to be the
+// read-only one, and the redirect has to be the calendar's own — sharing Gmail's
+// would put "which token store do I write" behind a lookup.
+func TestAuthorizeAsksForReadOnlyCalendar(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	ts, _ := connectedCalendar(t, g, nil)
+
+	resp, body := ts.do(t, "POST", "/api/v1/calendar/authorize", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("status %d, body %v", resp.StatusCode, body)
+	}
+	url, _ := body["url"].(string)
+	for _, want := range []string{
+		"scope=" + "https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fcalendar.readonly",
+		"%2Foauth%2Fcalendar%2Fcallback",
+		"access_type=offline",
+		"prompt=consent",
+	} {
+		if !strings.Contains(url, want) {
+			t.Errorf("the authorisation URL is missing %q:\n%s", want, url)
+		}
+	}
+	if strings.Contains(url, "calendar.events&") || strings.Contains(url, "auth%2Fcalendar&") {
+		t.Errorf("a writing scope was asked for:\n%s", url)
+	}
+}
+
+// Without a registered client there is nothing to connect through, and the answer
+// has to say that rather than "that already exists".
+func TestAuthorizeWithoutAClientSaysSo(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	resp, body := ts.do(t, "POST", "/api/v1/calendar/authorize", nil)
+	if resp.StatusCode != http.StatusConflict {
+		t.Fatalf("status %d, want 409", resp.StatusCode)
+	}
+	if body["code"] != CodeGmailNotConfigured {
+		t.Errorf("code = %v, want %q", body["code"], CodeGmailNotConfigured)
+	}
+}
+
+// The whole path, minus the consent screen: sync fetches the list, refreshes the
+// events of what was chosen, and the grid endpoint answers with them.
+func TestSyncFetchesTheChosenCalendarsAndTheGridReadsThem(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	g.events["fam"] = []map[string]any{
+		{"id": "e1", "summary": "Bestyrelsesmøde", "htmlLink": "https://calendar.google.com/e1",
+			"start": map[string]string{"dateTime": "2026-08-20T14:00:00+02:00"},
+			"end":   map[string]string{"dateTime": "2026-08-20T15:30:00+02:00"}},
+	}
+	g.events["kw@nolimit.dk"] = []map[string]any{
+		{"id": "e2", "summary": "Skal ikke med",
+			"start": map[string]string{"date": "2026-08-21"},
+			"end":   map[string]string{"date": "2026-08-22"}},
+	}
+	ts, user := connectedCalendar(t, g, nil)
+
+	// The list arrives on the first look at the settings page, which is what
+	// somebody would be doing at this point.
+	resp, body := ts.do(t, "GET", "/api/v1/calendar", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("status %d, body %v", resp.StatusCode, body)
+	}
+	if body["connected"] != true {
+		t.Fatalf("not reported as connected: %v", body)
+	}
+	if body["read_only"] != true {
+		t.Error("the connection did not say it is read-only; an interface that looks " +
+			"editable is one somebody will try to edit")
+	}
+	// The primary calendar's id is the address, which is how the account names
+	// itself without asking for a profile scope as well.
+	if body["account"] != "kw@nolimit.dk" {
+		t.Errorf("account = %v", body["account"])
+	}
+
+	list, _ := body["calendars"].([]any)
+	if len(list) != 2 {
+		t.Fatalf("got %d calendars, want 2: %v", len(list), body["calendars"])
+	}
+	var famID string
+	for _, raw := range list {
+		c := raw.(map[string]any)
+		if c["shown"] == true {
+			t.Errorf("%v was shown without anybody choosing it", c["name"])
+		}
+		if c["remote_id"] == "fam" {
+			famID = c["id"].(string)
+		}
+	}
+
+	// Nothing is chosen, so a sync writes nothing: an account with a dozen shared
+	// calendars must not fill somebody's grid the moment it is connected.
+	resp, body = ts.do(t, "POST", "/api/v1/calendar/sync", nil)
+	if resp.StatusCode != http.StatusOK || body["events"] != float64(0) {
+		t.Fatalf("an unchosen calendar was fetched: %d %v", resp.StatusCode, body)
+	}
+
+	resp, _ = ts.do(t, "PUT", "/api/v1/calendar/calendars",
+		map[string]any{"shown": []string{famID}})
+	if resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("choosing a calendar: status %d", resp.StatusCode)
+	}
+
+	resp, body = ts.do(t, "POST", "/api/v1/calendar/sync", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("sync: status %d, body %v", resp.StatusCode, body)
+	}
+	if body["events"] != float64(1) {
+		t.Fatalf("events = %v, want 1 — only the chosen calendar", body["events"])
+	}
+
+	resp, body = ts.do(t, "GET",
+		"/api/v1/calendar/events?from=2026-08-01&to=2026-08-31", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("events: status %d, body %v", resp.StatusCode, body)
+	}
+	events, _ := body["events"].([]any)
+	if len(events) != 1 {
+		t.Fatalf("got %d events, want 1: %v", len(events), body["events"])
+	}
+	event := events[0].(map[string]any)
+	if event["summary"] != "Bestyrelsesmøde" || event["start_day"] != "2026-08-20" {
+		t.Errorf("the event came back as %v", event)
+	}
+	// The calendar's colour rides along, or the chip has to ask for it per row.
+	if event["colour"] != "#0b8043" {
+		t.Errorf("colour = %v; two calendars in one grid are told apart by it", event["colour"])
+	}
+	// The window verdande holds is stated, so a grid paged past it can say so
+	// rather than showing an empty month it cannot explain.
+	if body["from"] == "" || body["to"] == "" {
+		t.Error("the answer does not say which window it covers")
+	}
+
+	// And it is one person's calendar. Somebody else on the same instance sees
+	// nothing of it.
+	other := ts.newUser(t, "andreas@example.dk", "Andreas")
+	_, body = other.do(t, "GET", "/api/v1/calendar/events?from=2026-08-01&to=2026-08-31", nil)
+	if events, _ := body["events"].([]any); len(events) != 0 {
+		t.Errorf("somebody else's calendar was visible: %v", body["events"])
+	}
+	_ = user
+}
+
+// A token that has not expired must not be traded in for another. Refreshing on
+// every run burns the quota and makes the failure mode of a bad refresh token
+// arrive on every sweep rather than once an hour.
+func TestAFreshTokenIsNotRefreshed(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	ts, user := connectedCalendar(t, g, nil)
+
+	account, err := ts.db.CalendarAccountFor(t.Context(), user.ID, "google")
+	if err != nil {
+		t.Fatal(err)
+	}
+	account.AccessToken = "ya29.still-good"
+	account.ExpiresAt = time.Now().Add(time.Hour)
+	if err := ts.db.SaveCalendarAccount(t.Context(), account); err != nil {
+		t.Fatal(err)
+	}
+
+	if resp, body := ts.do(t, "POST", "/api/v1/calendar/sync", nil); resp.StatusCode != http.StatusOK {
+		t.Fatalf("sync: status %d, body %v", resp.StatusCode, body)
+	}
+	if g.refreshes != 0 {
+		t.Errorf("the token was refreshed %d times while it was still good", g.refreshes)
+	}
+
+	// And an expired one is.
+	account.ExpiresAt = time.Now().Add(-time.Minute)
+	if err := ts.db.SaveCalendarAccount(t.Context(), account); err != nil {
+		t.Fatal(err)
+	}
+	if resp, body := ts.do(t, "POST", "/api/v1/calendar/sync", nil); resp.StatusCode != http.StatusOK {
+		t.Fatalf("sync: status %d, body %v", resp.StatusCode, body)
+	}
+	if g.refreshes != 1 {
+		t.Errorf("an expired token was refreshed %d times, want 1", g.refreshes)
+	}
+}
+
+// Access revoked at Google's end is a connection that has ended, not a failure to
+// retry. Polling it every fifteen minutes for ever is a week of log lines and a
+// connection the person believes they removed.
+func TestRevokedAccessDisconnects(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	ts, user := connectedCalendar(t, g, nil)
+
+	account, _ := ts.db.CalendarAccountFor(t.Context(), user.ID, "google")
+	account.AccessToken = "ya29.revoked"
+	account.ExpiresAt = time.Now().Add(time.Hour)
+	if err := ts.db.SaveCalendarAccount(t.Context(), account); err != nil {
+		t.Fatal(err)
+	}
+	g.refuse = true
+
+	resp, body := ts.do(t, "POST", "/api/v1/calendar/sync", nil)
+	if resp.StatusCode != StatusUpstreamRefused {
+		t.Fatalf("status %d, want %d — Google saying no is not this server breaking",
+			resp.StatusCode, StatusUpstreamRefused)
+	}
+	if body["code"] != CodeCalendarFailed {
+		t.Errorf("code = %v, want %q — `internal_error` is shown as a generic sentence, "+
+			"which throws the diagnosis away", body["code"], CodeCalendarFailed)
+	}
+
+	gone, err := ts.db.CalendarAccountFor(t.Context(), user.ID, "google")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if gone != nil {
+		t.Error("the connection is still there after Google refused the token")
+	}
+
+	// And the failure is in the error log, which is the point of having one: a
+	// sync that has been failing for a week is invisible otherwise.
+	_, logged := ts.do(t, "GET", "/api/v1/errors", nil)
+	rows, _ := logged["errors"].([]any)
+	if len(rows) == 0 {
+		t.Fatal("nothing was recorded")
+	}
+	if row := rows[0].(map[string]any); row["what"] != "sync calendar" {
+		t.Errorf("what = %v, want \"sync calendar\"", row["what"])
+	}
+}
+
+// Disconnecting takes the copy with it. None of it is the person's own work, and
+// a calendar left behind after the account is gone is a frozen lie about a day.
+func TestDisconnectingRemovesTheEvents(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	g.events["fam"] = []map[string]any{
+		{"id": "e1", "summary": "Møde",
+			"start": map[string]string{"date": "2026-08-20"},
+			"end":   map[string]string{"date": "2026-08-21"}},
+	}
+	ts, user := connectedCalendar(t, g, nil)
+
+	ts.do(t, "GET", "/api/v1/calendar", nil)
+	account, _ := ts.db.CalendarAccountFor(t.Context(), user.ID, "google")
+	calendars, _ := ts.db.Calendars(t.Context(), account.ID)
+	var famID string
+	for _, c := range calendars {
+		if c.RemoteID == "fam" {
+			famID = c.ID
+		}
+	}
+	ts.do(t, "PUT", "/api/v1/calendar/calendars", map[string]any{"shown": []string{famID}})
+	ts.do(t, "POST", "/api/v1/calendar/sync", nil)
+
+	if resp, _ := ts.do(t, "DELETE", "/api/v1/calendar", nil); resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("disconnect: status %d", resp.StatusCode)
+	}
+
+	_, body := ts.do(t, "GET", "/api/v1/calendar/events?from=2026-08-01&to=2026-08-31", nil)
+	if events, _ := body["events"].([]any); len(events) != 0 {
+		t.Errorf("events survived the disconnect: %v", body["events"])
+	}
+	_, body = ts.do(t, "GET", "/api/v1/calendar", nil)
+	if body["connected"] != false {
+		t.Error("still reported as connected")
+	}
+}
+
+// A range that is not a date has to be refused, not guessed at. An empty `from`
+// compared as a string matches everything before it, which is a grid quietly
+// answering a question nobody asked.
+func TestEventsRefuseARangeThatIsNotADate(t *testing.T) {
+	g := newFakeCalendarGoogle(t)
+	ts, _ := connectedCalendar(t, g, nil)
+
+	for _, query := range []string{
+		"", "?from=2026-08-01", "?from=&to=2026-08-31",
+		"?from=i+morgen&to=2026-08-31", "?from=2026-08-01&to=2026-13-40",
+	} {
+		resp, body := ts.do(t, "GET", "/api/v1/calendar/events"+query, nil)
+		if resp.StatusCode != http.StatusUnprocessableEntity {
+			t.Errorf("%q: status %d, want 422 (%v)", query, resp.StatusCode, body)
+		}
+	}
+}
+
+// A person with no calendar connected still opens the view. It has to answer, and
+// answer emptily, rather than fail — the grid is also verdande's own tasks, and
+// they must not disappear because Google is not set up.
+func TestTheGridAnswersWithNoCalendarConnected(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	resp, body := ts.do(t, "GET", "/api/v1/calendar/events?from=2026-08-01&to=2026-08-31", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("status %d, body %v", resp.StatusCode, body)
+	}
+	if events, ok := body["events"].([]any); !ok || len(events) != 0 {
+		t.Errorf("events = %v, want an empty list rather than null", body["events"])
+	}
+
+	resp, body = ts.do(t, "GET", "/api/v1/calendar", nil)
+	if resp.StatusCode != http.StatusOK || body["connected"] != false {
+		t.Fatalf("status %d, body %v", resp.StatusCode, body)
+	}
+	if body["has_client"] != false {
+		t.Error("an instance with no OAuth registration claimed to have one; the page " +
+			"would offer a button that answers 409")
+	}
+	// Spelled out even when nothing is connected: it is what has to be registered
+	// with Google, and it is the single most likely thing to be wrong.
+	if uri, _ := body["redirect_uri"].(string); !strings.HasSuffix(uri, "/oauth/calendar/callback") {
+		t.Errorf("redirect_uri = %q", uri)
+	}
+}
+
+// A slow Google must not hold the request open until something in front of this
+// server gives up on it and answers with its own HTML.
+func TestASlowCalendarDoesNotHangTheRequest(t *testing.T) {
+	blocked := make(chan struct{})
+	defer close(blocked)
+	slow := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		select {
+		case <-blocked:
+		case <-r.Context().Done():
+		}
+	}))
+	defer slow.Close()
+
+	ts := newTestServerWith(t, func(cfg *config.Config) {
+		cfg.GmailClientID = "test-client"
+		cfg.GmailClientSecret = "test-secret"
+		cfg.GoogleTokenURL = slow.URL + "/token"
+		cfg.CalendarAPIURL = slow.URL
+		// A second rather than the real twenty-five: what is being tested is that a
+		// budget is applied at all, and spending the production one on every run
+		// would put half a minute into the suite to learn the same thing.
+		cfg.CalendarSyncBudget = time.Second
+	})
+	ts.bootstrap(t)
+
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := ts.db.SaveCalendarAccount(t.Context(), &store.CalendarAccount{
+		UserID: user.ID, Provider: "google", RefreshToken: "r",
+		AccessToken: "a", ExpiresAt: time.Now().Add(time.Hour),
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	// Its own client, because the shared one gives up after ten seconds — which is
+	// shorter than the budget being tested and would prove nothing either way.
+	req, _ := http.NewRequest("POST", ts.URL+"/api/v1/calendar/sync", nil)
+	req.Header.Set("Sec-Fetch-Site", "same-origin")
+	for _, c := range ts.client.Jar.Cookies(mustParse(t, ts.URL)) {
+		req.AddCookie(c)
+	}
+
+	started := time.Now()
+	resp, err := (&http.Client{Timeout: 90 * time.Second}).Do(req)
+	if err != nil {
+		t.Fatalf("the request never came back (%v after %s); the sync has no budget of its "+
+			"own, and whatever sits in front of this server answers for it",
+			err, time.Since(started).Round(time.Second))
+	}
+	defer resp.Body.Close()
+
+	if took := time.Since(started); took > 10*time.Second {
+		t.Errorf("the sync took %s; the budget is not being applied", took.Round(time.Second))
+	}
+	if resp.StatusCode >= 500 {
+		t.Errorf("status %d — a slow Google is not this server breaking", resp.StatusCode)
+	}
+}

internal/httpapi/comments_handlers.go

@@ -225,17 +225,32 @@ func (s *Server) handleUploadAttachment(w http.ResponseWriter, r *http.Request)
 
 	a := &store.Attachment{
 		TaskID:     taskID,
-		CommentID:  r.FormValue("comment_id"),
 		Filename:   safeFilename(header.Filename),
 		MimeType:   detectMime(header.Filename, header.Header.Get("Content-Type")),
 		Size:       size,
 		Path:       stored,
 		UploadedBy: user.ID,
 	}
+
 	// An attachment hangs on a task or on a comment, never both — the schema says
 	// so, and the form may have supplied a comment id.
-	if a.CommentID != "" {
-		a.TaskID = ""
+	//
+	// The comment is resolved and checked against the task in the URL. It was
+	// taken on trust, which meant the permission check above guarded the *task
+	// somebody named* while the file landed on *whatever comment they asked for*.
+	// A viewer on a shared project can list its comments, so she could read an id
+	// there and then post to a task in her own project with that id attached: the
+	// file appears under somebody else's comment, in a project she may only read,
+	// and is served to everyone in it from this origin under a name she chose.
+	// Somebody removed from a project could do the same with ids they kept.
+	if id := r.FormValue("comment_id"); id != "" {
+		owner, err := s.db.CommentTask(r.Context(), id)
+		if err != nil || owner != taskID {
+			os.Remove(filepath.Join(s.cfg.FilesDir(), stored))
+			writeError(w, http.StatusNotFound, CodeNotFound, "not found")
+			return
+		}
+		a.CommentID, a.TaskID = id, ""
 	}
 
 	if err := s.db.CreateAttachment(r.Context(), a); err != nil {

internal/httpapi/csp_header_test.go

@@ -0,0 +1,66 @@
+package httpapi
+
+import (
+	"strings"
+	"testing"
+)
+
+// Politikken skal stå i svaret, og den skal stadig være stram.
+//
+// Den er dét, der gjorde en XSS i noteeditoren til indsat HTML frem for kørt
+// script, dengang en `alt` slap uden om sin undslupning — og der var ingen prøve
+// på, at headeren overhovedet blev sendt. En sikkerhedsforanstaltning, ingen
+// prøve rører, er en, der kan forsvinde i en oprydning uden at nogen opdager det.
+func TestEveryResponseCarriesAStrictCSP(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	resp, _ := ts.do(t, "GET", "/api/v1/auth/me", nil)
+	policy := resp.Header.Get("Content-Security-Policy")
+	if policy == "" {
+		t.Fatal("der er ingen Content-Security-Policy på svaret")
+	}
+
+	// Målt på script-src alene, og ikke på hele politikken.
+	//
+	// `style-src 'unsafe-inline'` står der med vilje og er forklaret i csp.go:
+	// Svelte skriver stilarter ind i dokumentet, og en indsat stil kan gøre en
+	// side grim, ikke køre kode. Det er script-src, der afgør, om indsat HTML
+	// bliver til et kørende script, og det er derfor den, prøven ser på.
+	scriptSrc := ""
+	for _, part := range strings.Split(policy, ";") {
+		if strings.HasPrefix(strings.TrimSpace(part), "script-src") {
+			scriptSrc = part
+		}
+	}
+	if scriptSrc == "" {
+		t.Fatal("der er ingen script-src i politikken")
+	}
+	for _, forbidden := range []string{"'unsafe-inline'", "'unsafe-eval'", "'unsafe-hashes'", "'strict-dynamic'"} {
+		if strings.Contains(scriptSrc, forbidden) {
+			t.Errorf("script-src indeholder %s: %q", forbidden, strings.TrimSpace(scriptSrc))
+		}
+	}
+
+	// Og de fire, der skal være der.
+	for _, want := range []string{"script-src", "object-src 'none'", "base-uri 'none'", "frame-ancestors 'none'"} {
+		if !strings.Contains(policy, want) {
+			t.Errorf("politikken mangler %q", want)
+		}
+	}
+}
+
+// Et bilag må ikke kunne blive læst som et dokument.
+func TestAttachmentsAreNotServedAsDocuments(t *testing.T) {
+	for _, mime := range []string{"image/svg+xml", "text/html", "application/xhtml+xml", "text/xml"} {
+		if inlineImage[mime] {
+			t.Errorf("%s står på listen over det, der vises inline", mime)
+		}
+	}
+	// Og listen skal faktisk kunne noget, ellers virker billeder i noter ikke.
+	for _, mime := range []string{"image/png", "image/jpeg"} {
+		if !inlineImage[mime] {
+			t.Errorf("%s mangler; billeder i noter kan ikke vises", mime)
+		}
+	}
+}

internal/httpapi/errors.go

@@ -63,6 +63,11 @@ const (
 	// diagnosis — a wrong app password, a host that is not listening, a folder that
 	// is not there — and it says it in its own words.
 	CodeMailboxFailed = "mailbox_failed"
+
+	// And for a calendar read from Google. Its own code rather than the Gmail one,
+	// because the sentence the interface wraps it in names where it came from — and
+	// "Gmail said no" about a calendar sends somebody to the wrong panel.
+	CodeCalendarFailed = "calendar_failed"
 )
 
 // errLastAdmin is returned to the caller that already wrote the response, so it

internal/httpapi/gmail_handlers.go

@@ -25,18 +25,16 @@ const defaultGmailSyncBudget = 25 * time.Second
 // to redeploy a container to paste a client id.
 // gmailClient is the API client, pointed wherever the config says Google is.
 func (s *Server) gmailClient(accessToken string) *gmail.Client {
-	return gmail.NewClient(accessToken).At(gmail.Endpoints{
-		Token: s.cfg.GmailTokenURL, API: s.cfg.GmailAPIURL,
-	})
+	return gmail.NewClient(accessToken).At(s.cfg.GmailAPIURL)
 }
 
 func (s *Server) gmailConfig(ctx context.Context) gmail.Config {
-	endpoints := gmail.Endpoints{Token: s.cfg.GmailTokenURL, API: s.cfg.GmailAPIURL}
 	cfg := gmail.Config{
 		ClientID:     s.cfg.GmailClientID,
 		ClientSecret: s.cfg.GmailClientSecret,
 		RedirectURL:  s.cfg.GmailRedirectURL(),
-		Endpoints:    endpoints,
+		Scope:        gmail.Scope,
+		TokenURL:     s.cfg.GoogleTokenURL,
 	}
 	if cfg.ClientID != "" && cfg.ClientSecret != "" {
 		return cfg
@@ -251,6 +249,17 @@ func (s *Server) handleGmailSyncNow(w http.ResponseWriter, r *http.Request) {
 //
 // Returns how many were made. Also used by the background job, which is why it
 // takes a context and a user rather than a request.
+// maxGmailSeen bounds the list of message ids already turned into tasks.
+//
+// Gmail is read-only here — the scope is `gmail.readonly`, deliberately, because
+// asking for less is what makes the consent screen something a person can agree
+// to — so verdande cannot unstar a message the way the IMAP side unflags one. The
+// record has to live on this side, and a record on this side has to be bounded.
+//
+// A count rather than the previous rule of "whatever Gmail returned this run",
+// which was not a bound but a way of forgetting.
+const maxGmailSeen = 2000
+
 func (s *Server) SyncGmail(ctx context.Context, user *store.User) (int, error) {
 	box, err := s.db.MailboxOfKind(ctx, user.ID, "gmail")
 	if err != nil {
@@ -260,7 +269,7 @@ func (s *Server) SyncGmail(ctx context.Context, user *store.User) (int, error) {
 		// The same guard as the mailboxes read over IMAP, for the same reason: a
 		// press of "Fetch now" during the sweep would otherwise make two runs from
 		// one marker.
-		unlock := s.lockMailbox(box.ID)
+		unlock := s.lockSync(box.ID)
 		defer unlock()
 		if fresh, err := s.db.MailboxOfKind(ctx, user.ID, "gmail"); err == nil && fresh != nil {
 			box = fresh
@@ -362,14 +371,35 @@ func (s *Server) SyncGmail(ctx context.Context, user *store.User) (int, error) {
 	}
 
 	if created > 0 {
-		// Only the ids Gmail still returns are kept, so the list cannot grow
-		// without bound as messages fall out of the 30-day window.
-		keep := make([]string, 0, len(ids))
+		// The ids this run saw, on top of the ones already known.
+		//
+		// It used to keep only the ids Gmail returned *this run*, to stop the list
+		// growing without bound. That threw away every id that was not on the page
+		// it happened to fetch — and the query is capped, so on an account with more
+		// starred mail than one page, each run forgot the messages the last one had
+		// remembered. They came back unseen, and became tasks again, and again. That
+		// is the duplicate.
+		//
+		// Bounded by a count instead. Newest first, and a ceiling far above any
+		// plausible page, so an id leaves the list because it is old rather than
+		// because it was not in the last answer.
+		keep := make([]string, 0, len(ids)+len(box.Seen))
+		added := map[string]bool{}
 		for _, id := range ids {
-			if seen[id] {
+			if seen[id] && !added[id] {
 				keep = append(keep, id)
+				added[id] = true
 			}
 		}
+		for _, id := range box.Seen {
+			if !added[id] {
+				keep = append(keep, id)
+				added[id] = true
+			}
+		}
+		if len(keep) > maxGmailSeen {
+			keep = keep[:maxGmailSeen]
+		}
 		box.Seen = keep
 		box.LastSyncAt = time.Now()
 		if err := s.db.SaveMailbox(ctx, box); err != nil {

internal/httpapi/gmail_test.go

@@ -29,7 +29,7 @@ func TestAGmailFailureSaysWhatGmailSaidAndIsRecorded(t *testing.T) {
 	ts := newTestServerWith(t, func(cfg *config.Config) {
 		cfg.GmailClientID = "test-client"
 		cfg.GmailClientSecret = "test-secret"
-		cfg.GmailTokenURL = google.URL + "/token"
+		cfg.GoogleTokenURL = google.URL + "/token"
 		cfg.GmailAPIURL = google.URL
 	})
 	ts.bootstrap(t)
@@ -97,7 +97,7 @@ func TestASlowMailboxDoesNotHangTheRequest(t *testing.T) {
 	ts := newTestServerWith(t, func(cfg *config.Config) {
 		cfg.GmailClientID = "test-client"
 		cfg.GmailClientSecret = "test-secret"
-		cfg.GmailTokenURL = google.URL + "/token"
+		cfg.GoogleTokenURL = google.URL + "/token"
 		cfg.GmailAPIURL = google.URL
 		// A second rather than the real twenty-five: what is being tested is that a
 		// budget is applied at all, and spending the production one on every run

internal/httpapi/integrations_handlers.go

@@ -11,6 +11,7 @@ import (
 	"github.com/kristianwind/verdande/internal/ai"
 	"github.com/kristianwind/verdande/internal/auth"
 	"github.com/kristianwind/verdande/internal/quickadd"
+	"github.com/kristianwind/verdande/internal/safedial"
 	"github.com/kristianwind/verdande/internal/store"
 )
 
@@ -214,6 +215,24 @@ func (s *Server) handleSetAISettings(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
+	// The address is checked where it is typed, as well as when it is dialled.
+	//
+	// safedial refuses a private destination at connect time and is the actual
+	// wall; this is so the person entering it hears about it now rather than
+	// discovering it later as a failed summary. A scheme other than http or https
+	// is refused outright — file:// and gopher:// are not providers.
+	if url := strings.TrimSpace(req.BaseURL); url != "" {
+		if !strings.HasPrefix(url, "http://") && !strings.HasPrefix(url, "https://") {
+			writeFieldErrors(w, map[string]string{"base_url": "must start with http:// or https://"})
+			return
+		}
+		if reason := safedial.CheckURL(url); reason != "" {
+			writeFieldErrors(w, map[string]string{"base_url": reason})
+			return
+		}
+		req.BaseURL = url
+	}
+
 	existing, err := s.aiConfig(r, user.ID)
 	if err != nil {
 		s.internal(w, r, "ai settings", err)

internal/httpapi/integrations_test.go

@@ -2,13 +2,18 @@ package httpapi
 
 import (
 	"encoding/json"
+	"errors"
+	"fmt"
 	"io"
 	"net/http"
+	"net/http/httptest"
 	"net/url"
 	"strings"
 	"testing"
+	"time"
 
 	"github.com/kristianwind/verdande/internal/config"
+	"github.com/kristianwind/verdande/internal/store"
 )
 
 // --- MCP -----------------------------------------------------------------------------
@@ -250,6 +255,116 @@ func (ts *testServer) apiToken(t *testing.T, userID string) string {
 	return token
 }
 
+// The promise of sharing has to hold at every door, not just the API's.
+//
+// TestViewerCannotMutateThroughTheAPI asserts a viewer cannot write, and it
+// asserts it against /api/v1. CalDAV is a second door into the same tasks, and it
+// authorised differently: it checked the caller's standing in the project *named
+// in the URL* and then took the task id from the URL as well. Nothing tied the
+// two together.
+//
+// So a viewer on somebody else's project could name a project of their own —
+// their Inbox, which everybody owns and everybody may edit — and put the other
+// project's task id in the filename. The project check passed on the Inbox, and
+// the write landed on a task in a project they may only read.
+func TestAViewerCannotWriteThroughCalDAV(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+	ts.newUser(t, "viewer@example.dk", "Viewer")
+
+	_, project := ts.do(t, "POST", "/api/v1/projects", map[string]any{"name": "Delt"})
+	projectID := project["id"].(string)
+	_, task := ts.do(t, "POST", "/api/v1/tasks", map[string]any{
+		"content": "en opgave", "project_id": projectID,
+	})
+	taskID := task["id"].(string)
+
+	if resp, body := ts.do(t, "POST", "/api/v1/projects/"+projectID+"/invites", map[string]any{
+		"email": "viewer@example.dk", "role": "viewer",
+	}); resp.StatusCode != http.StatusOK {
+		t.Fatalf("invite: %d %v", resp.StatusCode, body)
+	}
+
+	them, err := ts.db.UserByEmail(t.Context(), "viewer@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	token := ts.apiToken(t, them.ID)
+
+	// Their own Inbox: a project they own and may edit, and the one every account
+	// has. It is what makes the URL check pass.
+	projects, err := ts.db.ListProjects(t.Context(), them.ID, false)
+	if err != nil {
+		t.Fatal(err)
+	}
+	var inbox string
+	for _, p := range projects {
+		if p.IsInbox {
+			inbox = p.ID
+		}
+	}
+	if inbox == "" {
+		t.Fatal("the viewer has no inbox")
+	}
+
+	vtodo := "BEGIN:VCALENDAR\r\nVERSION:2.0\r\nBEGIN:VTODO\r\nUID:" + taskID +
+		"\r\nSUMMARY:kapret\r\nSTATUS:COMPLETED\r\nEND:VTODO\r\nEND:VCALENDAR\r\n"
+
+	resp := ts.dav(t, "PUT", "/caldav/"+them.ID+"/"+inbox+"/"+taskID+".ics",
+		token, them.Email, vtodo)
+	if resp.StatusCode < 400 {
+		t.Errorf("a viewer wrote through CalDAV: status %d", resp.StatusCode)
+	}
+
+	// And the task is as its owner left it. The status alone is not the assertion:
+	// the question is whether anything moved.
+	_, after := ts.do(t, "GET", "/api/v1/tasks/"+taskID, nil)
+	if after["content"] != "en opgave" {
+		t.Errorf("the task was renamed by a viewer: %v", after["content"])
+	}
+	if after["completed"] != false {
+		t.Error("the task was completed by a viewer")
+	}
+}
+
+// A task may only be addressed under the project it is actually in.
+//
+// The same hole seen from the other side: an editor on two projects could reach a
+// task in one of them through the other's URL. Nothing is escalated by it — they
+// may edit both — but it is the check that was missing, and a URL that lies about
+// where a resource lives is how the viewer case got in.
+func TestCalDAVRefusesATaskFromAnotherProject(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	token := ts.apiToken(t, user.ID)
+
+	_, one := ts.do(t, "POST", "/api/v1/projects", map[string]any{"name": "Et"})
+	_, two := ts.do(t, "POST", "/api/v1/projects", map[string]any{"name": "To"})
+	_, task := ts.do(t, "POST", "/api/v1/tasks", map[string]any{
+		"content": "hører til Et", "project_id": one["id"].(string),
+	})
+	taskID := task["id"].(string)
+
+	vtodo := "BEGIN:VCALENDAR\r\nVERSION:2.0\r\nBEGIN:VTODO\r\nUID:" + taskID +
+		"\r\nSUMMARY:flyttet i smug\r\nEND:VTODO\r\nEND:VCALENDAR\r\n"
+
+	resp := ts.dav(t, "PUT", "/caldav/"+user.ID+"/"+two["id"].(string)+"/"+taskID+".ics",
+		token, user.Email, vtodo)
+	if resp.StatusCode < 400 {
+		t.Errorf("a task was written under the wrong project: status %d", resp.StatusCode)
+	}
+
+	_, after := ts.do(t, "GET", "/api/v1/tasks/"+taskID, nil)
+	if after["content"] != "hører til Et" {
+		t.Errorf("the task changed: %v", after["content"])
+	}
+}
+
 func TestCalDAVRequiresCredentials(t *testing.T) {
 	ts := newTestServer(t)
 	ts.bootstrap(t)
@@ -562,6 +677,48 @@ func TestAIKeyIsNeverReturned(t *testing.T) {
 	}
 }
 
+// Connecting something else must not delete the AI provider.
+//
+// Reported from use as "the AI settings are not saved": they were saved, and then
+// taken. `user_settings` had `user_id` as its whole primary key, and the upsert
+// renamed `scope` on conflict — so the `ai`, `gmail` and `calendar` scopes shared
+// one row and each write deleted the other two. Every write returned 204.
+//
+// Through the handlers, because that is where it was seen. The shape is asserted
+// against the store as well: see TestOnePersonCanHaveSettingsForMoreThanOneThing.
+func TestTheAIProviderSurvivesConnectingSomethingElse(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	ts.do(t, "PUT", "/api/v1/ai/settings", map[string]any{
+		"provider": "anthropic", "model": "claude-sonnet-5", "api_key": "sk-hemmelig-nøgle",
+	})
+
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	// What the calendar and the mailbox handlers do when a handshake starts: write
+	// their own scope for the same person.
+	for _, scope := range []string{"calendar", "gmail"} {
+		if err := ts.db.SetUserSettings(t.Context(), user.ID, scope,
+			map[string]any{"pkce_state": "s-" + scope}); err != nil {
+			t.Fatalf("save %s: %v", scope, err)
+		}
+	}
+
+	_, body := ts.do(t, "GET", "/api/v1/ai/settings", nil)
+	if body["provider"] != "anthropic" {
+		t.Errorf("the provider is %v after connecting two other things, want anthropic", body["provider"])
+	}
+	if body["model"] != "claude-sonnet-5" {
+		t.Errorf("the model is %v", body["model"])
+	}
+	if body["has_key"] != true {
+		t.Error("the key was taken with it")
+	}
+}
+
 func TestGmailSettings(t *testing.T) {
 	ts := newTestServer(t)
 	ts.bootstrap(t)
@@ -1316,3 +1473,172 @@ func TestTheMailAddressCanBeRotated(t *testing.T) {
 		t.Error("the second rotation returned the same address")
 	}
 }
+
+// "Der kommer ingen notifikationer" må kunne besvares.
+//
+// Alt om det var usynligt: indstillingssiden sagde "til for denne enhed", så snart
+// der fandtes et abonnement, og en push, tjenesten afviste, gik i en Debug-linje,
+// som ikke skrives på en instans, der kører på standardniveau. Man kunne ikke
+// skelne "browseren tilmeldte sig aldrig" fra "tjenesten afviste den" fra "der er
+// ikke noget forfaldent endnu" — tre forskellige problemer med ét udseende.
+func TestATestNotificationSaysWhatHappenedToIt(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	// Ingen enheder: svaret skal sige nul frem for at lade som om det gik godt.
+	resp, body := ts.do(t, "POST", "/api/v1/push/test", map[string]any{
+		"title": "Prøve", "body": "kom den frem?",
+	})
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("status %d, body %v", resp.StatusCode, body)
+	}
+	if body["subscriptions"] != float64(0) {
+		t.Errorf("subscriptions = %v uden tilmeldte enheder", body["subscriptions"])
+	}
+
+	// En enhed, der peger på en tjeneste, som ikke findes. Den skal komme tilbage
+	// som et navngivent afslag og ikke som tavshed.
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if err := ts.db.SavePushSubscription(t.Context(), &store.PushSubscription{
+		UserID:   user.ID,
+		Endpoint: "https://push.example.invalid/abc",
+		P256dh:   "BEl62iUYgUivxIkv69yViEuiBIa-Ib9-SkvMeAtA3LFgDzkrxZJjSgSnfckjBJuBkr3qBUYIHBQFLXYp5Nksh8U",
+		Auth:     "aUZmY2xDS2FQUXJJVmJIeA",
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	_, body = ts.do(t, "POST", "/api/v1/push/test", map[string]any{"title": "Prøve"})
+	if body["subscriptions"] != float64(1) {
+		t.Fatalf("subscriptions = %v", body["subscriptions"])
+	}
+	failed, _ := body["failed"].([]any)
+	if len(failed) != 1 {
+		t.Fatalf("failed har %d poster, want 1 — afslaget blev slugt", len(failed))
+	}
+	first, _ := failed[0].(map[string]any)
+	if first["service"] != "push.example.invalid" {
+		t.Errorf("service = %v", first["service"])
+	}
+	if first["reason"] == "" || first["reason"] == nil {
+		t.Error("afslaget kom uden en grund")
+	}
+	// Og aldrig hele adressen: resten af den er den legitimation, tjenesten gav
+	// enheden, og den hører hverken i et svar, en log eller et skærmbillede.
+	if strings.Contains(fmt.Sprint(first), "/abc") {
+		t.Errorf("endepunktet lækkede med ud: %v", first)
+	}
+}
+
+// Et abonnement på en kalender, man kun har en adresse til.
+//
+// Bedt om som "abonnering af flere kalendere (ics og webcal?)". Google-vejen er
+// OAuth og én konto pr. person; det her er en adresse, og man har flere. Det er
+// også den eneste vej for en privat konto, som en Internal-registrering afviser
+// med `org_internal`.
+//
+// Prøven går uden om handleren og ind i det, den kalder. Grunden er væggen:
+// `httptest` lytter på loopback, og safedial nægter — både når adressen skrives
+// ned og igen når den hentes. Det er rigtigt, og det skal ikke kunne slås fra fra
+// en prøve. Så adressereglerne prøves for sig i TestWebcalIsRewrittenToHTTPS, og
+// her prøves resten af vejen: hentning, format, navn, gitteret og afmeldingen.
+func TestSubscribingToACalendarByAddress(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	day := time.Now().AddDate(0, 0, 2).Format("20060102")
+	feedBody := "BEGIN:VCALENDAR\r\nVERSION:2.0\r\nX-WR-CALNAME:Helligdage\r\n" +
+		"BEGIN:VEVENT\r\nUID:en@example.dk\r\n" +
+		"DTSTART;VALUE=DATE:" + day + "\r\nDTEND;VALUE=DATE:" + day + "\r\n" +
+		"SUMMARY:Grundlovsdag\r\nEND:VEVENT\r\nEND:VCALENDAR\r\n"
+
+	var hits int
+	feed := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		hits++
+		w.Header().Set("Content-Type", "text/calendar")
+		w.Write([]byte(feedBody))
+	}))
+	defer feed.Close()
+	ts.api.icsFetch = &http.Client{Timeout: 10 * time.Second}
+
+	user, err := ts.db.UserByEmail(t.Context(), "kristian@example.dk")
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Tilføjet under sin adresse, som er alt der er at gå efter, før den er hentet.
+	if _, err := ts.db.SubscribeCalendar(t.Context(), user.ID, feed.URL, feed.URL); err != nil {
+		t.Fatal(err)
+	}
+	if _, err := ts.api.syncSubscriptions(t.Context(), user); err != nil {
+		t.Fatal(err)
+	}
+	if hits == 0 {
+		t.Fatal("adressen blev aldrig hentet")
+	}
+
+	// Navnet kommer fra filen. X-WR-CALNAME er det, alle udgivere bruger; der
+	// findes ingen standardegenskab for det.
+	_, status := ts.do(t, "GET", "/api/v1/calendar", nil)
+	subs, _ := status["subscriptions"].([]any)
+	if len(subs) != 1 {
+		t.Fatalf("%d abonnementer i svaret", len(subs))
+	}
+	first, _ := subs[0].(map[string]any)
+	if first["account"] != "Helligdage" {
+		t.Errorf("navnet er %v, want Helligdage", first["account"])
+	}
+
+	// Og begivenheden står i gitteret.
+	from := time.Now().AddDate(0, 0, -1).Format("2006-01-02")
+	to := time.Now().AddDate(0, 0, 7).Format("2006-01-02")
+	_, events := ts.do(t, "GET", "/api/v1/calendar/events?from="+from+"&to="+to, nil)
+	list, _ := events["events"].([]any)
+	if len(list) != 1 {
+		t.Fatalf("%d begivenheder efter abonnementet", len(list))
+	}
+	if e, _ := list[0].(map[string]any); e["summary"] != "Grundlovsdag" {
+		t.Errorf("begivenheden er %v", e["summary"])
+	}
+
+	// Den samme adresse to gange er den samme kalender hentet to gange, og der er
+	// ingen måde at sige hvilken der gælder.
+	if _, err := ts.db.SubscribeCalendar(t.Context(), user.ID, feed.URL, ""); !errors.Is(err, store.ErrDuplicate) {
+		t.Errorf("den samme adresse blev accepteret to gange: %v", err)
+	}
+
+	// Fjernet igen, og begivenhederne går med.
+	resp, _ := ts.do(t, "DELETE", "/api/v1/calendar/subscriptions/"+first["id"].(string), nil)
+	if resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("unsubscribe: status %d", resp.StatusCode)
+	}
+	_, events = ts.do(t, "GET", "/api/v1/calendar/events?from="+from+"&to="+to, nil)
+	if list, _ := events["events"].([]any); len(list) != 0 {
+		t.Errorf("%d begivenheder blev tilbage efter afmeldingen", len(list))
+	}
+}
+
+// webcal:// er https med et andet skema foran. Det er ikke en protokol, noget
+// taler — det er en konvention, der får et skrivebord til at give adressen til
+// kalenderprogrammet i stedet for til browseren, og alle udgivere skriver deres
+// links sådan. At afvise den ville være at afvise folk på en teknikalitet.
+func TestWebcalIsRewrittenToHTTPS(t *testing.T) {
+	got, err := normaliseCalendarURL("webcal://example.dk/kalender.ics")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if got != "https://example.dk/kalender.ics" {
+		t.Errorf("blev til %q", got)
+	}
+
+	// Og en adresse indad afvises, hvor den skrives — ikke først når den hentes.
+	if _, err := normaliseCalendarURL("http://127.0.0.1:8080/x.ics"); err == nil {
+		t.Error("en loopback-adresse blev godtaget")
+	}
+	if _, err := normaliseCalendarURL("file:///etc/passwd"); err == nil {
+		t.Error("et skema, der ikke er http, blev godtaget")
+	}
+}

internal/httpapi/mailbox_handlers.go

@@ -143,7 +143,7 @@ func (s *Server) SyncMailbox(ctx context.Context, user *store.User, m *store.Mai
 	// One reader of this mailbox at a time. Held for the whole run, marker and all:
 	// releasing before the marker is written would leave exactly the gap this is
 	// here to close.
-	unlock := s.lockMailbox(m.ID)
+	unlock := s.lockSync(m.ID)
 	defer unlock()
 
 	// Re-read under the lock. The row that came in may have been fetched before the
@@ -175,6 +175,10 @@ func (s *Server) SyncMailbox(ctx context.Context, user *store.User, m *store.Mai
 
 	created := 0
 	highest := m.LastUID
+	// The uids of the mails that actually became tasks. Only those get unflagged,
+	// and only after the task is committed: a flag removed from a mail whose task
+	// was never written is a mail nobody will ever see again.
+	done := make([]uint32, 0, len(messages))
 	for _, msg := range messages {
 		if ctx.Err() != nil {
 			// What was made so far is already committed and counted, and the marker
@@ -204,12 +208,27 @@ func (s *Server) SyncMailbox(ctx context.Context, user *store.User, m *store.Mai
 			continue
 		}
 		created++
+		done = append(done, msg.UID)
 		if msg.UID > highest {
 			highest = msg.UID
 		}
 		s.hub.Publish(projectID, "task.created", toTaskJSON(*task))
 	}
 
+	// The flag comes off what became a task. This is what actually stops a mail
+	// arriving twice; the marker below is now a second line rather than the only
+	// one. See imap.Unflag for why the record belongs in the mailbox and not here.
+	//
+	// Outside the request's context, like the marker: a cancelled request that
+	// already created tasks must still take the flags off them, or the next run
+	// makes them all again.
+	if err := client.Unflag(done); err != nil {
+		// Not fatal, and not the caller's problem. A folder somebody opened
+		// read-only, or a server that refuses the STORE, still gets the uid marker
+		// underneath — worse, but not broken.
+		s.log.Warn("could not unflag imported mail", "err", err, "mailbox", m.ID, "count", len(done))
+	}
+
 	// A run that got all the way through moves the marker to the whole of what the
 	// search considered, not merely to the last message it happened to read. That
 	// is what stops the same mail becoming a task again on the next run when the
@@ -231,8 +250,14 @@ func (s *Server) SyncMailbox(ctx context.Context, user *store.User, m *store.Mai
 	return created, nil
 }
 
-// lockMailbox serialises one mailbox against itself and returns the release.
-func (s *Server) lockMailbox(id string) func() {
+// lockSync serialises one thing that polls against itself and returns the release.
+//
+// A mailbox to begin with; a calendar connection now takes the same lock under a
+// key of its own. Two runs at once — somebody pressing "fetch now" while the sweep
+// is in the middle of the same thing — both start from the state as it was, and
+// the loser writes last. For a mailbox that made forty copies of one mail in a
+// second; for a calendar it would be a window deleted after it was refilled.
+func (s *Server) lockSync(id string) func() {
 	value, _ := s.syncing.LoadOrStore(id, &sync.Mutex{})
 	mu := value.(*sync.Mutex)
 	mu.Lock()

internal/httpapi/notes_handlers.go

@@ -6,7 +6,9 @@ import (
 	"io"
 	"mime"
 	"net/http"
+	"os"
 	"path"
+	"path/filepath"
 	"sort"
 	"strings"
 	"time"
@@ -46,7 +48,7 @@ func (s *Server) handleListNotes(w http.ResponseWriter, r *http.Request) {
 			s.internal(w, r, "search notes", err)
 			return
 		}
-		writeJSON(w, http.StatusOK, map[string]any{"notes": s.readable(r, found)})
+		writeJSON(w, http.StatusOK, map[string]any{"notes": briefly(s.readable(r, found))})
 		return
 	}
 
@@ -60,7 +62,20 @@ func (s *Server) handleListNotes(w http.ResponseWriter, r *http.Request) {
 			s.internal(w, r, "notes in project", err)
 			return
 		}
-		writeJSON(w, http.StatusOK, map[string]any{"notes": found})
+		writeJSON(w, http.StatusOK, map[string]any{"notes": briefly(found)})
+		return
+	}
+
+	// The archive is the same list asked the other way round, so it is the same
+	// endpoint with a flag rather than a second one that would have to be kept in
+	// step with this.
+	if r.URL.Query().Get("archived") != "" {
+		found, err := s.db.ArchivedNotes(r.Context(), user.ID)
+		if err != nil {
+			s.internal(w, r, "list archived notes", err)
+			return
+		}
+		writeJSON(w, http.StatusOK, map[string]any{"notes": briefly(found)})
 		return
 	}
 
@@ -72,7 +87,41 @@ func (s *Server) handleListNotes(w http.ResponseWriter, r *http.Request) {
 		s.internal(w, r, "list notes", err)
 		return
 	}
-	writeJSON(w, http.StatusOK, map[string]any{"notes": found})
+	writeJSON(w, http.StatusOK, map[string]any{"notes": briefly(found)})
+}
+
+// briefly replaces each note's body with the first of it.
+//
+// The list shows a title and about ninety characters. It was being sent whole
+// notes to do that: with twelve hundred of them the response was 845 KB, of which
+// ninety-eight per cent was body text nobody would read on that screen — fetched,
+// parsed and held in memory on every visit.
+//
+// Body is emptied rather than shortened, and that is the important half. A field
+// called `body` that holds a *piece* of the body is a trap: the editor would open
+// it, the debounce would save it, and the note would be cut down to its own
+// preview by nothing more than being looked at. Empty is a shape the client can
+// check, and it does — see the notes page, which fetches the whole note when it
+// opens one.
+//
+// Four hundred characters rather than ninety: the client strips Markdown before
+// it shows anything, and a heading marker and a bold word are characters that
+// vanish. Cheap insurance against a preview that ends up shorter than the line it
+// is meant to fill.
+const previewChars = 400
+
+func briefly(notes []store.Note) []store.Note {
+	out := make([]store.Note, 0, len(notes))
+	for _, n := range notes {
+		body := []rune(n.Body)
+		if len(body) > previewChars {
+			body = body[:previewChars]
+		}
+		n.Preview = string(body)
+		n.Body = ""
+		out = append(out, n)
+	}
+	return out
 }
 
 // readable drops what this person may not see.
@@ -292,6 +341,14 @@ func noteFilename(title string) string {
 // a few megabytes — and small enough that a wrong file is refused rather than read.
 const maxNoteImportBytes = 64 << 20
 
+// What the archive may become once it is unpacked, and how many entries it may
+// hold. See handleImportNotes: the upload cap bounds what arrives, and says
+// nothing at all about what it expands to.
+const (
+	maxNoteImportUnpacked = 2 << 30
+	maxNoteImportFiles    = 20_000
+)
+
 // handleImportNotes reads a zip of Markdown files and makes a note of each.
 //
 // The mirror of the export, and deliberately the same shape: whatever comes out of
@@ -330,24 +387,57 @@ func (s *Server) handleImportNotes(w http.ResponseWriter, r *http.Request) {
 	// note is read its pictures already have addresses to be rewritten to — a note
 	// and its images arrive in whatever order the archive happens to hold them, and
 	// one pass would leave half the links pointing at nothing.
+	// The archive as a whole, not just the upload.
+	//
+	// The 64 MB cap is on what arrives; deflate reaches about 1032:1, so those 64
+	// MB can become sixty-six gigabytes on the way out — written into the same
+	// volume the database lives on. The upload limit says nothing about that, and
+	// nothing else did either.
+	//
+	// Two ceilings: one per file, and one across the archive. A single enormous
+	// photograph and ten thousand small ones are different attacks with the same
+	// ending.
+	if len(zr.File) > maxNoteImportFiles {
+		writeError(w, http.StatusRequestEntityTooLarge, CodePayloadTooLarge,
+			"that archive holds too many files")
+		return
+	}
+	budget := int64(maxNoteImportUnpacked)
+
 	stored := map[string]string{}
 	for _, f := range zr.File {
 		name := f.Name
 		if f.FileInfo().IsDir() || skipFile(name) || strings.EqualFold(path.Ext(name), ".md") {
 			continue
 		}
+		// The header's own claim first, which costs nothing to check and rejects
+		// the obvious case before a byte is written. It is only a claim, so the
+		// reader below is limited as well.
+		if f.UncompressedSize64 > maxUploadBytes {
+			s.log.Warn("import: a file in the archive is too large", "file", name)
+			skipped++
+			continue
+		}
 		rc, err := f.Open()
 		if err != nil {
 			continue
 		}
-		hashed, size, err := s.storeUpload(rc)
+		hashed, size, err := s.storeUpload(io.LimitReader(rc, min(budget, maxUploadBytes)+1))
 		rc.Close()
 		if err != nil {
 			s.log.Warn("import note file", "err", err, "file", name)
 			continue
 		}
+		if size > maxUploadBytes || size > budget {
+			// Over the line once it was actually read: remove what was written
+			// rather than keeping a truncated file that would open as nothing.
+			os.Remove(filepath.Join(s.cfg.FilesDir(), hashed))
+			s.log.Warn("import: archive is larger unpacked than allowed", "file", name)
+			skipped++
+			continue
+		}
+		budget -= size
 		stored[name] = hashed
-		_ = size
 		files++
 	}
 
@@ -579,3 +669,139 @@ func mimeOf(name string) string {
 	}
 	return "application/octet-stream"
 }
+
+// handleUploadNoteAttachment takes a file into a note.
+//
+// The import could already do this, and nothing else could: a note brought in
+// from Apple Notes arrived with its pictures, and a note written here could not
+// be given one. Pasting a screenshot into a note is the most ordinary thing there
+// is, and it did nothing at all.
+//
+// Editor rights on the note, checked the same way every other write to it is. The
+// answer for "no such note" and "not yours" is the same, so an id cannot be
+// probed by watching which one comes back.
+func (s *Server) handleUploadNoteAttachment(w http.ResponseWriter, r *http.Request) {
+	n, err := s.db.Note(r.Context(), chi.URLParam(r, "noteID"))
+	if err != nil {
+		s.internal(w, r, "get note", err)
+		return
+	}
+	if n == nil || !s.mayTouchNote(r, n, store.RoleEditor) {
+		writeError(w, http.StatusNotFound, CodeNotFound, "no such note")
+		return
+	}
+
+	r.Body = http.MaxBytesReader(w, r.Body, maxUploadBytes+(1<<20))
+	if err := r.ParseMultipartForm(8 << 20); err != nil {
+		writeError(w, http.StatusRequestEntityTooLarge, CodePayloadTooLarge,
+			"that file is too large (25 MB maximum)")
+		return
+	}
+	defer r.MultipartForm.RemoveAll()
+
+	file, header, err := r.FormFile("file")
+	if err != nil {
+		writeFieldErrors(w, map[string]string{"file": "required"})
+		return
+	}
+	defer file.Close()
+
+	if header.Size > maxUploadBytes {
+		writeError(w, http.StatusRequestEntityTooLarge, CodePayloadTooLarge,
+			"that file is too large (25 MB maximum)")
+		return
+	}
+
+	stored, size, err := s.storeUpload(file)
+	if err != nil {
+		s.internal(w, r, "store upload", err)
+		return
+	}
+
+	user := userFrom(r.Context())
+	a := &store.Attachment{
+		NoteID:     n.ID,
+		Filename:   safeFilename(header.Filename),
+		MimeType:   detectMime(header.Filename, header.Header.Get("Content-Type")),
+		Size:       size,
+		Path:       stored,
+		UploadedBy: user.ID,
+	}
+	if err := s.db.CreateAttachment(r.Context(), a); err != nil {
+		os.Remove(filepath.Join(s.cfg.FilesDir(), stored))
+		s.internal(w, r, "record attachment", err)
+		return
+	}
+	writeJSON(w, http.StatusCreated, toAttachmentJSON(*a))
+}
+
+// --- putting notes away, one or many ---------------------------------------------
+
+type noteBulkRequest struct {
+	IDs      []string `json:"ids"`
+	Archived *bool    `json:"archived"`
+	Delete   bool     `json:"delete"`
+}
+
+// maxNoteBulk bounds one call. Enough to select a screenful several times over,
+// and few enough that the loop below cannot become somebody's afternoon.
+const maxNoteBulk = 500
+
+// handleNoteBulk archives, unarchives or deletes a set of notes in one request.
+//
+// One call rather than one per note, because the thing being asked is one thing:
+// "put these away". Fifty separate requests would be fifty round trips, fifty
+// chances for half of them to fail, and a list that redraws fifty times — and
+// after an import of twelve hundred notes, selecting fifty is the ordinary case,
+// not the extreme one.
+//
+// Each note is still checked on its own. A set is a convenience for the caller
+// and must never become a way to touch something they could not have touched one
+// at a time.
+func (s *Server) handleNoteBulk(w http.ResponseWriter, r *http.Request) {
+	var req noteBulkRequest
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+	if len(req.IDs) == 0 {
+		writeFieldErrors(w, map[string]string{"ids": "required"})
+		return
+	}
+	if len(req.IDs) > maxNoteBulk {
+		writeFieldErrors(w, map[string]string{"ids": "too many at once"})
+		return
+	}
+
+	done, skipped := 0, 0
+	for _, id := range req.IDs {
+		n, err := s.db.Note(r.Context(), id)
+		if err != nil {
+			s.internal(w, r, "get note", err)
+			return
+		}
+		// Not an error, and not a 404 for the whole call: a set can contain a note
+		// somebody else deleted a second ago, and the other forty-nine should still
+		// be put away. Counted so the answer is honest about it.
+		if n == nil || !s.mayTouchNote(r, n, store.RoleEditor) {
+			skipped++
+			continue
+		}
+
+		switch {
+		case req.Delete:
+			err = s.db.DeleteNote(r.Context(), id)
+		case req.Archived != nil:
+			err = s.db.SetNoteArchived(r.Context(), id, *req.Archived)
+		default:
+			writeFieldErrors(w, map[string]string{"archived": "say archived or delete"})
+			return
+		}
+		if err != nil {
+			s.internal(w, r, "note bulk", err)
+			return
+		}
+		done++
+	}
+
+	writeJSON(w, http.StatusOK, map[string]any{"done": done, "skipped": skipped})
+}

internal/httpapi/notes_import_limits_test.go

@@ -0,0 +1,77 @@
+package httpapi
+
+import (
+	"archive/zip"
+	"bytes"
+	"io"
+	"testing"
+)
+
+// En zip-bombe skal ikke kunne skrive gigabyte ned på den disk, databasen ligger
+// på.
+//
+// Uploaden er begrænset til 64 MB, og det siger ingenting om, hvad de 64 MB
+// bliver til: deflate når omkring 1032:1, så de kan pakkes ud til seksogtres
+// gigabyte. Prøven bygger en rigtig lille bombe — fire megabyte nuller, som
+// pakker til nogle få kilobyte — og ser efter, at loftet på det udpakkede holder.
+//
+// Skrevet imod tallene frem for imod endepunktet, fordi det, der skal holde, er
+// forholdet: en fil, hvis udpakkede størrelse overstiger loftet, må ikke læses
+// færdig.
+func TestAZipBombIsRefusedByItsUnpackedSize(t *testing.T) {
+	var buf bytes.Buffer
+	zw := zip.NewWriter(&buf)
+	w, err := zw.Create("bombe.bin")
+	if err != nil {
+		t.Fatal(err)
+	}
+	// Nuller pakker næsten helt væk. Fire megabyte er nok til at vise forholdet
+	// uden at prøven selv bruger en gigabyte.
+	if _, err := w.Write(make([]byte, 4<<20)); err != nil {
+		t.Fatal(err)
+	}
+	if err := zw.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	packed := int64(buf.Len())
+	if packed > 64<<10 {
+		t.Fatalf("prøvens egen bombe pakkede kun til %d bytes; den viser ikke noget", packed)
+	}
+
+	zr, err := zip.NewReader(bytes.NewReader(buf.Bytes()), packed)
+	if err != nil {
+		t.Fatal(err)
+	}
+	f := zr.File[0]
+	if f.UncompressedSize64 != 4<<20 {
+		t.Fatalf("udpakket størrelse = %d", f.UncompressedSize64)
+	}
+
+	// Det, importen gør: læs højst det, budgettet tillader, plus én byte — så en
+	// fil, der er præcis på grænsen, kan skelnes fra en, der er over.
+	rc, err := f.Open()
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer rc.Close()
+	budget := int64(1 << 20)
+	n, err := io.Copy(io.Discard, io.LimitReader(rc, budget+1))
+	if err != nil {
+		t.Fatal(err)
+	}
+	if n <= budget {
+		t.Fatalf("læste %d bytes af en fil på 4 MB med et budget på %d — loftet fanger den ikke", n, budget)
+	}
+}
+
+// Tallene skal stå i et forhold, der giver mening: en enkelt fil må aldrig kunne
+// bruge hele arkivets budget, og arkivet må ikke være mindre end én fil.
+func TestTheImportCeilingsAgreeWithEachOther(t *testing.T) {
+	if maxNoteImportUnpacked <= maxUploadBytes {
+		t.Error("arkivets budget er ikke større end én enkelt fil")
+	}
+	if maxNoteImportFiles < 2000 {
+		t.Error("filloftet er under det, en rigtig Apple Noter-eksport indeholder")
+	}
+}

internal/httpapi/notes_test.go

@@ -328,12 +328,19 @@ func TestAnImportedNoteKeepsItsPictures(t *testing.T) {
 		t.Fatalf("imported %d notes and %d files, want 2 and 1", out.Created, out.Files)
 	}
 
+	// The list carries a preview, not the whole note — see briefly(). The body is
+	// fetched by id, which is what the interface does when it opens one.
 	_, listed := ts.do(t, "GET", "/api/v1/notes?q=kvittering", nil)
 	notes := listed["notes"].([]any)
 	if len(notes) != 1 {
 		t.Fatalf("found %d notes", len(notes))
 	}
-	body := notes[0].(map[string]any)["body"].(string)
+	noteID := notes[0].(map[string]any)["id"].(string)
+	if got := notes[0].(map[string]any)["body"]; got != nil && got != "" {
+		t.Errorf("the list carried a body: %v — an editor handed that would save the truncation", got)
+	}
+	_, whole := ts.do(t, "GET", "/api/v1/notes/"+noteID, nil)
+	body := whole["body"].(string)
 
 	// The link points at the file where it now lives, not at a path inside an
 	// archive nobody kept.

internal/httpapi/notifications_handlers.go

@@ -1,12 +1,16 @@
 package httpapi
 
 import (
+	"errors"
 	"net/http"
+	"net/url"
+	"strings"
 	"time"
 
 	"github.com/go-chi/chi/v5"
 
 	"github.com/kristianwind/verdande/internal/push"
+	"github.com/kristianwind/verdande/internal/safedial"
 	"github.com/kristianwind/verdande/internal/store"
 )
 
@@ -104,18 +108,44 @@ func (s *Server) PushToUser(userID, title, body, projectID string) {
 	s.pushToUser(userID, title, body, projectID)
 }
 
-func (s *Server) pushToUser(userID, title, body, projectID string) {
+// pushOutcome is what happened, per device.
+//
+// It exists because "nothing arrives" was unanswerable. A push that a service
+// refuses is the ordinary failure here — a stale subscription, a VAPID subject the
+// service will not take, a browser that revoked permission without telling the
+// server — and every one of them looked exactly like silence.
+type pushOutcome struct {
+	Subscriptions int           `json:"subscriptions"`
+	Sent          int           `json:"sent"`
+	Failed        []pushFailure `json:"failed"`
+}
+
+// pushFailure names the service and quotes what it said.
+//
+// The host and not the endpoint: the rest of that URL is the credential the push
+// service issued for this device, and it does not belong in an answer, a log line
+// or a screenshot.
+type pushFailure struct {
+	Service string `json:"service"`
+	Reason  string `json:"reason"`
+}
+
+func (s *Server) pushToUser(userID, title, body, projectID string) pushOutcome {
 	ctx, cancel := contextWithBackgroundTimeout(30 * time.Second)
 	defer cancel()
 
+	var out pushOutcome
 	subs, err := s.db.ListPushSubscriptions(ctx, userID)
 	if err != nil || len(subs) == 0 {
-		return
+		return out
 	}
+	out.Subscriptions = len(subs)
+
 	public, private, err := s.vapidKeys(ctx)
 	if err != nil {
 		s.log.Warn("vapid keys", "err", err)
-		return
+		out.Failed = append(out.Failed, pushFailure{Service: "verdande", Reason: err.Error()})
+		return out
 	}
 
 	payload := push.Payload{Title: title, Body: body, URL: "/projekt/" + projectID}
@@ -126,13 +156,101 @@ func (s *Server) pushToUser(userID, title, body, projectID string) {
 			Public: public, Private: private, Subject: "mailto:" + s.cfg.SMTP.From,
 		})
 		if err == nil {
+			out.Sent++
 			continue
 		}
-		s.log.Debug("web push failed", "err", err, "endpoint", sub.Endpoint)
+		// Warn, not Debug.
+		//
+		// This was the one line that would have said why no notification ever
+		// arrived, and at Debug it is not printed on an instance running at the
+		// default level. A delivery that fails silently and a delivery that does
+		// not happen look the same from the outside, and the difference is the
+		// whole diagnosis.
+		reason := pushReason(sub.Endpoint, err)
+		s.log.Warn("web push refused", "reason", reason, "service", pushHost(sub.Endpoint), "user", userID)
+		out.Failed = append(out.Failed, pushFailure{
+			Service: pushHost(sub.Endpoint),
+			Reason:  reason,
+		})
 		if recErr := s.db.RecordPushFailure(ctx, sub.Endpoint, push.IsGone(err)); recErr != nil {
 			s.log.Warn("record push failure", "err", recErr)
 		}
 	}
+	return out
+}
+
+// pushReason is what went wrong, with the endpoint taken back out of it.
+//
+// Go wraps a transport failure as `*url.Error`, and its message is `Post
+// "https://fcm.googleapis.com/fcm/send/dQw4w9…": dial tcp: …` — the whole
+// endpoint, which is the credential the push service issued for that device. It
+// would have gone into the answer, the log line and any screenshot of either. A
+// test written to prove the host was reported instead is what found it.
+//
+// Unwrapped first, so the cause is what is quoted. Then the endpoint is scrubbed
+// anyway: the next error type to carry it will not come with a warning.
+func pushReason(endpoint string, err error) string {
+	var urlErr *url.Error
+	if errors.As(err, &urlErr) && urlErr.Err != nil {
+		err = urlErr.Err
+	}
+	reason := err.Error()
+	if endpoint != "" {
+		reason = strings.ReplaceAll(reason, endpoint, pushHost(endpoint))
+	}
+	return reason
+}
+
+// pushHost is the service an endpoint belongs to, and nothing else from it.
+func pushHost(endpoint string) string {
+	u, err := url.Parse(endpoint)
+	if err != nil || u.Host == "" {
+		return "ukendt"
+	}
+	return u.Host
+}
+
+// handlePushTest sends one notification to this account's devices and says what
+// happened to it.
+//
+// Reported as "der kommer ikke nogen". Everything about that was invisible: the
+// settings page said "on for this device" as soon as a subscription existed, and a
+// refusal from the push service went into a Debug line nobody sees. There was no
+// way to tell "the browser never subscribed" from "the service refused it" from
+// "nothing has come due yet" — three different problems with one appearance.
+//
+// It goes through the same path as a real notification rather than a shortcut, so
+// what it proves is the thing that actually has to work.
+//
+// The words come from the client, which is where the dictionaries are. A server
+// that carries its own copy of "This is a test" carries it in one language, and
+// this one deliberately keeps its prose in the interface — see
+// `internal/i18n`. The text is only ever shown back to the person who asked for
+// it, on their own devices.
+func (s *Server) handlePushTest(w http.ResponseWriter, r *http.Request) {
+	var req struct {
+		Title string `json:"title"`
+		Body  string `json:"body"`
+	}
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+	title := strings.TrimSpace(req.Title)
+	if title == "" {
+		title = "verdande"
+	}
+	// Bounded: this is written into a notification on somebody's lock screen, and
+	// a megabyte of title is not a notification.
+	if len(title) > 200 {
+		title = title[:200]
+	}
+	body := strings.TrimSpace(req.Body)
+	if len(body) > 500 {
+		body = body[:500]
+	}
+
+	user := userFrom(r.Context())
+	writeJSON(w, http.StatusOK, s.pushToUser(user.ID, title, body, ""))
 }
 
 func (s *Server) vapidKeys(ctx contextLike) (string, string, error) {
@@ -170,6 +288,23 @@ func (s *Server) handleSubscribePush(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
+	// The endpoint is an address this server will POST to, on a schedule the
+	// caller chooses — set a reminder a minute out and the push job fetches it,
+	// again every minute. Unchecked, that is a way to make the instance reach
+	// anything it can route to, on demand.
+	//
+	// https only: a push service is on the public internet and speaks TLS. Nothing
+	// legitimate needs http here, and allowing it is the difference between a
+	// blind request and a readable one.
+	if !strings.HasPrefix(req.Endpoint, "https://") {
+		writeFieldErrors(w, map[string]string{"endpoint": "must be an https address"})
+		return
+	}
+	if reason := safedial.CheckURL(req.Endpoint); reason != "" {
+		writeFieldErrors(w, map[string]string{"endpoint": reason})
+		return
+	}
+
 	err := s.db.SavePushSubscription(r.Context(), &store.PushSubscription{
 		UserID:    userFrom(r.Context()).ID,
 		Endpoint:  req.Endpoint,
@@ -189,7 +324,10 @@ func (s *Server) handleUnsubscribePush(w http.ResponseWriter, r *http.Request) {
 	if err := decodeJSON(w, r, &req); err != nil {
 		return
 	}
-	if err := s.db.DeletePushSubscription(r.Context(), req.Endpoint); err != nil {
+	// Scoped to the caller. It was not, so anybody who learned somebody else's
+	// endpoint could unsubscribe their device — a small thing to do to somebody
+	// and no reason at all to be able to.
+	if err := s.db.DeletePushSubscription(r.Context(), userFrom(r.Context()).ID, req.Endpoint); err != nil {
 		s.internal(w, r, "delete push subscription", err)
 		return
 	}

internal/httpapi/panel_test.go

@@ -102,7 +102,7 @@ func TestTwoRunsOfOneMailboxDoNotOverlap(t *testing.T) {
 	// Stands in for a run: takes the lock, notes how many are inside, and holds it
 	// until told to let go.
 	run := func(done chan<- struct{}) {
-		unlock := srv.lockMailbox("box-1")
+		unlock := srv.lockSync("box-1")
 		mu.Lock()
 		inside++
 		if inside > most {
@@ -136,10 +136,10 @@ func TestTwoRunsOfOneMailboxDoNotOverlap(t *testing.T) {
 
 	// And two different mailboxes must not wait for each other: one slow host is
 	// not everybody's problem.
-	a := srv.lockMailbox("box-a")
+	a := srv.lockSync("box-a")
 	done := make(chan struct{})
 	go func() {
-		srv.lockMailbox("box-b")()
+		srv.lockSync("box-b")()
 		close(done)
 	}()
 	select {

internal/httpapi/projects_handlers.go

@@ -356,6 +356,34 @@ func (s *Server) handleCreateSection(w http.ResponseWriter, r *http.Request) {
 	writeJSON(w, http.StatusCreated, sectionJSON{sec.ID, sec.ProjectID, sec.Name, sec.SortOrder})
 }
 
+// handleReorderSections sets the order of the sections inside one project.
+//
+// Under /projects/{projectID}/ rather than beside the other section routes, which
+// hang off a section id: the thing being changed here is the project's list, not
+// any one section in it. That also puts it behind the same project permission
+// check as everything else in this subtree, instead of resolving a section to its
+// project one id at a time and hoping they all named the same one.
+func (s *Server) handleReorderSections(w http.ResponseWriter, r *http.Request) {
+	var req reorderProjectsRequest
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
+	// The same ceiling as the projects list, and a bound on the transaction rather
+	// than an opinion about how many sections a project may have.
+	if len(req.IDs) > 500 {
+		writeFieldErrors(w, map[string]string{"ids": "at most 500 at a time"})
+		return
+	}
+
+	projectID := chi.URLParam(r, "projectID")
+	if err := s.db.ReorderSections(r.Context(), projectID, req.IDs); err != nil {
+		s.internal(w, r, "reorder sections", err)
+		return
+	}
+	s.activity(r, projectID, "", "section.updated", nil)
+	w.WriteHeader(http.StatusNoContent)
+}
+
 // sectionAccess resolves a section to its project and checks the caller may edit
 // it. Sections have no permissions of their own — they belong to a project, and
 // that is where access is decided.

internal/httpapi/projects_test.go

@@ -327,6 +327,71 @@ func TestAProjectCanHaveSeveralSections(t *testing.T) {
 	}
 }
 
+// Sections can be dragged into another order, and the order is the project's.
+//
+// The list is what a project's plan *is* — "Planlægning" before "I gang" is the
+// order the work happens in — and until this endpoint existed the only way to
+// change it was to delete a section and make it again, which moves its tasks to
+// the unsectioned area on the way past.
+func TestSectionsCanBeReordered(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	_, project := ts.do(t, "POST", "/api/v1/projects", map[string]any{"name": "Ombygning"})
+	id := project["id"].(string)
+
+	ids := map[string]string{}
+	names := []string{"Planlægning", "I gang", "Til gennemsyn"}
+	for _, name := range names {
+		_, body := ts.do(t, "POST", "/api/v1/projects/"+id+"/sections", map[string]any{"name": name})
+		ids[name] = body["id"].(string)
+	}
+
+	want := []string{"Til gennemsyn", "Planlægning", "I gang"}
+	order := make([]string, len(want))
+	for i, name := range want {
+		order[i] = ids[name]
+	}
+
+	resp, body := ts.do(t, "POST", "/api/v1/projects/"+id+"/sections/reorder",
+		map[string]any{"ids": order})
+	if resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("reorder: status %d, body %v", resp.StatusCode, body)
+	}
+
+	_, listed := ts.do(t, "GET", "/api/v1/projects/"+id+"/sections", nil)
+	got, _ := listed["sections"].([]any)
+	if len(got) != len(want) {
+		t.Fatalf("the project has %d sections, want %d", len(got), len(want))
+	}
+	for i, raw := range got {
+		if name := raw.(map[string]any)["name"]; name != want[i] {
+			t.Errorf("section %d is %v, want %q", i, name, want[i])
+		}
+	}
+
+	// A section belonging to somebody else's project is not moved by naming it
+	// here. The ids in the body are not checked one by one — the WHERE clause is
+	// what refuses them — so this is the assertion that it actually does.
+	_, other := ts.do(t, "POST", "/api/v1/projects", map[string]any{"name": "Sommerhus"})
+	otherID := other["id"].(string)
+	_, foreign := ts.do(t, "POST", "/api/v1/projects/"+otherID+"/sections",
+		map[string]any{"name": "Havearbejde"})
+	foreignID := foreign["id"].(string)
+
+	resp, _ = ts.do(t, "POST", "/api/v1/projects/"+id+"/sections/reorder",
+		map[string]any{"ids": []string{foreignID}})
+	if resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("reorder with a foreign id: status %d", resp.StatusCode)
+	}
+
+	_, stillThere := ts.do(t, "GET", "/api/v1/projects/"+otherID+"/sections", nil)
+	elsewhere, _ := stillThere["sections"].([]any)
+	if len(elsewhere) != 1 || elsewhere[0].(map[string]any)["name"] != "Havearbejde" {
+		t.Errorf("the other project's section did not survive: %v", elsewhere)
+	}
+}
+
 // A project can be created straight into a group.
 //
 // `group_id` was declared on the request type, documented on it, and applied by

internal/httpapi/server.go

@@ -63,7 +63,15 @@ type Server struct {
 	// the endpoints that check a secret are limited separately and more tightly
 	// than the rest of the API.
 	loginLimiter *limiter
+	// Sat kun af prøver; se icsClient().
+	icsFetch     *http.Client
+	mailLimiter  *limiter
 	resetLimiter *limiter
+	// Second factors are counted against the *account*, not the address. An
+	// address bucket is defeated by anybody with more than one address, which is
+	// the ordinary case — and a six-digit code is only safe while guesses are
+	// scarce, so the scarcity has to be counted against the thing being attacked.
+	totpLimiter *limiter
 
 	// Computed once from the built index.html; see csp.go.
 	csp string
@@ -80,7 +88,16 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 		mail:         mail.New(cfg.SMTP, cfg.BaseURL, log),
 		hub:          realtime.NewHub(log),
 		loginLimiter: newLimiter(10, 15*time.Minute),
+		// Generous on purpose: the caller is the mail server, one address delivering
+		// everything, and a personal instance gets a handful of forwarded mails a
+		// day. Sixty a minute never touches real delivery and still puts a ceiling
+		// on the one public write path that has none.
+		mailLimiter:  newLimiter(60, time.Minute),
 		resetLimiter: newLimiter(5, time.Hour),
+		// Tighter than the password bucket, and for a whole hour. Somebody typing
+		// their own code gets it right on the first or second try; ten in an hour is
+		// generous for a person and hopeless for a search through a million.
+		totpLimiter: newLimiter(10, time.Hour),
 	}
 	s.csp = contentSecurityPolicy(scriptHashes(web))
 	if w, err := auth.NewWebAuthn(cfg.BaseURL, "verdande"); err == nil {
@@ -113,7 +130,15 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 
 	// Inbound mail. Delivered by the mail server rather than by a browser, and
 	// authenticated by the token in the recipient address.
-	r.Post("/inbound/mail", s.handleInboundMail)
+	//
+	// Rate limited for the same reason the beacon is: it is reachable by anybody
+	// who can route to this instance. It is the weaker case of the two — the
+	// beacon writes a row for a stranger, this one writes nothing without a
+	// 256-bit address it has to already know — but an address is shared with mail
+	// servers and printed on a settings page, and once it is known there was no
+	// ceiling at all on what could be made with it.
+	r.Method(http.MethodPost, "/inbound/mail",
+		s.rateLimit(s.mailLimiter, http.HandlerFunc(s.handleInboundMail)))
 
 	// MCP with the token in the query string, for clients that cannot send a
 	// header. Claude's custom-connector dialog takes a URL and nothing else, so
@@ -133,6 +158,12 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 	r.Group(func(r chi.Router) {
 		r.Use(s.requireAuth)
 		r.Get("/oauth/gmail/callback", s.handleGmailCallback)
+		// The calendar comes back on a URI of its own rather than sharing Gmail's.
+		// One callback deciding from stored state which of the two flows had
+		// returned would put the choice of *which token store to write* behind a
+		// lookup — and when that goes wrong it goes wrong by overwriting a working
+		// Gmail connection with calendar tokens.
+		r.Get("/oauth/calendar/callback", s.handleCalendarCallback)
 	})
 	r.Route("/caldav", s.caldavRoutes)
 
@@ -143,6 +174,19 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 			writeJSON(w, http.StatusOK, map[string]string{"status": "ok"})
 		})
 
+		// The beacon's receiving end, and the number it produces. Both without a
+		// session, because the caller is a stranger's installation — that is the
+		// whole feature. Rate limited like the login routes: it is the one endpoint
+		// anybody on the internet can write to, and a row per request is a row per
+		// request whether or not the request is honest.
+		//
+		// Refused with a 404 unless this instance has been told it is the collector,
+		// so an ordinary installation is not a spam sink that answers 200 to
+		// anything.
+		r.Method(http.MethodPost, "/beacon",
+			s.rateLimit(s.loginLimiter, http.HandlerFunc(s.handleBeaconPing)))
+		r.Get("/beacon/count", s.handleBeaconCount)
+
 		// Reachable without being signed in — these are how you become signed in.
 		r.Route("/auth", func(r chi.Router) {
 			r.Get("/setup", s.handleSetupState)
@@ -181,11 +225,7 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 				r.Put("/nav-order", s.handleSetNavOrder)
 				r.Post("/logout", s.handleLogout)
 				r.Post("/password/change", s.handleChangePassword)
-				r.Post("/totp/setup", s.handleTOTPSetup)
-				r.Post("/totp/confirm", s.handleTOTPConfirm)
-				r.Post("/totp/disable", s.handleTOTPDisable)
 				r.Get("/recovery-codes", s.handleRecoveryCodesCount)
-				r.Post("/recovery-codes", s.handleRecoveryCodesRegenerate)
 
 				// Behind requireSession, like the API tokens and for the same
 				// reason: a leaked token must not be able to end the sessions of
@@ -195,6 +235,23 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 					r.Get("/sessions", s.handleListSessions)
 					r.Delete("/sessions/{sessionID}", s.handleDeleteSession)
 
+					// Turning the second factor on, and the codes that get you back
+					// in. Sessions only, and setup asks for the password.
+					//
+					// These sat outside this group, which meant the door opened with
+					// a token and closed only with a password: `disable` has always
+					// asked for the password, `setup` asked for nothing. A stolen
+					// token could therefore fetch the secret, load it into its own
+					// authenticator, confirm it, and keep the only copy of the
+					// recovery codes. The owner's next login would stop at a code
+					// they do not have, and they could not turn it off without
+					// logging in first. A theft becomes a tenancy — the same
+					// sentence written next door about passkeys, and the same rule.
+					r.Post("/totp/setup", s.handleTOTPSetup)
+					r.Post("/totp/confirm", s.handleTOTPConfirm)
+					r.Post("/totp/disable", s.handleTOTPDisable)
+					r.Post("/recovery-codes", s.handleRecoveryCodesRegenerate)
+
 					// Registering a key, and the list of them. Sessions only, for
 					// the same reason: a leaked token must not be able to add its
 					// own way in — that would turn a theft into a tenancy.
@@ -240,6 +297,9 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 				r.Get("/key", s.handlePushKey)
 				r.Post("/subscribe", s.handleSubscribePush)
 				r.Post("/unsubscribe", s.handleUnsubscribePush)
+				// One notification, to this account's own devices, reporting what
+				// each of them did with it. The answer to "der kommer ikke nogen".
+				r.Post("/test", s.handlePushTest)
 			})
 
 			r.Route("/attachments/{attachmentID}", func(r chi.Router) {
@@ -288,6 +348,13 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 				r.Get("/{noteID}", s.handleGetNote)
 				r.Patch("/{noteID}", s.handleUpdateNote)
 				r.Delete("/{noteID}", s.handleDeleteNote)
+				// A picture pasted or dropped into a note. The import could bring
+				// files in and nothing else could, which made the most ordinary act
+				// in a notes app — paste a screenshot — do nothing at all.
+				r.Post("/{noteID}/attachments", s.handleUploadNoteAttachment)
+				// Mange på én gang. Efter en import af tolv hundrede noter er det at
+				// vælge halvtreds det almindelige tilfælde, ikke det yderste.
+				r.Post("/bulk", s.handleNoteBulk)
 			})
 
 			r.Route("/mailboxes", func(r chi.Router) {
@@ -316,6 +383,25 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 				})
 			})
 
+			// A calendar read from somewhere else, laid under verdande's own dated
+			// tasks. Read-only, so there is no write here to gate: the events are a
+			// copy of somebody else's calendar, and the only thing a person decides
+			// is which of them to look at.
+			r.Route("/calendar", func(r chi.Router) {
+				r.Get("/", s.handleGetCalendar)
+				r.Put("/calendars", s.handleSetCalendars)
+				r.Delete("/", s.handleDisconnectCalendar)
+				r.Post("/authorize", s.handleCalendarAuthorize)
+				r.Post("/sync", s.handleCalendarSyncNow)
+				r.Get("/events", s.handleCalendarEvents)
+
+				// Abonnementer: en kalender, man kun har en adresse til. Flere pr.
+				// person, hvor Google er én — det er adressen, der er identiteten,
+				// hvor tokenet er det for OAuth.
+				r.Post("/subscriptions", s.handleSubscribeCalendar)
+				r.Delete("/subscriptions/{id}", s.handleUnsubscribeCalendar)
+			})
+
 			r.Route("/import", func(r chi.Router) {
 				r.Post("/todoist", s.handleImportTodoist)
 				r.Post("/csv", s.handleImportCSV)
@@ -365,6 +451,7 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 					r.Group(func(r chi.Router) {
 						r.Use(s.requireProject(store.RoleEditor))
 						r.Post("/sections", s.handleCreateSection)
+						r.Post("/sections/reorder", s.handleReorderSections)
 					})
 
 					r.Group(func(r chi.Router) {
@@ -390,6 +477,11 @@ func New(cfg *config.Config, db *store.DB, log *slog.Logger, web fs.FS) *Server
 				// `:latest` on the way, which is what makes a restart an update.
 				r.Get("/panel", s.handlePanelStatus)
 				r.Post("/panel/restart", s.handleRestartFromPanel)
+				// Who this instance reports to, and whether it does. An
+				// administrator's decision, not a person's: there is one
+				// installation and it either phones home or it does not.
+				r.Get("/beacon/settings", s.handleBeaconStatus)
+				r.Put("/beacon/settings", s.handleSetBeacon)
 				// A backup file is a complete copy of the database, so this whole
 				// group is sessions-only as well as administrators-only: a leaked
 				// token must not be able to download everybody's data.

internal/httpapi/tasks_test.go

@@ -6,6 +6,7 @@ import (
 	"io"
 	"mime/multipart"
 	"net/http"
+	"net/url"
 	"strings"
 	"testing"
 	"time"
@@ -1777,3 +1778,42 @@ func TestQuickAddSectionIDIsCheckedAgainstTheProject(t *testing.T) {
 		t.Errorf("content = %v", moved["content"])
 	}
 }
+
+// A word inside a word, on the task side.
+//
+// The notes search grew this first — a note holding `keepInventory` could not be
+// found by typing "keep inventory" — and a task is the same sentence with a
+// different table behind it. `SpeedAdmin` is one token, so "speed admin" asks for
+// two words that are nowhere.
+//
+// Having it in one of the two searches would be worse than having it in neither:
+// "it finds it in my notes but not in my tasks" is the next report.
+func TestATaskIsFoundByAWordInsideAWord(t *testing.T) {
+	ts := newTestServer(t)
+	ts.bootstrap(t)
+
+	ts.do(t, "POST", "/api/v1/tasks", map[string]any{
+		"content": "Opret kodeord til Kulturskolen (SpeedAdmin)",
+	})
+	ts.do(t, "POST", "/api/v1/tasks", map[string]any{"content": "Køb kaffe"})
+
+	for _, tc := range []struct {
+		query string
+		want  int
+	}{
+		{"speed admin", 1},
+		{"admin", 1},
+		{"speedadmin", 1},
+		{"kulturskolen", 1},
+		{"vandhane", 0},
+	} {
+		resp, body := ts.do(t, "GET", "/api/v1/search?q="+url.QueryEscape(tc.query), nil)
+		if resp.StatusCode != http.StatusOK {
+			t.Fatalf("%q: status %d", tc.query, resp.StatusCode)
+		}
+		tasks, _ := body["tasks"].([]any)
+		if len(tasks) != tc.want {
+			t.Errorf("%q found %d tasks, want %d", tc.query, len(tasks), tc.want)
+		}
+	}
+}

internal/httpapi/totp_handlers.go

@@ -20,7 +20,24 @@ type totpSetupResponse struct {
 }
 
 func (s *Server) handleTOTPSetup(w http.ResponseWriter, r *http.Request) {
+	// The password, the same as disable has always asked for.
+	//
+	// Symmetry is the point. Turning the second factor on is at least as
+	// consequential as turning it off — it decides which device gets to say yes
+	// from now on — and a door that opens more easily than it closes is not a
+	// door. Together with the move behind requireSession, this is what stops a
+	// stolen credential from enrolling its own authenticator and keeping the only
+	// copy of the recovery codes.
+	var req totpDisableRequest
+	if err := decodeJSON(w, r, &req); err != nil {
+		return
+	}
 	user := userFrom(r.Context())
+	if err := auth.VerifyPassword(user.PasswordHash, req.Password); err != nil {
+		writeFieldErrors(w, map[string]string{"password": "not correct"})
+		return
+	}
+
 	if user.TOTPEnabled {
 		writeError(w, http.StatusConflict, CodeConflict,
 			"two-factor authentication is already switched on")

internal/ics/ics.go

@@ -367,3 +367,180 @@ func unescape(s string) string {
 	r := strings.NewReplacer(`\n`, "\n", `\N`, "\n", `\,`, ",", `\;`, ";", `\\`, `\`)
 	return r.Replace(s)
 }
+
+// --- reading a whole calendar ------------------------------------------------------
+
+// Event is one VEVENT, reduced to what a grid needs — the same shape the Google
+// side already stores, so the two arrive in the database identically.
+//
+// Days at both ends, inclusive, because that is the question a grid asks: which
+// cells does this cover. StartsAt keeps the stamp as the file wrote it, offset and
+// all: it is what the chip shows, and reparsing it here would ask *this process*
+// what time zone it is in.
+type Event struct {
+	UID      string
+	Summary  string
+	Location string
+	URL      string
+	AllDay   bool
+	StartsAt string // RFC3339, empty for an all-day event
+	EndsAt   string
+	StartDay string // YYYY-MM-DD
+	EndDay   string
+}
+
+// Subscribed is what an .ics file somebody subscribed to says.
+//
+// Its own name rather than `Calendar`: that one is the *writing* side — what
+// verdande publishes as a feed — and the two are different documents that happen
+// to share a format. One name for both would be a promise that a round trip
+// through this package is lossless, which it is not and does not need to be.
+type Subscribed struct {
+	Name     string
+	TimeZone string
+	Events   []Event
+}
+
+// ParseCalendar reads an .ics file that somebody subscribed to.
+//
+// Forgiving in the same way ParseVTODO is, and for a sharper reason: this file was
+// written by somebody else's software and is fetched on a schedule with nobody
+// watching. One property this does not understand must not cost the other four
+// hundred events in the file — so anything unrecognised is skipped, and an event
+// without the two things a grid cannot do without (a uid and a start) is dropped
+// rather than stored half-made.
+//
+// # What it deliberately does not do
+//
+// **Recurrence is not expanded.** RRULE in a subscribed calendar means "and then
+// every Tuesday until 2031", and expanding it here would mean writing a year of
+// rows per rule and re-deciding what happens when the file changes. A recurring
+// event is stored as its first occurrence, which is honest — it is what the file
+// says on its own terms — and the day it matters, the answer is an expander shared
+// with `internal/recurrence`, not a second one written here.
+//
+// **Time zones are not resolved.** A DTSTART with TZID=Europe/Copenhagen is stored
+// with the offset the file carries or, when it carries none, as a local stamp. The
+// alternative is shipping a tz database and deciding what to do when the file
+// names a zone this build has never heard of. The day that matters, it is one
+// place.
+func ParseCalendar(body string) Subscribed {
+	var cal Subscribed
+	var current *Event
+	depth := 0
+
+	for _, line := range unfold(body) {
+		name, params, value := splitLine(line)
+		upper := strings.ToUpper(name)
+
+		switch upper {
+		case "BEGIN":
+			if strings.EqualFold(value, "VEVENT") {
+				current = &Event{}
+			}
+			depth++
+			continue
+		case "END":
+			depth--
+			if strings.EqualFold(value, "VEVENT") && current != nil {
+				if current.UID != "" && current.StartDay != "" {
+					cal.Events = append(cal.Events, *current)
+				}
+				current = nil
+			}
+			continue
+		}
+
+		if current == nil {
+			// Calendar-level properties. X-WR-CALNAME is what every publisher uses
+			// to say what the calendar is called; there is no standard property for
+			// it, which is why a non-standard one is read here.
+			switch upper {
+			case "X-WR-CALNAME":
+				cal.Name = unescape(value)
+			case "X-WR-TIMEZONE":
+				cal.TimeZone = value
+			}
+			continue
+		}
+
+		switch upper {
+		case "UID":
+			current.UID = value
+		case "SUMMARY":
+			current.Summary = unescape(value)
+		case "LOCATION":
+			current.Location = unescape(value)
+		case "URL":
+			current.URL = value
+		case "DTSTART":
+			current.AllDay = isDateOnly(params, value)
+			current.StartDay, current.StartsAt = readStamp(params, value)
+		case "DTEND":
+			day, at := readStamp(params, value)
+			current.EndsAt = at
+			// DTEND is exclusive for an all-day event: a one-day event ends on the
+			// next morning. Stored inclusive, or every all-day event would cover one
+			// cell too many — and a public holiday would eat the day after it.
+			if current.AllDay {
+				day = shiftDay(day, -1)
+			}
+			current.EndDay = day
+		}
+	}
+
+	// A missing DTEND means the event ends when it starts. Filled after the loop
+	// rather than inside it, because DTEND may arrive before DTSTART in a file
+	// nobody promised would be in order.
+	for i := range cal.Events {
+		if cal.Events[i].EndDay == "" || cal.Events[i].EndDay < cal.Events[i].StartDay {
+			cal.Events[i].EndDay = cal.Events[i].StartDay
+		}
+	}
+	return cal
+}
+
+// isDateOnly reports whether a stamp is a day rather than a moment. VALUE=DATE
+// says so outright; an eight-character value says it by being one.
+func isDateOnly(params, value string) bool {
+	if strings.Contains(strings.ToUpper(params), "VALUE=DATE") {
+		return true
+	}
+	return len(value) == 8 && !strings.ContainsAny(value, "T")
+}
+
+// readStamp returns the day and, for a timed event, the moment as RFC3339.
+//
+// Three shapes exist in the wild: 20260821 (a day), 20260821T140000Z (UTC), and
+// 20260821T140000 with a TZID parameter (a local time in a named zone). The third
+// is kept as written, without a zone: resolving it needs a tz database, and a
+// stamp that says 14:00 is closer to the truth than one this process has moved.
+func readStamp(params, value string) (day, at string) {
+	value = strings.TrimSpace(value)
+	if len(value) < 8 {
+		return "", ""
+	}
+	day = value[0:4] + "-" + value[4:6] + "-" + value[6:8]
+	if len(value) < 15 || value[8] != 'T' {
+		return day, ""
+	}
+
+	clock := value[9:11] + ":" + value[11:13] + ":" + value[13:15]
+	switch {
+	case strings.HasSuffix(value, "Z"):
+		return day, day + "T" + clock + "Z"
+	default:
+		// A local stamp, with or without a TZID. Written without an offset, which
+		// is what the interface already treats as "the clock the file said".
+		return day, day + "T" + clock
+	}
+}
+
+// shiftDay moves a YYYY-MM-DD by whole days.
+func shiftDay(day string, by int) string {
+	t, err := time.Parse("2006-01-02", day)
+	if err != nil {
+		return day
+	}
+	return t.AddDate(0, 0, by).Format("2006-01-02")
+}

internal/ics/ics_test.go

@@ -220,3 +220,101 @@ func TestFilename(t *testing.T) {
 		}
 	}
 }
+
+// En abonnementskalender, som den ser ud fra nogen andens software.
+//
+// Alle fire former, der findes i naturen, i én fil: en heldagsbegivenhed, en med
+// et UTC-stempel, en med TZID og lokal tid, og en over flere dage. Plus en, der
+// mangler det, et gitter ikke kan undvære, og en foldet linje — begge dele
+// kommer, og de kommer uden varsel.
+func TestParseCalendarReadsTheFourShapesThatExist(t *testing.T) {
+	body := "BEGIN:VCALENDAR\r\n" +
+		"VERSION:2.0\r\n" +
+		"X-WR-CALNAME:Helligdage\r\n" +
+		"BEGIN:VEVENT\r\n" +
+		"UID:heldag@example.dk\r\n" +
+		"DTSTART;VALUE=DATE:20260824\r\n" +
+		"DTEND;VALUE=DATE:20260825\r\n" +
+		"SUMMARY:Grundlovsdag\r\n" +
+		"END:VEVENT\r\n" +
+		"BEGIN:VEVENT\r\n" +
+		"UID:utc@example.dk\r\n" +
+		"DTSTART:20260825T120000Z\r\n" +
+		"DTEND:20260825T133000Z\r\n" +
+		`SUMMARY:Møde med en meget lang titel\, der er foldet over to lin` + "\r\n" + " jer\r\n" +
+		"END:VEVENT\r\n" +
+		"BEGIN:VEVENT\r\n" +
+		"UID:lokal@example.dk\r\n" +
+		"DTSTART;TZID=Europe/Copenhagen:20260826T090000\r\n" +
+		"DTEND;TZID=Europe/Copenhagen:20260826T100000\r\n" +
+		"SUMMARY:Tandlæge\r\n" +
+		"END:VEVENT\r\n" +
+		"BEGIN:VEVENT\r\n" +
+		"UID:flerdage@example.dk\r\n" +
+		"DTSTART;VALUE=DATE:20260901\r\n" +
+		"DTEND;VALUE=DATE:20260904\r\n" +
+		"SUMMARY:Ferie\r\n" +
+		"END:VEVENT\r\n" +
+		"BEGIN:VEVENT\r\n" +
+		"SUMMARY:Uden uid og uden start\r\n" +
+		"END:VEVENT\r\n" +
+		"END:VCALENDAR\r\n"
+
+	cal := ParseCalendar(body)
+
+	if cal.Name != "Helligdage" {
+		t.Errorf("kalenderens navn = %q", cal.Name)
+	}
+	if len(cal.Events) != 4 {
+		t.Fatalf("%d begivenheder, want 4 — den femte mangler både uid og start", len(cal.Events))
+	}
+
+	// Heldag. DTEND er eksklusiv i filen: en endagsbegivenhed slutter næste morgen.
+	// Gemmes inklusivt, ellers dækker hver eneste heldagsbegivenhed én celle for
+	// meget — og en helligdag æder dagen efter sig.
+	one := cal.Events[0]
+	if !one.AllDay || one.StartDay != "2026-08-24" || one.EndDay != "2026-08-24" {
+		t.Errorf("heldag: %+v", one)
+	}
+	if one.StartsAt != "" {
+		t.Errorf("en heldagsbegivenhed fik et klokkeslæt: %q", one.StartsAt)
+	}
+
+	// UTC, og den foldede linje samlet igen.
+	two := cal.Events[1]
+	if two.StartsAt != "2026-08-25T12:00:00Z" || two.EndsAt != "2026-08-25T13:30:00Z" {
+		t.Errorf("utc: %+v", two)
+	}
+	if two.Summary != "Møde med en meget lang titel, der er foldet over to linjer" {
+		t.Errorf("den foldede linje blev ikke samlet: %q", two.Summary)
+	}
+
+	// Lokal tid med TZID: klokkeslættet står, som filen skrev det. At flytte det
+	// her ville spørge *denne* proces, hvilken tidszone den er i.
+	three := cal.Events[2]
+	if three.StartsAt != "2026-08-26T09:00:00" {
+		t.Errorf("lokal: %q", three.StartsAt)
+	}
+
+	// Flere dage, begge ender inklusive.
+	four := cal.Events[3]
+	if four.StartDay != "2026-09-01" || four.EndDay != "2026-09-03" {
+		t.Errorf("ferie: %s til %s", four.StartDay, four.EndDay)
+	}
+}
+
+// En fil, der ikke er en kalender, må ikke koste noget. Den hentes på et skema
+// uden nogen, der kigger med, og en fejl her ville være en sweep, der stopper.
+func TestParseCalendarSurvivesRubbish(t *testing.T) {
+	for _, body := range []string{
+		"",
+		"ikke en kalender overhovedet",
+		"BEGIN:VCALENDAR\r\nEND:VCALENDAR\r\n",
+		"BEGIN:VEVENT\r\nDTSTART:x\r\nEND:VEVENT\r\n",
+		"BEGIN:VEVENT\r\nUID:a\r\nDTSTART:2026\r\nEND:VEVENT\r\n",
+	} {
+		if got := ParseCalendar(body); len(got.Events) != 0 {
+			t.Errorf("%q gav %d begivenheder", body, len(got.Events))
+		}
+	}
+}

internal/imap/imap.go

@@ -310,3 +310,61 @@ func cut(text string) string {
 	}
 	return text
 }
+
+// Unflag takes the flag off the messages that have become tasks.
+//
+// This is what stops the same mail arriving twice, and it replaces a marker that
+// could not be made to hold. The marker was a uid: remember the highest one seen,
+// start above it next time. That is correct only if uids are what the server says
+// they are, and it broke in every way such a scheme can — a fetch reply without a
+// uid, a UIDVALIDITY that changed under a rebuilt mailbox, two mailbox rows
+// pointing at the same folder and each keeping its own marker. Every one of those
+// produced the same symptom: the same mail became a task again, and again.
+//
+// Unflagging moves the record out of our database and into the mailbox, where the
+// question is asked. "Flagged" means "turn this into a task"; once it is a task,
+// the flag has done its job and is not true any more. Nothing has to be
+// remembered, so there is nothing to get out of step — a second sync, a second
+// mailbox row, or a database restored from a backup all find the same answer,
+// because the answer is in the mailbox.
+//
+// It is destructive in the small way that matters: it changes something in
+// somebody's mail. That is the trade, and it is the right way round — a flag is a
+// note to self that this needs doing, and it has been done.
+//
+// The folder is re-selected writable. Since() opens it read-only on purpose, and
+// a search should not need write access just because a later step might.
+func (c *Client) Unflag(uids []uint32) error {
+	if len(uids) == 0 {
+		return nil
+	}
+	if _, err := c.c.Select(c.folder, nil).Wait(); err != nil {
+		return fmt.Errorf("open %s for writing: %w", c.folder, err)
+	}
+
+	set := imap.UIDSetNum()
+	for _, u := range uids {
+		if u == 0 {
+			// No uid means nothing to address. The message keeps its flag and will
+			// be seen again — which is the failure this whole function exists to
+			// prevent, so it is worth saying rather than skipping quietly.
+			continue
+		}
+		set.AddNum(imap.UID(u))
+	}
+	if len(set) == 0 {
+		return nil
+	}
+
+	// Silent: the server has nothing to tell us that we would act on, and the
+	// untagged replies would only have to be collected and thrown away.
+	cmd := c.c.Store(set, &imap.StoreFlags{
+		Op:     imap.StoreFlagsDel,
+		Silent: true,
+		Flags:  []imap.Flag{imap.FlagFlagged},
+	}, nil)
+	if err := cmd.Close(); err != nil {
+		return fmt.Errorf("unflag in %s: %w", c.folder, err)
+	}
+	return nil
+}

internal/imap/roundtrip_test.go

@@ -316,3 +316,81 @@ func TestNothingAtOrBelowTheMarkerComesBack(t *testing.T) {
 		}
 	}
 }
+
+// Den samme mail må ikke kunne blive en opgave to gange.
+//
+// Prøven kører hele vejen imod en rigtig server: den finder de flagede mails,
+// tager flaget af dem, og spørger så forfra som en ny synkronisering ville — med
+// markøren sat til nul, altså det værst tænkelige tilfælde, hvor alt, hvad vi
+// selv havde husket, er væk. Der skal ikke komme noget tilbage.
+//
+// Det er det, markøren ikke kunne love. Den holdt kun, hvis uids var, hvad
+// serveren sagde, og fejlede på et fetch-svar uden uid, på en UIDVALIDITY, der
+// skiftede, og på to postkasse-rækker, der pegede samme sted og hver især førte
+// deres eget regnskab. Alle tre gav det samme: mailen blev en opgave igen.
+func TestAnUnflaggedMailDoesNotComeBack(t *testing.T) {
+	addr, pool := serveMailbox(t, []mail{
+		{subject: "Faktura 4711", from: "anders@example.dk", body: "Vedhæftet.", flagged: true},
+		{subject: "Nyhedsbrev", from: "noreply@example.com", body: "Læs mere", flagged: false},
+		{subject: "Kontrakt", from: "jura@example.dk", body: "Til underskrift", flagged: true},
+	})
+
+	client, err := Dial(Account{Host: addr, Username: "kw", Password: "hemmelig", RootCAs: pool})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer client.Close()
+
+	msgs, _, err := client.Since(0, 25)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(msgs) != 2 {
+		t.Fatalf("fandt %d flagede mails, ventede 2", len(msgs))
+	}
+
+	uids := make([]uint32, 0, len(msgs))
+	for _, m := range msgs {
+		uids = append(uids, m.UID)
+	}
+	if err := client.Unflag(uids); err != nil {
+		t.Fatalf("Unflag: %v", err)
+	}
+
+	// Forfra, og fra nul: selv en instans, der intet husker, må ikke finde dem igen.
+	again, _, err := client.Since(0, 25)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(again) != 0 {
+		t.Fatalf("%d mails kom igen efter at flaget var taget af — den første af dem er %q",
+			len(again), again[0].Subject)
+	}
+}
+
+// Et tomt kald må ikke åbne mappen skrivbar for ingenting.
+func TestUnflagWithNothingToDoIsQuiet(t *testing.T) {
+	addr, pool := serveMailbox(t, []mail{
+		{subject: "Faktura", from: "a@example.dk", body: "x", flagged: true},
+	})
+	client, err := Dial(Account{Host: addr, Username: "kw", Password: "hemmelig", RootCAs: pool})
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer client.Close()
+
+	if err := client.Unflag(nil); err != nil {
+		t.Errorf("Unflag(nil) = %v", err)
+	}
+	// Kun nuller: der er intet at adressere, og mailen skal beholde sit flag.
+	if err := client.Unflag([]uint32{0, 0}); err != nil {
+		t.Errorf("Unflag(nuller) = %v", err)
+	}
+	msgs, _, err := client.Since(0, 25)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(msgs) != 1 {
+		t.Errorf("mailen mistede sit flag, uden at nogen havde adresseret den")
+	}
+}

internal/jobs/jobs.go

@@ -41,10 +41,19 @@ type Runner struct {
 	// reason as SyncGmail: it owns turning a message into a task.
 	SyncMailbox func(ctx context.Context, user *store.User, m *store.Mailbox) (int, error)
 
+	// SyncCalendars refreshes one person's connected calendars. Supplied the same
+	// way and for the same reason: the HTTP layer owns the token refresh.
+	SyncCalendars func(ctx context.Context, user *store.User) (int, error)
+
 	// Push delivers a notification to somebody's devices. Supplied by the HTTP
 	// layer, which owns the VAPID keys and the subscription list.
 	Push func(userID, title, body, projectID string)
 
+	// SendBeacon reports this installation to the collector, if it is switched on
+	// and a day has passed. Supplied by the HTTP layer for the same reason as the
+	// syncs: that is where the instance settings and the version live.
+	SendBeacon func(ctx context.Context) error
+
 	wg sync.WaitGroup
 }
 
@@ -69,6 +78,17 @@ func (r *Runner) Start(ctx context.Context) {
 	// account for a to-do app. Ten minutes is fast enough for something somebody
 	// starred and slow enough not to look like abuse.
 	r.every(ctx, "mailboxes", 10*time.Minute, r.syncMailboxes)
+	// Fifteen minutes rather than ten. A calendar is not a mailbox: nothing here
+	// becomes a task, so being a quarter of an hour behind on a meeting next
+	// Tuesday costs nothing — and the "fetch now" button covers the one case where
+	// it matters, which is somebody who has just moved something and is looking at
+	// the grid to check. A person who has just connected does not wait for this
+	// either; connecting fetches the list, and the first sync is one click away.
+	r.every(ctx, "calendars", 15*time.Minute, r.syncCalendars)
+	// Checked hourly, sent at most daily — the same shape as the backup, and for
+	// the same reason: a container asleep at the scheduled minute would otherwise
+	// skip the day. SendBeacon decides whether a day has passed.
+	r.every(ctx, "beacon", time.Hour, r.sendBeacon)
 }
 
 func (r *Runner) Wait() { r.wg.Wait() }
@@ -103,6 +123,17 @@ func (r *Runner) every(ctx context.Context, name string, interval time.Duration,
 	}()
 }
 
+// sendBeacon reports that this installation exists, at most once a day.
+//
+// Every decision about what is sent, and why this is on by default rather than
+// compulsory, lives in internal/beacon. This is only the clock.
+func (r *Runner) sendBeacon(ctx context.Context) error {
+	if r.SendBeacon == nil {
+		return nil
+	}
+	return r.SendBeacon(ctx)
+}
+
 // --- reminders ------------------------------------------------------------------
 
 // sendDueReminders delivers everything that has come due since the last pass.
@@ -322,6 +353,44 @@ func (r *Runner) syncMailboxes(ctx context.Context) error {
 	return nil
 }
 
+// --- calendars ------------------------------------------------------------------------
+
+// syncCalendars refreshes everybody's connected calendars.
+//
+// Its own sweep rather than a branch inside the mailbox one. They share nothing
+// past the word "poll": that loop asks each mailbox how far it has read and makes
+// tasks, this one replaces a window of somebody else's events. Folding them
+// together would mean one interval for two questions that do not want the same
+// answer, and a Google that is down would count against a mailbox that is not.
+//
+// Nobody is waiting, so no budget of its own — the request handler has one because
+// a person is watching a spinner. One account that has gone quiet costs this run
+// its own timeout and nothing else: the loop moves on per person.
+func (r *Runner) syncCalendars(ctx context.Context) error {
+	if r.SyncCalendars == nil {
+		return nil
+	}
+	userIDs, err := r.db.UsersWithCalendars(ctx)
+	if err != nil {
+		return err
+	}
+	for _, id := range userIDs {
+		user, err := r.db.UserByID(ctx, id)
+		if err != nil || user == nil {
+			continue
+		}
+		written, err := r.SyncCalendars(ctx, user)
+		if err != nil {
+			r.log.Warn("calendar sync", "err", err, "user", id)
+			continue
+		}
+		if written > 0 {
+			r.log.Info("calendar refreshed", "user", id, "events", written)
+		}
+	}
+	return nil
+}
+
 // --- housekeeping -----------------------------------------------------------------
 
 func (r *Runner) emptyTrash(ctx context.Context) error {

internal/push/push.go

@@ -32,6 +32,8 @@ import (
 	"net/url"
 	"strings"
 	"time"
+
+	"github.com/kristianwind/verdande/internal/safedial"
 )
 
 type Subscription struct {
@@ -101,7 +103,19 @@ func Send(ctx context.Context, sub Subscription, payload Payload, vapid VAPID) e
 	req.Header.Set("Urgency", "normal")
 	req.Header.Set("Authorization", "vapid t="+token+", k="+vapid.Public)
 
-	resp, err := (&http.Client{Timeout: 15 * time.Second}).Do(req)
+	// safedial, not a plain client. The endpoint comes from the browser's
+	// subscription, which is to say from whoever is signed in, and this server
+	// POSTs to it on a schedule they choose — set a reminder a minute out and the
+	// job goes again every minute.
+	//
+	// The subscribe handler already refuses a private address, and that check is
+	// not enough on its own: it runs on the URL as written, and a name it approves
+	// can answer 127.0.0.1 the next time it is looked up. safedial checks the
+	// *resolved* address at connect time, which is the only moment the address
+	// that will actually be used is known — and it checks every redirect hop the
+	// same way. The parse-time check stays, because it says no while somebody is
+	// looking at the screen.
+	resp, err := safedial.Client(15 * time.Second).Do(req)
 	if err != nil {
 		return err
 	}

internal/push/push_test.go

@@ -0,0 +1,66 @@
+package push
+
+import (
+	"context"
+	"crypto/ecdh"
+	"crypto/rand"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+)
+
+// A subscription must not be able to point the server at its own network.
+//
+// The endpoint is whatever the browser handed over, which is to say whatever the
+// person signed in handed over, and this server POSTs to it on a schedule they
+// choose. The subscribe handler refuses a private address when it is written
+// down — and that check runs on the URL as written, so a name it approves can
+// answer 127.0.0.1 the next time it is looked up. This is the other half: the
+// address that will actually be dialled, checked at the moment of dialling.
+func TestSendRefusesToReachInside(t *testing.T) {
+	reached := false
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		reached = true
+		w.WriteHeader(http.StatusCreated)
+	}))
+	defer srv.Close()
+	// httptest listens on 127.0.0.1, which is exactly the destination in question.
+	if !strings.Contains(srv.URL, "127.0.0.1") {
+		t.Skipf("the test server is not on loopback: %s", srv.URL)
+	}
+
+	err := Send(context.Background(), testSubscription(t, srv.URL), Payload{Title: "hej"}, testVAPID(t))
+	if err == nil {
+		t.Fatal("Send reached a loopback address without complaining")
+	}
+	if reached {
+		t.Error("the request arrived: the wall let it through and only the answer failed")
+	}
+}
+
+func testSubscription(t *testing.T, endpoint string) Subscription {
+	t.Helper()
+	key, err := ecdh.P256().GenerateKey(rand.Reader)
+	if err != nil {
+		t.Fatal(err)
+	}
+	auth := make([]byte, 16)
+	if _, err := rand.Read(auth); err != nil {
+		t.Fatal(err)
+	}
+	return Subscription{
+		Endpoint: endpoint,
+		P256dh:   b64.EncodeToString(key.PublicKey().Bytes()),
+		Auth:     b64.EncodeToString(auth),
+	}
+}
+
+func testVAPID(t *testing.T) VAPID {
+	t.Helper()
+	pub, priv, err := GenerateVAPIDKeys()
+	if err != nil {
+		t.Fatal(err)
+	}
+	return VAPID{Public: pub, Private: priv, Subject: "mailto:kristian@example.dk"}
+}

internal/quickadd/quickadd.go

@@ -650,8 +650,18 @@ func pad2(n int) string {
 // The trailing group is bounded rather than greedy: without a limit, "hver mandag
 // kl 10 p1" would hand the whole tail to the recurrence parser, which would refuse
 // all of it and leave the task not repeating at all.
+//
+// Mellemleddet må være et ord og ikke kun et tal. "hver anden tirsdag" er den måde,
+// folk siger det på — recurrence-pakken har oversat "hver anden" til "hver 2."
+// siden den blev skrevet — men her stoppede mønsteret efter "anden" og gav den
+// sætning videre. Den er ikke en regel, så "hver anden" blev stående i titlen og
+// "tirsdag" blev en forfaldsdato. Sløjfen nedenfor kan kun *fjerne* ord fra
+// enden; den kan ikke hente et, mønsteret aldrig tog med.
+//
+// Fundet af en prøve, der brugte vendingen som selvfølgelig og opdagede, at den
+// ikke var det.
 var reRepeat = regexp.MustCompile(
-	`(?i)(?:^|\s)((?:hver|hvert|every|each)\s+(?:\d+\.?\s+)?[\p{L}]+(?:\s*(?:,|\bog\b|\band\b)\s*[\p{L}]+)*` +
+	`(?i)(?:^|\s)((?:hver|hvert|every|each)\s+(?:\d+\.?\s+|(?:anden|andet|tredje|fjerde|femte|sjette|other)\s+)?[\p{L}]+(?:\s*(?:,|\bog\b|\band\b)\s*[\p{L}]+)*` +
 		`|hverdage|ugedage|weekdays?|dagligt|daily|ugentligt|weekly|månedligt|maanedligt|monthly|årligt|aarligt|yearly|annually` +
 		`|den\s+\d{1,2}\.?\s+i\s+måneden|den\s+\d{1,2}\.?\s+i\s+maaneden)(?:\s|$)`)
 

internal/quickadd/quickadd_test.go

@@ -603,3 +603,36 @@ func TestASectionIsWrittenWithASlash(t *testing.T) {
 		}
 	}
 }
+
+// "hver anden tirsdag" er den måde, folk siger det på.
+//
+// recurrence-pakken har oversat "hver anden" til "hver 2." siden den blev skrevet,
+// men sætningen nåede aldrig frem til den: mønsteret her tog kun ét ord efter
+// "hver", så det, der blev givet videre, var "hver anden" — ikke en regel. Så blev
+// de to ord stående i titlen, og "tirsdag" blev til en forfaldsdato. Opgaven
+// gentog sig ikke, og der stod ikke noget om det.
+//
+// Fundet af en prøve, der brugte vendingen som selvfølgelig og opdagede, at den
+// ikke var det.
+func TestEveryOtherWeekdayIsARepetition(t *testing.T) {
+	for _, tc := range []struct {
+		text string
+		rule string
+		left string
+	}{
+		{"vande planterne hver anden tirsdag", "FREQ=WEEKLY;BYDAY=TU;INTERVAL=2", "vande planterne"},
+		{"vande planterne hver 2. tirsdag", "FREQ=WEEKLY;BYDAY=TU;INTERVAL=2", "vande planterne"},
+		{"water the plants every other tuesday", "FREQ=WEEKLY;BYDAY=TU;INTERVAL=2", "water the plants"},
+		{"gå tur hver tredje søndag", "FREQ=WEEKLY;BYDAY=SU;INTERVAL=3", "gå tur"},
+		// Den, der virkede i forvejen, skal blive ved med at gøre det.
+		{"tømme opvasker hver mandag", "FREQ=WEEKLY;BYDAY=MO", "tømme opvasker"},
+	} {
+		got := Parse(tc.text, time.Date(2026, 8, 20, 9, 0, 0, 0, time.UTC), "da")
+		if got.Recurrence != tc.rule {
+			t.Errorf("%q gav reglen %q, want %q", tc.text, got.Recurrence, tc.rule)
+		}
+		if got.Content != tc.left {
+			t.Errorf("%q efterlod titlen %q, want %q", tc.text, got.Content, tc.left)
+		}
+	}
+}

internal/safedial/safedial.go

@@ -0,0 +1,136 @@
+// Package safedial refuses to open a connection to this machine or to the
+// network it sits on.
+//
+// The problem it solves is the same one three times: somewhere a user supplies a
+// URL or a host, and the server fetches it. An AI provider's base address, a push
+// endpoint, an IMAP host, a beacon collector. Each of them is a feature, and each
+// of them is also a way to ask the server to reach something the caller cannot —
+// the panel next door, a metadata endpoint, a database on the same bridge — and
+// to report back whether it answered. On a homelab that is the whole point of the
+// box being where it is.
+//
+// # Why the check is on the address and not on the URL
+//
+// Parsing the URL and refusing "localhost" or "127.0.0.1" is the obvious version
+// and it does not work: a name resolves to whatever its owner says, so
+// `evil.example.com` can answer 127.0.0.1 — and can answer differently the second
+// time, after a parse-time check has passed. That is DNS rebinding, and the only
+// place it can be caught is after resolution and before the connection, which is
+// exactly where net.Dialer's Control hook runs.
+//
+// # What is refused
+//
+// Loopback, link-local (including 169.254.169.254, which is where cloud metadata
+// lives), the private ranges, unique-local IPv6, and the unspecified address.
+// Everything else is allowed: this is not an allowlist of the internet, it is a
+// wall around the inside.
+package safedial
+
+import (
+	"fmt"
+	"net"
+	"net/http"
+	"net/url"
+	"syscall"
+	"time"
+)
+
+// ErrPrivate is returned when a connection was refused for where it pointed. It
+// deliberately does not say which address it resolved to: the answer is itself
+// the thing an attacker is fishing for.
+type ErrPrivate struct{ Host string }
+
+func (e *ErrPrivate) Error() string {
+	return fmt.Sprintf("refusing to connect to %s: it resolves to an address on this machine or its private network", e.Host)
+}
+
+// Blocked reports whether an address must not be dialled.
+func Blocked(ip net.IP) bool {
+	return ip == nil ||
+		ip.IsLoopback() ||
+		ip.IsUnspecified() ||
+		ip.IsLinkLocalUnicast() ||
+		ip.IsLinkLocalMulticast() ||
+		ip.IsInterfaceLocalMulticast() ||
+		ip.IsMulticast() ||
+		ip.IsPrivate()
+}
+
+// Control is a net.Dialer Control function that refuses a private destination.
+//
+// It runs after the name has been resolved and before the socket is connected,
+// which is the only moment where the address that will actually be used is known.
+func Control(network, address string, _ syscall.RawConn) error {
+	host, _, err := net.SplitHostPort(address)
+	if err != nil {
+		return err
+	}
+	ip := net.ParseIP(host)
+	if Blocked(ip) {
+		return &ErrPrivate{Host: host}
+	}
+	return nil
+}
+
+// Dialer is a net.Dialer that will not reach inside.
+func Dialer(timeout time.Duration) *net.Dialer {
+	return &net.Dialer{Timeout: timeout, Control: Control}
+}
+
+// Client is an http.Client for fetching something a user named.
+//
+// Redirects are followed but capped, and every hop goes through the same dialer —
+// a redirect to 127.0.0.1 is the oldest way around a check that only looked at the
+// URL somebody typed.
+func Client(timeout time.Duration) *http.Client {
+	return &http.Client{
+		Timeout: timeout,
+		Transport: &http.Transport{
+			DialContext:           Dialer(timeout).DialContext,
+			TLSHandshakeTimeout:   timeout,
+			ResponseHeaderTimeout: timeout,
+		},
+		CheckRedirect: func(req *http.Request, via []*http.Request) error {
+			if len(via) >= 5 {
+				return fmt.Errorf("too many redirects")
+			}
+			return nil
+		},
+	}
+}
+
+// AllowPrivate is the same client without the wall, for a destination the
+// operator configured rather than a user — the panel, and nothing else. Named so
+// that using it is a decision somebody wrote down.
+func AllowPrivate(timeout time.Duration) *http.Client {
+	return &http.Client{Timeout: timeout}
+}
+
+// CheckURL reports why an address must not be fetched, or "" if it may be.
+//
+// A convenience for saying no at the moment somebody types an address, rather
+// than later when a request quietly fails. It is NOT the protection — a name can
+// resolve one way now and another way when it is dialled, which is why Control
+// exists and runs every time. Anything this rejects is certainly wrong; anything
+// it accepts is merely not obviously wrong.
+func CheckURL(raw string) string {
+	u, err := url.Parse(raw)
+	if err != nil || u.Host == "" {
+		return "that is not a URL"
+	}
+	host := u.Hostname()
+	if ip := net.ParseIP(host); ip != nil && Blocked(ip) {
+		return "that address is on this machine or its private network"
+	}
+	// A bare name is looked up once so an obvious "localhost" is caught here. A
+	// failure to resolve is not an error: the host may simply not exist yet, and
+	// refusing to save an address because DNS was slow would be its own bug.
+	if ips, err := net.LookupIP(host); err == nil {
+		for _, ip := range ips {
+			if Blocked(ip) {
+				return "that address is on this machine or its private network"
+			}
+		}
+	}
+	return ""
+}

internal/safedial/safedial_test.go

@@ -0,0 +1,55 @@
+package safedial
+
+import (
+	"net"
+	"net/http"
+	"net/http/httptest"
+	"testing"
+	"time"
+)
+
+func TestBlockedCoversTheInside(t *testing.T) {
+	blocked := []string{
+		"127.0.0.1", "::1", "0.0.0.0",
+		// Skyens metadata-endepunkt. Den ene adresse, enhver SSRF sigter efter.
+		"169.254.169.254",
+		"10.0.0.5", "172.16.0.1", "192.168.1.10", "fd00::1", "fe80::1",
+	}
+	for _, s := range blocked {
+		if !Blocked(net.ParseIP(s)) {
+			t.Errorf("Blocked(%s) = false", s)
+		}
+	}
+	for _, s := range []string{"93.184.216.34", "8.8.8.8", "2606:2800:220:1:248:1893:25c8:1946"} {
+		if Blocked(net.ParseIP(s)) {
+			t.Errorf("Blocked(%s) = true, men det er en almindelig adresse", s)
+		}
+	}
+}
+
+// En httptest-server lytter på loopback, hvilket gør den til den perfekte prøve:
+// den findes, den svarer, og den må ikke kunne nås.
+func TestClientRefusesLoopback(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		t.Error("forbindelsen blev oprettet; den skulle være afvist")
+	}))
+	defer srv.Close()
+
+	_, err := Client(3 * time.Second).Get(srv.URL)
+	if err == nil {
+		t.Fatal("en forespørgsel til loopback skal afvises")
+	}
+}
+
+func TestAllowPrivateStillReaches(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.WriteHeader(http.StatusNoContent)
+	}))
+	defer srv.Close()
+
+	resp, err := AllowPrivate(3 * time.Second).Get(srv.URL)
+	if err != nil {
+		t.Fatalf("den skal kunne nå det, operatøren selv har peget på: %v", err)
+	}
+	resp.Body.Close()
+}

internal/store/beacon.go

@@ -0,0 +1,138 @@
+package store
+
+import (
+	"context"
+	"time"
+)
+
+// The collector's side of the beacon: what has reported in, and how many.
+//
+// Every method here is about counting installations, never about identifying one.
+// There is no lookup by anything but the id an installation made up for itself,
+// and there is nothing to join it against.
+
+// BeaconStats is the whole of what the collector can say.
+type BeaconStats struct {
+	Total      int            `json:"total"`
+	Active7d   int            `json:"active_7d"`
+	Active30d  int            `json:"active_30d"`
+	ByVersion  map[string]int `json:"by_version"`
+	LastPingAt *time.Time     `json:"last_ping_at,omitempty"`
+}
+
+// maxBeaconInstalls bounds the table.
+//
+// The receiving endpoint is unauthenticated — it has to be, since the caller is a
+// stranger's installation — and it inserts a row keyed on an id the caller made
+// up. Anybody can therefore mint ids and grow this table until the volume the
+// database lives on is full. A ceiling turns that from a disk-fill into a number
+// that is merely wrong, which is the difference between an outage and an
+// annoyance. A hundred thousand is far above any plausible real count and far
+// below anything that matters on disk.
+const maxBeaconInstalls = 100_000
+
+// RecordBeacon writes down that an installation is alive, and on which version.
+//
+// `first_seen` is kept from the original row on purpose: it is the only thing
+// that distinguishes an installation that has been running for a year from one
+// that appeared this morning, and it costs nothing to keep.
+func (db *DB) RecordBeacon(ctx context.Context, instanceID, version string) error {
+	now := time.Now().Unix()
+
+	// An installation already known keeps reporting whatever the ceiling says: the
+	// cap must never stop counting the real ones, only stop new rows from being
+	// invented without limit.
+	var known int
+	if err := db.QueryRowContext(ctx,
+		`SELECT count(*) FROM beacon_installs WHERE instance_id = ?`, instanceID).Scan(&known); err != nil {
+		return err
+	}
+	if known == 0 {
+		var total int
+		if err := db.QueryRowContext(ctx,
+			`SELECT count(*) FROM beacon_installs`).Scan(&total); err != nil {
+			return err
+		}
+		if total >= maxBeaconInstalls {
+			// Not an error to the caller. Their installation did nothing wrong, and
+			// there is nothing for them to do about it.
+			return nil
+		}
+	}
+
+	_, err := db.ExecContext(ctx, `
+		INSERT INTO beacon_installs (instance_id, version, first_seen, last_seen)
+		VALUES (?, ?, ?, ?)
+		ON CONFLICT (instance_id) DO UPDATE SET
+		    version   = excluded.version,
+		    last_seen = excluded.last_seen`,
+		instanceID, version, now, now)
+	return err
+}
+
+// BeaconCount is how many installations have reported within the window.
+//
+// A window rather than a running total, and that is not a detail. A total only
+// ever climbs: every instance anybody ever started for ten minutes stays in it
+// forever, so the number stops meaning "how many people use this" and starts
+// meaning "how many people have ever tried it". The question being asked is the
+// first one.
+func (db *DB) BeaconCount(ctx context.Context, within time.Duration) (int, error) {
+	var n int
+	err := db.QueryRowContext(ctx,
+		`SELECT count(*) FROM beacon_installs WHERE last_seen >= ?`,
+		time.Now().Add(-within).Unix()).Scan(&n)
+	return n, err
+}
+
+// BeaconSummary is everything the settings page shows.
+func (db *DB) BeaconSummary(ctx context.Context) (BeaconStats, error) {
+	out := BeaconStats{ByVersion: map[string]int{}}
+
+	if err := db.QueryRowContext(ctx,
+		`SELECT count(*) FROM beacon_installs`).Scan(&out.Total); err != nil {
+		return out, err
+	}
+
+	var err error
+	if out.Active7d, err = db.BeaconCount(ctx, 7*24*time.Hour); err != nil {
+		return out, err
+	}
+	if out.Active30d, err = db.BeaconCount(ctx, 30*24*time.Hour); err != nil {
+		return out, err
+	}
+
+	// Versions of the ones still alive, not of everything ever seen: an instance
+	// that stopped reporting two years ago should not still be counted as running
+	// the version it was on when it stopped.
+	rows, err := db.QueryContext(ctx, `
+		SELECT version, count(*) FROM beacon_installs
+		 WHERE last_seen >= ? AND version <> ''
+		 GROUP BY version`, time.Now().Add(-30*24*time.Hour).Unix())
+	if err != nil {
+		return out, err
+	}
+	defer rows.Close()
+	for rows.Next() {
+		var v string
+		var n int
+		if err := rows.Scan(&v, &n); err != nil {
+			return out, err
+		}
+		out.ByVersion[v] = n
+	}
+	if err := rows.Err(); err != nil {
+		return out, err
+	}
+
+	var last *int64
+	if err := db.QueryRowContext(ctx,
+		`SELECT max(last_seen) FROM beacon_installs`).Scan(&last); err != nil {
+		return out, err
+	}
+	if last != nil {
+		t := time.Unix(*last, 0)
+		out.LastPingAt = &t
+	}
+	return out, nil
+}

internal/store/calendars.go

@@ -0,0 +1,531 @@
+package store
+
+import (
+	"context"
+	"database/sql"
+	"errors"
+	"fmt"
+	"strings"
+	"time"
+)
+
+// CalendarAccount is one connected account belonging to one person.
+type CalendarAccount struct {
+	ID       string `json:"id"`
+	UserID   string `json:"-"`
+	Provider string `json:"provider"`
+	Account  string `json:"account,omitempty"`
+
+	// URL er tom for en OAuth-konto og adressen for et abonnement. Den er
+	// identiteten på et abonnement, som tokenet er det for Google: to rækker med
+	// den samme adresse for den samme person ville være den samme kalender hentet
+	// to gange, uden nogen måde at sige hvilken der gælder.
+	URL string `json:"url,omitempty"`
+
+	RefreshToken string    `json:"-"` // never leaves the server
+	AccessToken  string    `json:"-"`
+	ExpiresAt    time.Time `json:"-"`
+
+	LastSyncAt time.Time `json:"last_sync_at,omitempty"`
+	CreatedAt  time.Time `json:"created_at"`
+}
+
+// Calendar is one calendar inside an account, shown or not.
+type Calendar struct {
+	ID        string `json:"id"`
+	AccountID string `json:"-"`
+	RemoteID  string `json:"remote_id"`
+	Name      string `json:"name"`
+	Colour    string `json:"colour,omitempty"`
+	TimeZone  string `json:"time_zone,omitempty"`
+	Primary   bool   `json:"primary"`
+	// Writable says the connected account could write to it. Nothing does — the
+	// scope is read-only — and it is stored so the day something can, the interface
+	// knows which calendars could even be offered.
+	Writable  bool `json:"writable"`
+	Shown     bool `json:"shown"`
+	SortOrder int  `json:"-"`
+}
+
+// CalendarEvent is one occurrence, in the shape a grid asks about: days at both
+// ends, inclusive.
+type CalendarEvent struct {
+	ID         string `json:"id"`
+	CalendarID string `json:"calendar_id"`
+	RemoteID   string `json:"-"`
+	Summary    string `json:"summary"`
+	StartsAt   string `json:"starts_at,omitempty"`
+	EndsAt     string `json:"ends_at,omitempty"`
+	StartDay   string `json:"start_day"`
+	EndDay     string `json:"end_day"`
+	AllDay     bool   `json:"all_day"`
+	Location   string `json:"location,omitempty"`
+	URL        string `json:"url,omitempty"`
+
+	// CalendarName and Colour ride along with the event rather than being looked up
+	// per row by the caller. Same reasoning as the assignee names on a task list:
+	// the chip has to be drawn in the calendar's own colour, and a second request
+	// per event to learn it is a request per event.
+	CalendarName string `json:"calendar_name,omitempty"`
+	Colour       string `json:"colour,omitempty"`
+}
+
+const calendarAccountColumns = `id, user_id, provider, account, refresh_token,
+	access_token, expires_at, last_sync_at, created_at`
+
+// Med adressen. Skrivningen tilføjer den selv, fordi indsættelsen navngiver
+// søjlerne; læsningen skal bede om den.
+const calendarAccountRead = calendarAccountColumns + `, url`
+
+// CalendarAccountFor returns a person's connection to one provider, or nil.
+func (db *DB) CalendarAccountFor(ctx context.Context, userID, provider string) (*CalendarAccount, error) {
+	// `url = ''`: en OAuth-konto. Uden det ville et abonnement hos den samme
+	// udbyder kunne svare på spørgsmålet "hvad er personen forbundet til med
+	// Google", og et abonnement har hverken token eller konto at svare med.
+	row := db.QueryRowContext(ctx, `SELECT `+calendarAccountRead+`
+		FROM calendar_accounts WHERE user_id = ? AND provider = ? AND url = ''`, userID, provider)
+	a, err := db.scanCalendarAccount(row)
+	if err == sql.ErrNoRows {
+		return nil, nil
+	}
+	if err != nil {
+		return nil, err
+	}
+	return &a, nil
+}
+
+// SaveCalendarAccount writes one, sealing the two columns that must not travel
+// with a backup. Called by hand rather than by the settings tables' own seal,
+// which does not reach this table.
+func (db *DB) SaveCalendarAccount(ctx context.Context, a *CalendarAccount) error {
+	if a.ID == "" {
+		a.ID = NewID()
+	}
+	if a.Provider == "" {
+		a.Provider = "google"
+	}
+	if a.CreatedAt.IsZero() {
+		a.CreatedAt = time.Now()
+	}
+
+	refresh, err := db.sealValue(a.RefreshToken)
+	if err != nil {
+		return fmt.Errorf("seal refresh_token: %w", err)
+	}
+	access, err := db.sealValue(a.AccessToken)
+	if err != nil {
+		return fmt.Errorf("seal access_token: %w", err)
+	}
+
+	// The conflict is on the pair, not on the id: a reconnect builds a fresh
+	// struct with no id in it, and matching on the id would insert a second row
+	// that the unique index then refuses — which reads to the person as "connecting
+	// failed" when what happened is that it already had.
+	// `WHERE url = ''` fordi indekset er delvist siden 0026.
+	//
+	// En OAuth-konto er én pr. person og udbyder — det er tokenet, der er
+	// identiteten — mens et abonnement er én pr. adresse, og man har flere. De to
+	// bor i samme tabel, fordi et abonnement *er* en kalender med begivenheder i,
+	// og den halvdel er den samme. Målet for konflikten skal sige, hvilken af de to
+	// slags rækker der tales om, ellers ved SQLite det ikke.
+	_, err = db.ExecContext(ctx, `
+		INSERT INTO calendar_accounts (`+calendarAccountColumns+`, url)
+		VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+		ON CONFLICT (user_id, provider) WHERE url = '' DO UPDATE SET
+		    account = excluded.account,
+		    refresh_token = excluded.refresh_token,
+		    access_token = excluded.access_token,
+		    expires_at = excluded.expires_at,
+		    last_sync_at = excluded.last_sync_at`,
+		a.ID, a.UserID, a.Provider, a.Account, refresh, access,
+		unixOrZero(a.ExpiresAt), unixOrZero(a.LastSyncAt), a.CreatedAt.Unix(), a.URL)
+	if err != nil {
+		return err
+	}
+	if a.URL != "" {
+		return nil
+	}
+
+	// The row that is actually there wins. On a reconnect the insert above updated
+	// an existing row, so the id in hand is one nothing points at — and the
+	// calendars are hung off the id, so writing them against it would attach them
+	// to nothing.
+	return db.QueryRowContext(ctx,
+		`SELECT id FROM calendar_accounts WHERE user_id = ? AND provider = ? AND url = ''`,
+		a.UserID, a.Provider).Scan(&a.ID)
+}
+
+// MarkCalendarSynced records how far a run got. Written separately from the rest
+// so a sync never rewrites the credentials it is holding in memory.
+func (db *DB) MarkCalendarSynced(ctx context.Context, accountID string, at time.Time) error {
+	_, err := db.ExecContext(ctx,
+		`UPDATE calendar_accounts SET last_sync_at = ? WHERE id = ?`, at.Unix(), accountID)
+	return err
+}
+
+// DeleteCalendarAccount disconnects one. The calendars and their events go with it
+// — unlike the tasks a mailbox made, none of this is the person's own work; it is
+// a copy of somebody else's calendar and it has no meaning without the account.
+func (db *DB) DeleteCalendarAccount(ctx context.Context, userID, id string) error {
+	_, err := db.ExecContext(ctx,
+		`DELETE FROM calendar_accounts WHERE id = ? AND user_id = ?`, id, userID)
+	return err
+}
+
+// UsersWithCalendars returns everybody the sweep has to visit.
+func (db *DB) UsersWithCalendars(ctx context.Context) ([]string, error) {
+	rows, err := db.QueryContext(ctx, `SELECT DISTINCT user_id FROM calendar_accounts`)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	var ids []string
+	for rows.Next() {
+		var id string
+		if err := rows.Scan(&id); err != nil {
+			return nil, err
+		}
+		ids = append(ids, id)
+	}
+	return ids, rows.Err()
+}
+
+// Calendars returns what one account holds, in the order the provider listed them.
+func (db *DB) Calendars(ctx context.Context, accountID string) ([]Calendar, error) {
+	rows, err := db.QueryContext(ctx, `
+		SELECT id, account_id, remote_id, name, colour, time_zone, primary_one,
+		       writable, shown, sort_order
+		FROM calendars WHERE account_id = ? ORDER BY sort_order, name`, accountID)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	var out []Calendar
+	for rows.Next() {
+		var c Calendar
+		if err := rows.Scan(&c.ID, &c.AccountID, &c.RemoteID, &c.Name, &c.Colour,
+			&c.TimeZone, &c.Primary, &c.Writable, &c.Shown, &c.SortOrder); err != nil {
+			return nil, err
+		}
+		out = append(out, c)
+	}
+	return out, rows.Err()
+}
+
+// ReplaceCalendars writes the list the provider just gave, keeping what the person
+// chose.
+//
+// The choice survives because it is not in the list: Google says which calendars
+// exist and what they are called, and only verdande knows which of them somebody
+// wants to look at. Merged rather than deleted and rewritten, so ticking a calendar
+// on does not come undone the next time the sweep runs.
+//
+// A calendar that has disappeared from the account is removed, and its events go
+// with it — a calendar somebody was unshared from should stop being drawn, not
+// stay behind as a frozen copy of what it held on the last good day.
+func (db *DB) ReplaceCalendars(ctx context.Context, accountID string, list []Calendar) error {
+	return db.Tx(ctx, func(tx *sql.Tx) error {
+		keep := map[string]bool{}
+		for i, c := range list {
+			keep[c.RemoteID] = true
+			// The insert cannot carry `shown`: on conflict it would overwrite the
+			// choice with the default. A new calendar starts hidden — an account
+			// with a dozen shared calendars in it should not fill somebody's grid
+			// the moment they connect it.
+			if _, err := tx.ExecContext(ctx, `
+				INSERT INTO calendars
+				    (id, account_id, remote_id, name, colour, time_zone, primary_one,
+				     writable, shown, sort_order)
+				VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+				ON CONFLICT (account_id, remote_id) DO UPDATE SET
+				    name = excluded.name,
+				    colour = excluded.colour,
+				    time_zone = excluded.time_zone,
+				    primary_one = excluded.primary_one,
+				    writable = excluded.writable,
+				    sort_order = excluded.sort_order`,
+				NewID(), accountID, c.RemoteID, c.Name, c.Colour, c.TimeZone,
+				c.Primary, c.Writable, c.Shown, i); err != nil {
+				return err
+			}
+		}
+
+		rows, err := tx.QueryContext(ctx,
+			`SELECT id, remote_id FROM calendars WHERE account_id = ?`, accountID)
+		if err != nil {
+			return err
+		}
+		var gone []string
+		for rows.Next() {
+			var id, remote string
+			if err := rows.Scan(&id, &remote); err != nil {
+				rows.Close()
+				return err
+			}
+			if !keep[remote] {
+				gone = append(gone, id)
+			}
+		}
+		// Closed before the next statement runs on this transaction. An open result
+		// set with a query behind it is the deadlock the pool comment in store.go
+		// exists to make slow rather than fatal.
+		rows.Close()
+		if err := rows.Err(); err != nil {
+			return err
+		}
+
+		for _, id := range gone {
+			if _, err := tx.ExecContext(ctx, `DELETE FROM calendars WHERE id = ?`, id); err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+}
+
+// ShowCalendars sets which of an account's calendars are drawn.
+//
+// The whole set in one write rather than one flip at a time, for the reason the
+// projects' order is written the same way: a person has a handful of calendars, the
+// question is "these ones", and a list that can only be built by several requests
+// can land half applied.
+func (db *DB) ShowCalendars(ctx context.Context, accountID string, shown []string) error {
+	return db.Tx(ctx, func(tx *sql.Tx) error {
+		if _, err := tx.ExecContext(ctx,
+			`UPDATE calendars SET shown = 0 WHERE account_id = ?`, accountID); err != nil {
+			return err
+		}
+		for _, id := range shown {
+			// Scoped in the SQL rather than checked first: an id read in one step
+			// and written in the next is a window, and the account is the whole of
+			// the permission here.
+			if _, err := tx.ExecContext(ctx,
+				`UPDATE calendars SET shown = 1 WHERE id = ? AND account_id = ?`,
+				id, accountID); err != nil {
+				return err
+			}
+		}
+		// Events of a calendar nobody is looking at are dead weight in a table the
+		// grid queries. They come back with the next sync if it is turned on again.
+		_, err := tx.ExecContext(ctx, `
+			DELETE FROM calendar_events WHERE calendar_id IN
+			    (SELECT id FROM calendars WHERE account_id = ? AND shown = 0)`, accountID)
+		return err
+	})
+}
+
+// ReplaceCalendarEvents swaps one calendar's events inside a window.
+//
+// Replaced rather than merged, and no sync token. Google offers incremental sync,
+// and it would save bytes verdande is not short of while costing the one thing it
+// cannot afford: a token that has to be right, forever, for the grid not to be
+// quietly wrong. A window that is deleted and rewritten cannot drift — an event
+// somebody moved out of the window disappears because it is not in the answer, not
+// because verdande remembered to delete it.
+//
+// Bounded by the window at both ends, so a sync of September does not throw away
+// August.
+func (db *DB) ReplaceCalendarEvents(ctx context.Context, calendarID, from, to string, events []CalendarEvent) error {
+	return db.Tx(ctx, func(tx *sql.Tx) error {
+		if _, err := tx.ExecContext(ctx, `
+			DELETE FROM calendar_events
+			WHERE calendar_id = ? AND end_day >= ? AND start_day <= ?`,
+			calendarID, from, to); err != nil {
+			return err
+		}
+		for _, e := range events {
+			if _, err := tx.ExecContext(ctx, `
+				INSERT INTO calendar_events
+				    (id, calendar_id, remote_id, summary, starts_at, ends_at,
+				     start_day, end_day, all_day, location, url)
+				VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)`,
+				NewID(), calendarID, e.RemoteID, e.Summary, e.StartsAt, e.EndsAt,
+				e.StartDay, e.EndDay, e.AllDay, e.Location, e.URL); err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+}
+
+// CalendarEvents returns everything one person can see between two days,
+// inclusive at both ends.
+//
+// Scoped to the caller in the join rather than checked afterwards: the events of
+// somebody else's calendar are somebody else's, and a filter applied after the read
+// is a filter that can be forgotten.
+func (db *DB) CalendarEvents(ctx context.Context, userID, from, to string) ([]CalendarEvent, error) {
+	rows, err := db.QueryContext(ctx, `
+		SELECT e.id, e.calendar_id, e.remote_id, e.summary, e.starts_at, e.ends_at,
+		       e.start_day, e.end_day, e.all_day, e.location, e.url, c.name, c.colour
+		FROM calendar_events e
+		JOIN calendars c ON c.id = e.calendar_id
+		JOIN calendar_accounts a ON a.id = c.account_id
+		WHERE a.user_id = ? AND c.shown = 1 AND e.end_day >= ? AND e.start_day <= ?
+		ORDER BY e.start_day, e.all_day DESC, e.starts_at`, userID, from, to)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	var out []CalendarEvent
+	for rows.Next() {
+		var e CalendarEvent
+		if err := rows.Scan(&e.ID, &e.CalendarID, &e.RemoteID, &e.Summary, &e.StartsAt,
+			&e.EndsAt, &e.StartDay, &e.EndDay, &e.AllDay, &e.Location, &e.URL,
+			&e.CalendarName, &e.Colour); err != nil {
+			return nil, err
+		}
+		out = append(out, e)
+	}
+	return out, rows.Err()
+}
+
+func (db *DB) scanCalendarAccount(row scanner) (CalendarAccount, error) {
+	var a CalendarAccount
+	var expires, lastSync, created int64
+	err := row.Scan(&a.ID, &a.UserID, &a.Provider, &a.Account, &a.RefreshToken,
+		&a.AccessToken, &expires, &lastSync, &created, &a.URL)
+	if err != nil {
+		return a, err
+	}
+	for _, field := range []*string{&a.RefreshToken, &a.AccessToken} {
+		plain, err := db.unsealValue(*field)
+		if err != nil {
+			return a, err
+		}
+		*field = plain
+	}
+	a.ExpiresAt = timeOrZero(expires)
+	a.LastSyncAt = timeOrZero(lastSync)
+	a.CreatedAt = time.Unix(created, 0)
+	return a, nil
+}
+
+// --- abonnementer ---------------------------------------------------------------
+
+// SubscribeCalendar adds a calendar somebody only has an address to.
+//
+// One row in `calendar_accounts` with a url and no tokens, and one row in
+// `calendars` hung off it — because a subscription *is* one calendar, where an
+// OAuth account is a list of them somebody picks from. Written together so a
+// subscription is never a row with nothing under it: the sweep looks for accounts,
+// and one without a calendar would be fetched for ever and stored nowhere.
+//
+// Shown from the start. Ticking a box after adding an address is a second step for
+// a decision the person already made by adding it.
+func (db *DB) SubscribeCalendar(ctx context.Context, userID, url, name string) (*CalendarAccount, error) {
+	account := &CalendarAccount{
+		ID:        NewID(),
+		UserID:    userID,
+		Provider:  "ics",
+		Account:   name,
+		URL:       url,
+		CreatedAt: time.Now(),
+	}
+
+	err := db.Tx(ctx, func(tx *sql.Tx) error {
+		if _, err := tx.ExecContext(ctx, `
+			INSERT INTO calendar_accounts (id, user_id, provider, account, url, created_at)
+			VALUES (?, ?, 'ics', ?, ?, ?)`,
+			account.ID, userID, name, url, account.CreatedAt.Unix()); err != nil {
+			return err
+		}
+		_, err := tx.ExecContext(ctx, `
+			INSERT INTO calendars (id, account_id, remote_id, name, shown)
+			VALUES (?, ?, ?, ?, 1)`,
+			NewID(), account.ID, url, name)
+		return err
+	})
+	if isUniqueViolation(err) {
+		return nil, ErrDuplicate
+	}
+	if err != nil {
+		return nil, err
+	}
+	return account, nil
+}
+
+// Subscriptions lists the calendars a person subscribes to by address.
+func (db *DB) Subscriptions(ctx context.Context, userID string) ([]CalendarAccount, error) {
+	rows, err := db.QueryContext(ctx, `SELECT `+calendarAccountRead+`
+		FROM calendar_accounts WHERE user_id = ? AND url <> ''
+		ORDER BY created_at`, userID)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	out := []CalendarAccount{}
+	for rows.Next() {
+		a, err := db.scanCalendarAccount(rows)
+		if err != nil {
+			return nil, err
+		}
+		out = append(out, a)
+	}
+	return out, rows.Err()
+}
+
+// AllSubscriptions is every subscription on the instance, for the sweep.
+//
+// The sweep runs for the server rather than for a session, so it cannot ask "whose
+// is this" the way a handler can — it is handed the user id along with the row and
+// writes against that.
+func (db *DB) AllSubscriptions(ctx context.Context) ([]CalendarAccount, error) {
+	rows, err := db.QueryContext(ctx, `SELECT `+calendarAccountRead+`
+		FROM calendar_accounts WHERE url <> '' ORDER BY last_sync_at`)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+
+	out := []CalendarAccount{}
+	for rows.Next() {
+		a, err := db.scanCalendarAccount(rows)
+		if err != nil {
+			return nil, err
+		}
+		out = append(out, a)
+	}
+	return out, rows.Err()
+}
+
+// RenameSubscribedCalendar writes the name the file calls itself.
+//
+// A subscription is added with the address as its name, because that is all there
+// is to go on before it has been fetched once. The file usually says better —
+// X-WR-CALNAME is what every publisher uses — and taking it means the list says
+// "Danske helligdage" instead of a URL nobody reads.
+//
+// Only when the person has not renamed it themselves: a name they typed is an
+// instruction, and a name from the file is a guess.
+func (db *DB) RenameSubscribedCalendar(ctx context.Context, accountID, name string) error {
+	if strings.TrimSpace(name) == "" {
+		return nil
+	}
+	return db.Tx(ctx, func(tx *sql.Tx) error {
+		if _, err := tx.ExecContext(ctx, `
+			UPDATE calendar_accounts SET account = ?
+			WHERE id = ? AND (account = '' OR account = url)`, name, accountID); err != nil {
+			return err
+		}
+		_, err := tx.ExecContext(ctx, `
+			UPDATE calendars SET name = ?
+			WHERE account_id = ? AND (name = '' OR name = remote_id)`, name, accountID)
+		return err
+	})
+}
+
+// SubscribedCalendarID is the single calendar row under a subscription.
+func (db *DB) SubscribedCalendarID(ctx context.Context, accountID string) (string, error) {
+	var id string
+	err := db.QueryRowContext(ctx,
+		`SELECT id FROM calendars WHERE account_id = ? LIMIT 1`, accountID).Scan(&id)
+	if errors.Is(err, sql.ErrNoRows) {
+		return "", ErrNotFound
+	}
+	return id, err
+}

internal/store/calendars_test.go

@@ -0,0 +1,358 @@
+package store
+
+import (
+	"context"
+	"strings"
+	"testing"
+	"time"
+)
+
+func connectedAccount(t *testing.T, db *DB, userID string) *CalendarAccount {
+	t.Helper()
+	a := &CalendarAccount{
+		UserID: userID, Provider: "google", Account: "kw@nolimit.dk",
+		RefreshToken: "1//0-a-real-looking-refresh-token", AccessToken: "ya29.access",
+		ExpiresAt: time.Now().Add(time.Hour),
+	}
+	if err := db.SaveCalendarAccount(context.Background(), a); err != nil {
+		t.Fatal(err)
+	}
+	return a
+}
+
+// The same claim the mailbox tokens have to meet: what is written must not be
+// readable in the file a backup copies.
+func TestCalendarTokensAreNotStoredInTheClear(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+
+	var refresh, access string
+	if err := db.QueryRowContext(ctx,
+		`SELECT refresh_token, access_token FROM calendar_accounts WHERE id = ?`,
+		a.ID).Scan(&refresh, &access); err != nil {
+		t.Fatal(err)
+	}
+	if strings.Contains(refresh, "real-looking") || strings.Contains(access, "ya29") {
+		t.Fatalf("a token is in the row as written: %q %q", refresh, access)
+	}
+	// The address stays readable, which is the reason for sealing named columns
+	// rather than the whole row.
+	var account string
+	if err := db.QueryRowContext(ctx,
+		`SELECT account FROM calendar_accounts WHERE id = ?`, a.ID).Scan(&account); err != nil {
+		t.Fatal(err)
+	}
+	if account != "kw@nolimit.dk" {
+		t.Errorf("the address was sealed too: %q", account)
+	}
+
+	back, err := db.CalendarAccountFor(ctx, userID, "google")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if back.RefreshToken != a.RefreshToken || back.AccessToken != a.AccessToken {
+		t.Errorf("the tokens did not survive the round trip: %+v", back)
+	}
+}
+
+// Connecting again must replace what was there, not fail. It is the same OAuth
+// flow run a second time, and the person's word for it is "connect", not "add".
+func TestConnectingAgainReplacesTheAccount(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	first := connectedAccount(t, db, userID)
+
+	second := &CalendarAccount{
+		UserID: userID, Provider: "google", Account: "kw@nolimit.dk",
+		RefreshToken: "1//0-fresh", AccessToken: "ya29.fresh",
+	}
+	if err := db.SaveCalendarAccount(ctx, second); err != nil {
+		t.Fatalf("connecting a second time failed: %v", err)
+	}
+	// And the id in hand is the row that is actually there. The calendars hang off
+	// it, so an id nothing points at would attach them to nothing.
+	if second.ID != first.ID {
+		t.Errorf("the second connect came back with id %q, want the existing %q",
+			second.ID, first.ID)
+	}
+
+	back, err := db.CalendarAccountFor(ctx, userID, "google")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if back.RefreshToken != "1//0-fresh" {
+		t.Errorf("the token came back as %q", back.RefreshToken)
+	}
+}
+
+// Google says which calendars exist; only verdande knows which ones somebody wants
+// to look at. A refresh that came back with the same list must not undo the choice.
+func TestRefreshingTheListKeepsWhatWasChosen(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+
+	list := []Calendar{
+		{RemoteID: "primary", Name: "kw@nolimit.dk", Colour: "#3f51b5", Primary: true},
+		{RemoteID: "fam", Name: "Familien", Colour: "#0b8043"},
+		{RemoteID: "old", Name: "Et delt"},
+	}
+	if err := db.ReplaceCalendars(ctx, a.ID, list); err != nil {
+		t.Fatal(err)
+	}
+
+	// Nothing is shown to begin with: an account with a dozen shared calendars in
+	// it should not fill somebody's grid the moment they connect it.
+	stored, err := db.Calendars(ctx, a.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(stored) != 3 {
+		t.Fatalf("got %d calendars, want 3", len(stored))
+	}
+	for _, c := range stored {
+		if c.Shown {
+			t.Errorf("%q was shown without anybody asking for it", c.Name)
+		}
+	}
+
+	var famID string
+	for _, c := range stored {
+		if c.RemoteID == "fam" {
+			famID = c.ID
+		}
+	}
+	if err := db.ShowCalendars(ctx, a.ID, []string{famID}); err != nil {
+		t.Fatal(err)
+	}
+
+	// The sweep runs again, and one calendar has been renamed and one has gone.
+	if err := db.ReplaceCalendars(ctx, a.ID, []Calendar{
+		{RemoteID: "primary", Name: "kw@nolimit.dk", Colour: "#3f51b5", Primary: true},
+		{RemoteID: "fam", Name: "Familien og hunden", Colour: "#0b8043"},
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	stored, err = db.Calendars(ctx, a.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(stored) != 2 {
+		t.Fatalf("got %d calendars; one that has left the account should stop being drawn",
+			len(stored))
+	}
+	for _, c := range stored {
+		switch c.RemoteID {
+		case "fam":
+			if !c.Shown {
+				t.Error("the choice was undone by a refresh of the list")
+			}
+			if c.Name != "Familien og hunden" {
+				t.Errorf("the new name did not arrive: %q", c.Name)
+			}
+			if c.ID != famID {
+				t.Error("the row was rebuilt rather than updated; its id is what was chosen")
+			}
+		case "primary":
+			if c.Shown {
+				t.Error("a calendar nobody chose became shown")
+			}
+		}
+	}
+}
+
+// A window is replaced, not merged. An event moved out of it has to disappear
+// because it is absent from the answer — and the months on either side must not go
+// with it.
+func TestReplacingAWindowLeavesTheMonthsAroundItAlone(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+
+	if err := db.ReplaceCalendars(ctx, a.ID, []Calendar{{RemoteID: "primary", Name: "Min"}}); err != nil {
+		t.Fatal(err)
+	}
+	stored, _ := db.Calendars(ctx, a.ID)
+	cal := stored[0]
+	if err := db.ShowCalendars(ctx, a.ID, []string{cal.ID}); err != nil {
+		t.Fatal(err)
+	}
+
+	july := CalendarEvent{RemoteID: "j", Summary: "Juli", StartDay: "2026-07-15", EndDay: "2026-07-15"}
+	august := CalendarEvent{RemoteID: "a", Summary: "August", StartDay: "2026-08-15", EndDay: "2026-08-15"}
+	if err := db.ReplaceCalendarEvents(ctx, cal.ID, "2026-07-01", "2026-07-31",
+		[]CalendarEvent{july}); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendarEvents(ctx, cal.ID, "2026-08-01", "2026-08-31",
+		[]CalendarEvent{august}); err != nil {
+		t.Fatal(err)
+	}
+
+	got, err := db.CalendarEvents(ctx, userID, "2026-07-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 2 {
+		t.Fatalf("got %d events, want both months: %+v", len(got), got)
+	}
+
+	// August is synced again and the meeting has been cancelled. July stays.
+	if err := db.ReplaceCalendarEvents(ctx, cal.ID, "2026-08-01", "2026-08-31", nil); err != nil {
+		t.Fatal(err)
+	}
+	got, err = db.CalendarEvents(ctx, userID, "2026-07-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 1 || got[0].Summary != "Juli" {
+		t.Fatalf("got %+v; syncing August threw away July", got)
+	}
+}
+
+// The grid asks "which cells does this cover", so an event that straddles the edge
+// of the question has to come back.
+func TestAnEventThatStraddlesTheWindowIsFound(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+	if err := db.ReplaceCalendars(ctx, a.ID, []Calendar{{RemoteID: "c", Name: "Min", Colour: "#0b8043"}}); err != nil {
+		t.Fatal(err)
+	}
+	stored, _ := db.Calendars(ctx, a.ID)
+	if err := db.ShowCalendars(ctx, a.ID, []string{stored[0].ID}); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendarEvents(ctx, stored[0].ID, "2026-08-01", "2026-09-30",
+		[]CalendarEvent{
+			{RemoteID: "trip", Summary: "Aarhus", StartDay: "2026-08-28", EndDay: "2026-09-02", AllDay: true},
+		}); err != nil {
+		t.Fatal(err)
+	}
+
+	// A week that contains neither end of it.
+	got, err := db.CalendarEvents(ctx, userID, "2026-08-31", "2026-09-01")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 1 {
+		t.Fatalf("got %+v; an event covering the whole week was not found", got)
+	}
+	// The calendar's own colour rides along, or the chip has to ask for it per row.
+	if got[0].Colour != "#0b8043" || got[0].CalendarName != "Min" {
+		t.Errorf("the calendar did not come with the event: %+v", got[0])
+	}
+}
+
+// Somebody else's calendar is somebody else's.
+func TestOnlyTheOwnersEventsComeBack(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	other := &User{Email: "andreas@example.dk", Name: "Andreas"}
+	if err := db.CreateUser(ctx, other, "Indbakke"); err != nil {
+		t.Fatal(err)
+	}
+	theirs := &CalendarAccount{UserID: other.ID, Provider: "google", RefreshToken: "r"}
+	if err := db.SaveCalendarAccount(ctx, theirs); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendars(ctx, theirs.ID, []Calendar{{RemoteID: "c", Name: "Deres"}}); err != nil {
+		t.Fatal(err)
+	}
+	stored, _ := db.Calendars(ctx, theirs.ID)
+	if err := db.ShowCalendars(ctx, theirs.ID, []string{stored[0].ID}); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendarEvents(ctx, stored[0].ID, "2026-08-01", "2026-08-31",
+		[]CalendarEvent{{RemoteID: "x", Summary: "Privat", StartDay: "2026-08-10", EndDay: "2026-08-10"}}); err != nil {
+		t.Fatal(err)
+	}
+
+	got, err := db.CalendarEvents(ctx, userID, "2026-08-01", "2026-08-31")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(got) != 0 {
+		t.Fatalf("somebody else's calendar came back: %+v", got)
+	}
+}
+
+// A calendar turned off stops being drawn, and stops costing the grid anything.
+func TestTurningACalendarOffRemovesItsEvents(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+	if err := db.ReplaceCalendars(ctx, a.ID, []Calendar{{RemoteID: "c", Name: "Min"}}); err != nil {
+		t.Fatal(err)
+	}
+	stored, _ := db.Calendars(ctx, a.ID)
+	if err := db.ShowCalendars(ctx, a.ID, []string{stored[0].ID}); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendarEvents(ctx, stored[0].ID, "2026-08-01", "2026-08-31",
+		[]CalendarEvent{{RemoteID: "x", Summary: "Møde", StartDay: "2026-08-10", EndDay: "2026-08-10"}}); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := db.ShowCalendars(ctx, a.ID, nil); err != nil {
+		t.Fatal(err)
+	}
+	var count int
+	if err := db.QueryRowContext(ctx, `SELECT count(*) FROM calendar_events`).Scan(&count); err != nil {
+		t.Fatal(err)
+	}
+	if count != 0 {
+		t.Errorf("%d events of a calendar nobody is looking at are still in the table", count)
+	}
+}
+
+// Disconnecting takes the calendars and the events with it. Unlike the tasks a
+// mailbox made, none of this is the person's own work — it is a copy of somebody
+// else's calendar and it means nothing without the account.
+func TestDisconnectingTakesEverythingWithIt(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+	a := connectedAccount(t, db, userID)
+	if err := db.ReplaceCalendars(ctx, a.ID, []Calendar{{RemoteID: "c", Name: "Min"}}); err != nil {
+		t.Fatal(err)
+	}
+	stored, _ := db.Calendars(ctx, a.ID)
+	if err := db.ShowCalendars(ctx, a.ID, []string{stored[0].ID}); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.ReplaceCalendarEvents(ctx, stored[0].ID, "2026-08-01", "2026-08-31",
+		[]CalendarEvent{{RemoteID: "x", Summary: "Møde", StartDay: "2026-08-10", EndDay: "2026-08-10"}}); err != nil {
+		t.Fatal(err)
+	}
+
+	if err := db.DeleteCalendarAccount(ctx, userID, a.ID); err != nil {
+		t.Fatal(err)
+	}
+	for _, table := range []string{"calendar_accounts", "calendars", "calendar_events"} {
+		var count int
+		if err := db.QueryRowContext(ctx, `SELECT count(*) FROM `+table).Scan(&count); err != nil {
+			t.Fatal(err)
+		}
+		if count != 0 {
+			t.Errorf("%s still holds %d rows after disconnecting", table, count)
+		}
+	}
+}
+
+// The sweep finds people by the presence of an account, so it has to keep working
+// now that the credential in it is ciphertext.
+func TestTheSweepFindsAConnectedCalendar(t *testing.T) {
+	db, userID := sealedStore(t)
+	connectedAccount(t, db, userID)
+
+	users, err := db.UsersWithCalendars(context.Background())
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(users) != 1 || users[0] != userID {
+		t.Fatalf("the sweep found %v", users)
+	}
+}

internal/store/mailboxes_test.go

@@ -195,10 +195,14 @@ func TestAGmailConnectedBeforeTheMoveSurvivesIt(t *testing.T) {
 	}
 	values := `{"refresh_token":"` + sealed + `","trigger":"label","label":"Verdande",` +
 		`"email":"kw@nolimit.dk","seen":["abc","def"],"project_id":"p-1"}`
+	// The conflict target is (user_id, scope) since 0025 — before that the table
+	// held one row per person and the upsert renamed its scope, which is the bug
+	// 0025 exists to undo. Nothing conflicts here; the clause only has to be legal
+	// against the schema this test is running on.
 	if _, err := db.ExecContext(context.Background(), `
 		INSERT INTO user_settings (user_id, scope, values_json, updated_at)
 		VALUES (?, 'gmail', ?, 0)
-		ON CONFLICT (user_id) DO UPDATE SET scope = 'gmail', values_json = excluded.values_json`,
+		ON CONFLICT (user_id, scope) DO UPDATE SET values_json = excluded.values_json`,
 		userID, values); err != nil {
 		t.Fatal(err)
 	}

internal/store/migrations/0022_beacon.sql

@@ -0,0 +1,26 @@
+-- One row per installation that has ever reported in.
+--
+-- This table only exists on the instance that is acting as the collector. Every
+-- other instance has it too, empty, because the schema is the same everywhere —
+-- and because "am I the collector" is a setting, not a build.
+--
+-- What is deliberately NOT here is the point of the whole feature: there is no
+-- column for an IP address, a hostname, a domain, an account, a count of tasks,
+-- or anything else. The two things stored are the two things sent, and the table
+-- is the promise written down where it cannot drift from the prose — a column
+-- added here later is a promise broken, and it should be as awkward to add as
+-- possible.
+--
+-- `instance_id` is the primary key rather than a rowid alias, because the same
+-- installation reporting daily must update its row rather than grow a new one.
+-- What is wanted is "how many installations", not "how many pings".
+CREATE TABLE beacon_installs (
+    instance_id TEXT PRIMARY KEY,
+    version     TEXT NOT NULL DEFAULT '',
+    first_seen  INTEGER NOT NULL,
+    last_seen   INTEGER NOT NULL
+);
+
+-- The count is always "seen in the last N days", so last_seen is what every query
+-- sorts and filters on.
+CREATE INDEX beacon_by_last_seen ON beacon_installs (last_seen DESC);

internal/store/migrations/0023_calendars.sql

@@ -0,0 +1,111 @@
+-- Calendars read from somewhere else: the account, the calendars in it, and the
+-- events they hold.
+--
+-- Not in `mailboxes`, and it was the first thing considered. The two look alike
+-- from a distance — a person connects a Google account over OAuth, tokens are
+-- sealed, a sweep polls it — and every column past that disagrees. `trigger_kind`,
+-- `seen`, `last_uid` and `project_id` all answer "what turns a message into a
+-- task", and nothing here turns into a task. What is here instead is a *set*:
+-- one account holds many calendars, and a person picks which of them to look at.
+-- `mailboxes` has no room for a one-to-many, and giving it one would push the
+-- other kind's columns onto rows that can never use them.
+--
+-- The decisive part is smaller and more practical. `mailboxes.kind` carries
+-- `CHECK (kind IN ('gmail', 'imap'))`, and SQLite cannot alter a CHECK — adding a
+-- third value means rebuilding the table with foreign keys off, which is the
+-- expensive, careful thing migration 0021 had to do for a reason. Paying that to
+-- store something that shares no column with the table is the wrong trade.
+--
+-- Hence no CHECK on `provider` here. CalDAV is the next source to read, and it
+-- should cost a row, not a rebuild.
+
+-- One connected account per person per provider. The tokens live here rather than
+-- on each calendar: a refresh token copied onto ten rows is ten copies of one
+-- secret, ten refreshes racing each other, and ten places to miss when access is
+-- revoked.
+--
+-- `refresh_token` and `access_token` are written sealed; see internal/secret. The
+-- store's own seal/unseal only covers the settings tables, so this table's methods
+-- call them by hand — the same fact worth knowing that migration 0014 wrote down.
+CREATE TABLE calendar_accounts (
+    id            TEXT PRIMARY KEY,
+    user_id       TEXT NOT NULL REFERENCES users (id) ON DELETE CASCADE,
+    provider      TEXT NOT NULL DEFAULT 'google',
+    -- The address the account reports, so the settings page can say which one is
+    -- connected rather than only that one is.
+    account       TEXT NOT NULL DEFAULT '',
+
+    refresh_token TEXT NOT NULL DEFAULT '',
+    access_token  TEXT NOT NULL DEFAULT '',
+    expires_at    INTEGER NOT NULL DEFAULT 0,
+
+    last_sync_at  INTEGER NOT NULL DEFAULT 0,
+    created_at    INTEGER NOT NULL
+);
+
+-- Connecting again replaces what was there, the way Gmail's does: the OAuth flow
+-- is one account at a time, so a second row for the same provider would be a
+-- second set of tokens for the same grant and no way to say which is live.
+CREATE UNIQUE INDEX calendar_accounts_once ON calendar_accounts (user_id, provider);
+
+-- One row per calendar the account can see, kept whether or not it is shown.
+--
+-- `shown` rather than deleting the ones that are not: a person who turns a calendar
+-- off and on again should get the same row back, and the list is refreshed from
+-- Google on every sync — so a calendar that only existed while it was ticked would
+-- lose its choice every time somebody added a new one.
+--
+-- `colour` is Google's own hex for the calendar. Not one of verdande's ten named
+-- colours, which is the rule everywhere else in this database: those are names so
+-- the interface can restyle them per theme, and they are *verdande's* palette. A
+-- Google calendar is told apart by the colour the person already knows it by, and
+-- that colour belongs to Google.
+CREATE TABLE calendars (
+    id         TEXT PRIMARY KEY,
+    account_id TEXT NOT NULL REFERENCES calendar_accounts (id) ON DELETE CASCADE,
+    -- The id Google knows it by. Long, and an address for the primary calendar.
+    remote_id  TEXT NOT NULL,
+    name       TEXT NOT NULL DEFAULT '',
+    colour     TEXT NOT NULL DEFAULT '',
+    time_zone  TEXT NOT NULL DEFAULT '',
+    primary_one INTEGER NOT NULL DEFAULT 0,
+    writable   INTEGER NOT NULL DEFAULT 0,
+    shown      INTEGER NOT NULL DEFAULT 0,
+    sort_order INTEGER NOT NULL DEFAULT 0
+);
+
+CREATE UNIQUE INDEX calendars_once ON calendars (account_id, remote_id);
+
+-- The events themselves, cached.
+--
+-- Cached rather than fetched per view, and the reason is what a month grid does:
+-- it pages. Asking Google on every arrow press means a token refresh, a round trip
+-- and a spinner for a question that was answered ninety seconds ago — and a
+-- calendar that is slower than the calendar it is showing is not worth having.
+-- The window is refreshed by the sweep and by the "fetch now" button.
+--
+-- Days rather than instants, and both ends inclusive. A grid asks one question —
+-- "which cells does this cover" — and `start_day <= ? AND end_day >= ?` is that
+-- question. `starts_at` is kept as Google wrote it, offset and all, because it is
+-- what the chip shows and reparsing it here would ask *this process* what time
+-- zone it is in. An all-day event has neither: it has a day, not a moment, and
+-- inventing midnight for it is how an event moves a day when a container moves
+-- time zone.
+CREATE TABLE calendar_events (
+    id          TEXT PRIMARY KEY,
+    calendar_id TEXT NOT NULL REFERENCES calendars (id) ON DELETE CASCADE,
+    remote_id   TEXT NOT NULL,
+    summary     TEXT NOT NULL DEFAULT '',
+    starts_at   TEXT NOT NULL DEFAULT '',
+    ends_at     TEXT NOT NULL DEFAULT '',
+    start_day   TEXT NOT NULL,
+    end_day     TEXT NOT NULL,
+    all_day     INTEGER NOT NULL DEFAULT 0,
+    location    TEXT NOT NULL DEFAULT '',
+    url         TEXT NOT NULL DEFAULT ''
+);
+
+-- The one query there is: everything a person can see, between two days. Ordered
+-- by the account so the join has somewhere to start, then by the days the grid
+-- compares against.
+CREATE INDEX calendar_events_by_day ON calendar_events (calendar_id, start_day, end_day);

internal/store/migrations/0024_note_archive.sql

@@ -0,0 +1,23 @@
+-- A note can be put away without being thrown away.
+--
+-- The trash already exists and is the wrong tool. It says "this was a mistake,
+-- and in thirty days it is gone" — so anything worth keeping cannot go in it, and
+-- an archive of twelve hundred imported notes is exactly the case where most of
+-- them are worth keeping and almost none are worth looking at. Without somewhere
+-- to put them, the list is the whole archive forever, and finding this week's note
+-- means reading past a decade.
+--
+-- A column rather than a folder. Verdande already refused a second hierarchy for
+-- notes — see 0018 — and archiving is not filing: it is one bit about whether a
+-- note is current, and it has to be answerable next to `deleted_at`, which is the
+-- other bit of the same kind.
+--
+-- Nullable timestamp rather than a boolean, for the same reason `deleted_at` is
+-- one: "when was this put away" answers questions a flag cannot, and it costs the
+-- same byte.
+ALTER TABLE notes ADD COLUMN archived_at INTEGER;
+
+-- The list asks "not deleted, not archived, newest first" on every load, and with
+-- twelve hundred notes that is the query that has to be quick.
+CREATE INDEX notes_current ON notes (archived_at, deleted_at, updated_at DESC)
+    WHERE deleted_at IS NULL;

internal/store/migrations/0025_user_settings_per_scope.sql

@@ -0,0 +1,37 @@
+-- One row per person *and scope*, which is what the table always meant.
+--
+-- `user_settings` was created with `user_id` as its whole primary key, and its
+-- upsert rewrote `scope` on conflict. So a person had exactly one settings row,
+-- and whichever scope was written last was the only one that survived: saving an
+-- AI provider deleted the Gmail connection, and connecting a calendar deleted the
+-- AI provider. Nothing reported it — the write succeeded, and the loss was in the
+-- row it landed on.
+--
+-- This was seen coming twice. Migration 0011 gave the instance's settings their
+-- own table partly for this reason and wrote the warning down in as many words;
+-- 0014 moved mailboxes out and said the warning "was nearly walked into a second
+-- time". The third time was the calendar, which added a `calendar` scope beside
+-- `ai` and `gmail` — and walked in.
+--
+-- The lesson worth keeping: a warning in a migration is read by whoever is editing
+-- *that* migration, and nobody was. The shape is fixed here so there is nothing
+-- left to warn about.
+--
+-- SQLite cannot change a primary key in place, so the table is rebuilt. Existing
+-- rows are copied as they are: each person has at most one, and it holds whatever
+-- scope won last. What the earlier writes lost is already gone and cannot be
+-- recovered here — the values were overwritten, not orphaned.
+CREATE TABLE user_settings_new (
+    user_id     TEXT NOT NULL REFERENCES users (id) ON DELETE CASCADE,
+    scope       TEXT NOT NULL DEFAULT 'general',
+    values_json TEXT NOT NULL DEFAULT '{}',
+    updated_at  INTEGER NOT NULL,
+    PRIMARY KEY (user_id, scope)
+);
+
+INSERT INTO user_settings_new (user_id, scope, values_json, updated_at)
+SELECT user_id, scope, values_json, updated_at FROM user_settings;
+
+DROP TABLE user_settings;
+
+ALTER TABLE user_settings_new RENAME TO user_settings;

internal/store/migrations/0026_calendar_subscriptions.sql

@@ -0,0 +1,30 @@
+-- Abonnement på en kalender, man kun har en adresse til.
+--
+-- Google-vejen er OAuth: én konto pr. person, hvor tokenet er identiteten, og
+-- `calendar_accounts_once` sørgede for netop det — én række pr. (person,
+-- udbyder). Et abonnement er en anden slags ting. Der er intet token og ingen
+-- konto, kun en adresse, og man har flere: helligdage, en fodboldklub, en
+-- fælleskalender fra arbejdet. Det er derfor det blev bedt om som "flere
+-- kalendere".
+--
+-- Samme tabel frem for en ny. Et abonnement er én kalender med begivenheder i, og
+-- det er præcis det, `calendars` og `calendar_events` allerede beskriver — med
+-- vinduet, `shown`, farven og hele fejemaskinen omkring sig. En tabel til ville
+-- være en anden vej til det samme sted, og to veje til ét sted er to steder at
+-- rette den dag noget om begivenheder ændrer sig.
+--
+-- Adressen bliver identiteten, som tokenet er det for Google. To rækker med den
+-- samme adresse for den samme person ville være den samme kalender hentet to
+-- gange, og der er ingen måde at sige hvilken der gælder.
+ALTER TABLE calendar_accounts ADD COLUMN url TEXT NOT NULL DEFAULT '';
+
+-- Det gamle indeks holder stadig for OAuth-rækkerne — dem uden adresse — og
+-- slipper abonnementerne fri. Delvise indekser er den ene ting SQLite kan her,
+-- som en tabelombygning ellers skulle til for.
+DROP INDEX calendar_accounts_once;
+
+CREATE UNIQUE INDEX calendar_accounts_once
+    ON calendar_accounts (user_id, provider) WHERE url = '';
+
+CREATE UNIQUE INDEX calendar_subscription_once
+    ON calendar_accounts (user_id, url) WHERE url <> '';

internal/store/migrations/0027_note_bodies_are_sealed.sql

@@ -0,0 +1,75 @@
+-- verdande:rebuild-tables
+--
+-- Notekroppe forsegles.
+--
+-- Grunden er den samme som for postkassers kodeord og OAuth-tokens: en
+-- sikkerhedskopi kan hentes gennem fladen, og indtil nu bar hver eneste af dem
+-- hver eneste note i klartekst. Noter er dér, folk skriver kodeord og API-nøgler
+-- ned — det ved vi, fordi der ligger sådan nogle i denne installation.
+--
+-- # Hvorfor det ikke var gjort for længe siden
+--
+-- Kroppen var kopieret to steder mere, og begge kopier var klartekst:
+--
+--   * `fold` — en genereret søjle med title || ' ' || body, translittereret, så
+--     "grøn" kan findes ved at skrive "gron".
+--   * `notes_fts` — et FTS5-indeks over titel, krop og fold.
+--
+-- At forsegle kroppen og lade dem stå ville have været teater: en angriber med
+-- filen kunne læse hver note ud af indekset. Så de skal begge væk, og søgningen
+-- skal skrives om til at åbne teksten i Go.
+--
+-- # Hvad det koster, og hvornår formen er forkert
+--
+-- Søgningen læser nu hver note. Målt: tolv hundrede noter à 3,3 kB koster 33–43 ms
+-- pr. søgning, og fladen venter 250 ms efter sidste tastetryk, før den spørger — så
+-- det ligger inde i pausen. Tiden går ikke med at åbne konvolutterne, men med at
+-- folde og småskrive hver krop; det er lineært, så regn med omtrent 350 ms ved ti
+-- tusind noter. Dér er det den forkerte form, og svaret er da et *nøglet* indeks:
+-- en HMAC pr. ord under den samme nøgle. Bemærk hvad det koster, når det bliver
+-- aktuelt — et nøglet indeks kan slå et helt ord op og ikke andet, så præfiks- og
+-- delordssøgningen, der findes i dag, ville falde bort.
+--
+-- # Hvad der stadig ligger i klartekst
+--
+-- Alt andet. Opgavetitler, beskrivelser, kommentarer — og `note_links`, som er
+-- id'er udledt af kroppen, så en fil røber stadig *hvilke* projekter og opgaver en
+-- note peger på, men ikke hvad der står. Det er skrevet i SECURITY.md, og det er
+-- ikke løst her.
+--
+-- # Selve ombygningen
+--
+-- SQLite kan ikke fjerne en genereret søjle på plads, så tabellen bygges om.
+-- Rækkerne kopieres med kroppen, som den er: forseglingen sker i Go, hvor nøglen
+-- er, og en opstart efter denne migration går rækkerne igennem og lukker dem. Til
+-- den tid kan en søjle indeholde begge dele, og `v1:`-præfikset findes netop for
+-- at kunne læses uden at gætte.
+CREATE TABLE notes_sealed (
+    id          TEXT PRIMARY KEY,
+    project_id  TEXT REFERENCES projects (id) ON DELETE CASCADE,
+    title       TEXT NOT NULL DEFAULT '',
+    body        TEXT NOT NULL DEFAULT '',
+    created_by  TEXT REFERENCES users (id) ON DELETE SET NULL,
+    pinned      INTEGER NOT NULL DEFAULT 0,
+    created_at  INTEGER NOT NULL,
+    updated_at  INTEGER NOT NULL,
+    deleted_at  INTEGER,
+    archived_at INTEGER
+);
+
+INSERT INTO notes_sealed
+    (id, project_id, title, body, created_by, pinned, created_at, updated_at, deleted_at, archived_at)
+SELECT id, project_id, title, body, created_by, pinned, created_at, updated_at, deleted_at, archived_at
+FROM notes;
+
+DROP TRIGGER notes_fts_insert;
+DROP TRIGGER notes_fts_delete;
+DROP TRIGGER notes_fts_update;
+DROP TABLE notes_fts;
+DROP TABLE notes;
+
+ALTER TABLE notes_sealed RENAME TO notes;
+
+CREATE INDEX notes_by_project ON notes (project_id, deleted_at, updated_at DESC);
+CREATE INDEX notes_current ON notes (archived_at, deleted_at, updated_at DESC)
+    WHERE deleted_at IS NULL;

internal/store/notes.go

@@ -3,22 +3,35 @@ package store
 import (
 	"context"
 	"database/sql"
+	"fmt"
 	"regexp"
+	"sort"
 	"strings"
 	"time"
 )
 
 // Note is a page of Markdown belonging to a project, or to nobody.
 type Note struct {
-	ID        string     `json:"id"`
-	ProjectID string     `json:"project_id,omitempty"`
-	Title     string     `json:"title"`
-	Body      string     `json:"body"`
-	CreatedBy string     `json:"created_by,omitempty"`
-	Pinned    bool       `json:"pinned"`
-	CreatedAt time.Time  `json:"created_at"`
-	UpdatedAt time.Time  `json:"updated_at"`
-	DeletedAt *time.Time `json:"deleted_at,omitempty"`
+	ID         string     `json:"id"`
+	ProjectID  string     `json:"project_id,omitempty"`
+	Title      string     `json:"title"`
+	Body       string     `json:"body"`
+	CreatedBy  string     `json:"created_by,omitempty"`
+	Pinned     bool       `json:"pinned"`
+	CreatedAt  time.Time  `json:"created_at"`
+	UpdatedAt  time.Time  `json:"updated_at"`
+	DeletedAt  *time.Time `json:"deleted_at,omitempty"`
+	ArchivedAt *time.Time `json:"archived_at,omitempty"`
+
+	// Preview is the first of the body, for a list that shows one line of it.
+	//
+	// It exists so a list of twelve hundred notes does not have to carry twelve
+	// hundred whole notes: the list rendered 845 KB of JSON to show ninety
+	// characters of each. When it is set, Body is deliberately empty — see
+	// handleListNotes. That is not tidiness but a guard: an editor handed a
+	// truncated body would save the truncation, and the note would be cut down to
+	// its own preview.
+	Preview string `json:"preview,omitempty"`
 
 	// Links is what the body points at, read out of the text on the way in. Never
 	// set by a caller: it is derived, and a caller that set it would be describing
@@ -33,7 +46,16 @@ type NoteLink struct {
 }
 
 const noteColumns = `id, project_id, title, body, created_by, pinned,
-	created_at, updated_at, deleted_at`
+	created_at, updated_at, deleted_at, archived_at`
+
+// The same columns, said with the table in front of them.
+//
+// Needed only by the search, which joins `notes_fts` — and that table carries its
+// own `id` and `project_id` (UNINDEXED, so the row can be found from a match).
+// Unqualified, SQLite cannot tell which one is meant and refuses the query.
+const noteColumnsQualified = `notes.id, notes.project_id, notes.title, notes.body,
+	notes.created_by, notes.pinned, notes.created_at, notes.updated_at, notes.deleted_at,
+	notes.archived_at`
 
 // Reference syntax, and deliberately the same characters the quick-add parser
 // already uses. A `#Firma` in a note means the same project as a `#Firma` typed
@@ -69,6 +91,24 @@ func (db *DB) SaveNote(ctx context.Context, n *Note) error {
 	// it, and every caller gets the same answer because none of them can set it.
 	n.Title = firstLine(n.Body)
 
+	// Forseglet på vej ned, åbnet på vej op.
+	//
+	// Både titel og krop: titlen *er* første linje, så en forseglet krop ved siden
+	// af en læselig titel ville efterlade begyndelsen af hver eneste note i
+	// klartekst — og det er dér, folk skriver hvad noten handler om.
+	//
+	// `n` selv røres ikke. Kalderen har den note i hånden og læser videre i den
+	// bagefter; at forsegle den på plads ville give en note, hvis krop er
+	// ciffertekst, til den, der lige har gemt den.
+	title, err := db.sealValue(n.Title)
+	if err != nil {
+		return fmt.Errorf("seal note title: %w", err)
+	}
+	body, err := db.sealValue(n.Body)
+	if err != nil {
+		return fmt.Errorf("seal note body: %w", err)
+	}
+
 	return db.Tx(ctx, func(tx *sql.Tx) error {
 		var projectID any
 		if n.ProjectID != "" {
@@ -88,7 +128,7 @@ func (db *DB) SaveNote(ctx context.Context, n *Note) error {
 			    body = excluded.body,
 			    pinned = excluded.pinned,
 			    updated_at = excluded.updated_at`,
-			n.ID, projectID, n.Title, n.Body, createdBy, n.Pinned,
+			n.ID, projectID, title, body, createdBy, n.Pinned,
 			n.CreatedAt.Unix(), n.UpdatedAt.Unix()); err != nil {
 			return err
 		}
@@ -162,7 +202,7 @@ func firstLine(body string) string {
 func (db *DB) Note(ctx context.Context, id string) (*Note, error) {
 	row := db.QueryRowContext(ctx,
 		`SELECT `+noteColumns+` FROM notes WHERE id = ? AND deleted_at IS NULL`, id)
-	n, err := scanNote(row)
+	n, err := db.scanNote(row)
 	if err == sql.ErrNoRows {
 		return nil, nil
 	}
@@ -199,8 +239,23 @@ func (db *DB) LooseNotes(ctx context.Context, userID string) ([]Note, error) {
 // shared, and a list that dropped it at that moment would make sharing look like
 // deleting — which is exactly how it read before this existed.
 func (db *DB) AllNotes(ctx context.Context, userID string) ([]Note, error) {
-	return db.notesWhere(ctx, `
-		WHERE deleted_at IS NULL
+	return db.allNotes(ctx, userID, false)
+}
+
+// ArchivedNotes is the same list, the other way round: what has been put away.
+func (db *DB) ArchivedNotes(ctx context.Context, userID string) ([]Note, error) {
+	return db.allNotes(ctx, userID, true)
+}
+
+func (db *DB) allNotes(ctx context.Context, userID string, archived bool) ([]Note, error) {
+	// Archived notes are left out of the ordinary list rather than mixed in and
+	// greyed: the point of putting something away is not seeing it.
+	state := "notes.archived_at IS NULL"
+	if archived {
+		state = "notes.archived_at IS NOT NULL"
+	}
+	return db.notesJoin(ctx, noteColumnsQualified, `
+		WHERE `+state+` AND deleted_at IS NULL
 		  AND (
 		    (project_id IS NULL AND created_by = ?)
 		    OR project_id IN (
@@ -228,23 +283,104 @@ func (db *DB) NotesLinking(ctx context.Context, kind, targetID string) ([]Note,
 
 // SearchNotes matches title and body, in either spelling of a Danish word.
 func (db *DB) SearchNotes(ctx context.Context, query string, limit int) ([]Note, error) {
-	if strings.TrimSpace(query) == "" {
+	terms := SearchTerms(query)
+	if len(terms) == 0 {
 		return nil, nil
 	}
 	if limit <= 0 {
 		limit = 50
 	}
-	expr := MatchExprOver(query, "title", "body")
-	if expr == "" {
-		return nil, nil
+
+	// Hver note læses og åbnes.
+	//
+	// Der er ikke noget indeks at spørge længere, og det er med vilje: `notes_fts`
+	// og den genererede `fold`-søjle var begge klartekstkopier af kroppen, så en
+	// forseglet krop ved siden af dem ville have været teater — en fil ville stadig
+	// røbe hver note. Se 0027.
+	//
+	// Målt frem for gættet: tolv hundrede noter à 3,3 kB koster **33–43 ms** pr.
+	// søgning. Fladen venter 250 ms efter sidste tastetryk, før den spørger, så det
+	// ligger inde i pausen og ses ikke.
+	//
+	// Tiden går ikke med at åbne konvolutterne — fire megabyte AES-GCM er få
+	// millisekunder — men med at folde og småskrive hver krop for hver søgning.
+	// Det er lineært, så regn med omtrent 350 ms ved ti tusind noter, og dér er det
+	// den forkerte form. Svaret er da et *nøglet* indeks: en HMAC pr. ord under
+	// samme nøgle. Bemærk hvad det koster, når det bliver aktuelt — et nøglet
+	// indeks kan slå et helt ord op og ikke andet, så præfiks- og delordssøgningen,
+	// der findes i dag, ville falde bort.
+	//
+	// Papirkurven springes over i SQL frem for i løkken: det er den ene begrænsning,
+	// der kan bruge et indeks, og den fjerner alt hvad der er slettet uden at åbne
+	// det.
+	notes, err := db.notesWhere(ctx, `WHERE deleted_at IS NULL`)
+	if err != nil {
+		return nil, err
 	}
-	// By rowid, the way tasks do it: `id` is UNINDEXED in the table, so selecting
-	// on it would work and read the whole index to do it.
-	return db.notesWhere(ctx, `
-		WHERE deleted_at IS NULL
-		  AND rowid IN (SELECT rowid FROM notes_fts WHERE notes_fts MATCH ?)
-		ORDER BY updated_at DESC
-		LIMIT ?`, expr, limit)
+
+	type scored struct {
+		note  Note
+		score int
+	}
+	found := make([]scored, 0, limit)
+
+	for _, note := range notes {
+		score := scoreNote(note, terms)
+		if score == 0 {
+			continue
+		}
+		found = append(found, scored{note: note, score: score})
+	}
+
+	// Bedste først, og ved lige stand den senest rørte. Uden det andet led er
+	// rækkefølgen mellem to lige gode svar den, SQLite tilfældigvis gav dem, og den
+	// skifter mellem to søgninger på det samme.
+	sort.SliceStable(found, func(i, j int) bool {
+		if found[i].score != found[j].score {
+			return found[i].score > found[j].score
+		}
+		return found[i].note.UpdatedAt.After(found[j].note.UpdatedAt)
+	})
+
+	out := make([]Note, 0, min(limit, len(found)))
+	for i, f := range found {
+		if i >= limit {
+			break
+		}
+		out = append(out, f.note)
+	}
+	return out, nil
+}
+
+// scoreNote er hvor godt en note svarer på ordene, eller nul for slet ikke.
+//
+// Erstatter bm25, som forsvandt med indekset. Den er grovere, og det er værd at
+// vide hvor: bm25 vægter et ord efter hvor sjældent det er i hele samlingen, og
+// det kan ikke gøres uden en samling at tælle i. Det, der er beholdt, er den
+// halvdel, der betød mest — **titlen vejer ti gange kroppen**. Et ord i titlen er
+// hvad noten *handler om*; det samme ord én gang i en lang krop er en omtale.
+//
+// Alle ord skal findes. Det er den samme regel som før, hvor leddene var bundet
+// med AND.
+func scoreNote(note Note, terms []string) int {
+	title := FoldDanish(strings.ToLower(note.Title))
+	body := FoldDanish(strings.ToLower(note.Body))
+
+	score := 0
+	for _, term := range terms {
+		needle := FoldDanish(strings.ToLower(term))
+		inTitle := strings.Count(title, needle)
+		inBody := strings.Count(body, needle)
+		if inTitle == 0 && inBody == 0 {
+			// Ét ord, der ikke findes, er et nej for hele noten.
+			return 0
+		}
+		// Loft pr. ord, så en note, der nævner ordet fyrre gange, ikke slår en, der
+		// nævner alle ordene én gang. Det er den fejl, en ren optælling laver, og
+		// den er let at lave og svær at få øje på.
+		score += 10*min(inTitle, 3) + min(inBody, 5)
+	}
+	return score
 }
 
 // SetNoteTimes writes the dates a note had somewhere else.
@@ -278,6 +414,27 @@ func (db *DB) SetNoteTimes(ctx context.Context, id string, created, updated time
 	return err
 }
 
+// SetNoteArchived puts a note away, or takes it back out.
+//
+// Separate from DeleteNote because they mean different things and the difference
+// is the whole feature: the trash says "this was a mistake, and in thirty days it
+// is gone", and the archive says "this is finished, and I want to keep it". A
+// twelve-hundred-note import is mostly the second, and without somewhere to put
+// them the list is the whole archive forever.
+//
+// `updated_at` is deliberately not touched. Archiving is not editing, and a list
+// sorted by when things were last written should not be reordered by somebody
+// tidying up.
+func (db *DB) SetNoteArchived(ctx context.Context, id string, archived bool) error {
+	var at any
+	if archived {
+		at = time.Now().Unix()
+	}
+	_, err := db.ExecContext(ctx,
+		`UPDATE notes SET archived_at = ? WHERE id = ? AND deleted_at IS NULL`, at, id)
+	return err
+}
+
 // DeleteNote puts it in the trash. Nothing here removes a row.
 func (db *DB) DeleteNote(ctx context.Context, id string) error {
 	_, err := db.ExecContext(ctx,
@@ -295,7 +452,13 @@ func (db *DB) RestoreNote(ctx context.Context, id string) error {
 }
 
 func (db *DB) notesWhere(ctx context.Context, where string, args ...any) ([]Note, error) {
-	rows, err := db.QueryContext(ctx, `SELECT `+noteColumns+` FROM notes `+where, args...)
+	return db.notesJoin(ctx, noteColumns, where, args...)
+}
+
+// notesJoin is notesWhere with the column list spelled out, for the one caller
+// that joins another table and must therefore say which `id` it means.
+func (db *DB) notesJoin(ctx context.Context, columns, where string, args ...any) ([]Note, error) {
+	rows, err := db.QueryContext(ctx, `SELECT `+columns+` FROM notes `+where, args...)
 	if err != nil {
 		return nil, err
 	}
@@ -303,7 +466,7 @@ func (db *DB) notesWhere(ctx context.Context, where string, args ...any) ([]Note
 
 	var out []Note
 	for rows.Next() {
-		n, err := scanNote(rows)
+		n, err := db.scanNote(rows)
 		if err != nil {
 			return nil, err
 		}
@@ -331,14 +494,18 @@ func (db *DB) linksOf(ctx context.Context, noteID string) ([]NoteLink, error) {
 	return out, rows.Err()
 }
 
-func scanNote(row scanner) (Note, error) {
+// scanNote læser en række og åbner det, der er forseglet.
+//
+// En metode og ikke en funktion, fordi nøglen sidder på `db`. Det er den eneste
+// grund; alt andet i den er den samme aflæsning som før.
+func (db *DB) scanNote(row scanner) (Note, error) {
 	var n Note
 	var projectID, createdBy sql.NullString
 	var created, updated int64
-	var deleted sql.NullInt64
+	var deleted, archived sql.NullInt64
 
 	if err := row.Scan(&n.ID, &projectID, &n.Title, &n.Body, &createdBy,
-		&n.Pinned, &created, &updated, &deleted); err != nil {
+		&n.Pinned, &created, &updated, &deleted, &archived); err != nil {
 		return n, err
 	}
 	n.ProjectID = projectID.String
@@ -349,5 +516,22 @@ func scanNote(row scanner) (Note, error) {
 		t := time.Unix(deleted.Int64, 0)
 		n.DeletedAt = &t
 	}
+	if archived.Valid {
+		t := time.Unix(archived.Int64, 0)
+		n.ArchivedAt = &t
+	}
+
+	// Åbnet her, ét sted, så ingen kalder kan komme til at læse ciffertekst.
+	//
+	// En værdi skrevet før der var en nøgle kommer tilbage som sig selv — `v1:`
+	// findes netop for at kunne læse en blandet søjle — så en installation, der
+	// endnu ikke har fået sin bagudfyldning, virker imens.
+	for _, field := range []*string{&n.Title, &n.Body} {
+		plain, err := db.unsealValue(*field)
+		if err != nil {
+			return n, fmt.Errorf("open note %s: %w", n.ID, err)
+		}
+		*field = plain
+	}
 	return n, nil
 }

internal/store/notes_test.go

@@ -333,3 +333,103 @@ func TestATagFindsItsProjectWhateverTheCase(t *testing.T) {
 		t.Error("a note title did not match itself")
 	}
 }
+
+// En søgning skal svare med det, der passer bedst, øverst.
+//
+// Den sorterede efter `updated_at`, hvilket var usynligt med ti noter og
+// ubrugeligt med tolv hundrede: de blev alle importeret samme aften, så datoen
+// siger ingenting, og loftet skar så listen af ved halvtreds tilfældige rækker.
+// At søge "Tesla" svarede med en note om nginx først og noten, der hedder "Tesla
+// Model Y", bagefter — hvilket læses som en søgning, der ikke virker, fordi det
+// fra brugerens plads er præcis dét.
+func TestSearchAnswersWithTheBestMatchFirst(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	// Skrevet i den rækkefølge, der ville give det forkerte svar: den, der kun
+	// nævner ordet i kroppen, gemmes sidst og har derfor den nyeste dato.
+	for _, body := range []string{
+		"# Tesla Model Y\n\nBilen selv.\n",
+		"# Ladestandere\n\nNoter om Tesla og andre mærker.\n",
+		"# nginx proxy manager backup\n\nEn konfiguration. Tesla nævnes i forbifarten.\n",
+	} {
+		if err := db.SaveNote(ctx, &Note{CreatedBy: userID, Body: body}); err != nil {
+			t.Fatal(err)
+		}
+	}
+
+	found, err := db.SearchNotes(ctx, "Tesla", 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(found) != 3 {
+		t.Fatalf("fandt %d noter, ventede 3", len(found))
+	}
+	if found[0].Title != "Tesla Model Y" {
+		t.Errorf("øverst stod %q; den note, der hedder det, man søgte efter, skal stå først", found[0].Title)
+	}
+}
+
+// A word inside a word.
+//
+// Reported as "søgning har ikke det store held": a note holding
+// `gamerule keepInventory true` could not be found by typing "keep inventory".
+// The index is built of tokens, and `keepInventory` is one token — so the two
+// words are nowhere, and the search says there is nothing while the note is right
+// there.
+//
+// A decade of notes is full of these: commands, identifiers, file names. Every one
+// of them is several words the index thinks is one.
+func TestAWordInsideAWordIsStillFound(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	if err := db.SaveNote(ctx, &Note{
+		CreatedBy: userID,
+		Body:      "Minecraft Bedrock\n\ngamerule keepInventory true",
+	}); err != nil {
+		t.Fatal(err)
+	}
+	// A second note, so a fallback that simply returns everything would be caught.
+	if err := db.SaveNote(ctx, &Note{
+		CreatedBy: userID,
+		Body:      "Indkøb\n\nKaffe og filtre",
+	}); err != nil {
+		t.Fatal(err)
+	}
+
+	for _, q := range []string{
+		"keep inventory", // the two words, which is how somebody looks for it
+		"inventory",      // the tail on its own — no prefix query can reach this
+		"keepinventory",  // written as one, which the index does hold
+		"KEEP INVENTORY", // and case is not a question anybody is asking
+		"vacuum",         // not in either note
+	} {
+		found, err := db.SearchNotes(ctx, q, 10)
+		if err != nil {
+			t.Fatalf("%q: %v", q, err)
+		}
+		want := 1
+		if q == "vacuum" {
+			want = 0
+		}
+		if len(found) != want {
+			t.Errorf("%q found %d notes, want %d", q, len(found), want)
+			continue
+		}
+		if want == 1 && found[0].Title != "Minecraft Bedrock" {
+			t.Errorf("%q found %q", q, found[0].Title)
+		}
+	}
+
+	// The fast path is still the fast path: a word the index does hold must not be
+	// answered by the scan, or the ranking that was built for twelve hundred notes
+	// is quietly gone.
+	found, err := db.SearchNotes(ctx, "Bedrock", 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(found) != 1 {
+		t.Fatalf("an ordinary word found %d notes", len(found))
+	}
+}

internal/store/notifications.go

@@ -140,11 +140,17 @@ func (db *DB) SavePushSubscription(ctx context.Context, sub *PushSubscription) e
 	if sub.ID == "" {
 		sub.ID = NewID()
 	}
+	// The upsert no longer moves the row between accounts.
+	//
+	// `user_id = excluded.user_id` was in the DO UPDATE, so anybody who learned
+	// another person's endpoint could claim it by subscribing with it: the
+	// victim's device would then receive the claimant's notifications and none of
+	// its own. A re-subscribe from the same person still refreshes the keys, which
+	// is what the upsert is actually for.
 	_, err := db.ExecContext(ctx,
 		`INSERT INTO push_subscriptions (id, user_id, endpoint, p256dh, auth, user_agent, created_at)
 		 VALUES (?, ?, ?, ?, ?, ?, ?)
 		 ON CONFLICT (endpoint) DO UPDATE SET
-		     user_id = excluded.user_id,
 		     p256dh = excluded.p256dh,
 		     auth = excluded.auth,
 		     user_agent = excluded.user_agent,
@@ -175,7 +181,21 @@ func (db *DB) ListPushSubscriptions(ctx context.Context, userID string) ([]PushS
 	return out, rows.Err()
 }
 
-func (db *DB) DeletePushSubscription(ctx context.Context, endpoint string) error {
+// DeletePushSubscription removes one of this person's devices.
+//
+// Scoped by user as well as by endpoint. It was endpoint alone, which meant
+// anybody who learned somebody else's endpoint could silently unsubscribe their
+// phone — and the endpoint is the only thing about a subscription that ever
+// leaves the browser.
+func (db *DB) DeletePushSubscription(ctx context.Context, userID, endpoint string) error {
+	_, err := db.ExecContext(ctx,
+		`DELETE FROM push_subscriptions WHERE endpoint = ? AND user_id = ?`, endpoint, userID)
+	return err
+}
+
+// dropPushEndpoint removes a subscription the push service has declared dead.
+// Unexported so the only *request* path to a delete stays the scoped one.
+func (db *DB) dropPushEndpoint(ctx context.Context, endpoint string) error {
 	_, err := db.ExecContext(ctx, `DELETE FROM push_subscriptions WHERE endpoint = ?`, endpoint)
 	return err
 }
@@ -186,7 +206,10 @@ func (db *DB) DeletePushSubscription(ctx context.Context, endpoint string) error
 // silently unsubscribing somebody's phone.
 func (db *DB) RecordPushFailure(ctx context.Context, endpoint string, permanent bool) error {
 	if permanent {
-		return db.DeletePushSubscription(ctx, endpoint)
+		// Not scoped by user, and that is right here: this is the push service
+		// saying the endpoint is gone, not a person asking for something. There is
+		// no caller to check against — the caller is the delivery job.
+		return db.dropPushEndpoint(ctx, endpoint)
 	}
 	_, err := db.ExecContext(ctx,
 		`UPDATE push_subscriptions SET failures = failures + 1 WHERE endpoint = ?`, endpoint)
@@ -289,6 +312,11 @@ var secretFields = map[string]bool{
 	"access_token":  true,
 	"password":      true,
 	"client_secret": true,
+	// The AI provider's key. Missed when this list was written, and it is exactly
+	// the same kind of thing as the rest: a live credential, in a settings blob,
+	// in a database anybody with a backup file can read. Existing rows convert
+	// themselves the next time the settings are saved.
+	"api_key": true,
 }
 
 // seal returns a copy of values with the secret fields sealed. Without a key it
@@ -363,11 +391,15 @@ func (db *DB) SetUserSettings(ctx context.Context, userID, scope string, values
 	if err != nil {
 		return err
 	}
+	// On (user_id, scope), which is the key since 0025. It used to be on `user_id`
+	// alone and set `scope = excluded.scope` — so writing one scope did not add a
+	// row beside the others, it took theirs and renamed it. The `ai`, `gmail` and
+	// `calendar` settings each deleted the other two, and the write reported
+	// success every time.
 	_, err = db.ExecContext(ctx, `
 		INSERT INTO user_settings (user_id, scope, values_json, updated_at)
 		VALUES (?, ?, ?, ?)
-		ON CONFLICT (user_id) DO UPDATE SET
-		    scope = excluded.scope,
+		ON CONFLICT (user_id, scope) DO UPDATE SET
 		    values_json = excluded.values_json,
 		    updated_at = excluded.updated_at`,
 		userID, scope, string(raw), time.Now().Unix())

internal/store/projects.go

@@ -646,6 +646,35 @@ func (db *DB) ReorderProjects(ctx context.Context, userID string, ids []string)
 	})
 }
 
+// ReorderSections writes the order of one project's sections, top to bottom.
+//
+// The whole list at once, the same as ReorderProjects and for the same reasons: a
+// project has a handful of sections, so writing 0, 1, 2 … is one small
+// transaction that cannot run out of precision, and a list is impossible to get
+// into the half-ordered state a sequence of individual moves leaves behind when
+// one of them fails.
+//
+// Scoped to the project rather than to a user. A section has no owner of its own —
+// it belongs to a project, and whether the caller may rearrange it is a question
+// about that project, which the handler has already asked. Ids belonging to some
+// other project are skipped by the WHERE clause rather than refused, so a stale
+// list from a client that has just moved a section elsewhere cannot rearrange a
+// project the caller was never looking at.
+func (db *DB) ReorderSections(ctx context.Context, projectID string, ids []string) error {
+	return db.Tx(ctx, func(tx *sql.Tx) error {
+		now := time.Now().Unix()
+		for i, id := range ids {
+			if _, err := tx.ExecContext(ctx,
+				`UPDATE sections SET sort_order = ?, updated_at = ?
+				 WHERE id = ? AND project_id = ? AND deleted_at IS NULL`,
+				float64(i), now, id, projectID); err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+}
+
 // SectionByName resolves a "/name" from quick add within one project.
 //
 // Case-insensitive, and scoped to the project the task is landing in — a section

internal/store/search.go

@@ -20,6 +20,13 @@ func FoldDanish(s string) string {
 	return r.Replace(s)
 }
 
+// The ceilings on a search. See MatchExprOver for why they exist; both are far
+// above anything a person types and far below anything that costs.
+const (
+	maxSearchInput = 512
+	maxSearchTerms = 16
+)
+
 // MatchExpr turns what someone typed into the search box into an FTS5 MATCH
 // expression.
 //
@@ -51,6 +58,23 @@ func MatchExprOver(query string, columns ...string) string {
 		return ""
 	}
 
+	// Capped, because the expression is built from the input and then run inside
+	// SQLite on one of a handful of pooled connections.
+	//
+	// Every term becomes roughly fifty bytes of FTS5, and every one of them is a
+	// prefix query — which is what makes ⌘K feel like it searches as you type, and
+	// also what makes each one expensive. A quarter of a megabyte pasted into the
+	// box became a two-megabyte MATCH expression and half a minute inside SQLite,
+	// holding a connection the whole time. A few of those at once and the database
+	// is unavailable to everybody on the instance, logins included.
+	//
+	// Sixteen terms is far past anything a person types into a search box and
+	// nowhere near enough to hurt. Silently, because a message about a limit
+	// nobody has reached on purpose is noise.
+	if len(terms) > maxSearchTerms {
+		terms = terms[:maxSearchTerms]
+	}
+
 	var b strings.Builder
 	for i, term := range terms {
 		if i > 0 {
@@ -66,10 +90,29 @@ func MatchExprOver(query string, columns ...string) string {
 	return b.String()
 }
 
+// SearchTerms is the words a query asks about, capped the same way the MATCH
+// expression is.
+//
+// Exported because two searches need the same list: the FTS expression is built
+// from it, and so is the substring arm that finds a word living inside another
+// word. Two tokenizers would be two answers to "what did they type".
+func SearchTerms(query string) []string {
+	terms := tokenize(query)
+	if len(terms) > maxSearchTerms {
+		terms = terms[:maxSearchTerms]
+	}
+	return terms
+}
+
 // tokenize splits on everything that is not a letter or a digit. Splitting on
 // punctuation rather than whitespace alone means "q3/2026" searches for "q3" and
 // "2026" instead of for a term FTS5's own tokenizer would never have indexed.
 func tokenize(query string) []string {
+	// Trimmed before it is split, so a megabyte of text is not walked rune by rune
+	// only to have all but sixteen of the results thrown away.
+	if len(query) > maxSearchInput {
+		query = query[:maxSearchInput]
+	}
 	fields := strings.FieldsFunc(query, func(r rune) bool {
 		return !unicode.IsLetter(r) && !unicode.IsDigit(r)
 	})

internal/store/secrets_test.go

@@ -124,3 +124,272 @@ func sealedStore(t *testing.T) (*DB, string) {
 	}
 	return db, user.ID
 }
+
+// Three integrations, one person, and none of them may delete the others.
+//
+// `user_settings` had `user_id` as its whole primary key and an upsert that
+// rewrote `scope` on conflict, so every person had exactly one settings row and
+// the last write won the lot: saving an AI provider deleted the Gmail connection,
+// and connecting a calendar deleted the AI provider. The write reported success
+// each time, which is why it was reported as "the AI settings are not saved"
+// rather than as "connecting the calendar deleted them".
+//
+// Written against the store rather than the handlers, because the shape was the
+// bug: every caller was correct on its own.
+func TestOnePersonCanHaveSettingsForMoreThanOneThing(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	want := map[string]map[string]any{
+		"ai":       {"provider": "anthropic", "model": "claude-sonnet-5", "api_key": "sk-ant-xxx"},
+		"gmail":    {"refresh_token": "1//0-gmail", "label": "Verdande"},
+		"calendar": {"refresh_token": "1//0-calendar", "calendar_id": "primary"},
+	}
+	// In this order on purpose: the AI settings are written first and read last,
+	// so a scope that steals the row is caught by the one written earliest.
+	for _, scope := range []string{"ai", "gmail", "calendar"} {
+		if err := db.SetUserSettings(ctx, userID, scope, want[scope]); err != nil {
+			t.Fatalf("save %s: %v", scope, err)
+		}
+	}
+
+	for _, scope := range []string{"ai", "gmail", "calendar"} {
+		got, err := db.UserSettings(ctx, userID, scope)
+		if err != nil {
+			t.Fatalf("read %s: %v", scope, err)
+		}
+		for key, value := range want[scope] {
+			if got[key] != value {
+				t.Errorf("%s.%s is %v, want %v", scope, key, got[key], value)
+			}
+		}
+	}
+
+	// And a second write to one scope updates it rather than adding another row —
+	// the upsert still has to upsert.
+	if err := db.SetUserSettings(ctx, userID, "ai", map[string]any{"provider": "openai"}); err != nil {
+		t.Fatal(err)
+	}
+	var rows int
+	if err := db.QueryRowContext(ctx,
+		`SELECT count(*) FROM user_settings WHERE user_id = ?`, userID).Scan(&rows); err != nil {
+		t.Fatal(err)
+	}
+	if rows != 3 {
+		t.Errorf("the person has %d settings rows, want 3", rows)
+	}
+	back, err := db.UserSettings(ctx, userID, "ai")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if back["provider"] != "openai" {
+		t.Errorf("the second write did not land: %v", back["provider"])
+	}
+}
+
+// The upgrade path, which is the one nobody runs: an existing database with a row
+// in the old shape, opened by a binary that carries 0025.
+//
+// Every other test here starts empty and runs all the migrations in order, which
+// exercises the migration's SQL but not its *situation*. The situation is a table
+// that already holds somebody's settings, and the question is whether they come
+// through it and whether a second scope can be written afterwards.
+func TestSettingsSurviveTheUpgradeToPerScopeRows(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "test.db")
+
+	db, err := Open(path)
+	if err != nil {
+		t.Fatal(err)
+	}
+	ctx := context.Background()
+	user := &User{Email: "kristian@example.dk", Name: "Kristian"}
+	if err := db.CreateUser(ctx, user, "Indbakke"); err != nil {
+		t.Fatal(err)
+	}
+
+	// Wind the database back to how v0.33.3 left it: user_id as the whole primary
+	// key, and 0025 not yet applied.
+	for _, stmt := range []string{
+		`PRAGMA foreign_keys = OFF`,
+		`CREATE TABLE us_old (
+			user_id     TEXT PRIMARY KEY REFERENCES users (id) ON DELETE CASCADE,
+			scope       TEXT NOT NULL DEFAULT 'general',
+			values_json TEXT NOT NULL DEFAULT '{}',
+			updated_at  INTEGER NOT NULL
+		)`,
+		`INSERT INTO us_old (user_id, scope, values_json, updated_at)
+		 SELECT user_id, scope, values_json, updated_at FROM user_settings`,
+		`DROP TABLE user_settings`,
+		`ALTER TABLE us_old RENAME TO user_settings`,
+		`DELETE FROM schema_migrations WHERE name = '0025_user_settings_per_scope.sql'`,
+		`PRAGMA foreign_keys = ON`,
+	} {
+		if _, err := db.ExecContext(ctx, stmt); err != nil {
+			t.Fatalf("rewind (%s): %v", stmt, err)
+		}
+	}
+
+	// A Gmail connection, written the way the old upsert wrote one.
+	if _, err := db.ExecContext(ctx, `
+		INSERT INTO user_settings (user_id, scope, values_json, updated_at)
+		VALUES (?, 'gmail', '{"label":"Verdande"}', 0)`, user.ID); err != nil {
+		t.Fatal(err)
+	}
+	if err := db.Close(); err != nil {
+		t.Fatal(err)
+	}
+
+	// And now the new binary opens it.
+	db, err = Open(path)
+	if err != nil {
+		t.Fatalf("the upgrade refused to open the database: %v", err)
+	}
+	defer db.Close()
+
+	gmail, err := db.UserSettings(ctx, user.ID, "gmail")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if gmail["label"] != "Verdande" {
+		t.Errorf("the Gmail settings did not survive the migration: %v", gmail)
+	}
+
+	// The thing that was reported: set up the AI provider, and have it still be
+	// there afterwards.
+	if err := db.SetUserSettings(ctx, user.ID, "ai", map[string]any{
+		"provider": "anthropic", "model": "claude-sonnet-5",
+	}); err != nil {
+		t.Fatalf("saving the AI provider after the upgrade: %v", err)
+	}
+	ai, err := db.UserSettings(ctx, user.ID, "ai")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if ai["provider"] != "anthropic" {
+		t.Errorf("the AI provider is %v after the upgrade, want anthropic", ai["provider"])
+	}
+	again, err := db.UserSettings(ctx, user.ID, "gmail")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if again["label"] != "Verdande" {
+		t.Errorf("saving the AI provider took the Gmail settings: %v", again)
+	}
+}
+
+// Det, en note siger, må ikke kunne læses i filen, som en sikkerhedskopi kopierer.
+//
+// Grunden er den samme som for postkassers kodeord: en sikkerhedskopi kan hentes
+// gennem fladen. Noter er dér, folk skriver kodeord og API-nøgler ned — det ved vi,
+// fordi der ligger sådan nogle i den installation, det her blev bygget til.
+func TestANoteIsNotStoredInTheClear(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	const secretLine = "gamerule keepInventory true"
+	n := &Note{CreatedBy: userID, Body: "Minecraft Bedrock\n\n" + secretLine}
+	if err := db.SaveNote(ctx, n); err != nil {
+		t.Fatal(err)
+	}
+
+	var title, body string
+	if err := db.QueryRowContext(ctx,
+		`SELECT title, body FROM notes WHERE id = ?`, n.ID).Scan(&title, &body); err != nil {
+		t.Fatal(err)
+	}
+	if strings.Contains(body, secretLine) {
+		t.Errorf("kroppen står i klartekst i rækken: %s", body)
+	}
+	// Og titlen med. Titlen *er* første linje, så en forseglet krop ved siden af en
+	// læselig titel ville efterlade begyndelsen af hver eneste note i klartekst.
+	if strings.Contains(title, "Minecraft") {
+		t.Errorf("titlen står i klartekst i rækken: %s", title)
+	}
+
+	// Kalderens egen note er urørt: den har den i hånden og læser videre i den.
+	if n.Body != "Minecraft Bedrock\n\n"+secretLine {
+		t.Errorf("gemningen ændrede kalderens note: %q", n.Body)
+	}
+
+	// Læst tilbage er den sig selv igen.
+	back, err := db.Note(ctx, n.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if back.Body != n.Body || back.Title != "Minecraft Bedrock" {
+		t.Errorf("noten kom anderledes tilbage: %+v", back)
+	}
+
+	// Og søgningen virker stadig — den åbner teksten nu, hvor den før spurgte et
+	// indeks, der var en klartekstkopi af den.
+	found, err := db.SearchNotes(ctx, "keep inventory", 10)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if len(found) != 1 {
+		t.Fatalf("søgningen fandt %d noter i en forseglet samling", len(found))
+	}
+}
+
+// Noter skrevet før nøglen fandtes skal lukkes, ikke efterlades.
+//
+// Uden en bagudfyldning ville kroppene blive forseglet efterhånden som de bliver
+// *skrevet* — og et arkiv på tolv hundrede importerede noter, som ingen redigerer
+// igen, ville aldrig blive det. Det er præcis dét arkiv, der er værd at forsegle.
+func TestOldNotesAreSealedOnStartup(t *testing.T) {
+	db, userID := sealedStore(t)
+	ctx := context.Background()
+
+	// Skrevet uden om Go-laget, som en note fra før 0027 ligger.
+	if _, err := db.ExecContext(ctx, `
+		INSERT INTO notes (id, title, body, created_by, created_at, updated_at)
+		VALUES ('n-gammel', 'Immich DB', 'Immich DB\nkodeord: hemmeligt', ?, 100, 200)`,
+		userID); err != nil {
+		t.Fatal(err)
+	}
+
+	sealed, err := db.SealNotes(ctx)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if sealed != 1 {
+		t.Fatalf("%d noter blev lukket, want 1", sealed)
+	}
+
+	var body string
+	var updated int64
+	if err := db.QueryRowContext(ctx,
+		`SELECT body, updated_at FROM notes WHERE id = 'n-gammel'`).Scan(&body, &updated); err != nil {
+		t.Fatal(err)
+	}
+	if strings.Contains(body, "hemmeligt") {
+		t.Errorf("den gamle note ligger stadig åben: %s", body)
+	}
+	// `updated_at` er ikke rørt. Det betyder "hvornår skrev dette program noten
+	// sidst", og at lukke en konvolut om den er ikke at skrive den — en
+	// bagudfyldning, der stemplede tolv hundrede noter med i aften, ville kaste den
+	// rækkefølge væk, importen brugte en dag på at hente hjem.
+	if updated != 200 {
+		t.Errorf("bagudfyldningen flyttede updated_at til %d", updated)
+	}
+
+	// Og den kan læses.
+	back, err := db.Note(ctx, "n-gammel")
+	if err != nil {
+		t.Fatal(err)
+	}
+	if back == nil || !strings.Contains(back.Body, "hemmeligt") {
+		t.Fatalf("den gamle note kunne ikke læses tilbage: %+v", back)
+	}
+
+	// Kørt igen gør den ingenting: en note, der allerede er lukket, kendes på sit
+	// præfiks.
+	again, err := db.SealNotes(ctx)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if again != 0 {
+		t.Errorf("anden kørsel lukkede %d noter om igen", again)
+	}
+}

internal/store/store.go

@@ -38,6 +38,81 @@ type DB struct {
 // misplaced key look like a corrupt disk.
 func (db *DB) UseSecrets(box *secret.Box) { db.box = box }
 
+// SealNotes lukker de noter, der stadig ligger åbne.
+//
+// Kaldes én gang ved opstart, efter nøglen er sat. Uden den ville kroppene blive
+// forseglet efterhånden som de bliver *skrevet* — sådan som instansens
+// indstillinger gør det — og for et arkiv på tolv hundrede importerede noter, som
+// ingen redigerer igen, betyder det aldrig. En bagudfyldning er den eneste måde,
+// den her ændring rent faktisk beskytter noget på.
+//
+// Idempotent: en note, der allerede er forseglet, kendes på sit præfiks og røres
+// ikke. Så den kan køre ved hver opstart uden at koste andet end et gennemløb af
+// id'er og en længde.
+//
+// Skriver med SQL frem for gennem SaveNote. SaveNote udleder titlen af kroppen og
+// bygger links om, og begge dele er allerede rigtige her — det, der skal ske, er
+// alene, at de to søjler skal skiftes ud med sig selv, lukket.
+func (db *DB) SealNotes(ctx context.Context) (int, error) {
+	if db.box == nil {
+		return 0, nil
+	}
+
+	rows, err := db.QueryContext(ctx, `SELECT id, title, body FROM notes`)
+	if err != nil {
+		return 0, err
+	}
+	type open struct{ id, title, body string }
+	var pending []open
+	for rows.Next() {
+		var o open
+		if err := rows.Scan(&o.id, &o.title, &o.body); err != nil {
+			rows.Close()
+			return 0, err
+		}
+		if secret.IsSealed(o.body) && (o.title == "" || secret.IsSealed(o.title)) {
+			continue
+		}
+		pending = append(pending, o)
+	}
+	rows.Close()
+	if err := rows.Err(); err != nil {
+		return 0, err
+	}
+	if len(pending) == 0 {
+		return 0, nil
+	}
+
+	// Hele bunken i én transaktion. En halvt forseglet tabel er ikke i stykker —
+	// præfikset gør, at begge slags kan læses — men en afbrudt kørsel, der skal
+	// tages om, er lettere at ræsonnere om, når den enten skete eller ikke gjorde.
+	err = db.Tx(ctx, func(tx *sql.Tx) error {
+		for _, o := range pending {
+			title, err := db.sealValue(o.title)
+			if err != nil {
+				return err
+			}
+			body, err := db.sealValue(o.body)
+			if err != nil {
+				return err
+			}
+			// `updated_at` røres ikke. Det betyder "hvornår skrev dette program noten
+			// sidst", og at lukke en konvolut om den er ikke at skrive den — en
+			// bagudfyldning, der stemplede tolv hundrede noter med i aften, ville
+			// kaste den rækkefølge væk, importen brugte en dag på at hente hjem.
+			if _, err := tx.ExecContext(ctx,
+				`UPDATE notes SET title = ?, body = ? WHERE id = ?`, title, body, o.id); err != nil {
+				return err
+			}
+		}
+		return nil
+	})
+	if err != nil {
+		return 0, err
+	}
+	return len(pending), nil
+}
+
 // dsn is the connection string.
 //
 // SQLite is a single file, but it is being driven by an HTTP server with concurrent

internal/store/tasks.go

@@ -166,15 +166,42 @@ func (db *DB) ListTasks(ctx context.Context, userID string, f TaskFilter) ([]Tas
 		// labels are short, and their names are typed by the person searching for
 		// them, so there is no spelling to be generous about. The diacritic
 		// folding that matters for task text does not earn a second index here.
+		//
+		// The third arm is a word inside a word. An index is built of tokens, and
+		// `SpeedAdmin` is one of them — so "speed admin" asks for two words that are
+		// nowhere. A prefix query gets half of it (`speed*` matches, `admin*` cannot)
+		// and the two are joined with AND, so the half that works is thrown away.
+		//
+		// Notes got the same arm, and for the same reason. Having it in one of the
+		// two would be worse than having it in neither: "it finds it in my notes but
+		// not in my tasks" is a bug report waiting, and the answer to it would be
+		// this line.
+		//
+		// It scans, and it only ever runs when somebody typed something in the box —
+		// Today and Kommende do not set `Search`. Every term has to appear, so the
+		// arms are ANDed with each other while the whole thing is ORed with the two
+		// above.
+		inside := make([]string, 0, maxSearchTerms)
+		insideArgs := make([]any, 0, maxSearchTerms)
+		for _, term := range SearchTerms(f.Search) {
+			inside = append(inside, "t.fold LIKE ?")
+			insideArgs = append(insideArgs, "%"+FoldDanish(term)+"%")
+		}
+		text := ""
+		if len(inside) > 0 {
+			text = " OR (" + strings.Join(inside, " AND ") + ")"
+		}
+
 		where = append(where, `(
 			t.rowid IN (SELECT rowid FROM tasks_fts WHERE tasks_fts MATCH ?)
 			OR t.id IN (
 				SELECT tl.task_id FROM task_labels tl
 				JOIN labels l ON l.id = tl.label_id
 				WHERE lower(l.name) LIKE ?
-			)
+			)`+text+`
 		)`)
 		args = append(args, expr, "%"+strings.ToLower(strings.TrimSpace(f.Search))+"%")
+		args = append(args, insideArgs...)
 	}
 	if f.FilterSQL != "" {
 		where = append(where, "("+f.FilterSQL+")")

internal/store/users.go

@@ -16,6 +16,11 @@ var (
 	ErrNotFound      = errors.New("store: not found")
 	ErrEmailInUse    = errors.New("store: that email address already has an account")
 	ErrInviteInvalid = errors.New("store: invite is not valid")
+	// ErrDuplicate er en unik nøgle, der blev brudt, oversat til noget en handler
+	// kan spørge om. Drivereren udstiller ikke sin egen fejltype brugbart, og en
+	// handler, der matcher på tekst, er en handler, der går i stykker den dag
+	// driveren omformulerer sig.
+	ErrDuplicate = errors.New("store: already exists")
 )
 
 type User struct {
@@ -129,7 +134,14 @@ func (db *DB) scanUser(ctx context.Context, query string, args ...any) (*User, e
 		return nil, err
 	}
 
-	u.TOTPSecret = secret.String
+	// Opened on the way out. A row written before the key existed is returned as
+	// itself — see secret.Unseal — so this does not need a migration; a secret
+	// converts the next time it is written.
+	if plain, err := db.unsealValue(secret.String); err == nil {
+		u.TOTPSecret = plain
+	} else {
+		u.TOTPSecret = secret.String
+	}
 	u.TOTPEnabled = totpEnabled == 1
 	u.IsAdmin = isAdmin == 1
 	// A malformed value is read as "nothing folded" rather than as an error: the
@@ -433,9 +445,20 @@ func (db *DB) UpdateProfile(ctx context.Context, userID, name, timezone, locale
 }
 
 func (db *DB) SetTOTPSecret(ctx context.Context, userID, secret string, enabled bool) error {
+	// Sealed, like a mailbox password and for the same reason: the copies, not the
+	// server. A TOTP secret in a downloadable backup lets whoever holds the copy
+	// generate valid codes for every account on the instance — which turns the
+	// second factor from a defence into a file somebody can take home.
+	//
+	// It was missed when internal/secret was written, along with three other
+	// values; see SECURITY.md.
 	var value any
 	if secret != "" {
-		value = secret
+		sealed, err := db.sealValue(secret)
+		if err != nil {
+			return err
+		}
+		value = sealed
 	}
 	_, err := db.ExecContext(ctx,
 		`UPDATE users SET totp_secret = ?, totp_enabled = ?, updated_at = ? WHERE id = ?`,

mkdocs.yml

@@ -50,10 +50,12 @@ nav:
   - Using it:
       - Signing in: signing-in.md
       - Quick add: quick-add.md
+      - Notes: notes.md
       - Filters: filters.md
       - Importing from Todoist: import.md
   - Connecting things:
-      - Calendar and CalDAV: caldav.md
+      - Calendar: calendar.md
+      - Feeds and CalDAV: caldav.md
       - Mail to task: mail.md
       - Mailboxes: mailboxes.md
       - Claude (MCP): mcp.md

site/index.html

@@ -6,12 +6,12 @@
 		<title>verdande — self-hosted tasks, shared</title>
 		<meta
 			name="description"
-			content="Self-hosted tasks and projects, shared. One Go binary with SQLite inside it. Your mail as an inbox, CalDAV, passkeys, and an MCP connector for Claude."
+			content="Self-hosted tasks, notes and projects, shared. One Go binary with SQLite inside it. Your mail as an inbox, a CalDAV server, notes in Markdown, passkeys, and an MCP connector for Claude."
 		/>
 		<meta name="theme-color" content="#111312" />
 
 		<meta property="og:title" content="verdande" />
-		<meta property="og:description" content="Self-hosted tasks and projects, shared. One binary, SQLite inside it." />
+		<meta property="og:description" content="Self-hosted tasks, notes and projects, shared. One binary, SQLite inside it." />
 		<meta property="og:type" content="website" />
 		<meta property="og:image" content="screenshots/today.webp" />
 		<meta property="og:image:alt" content="The Today view: projects and filters beside a list of what is due." />
@@ -76,10 +76,10 @@
 			}
 
 			.theme-toggle {
-				position: fixed;
-				top: 1rem;
-				right: 1rem;
-				z-index: 10;
+				/* In the bar now rather than floating over the corner. Fixed, it sat
+				   on top of whatever happened to scroll under it — on a narrow window
+				   that was the first screenshot's own corner. */
+				flex: none;
 				width: 2.25rem;
 				height: 2.25rem;
 				display: grid;
@@ -141,10 +141,86 @@
 				padding: 0 1.5rem;
 			}
 
+			/* --- the bar that stays -------------------------------------------
+
+			   A page this long had its only way out at the very top: somebody who
+			   read as far as "What it will not do" had to scroll back past nine
+			   screenshots to reach the documentation. The bar is the way out from
+			   wherever you are.
+
+			   `sticky` and not `fixed`: sticky keeps its space in the flow, so
+			   nothing below has to be pushed down by a number that then has to be
+			   kept in step with the bar's own height.
+
+			   Full width, with the wrap *inside* it. Put the other way round, the
+			   background would end where the column ends and the page would scroll
+			   past it on both sides. */
+			.topbar {
+				position: sticky;
+				top: 0;
+				z-index: 20;
+				background: var(--ground);
+				border-bottom: 1px solid var(--line);
+			}
+
+			.bar {
+				display: flex;
+				align-items: center;
+				gap: 1.25rem;
+				padding: 0.625rem 1.5rem;
+			}
+
+			.bar-brand {
+				display: flex;
+				align-items: center;
+				gap: 0.5rem;
+				color: var(--ink);
+				text-decoration: none;
+				font-weight: 560;
+				letter-spacing: -0.01em;
+			}
+
+			.bar-brand .rune {
+				color: var(--accent);
+				font-size: 1.125rem;
+				line-height: 1;
+			}
+
+			.bar-links {
+				display: flex;
+				align-items: center;
+				gap: 1.25rem;
+				margin-left: auto;
+				font-size: 0.875rem;
+			}
+
+			.bar-links a {
+				color: var(--ink-muted);
+				text-decoration: none;
+			}
+
+			.bar-links a:hover {
+				color: var(--ink);
+			}
+
+			/* The three in-page links go first on a narrow screen: they are the ones
+			   a phone can reach by scrolling anyway. Documentation and Source cannot
+			   be reached any other way, so they stay. */
+			@media (max-width: 640px) {
+				.bar-links .inpage {
+					display: none;
+				}
+			}
+
+			/* Or the bar covers the heading you just jumped to. */
+			:target {
+				scroll-margin-top: 4.5rem;
+			}
+
 			/* --- hero --- */
 
-			header {
-				padding: 6rem 0 4rem;
+			.hero {
+				padding: 5rem 0 3.5rem;
 				text-align: center;
 			}
 
@@ -172,7 +248,7 @@
 			}
 
 			.norn {
-				margin: 2rem auto 0;
+				margin: 0 auto 1.75rem;
 				max-width: 34rem;
 				font-size: 0.9375rem;
 				color: var(--ink-faint);
@@ -289,44 +365,104 @@
 
 			/* --- features --- */
 
+			/* A hairline between sections rather than under every heading: the page
+			   is one column, and the line says where one subject stops. */
 			section {
-				padding: 3rem 0;
+				padding: 3.5rem 0;
+				border-top: 1px solid var(--line);
 			}
 
 			h2 {
+				display: flex;
+				align-items: baseline;
+				gap: 0.75rem;
 				font-size: 1.5rem;
 				font-weight: 560;
 				letter-spacing: -0.02em;
 				margin: 0 0 2rem;
 			}
 
+			/* A short accent rule in front of every h2. The sections had nothing to
+			   mark them but their own words, and at a glance the page read as one
+			   long run of paragraphs. */
+			h2::before {
+				content: '';
+				flex: none;
+				width: 1.125rem;
+				height: 2px;
+				border-radius: 1px;
+				background: var(--accent);
+				transform: translateY(-0.35em);
+			}
+
+			/* Three groups rather than eighteen cards in a row. Grouped, the list is
+			   read as three answers — what it does, what it talks to, what it
+			   promises — and any of the three can be skipped. */
+			.group-label {
+				margin: 2.5rem 0 1rem;
+				font-size: 0.75rem;
+				font-weight: 560;
+				text-transform: uppercase;
+				letter-spacing: 0.08em;
+				color: var(--ink-faint);
+			}
+
+			.group-label:first-of-type {
+				margin-top: 0;
+			}
+
+			/* Cards with their own edges, not cells in one table.
+			   The table shared a hairline grid, which meant the last row had to be
+			   full or it was a hole in the drawing — and it was a hole, for as long
+			   as there were ten of them. A card that stands alone on its row is a
+			   card; an empty cell is a mistake. */
 			.grid {
 				display: grid;
-				grid-template-columns: repeat(auto-fit, minmax(260px, 1fr));
-				gap: 1px;
-				background: var(--line);
-				border: 1px solid var(--line);
-				border-radius: var(--radius-lg);
-				overflow: hidden;
+				grid-template-columns: repeat(auto-fit, minmax(15rem, 1fr));
+				gap: 0.875rem;
 			}
 
 			.cell {
-				background: var(--ground);
-				padding: 1.5rem;
+				background: var(--surface);
+				border: 1px solid var(--line);
+				border-radius: var(--radius-lg);
+				padding: 1.125rem 1.25rem 1.25rem;
+				transition: border-color 150ms ease;
+			}
+
+			.cell:hover {
+				border-color: var(--line-strong);
 			}
 
 			.cell h3 {
-				margin: 0 0 0.5rem;
-				font-size: 1rem;
+				margin: 0 0 0.4375rem;
+				font-size: 0.9375rem;
 				font-weight: 560;
 			}
 
 			.cell p {
 				margin: 0;
-				font-size: 0.9375rem;
+				font-size: 0.875rem;
 				color: var(--ink-muted);
 			}
 
+			.cell code {
+				font-family: var(--font-mono);
+				font-size: 0.8125em;
+				color: var(--ink);
+			}
+
+			/* A card that has something to admit rather than something to offer.
+			   It is not a footnote and it is not fine print: it is the same size as
+			   everything else, and it says the same thing the README says. */
+			.cell .caveat {
+				margin: 0.5rem 0 0;
+				padding-top: 0.5rem;
+				border-top: 1px solid var(--line);
+				font-size: 0.8125rem;
+				color: var(--ink-faint);
+			}
+
 			/* --- install --- */
 
 			pre {
@@ -374,6 +510,168 @@
 				color: var(--ink-muted);
 			}
 
+			/* Four facts under the buttons, in the place where a landing page usually
+			   puts logos it has not earned. They are the four questions somebody
+			   self-hosting asks before anything else. */
+			ul.facts {
+				list-style: none;
+				display: flex;
+				flex-wrap: wrap;
+				justify-content: center;
+				gap: 0.5rem 1.25rem;
+				margin: 2rem 0 0;
+				padding: 0;
+				font-size: 0.8125rem;
+				color: var(--ink-faint);
+			}
+
+			ul.facts li {
+				display: flex;
+				align-items: center;
+				gap: 1.25rem;
+			}
+
+			/* The separator belongs between two facts, not after the last one — so it
+			   is drawn by the item that follows it. */
+			ul.facts li + li::before {
+				content: '';
+				width: 3px;
+				height: 3px;
+				border-radius: 50%;
+				background: var(--line-strong);
+			}
+
+			/* On a phone the row wraps, and a separator drawn by the item that follows
+			   it lands at the *start* of the new line — a dot with nothing before it.
+			   No separators there; the wrap is the separation. */
+			@media (max-width: 640px) {
+				ul.facts {
+					gap: 0.375rem 1.125rem;
+				}
+
+				ul.facts li + li::before {
+					display: none;
+				}
+			}
+
+			/* Two ways to run it, side by side, because they are alternatives rather
+			   than steps. Stacked, the second read as what you do after the first. */
+			.panels {
+				display: grid;
+				grid-template-columns: repeat(auto-fit, minmax(19rem, 1fr));
+				gap: 0.875rem;
+			}
+
+			.panel {
+				background: var(--surface);
+				border: 1px solid var(--line);
+				border-radius: var(--radius-lg);
+				padding: 1.25rem;
+			}
+
+			.panel h3 {
+				margin: 0 0 0.5rem;
+				font-size: 0.9375rem;
+				font-weight: 560;
+			}
+
+			/* Sunk rather than raised: inside a card, a block in --surface would be
+			   the same colour as the card it sits in. */
+			.panel pre {
+				background: var(--ground);
+				margin: 1rem 0 0;
+			}
+
+			.panel p {
+				margin: 0;
+				font-size: 0.875rem;
+				color: var(--ink-muted);
+			}
+
+			/* Versionsnummeret i sin egen søjle, så øjet kan løbe ned ad dem uden at
+			   læse sætningerne. `ch` frem for pixels: søjlen er lige så bred som de
+			   tal, der står i den, og den følger med, hvis skriften skifter. */
+			ul.lately {
+				list-style: none;
+				margin: 1.5rem 0 0;
+				padding: 0;
+			}
+
+			ul.lately li {
+				display: grid;
+				grid-template-columns: 5ch 1fr;
+				gap: 0 1rem;
+				padding: 0.875rem 0;
+				border-top: 1px solid var(--line);
+				font-size: 0.9375rem;
+				color: var(--ink-muted);
+			}
+
+			ul.lately .ver {
+				font-family: var(--font-mono);
+				font-size: 0.8125rem;
+				color: var(--ink-faint);
+				/* Tallet står på linje med den første tekstlinje og ikke over den. */
+				line-height: 1.7;
+			}
+
+			@media (max-width: 520px) {
+				ul.lately li {
+					grid-template-columns: 1fr;
+					gap: 0.25rem;
+				}
+			}
+
+			/* Risikoen står i kort, ikke i småt.
+			   Et forbehold sat med mindre skrift end resten er et forbehold, der er
+			   skrevet for at kunne peges på bagefter — og det er den ene ting, denne
+			   side ikke skal gøre. Samme størrelse som alt andet. */
+			.risk {
+				border: 1px solid var(--line-strong);
+				border-radius: var(--radius-lg);
+				padding: 1.5rem;
+				background: var(--surface);
+			}
+
+			.risk p {
+				margin: 0 0 1rem;
+				font-size: 0.9375rem;
+				color: var(--ink-muted);
+			}
+
+			.risk p:last-child {
+				margin-bottom: 0;
+			}
+
+			.risk strong {
+				color: var(--ink);
+			}
+
+			/* The non-goals were one paragraph of seven sentences, which is the shape
+			   nobody reads. Each is its own line now, and they can be counted. */
+			ul.nogo {
+				list-style: none;
+				display: grid;
+				grid-template-columns: repeat(auto-fit, minmax(16rem, 1fr));
+				gap: 0.5rem 1.75rem;
+				margin: 0;
+				padding: 0;
+				font-size: 0.9375rem;
+				color: var(--ink-muted);
+			}
+
+			ul.nogo li {
+				position: relative;
+				padding-left: 1.25rem;
+			}
+
+			ul.nogo li::before {
+				content: '—';
+				position: absolute;
+				left: 0;
+				color: var(--ink-faint);
+			}
+
 			.demo-note {
 				margin: 1rem 0 0;
 				font-family: var(--font);
@@ -524,27 +822,38 @@
 	</head>
 
 	<body>
-		<button class="theme-toggle" type="button" aria-label="Switch between light and dark" title="Light or dark">
-			<span aria-hidden="true">◐</span>
-		</button>
-		<div class="wrap">
-			<header>
+		<div class="topbar">
+			<div class="wrap bar">
+				<a class="bar-brand" href="#top">
+					<span class="rune" aria-hidden="true">ᚹ</span>
+					verdande
+				</a>
+				<nav class="bar-links" aria-label="Sections">
+					<a class="inpage" href="#screens">Screens</a>
+					<a class="inpage" href="#features">Features</a>
+					<a class="inpage" href="#running">Running it</a>
+					<a href="https://kristianwind.github.io/verdande/">Documentation</a>
+					<a href="https://github.com/kristianwind/verdande">Source</a>
+					<button class="theme-toggle" type="button" aria-label="Switch between light and dark" title="Light or dark">
+						<span aria-hidden="true">◐</span>
+					</button>
+				</nav>
+			</div>
+		</div>
+
+		<div class="wrap" id="top">
+			<header class="hero">
 				<span class="mark" aria-hidden="true">ᚹ</span>
 				<h1>verdande</h1>
 				<p class="lede">
-					Self-hosted tasks and projects, shared with the people you share the work with.
-					One binary, SQLite inside it.
+					Self-hosted tasks, notes and projects, shared with the people you share the work
+					with. One binary, SQLite inside it.
 				</p>
 				<p class="rune-note">
 					Built to be deployed as a <strong>Rune</strong> in
 					<a href="https://github.com/kristianwind/yggdrasil">Yggdrasil Panel</a> — upload
 					the manifest, fill in two fields, press start. It also runs anywhere Docker does.
 				</p>
-				<p class="norn">
-					Verdande is one of the three Norns — the one who spins <em>that which is
-					becoming</em>. Not what has been, and not what is fated to be: the thing you are
-					in the middle of. Which is what a to-do list is.
-				</p>
 
 				<div class="actions">
 					<a class="button primary" href="https://kristianwind.github.io/verdande/self-hosting/">
@@ -553,6 +862,14 @@
 					<a class="button" href="https://github.com/kristianwind/verdande">Source</a>
 					<a class="button" href="https://kristianwind.github.io/verdande/">Documentation</a>
 				</div>
+
+				<ul class="facts">
+					<li>One Go binary</li>
+					<li>SQLite inside it</li>
+					<li>amd64 and arm64</li>
+					<li>No account anywhere but yours</li>
+					<li>MIT</li>
+				</ul>
 			</header>
 
 			<!-- No screenshot: the quick-add parser is the thing worth showing, and it
@@ -584,7 +901,7 @@
 				</div>
 			</div>
 
-			<section>
+			<section id="screens">
 				<h2>What it looks like</h2>
 				<figure class="shot only-dark">
 					<img
@@ -621,13 +938,13 @@
 					</figure>
 					<figure class="shot">
 						<img
-							src="screenshots/today-light.webp"
-							alt="The same Today view in the light theme."
+							src="screenshots/notes.webp"
+							alt="Notes: a list of notes on the left, one open on the right with headings, bullet lists, a quotation and a shell command in a code block."
 							width="1400"
 							height="900"
 							loading="lazy"
 						/>
-						<figcaption>The light theme.</figcaption>
+						<figcaption>Notes. Rich text on top, Markdown underneath.</figcaption>
 					</figure>
 					<figure class="shot">
 						<img
@@ -642,7 +959,7 @@
 					<figure class="shot">
 						<img
 							src="screenshots/detail.webp"
-							alt="A task open in the detail drawer, with its description, due date, labels, sub-tasks and comments."
+							alt="A task open in the detail drawer, with its description, priority, due date, labels, sub-tasks and an attached file."
 							width="1400"
 							height="900"
 							loading="lazy"
@@ -652,12 +969,12 @@
 					<figure class="shot">
 						<img
 							src="screenshots/themes.webp"
-							alt="The settings page showing the five themes as swatches."
+							alt="The settings page showing five themes as swatches and, below them, four looks written in the typeface each one selects."
 							width="1400"
 							height="900"
 							loading="lazy"
 						/>
-						<figcaption>Five themes, all of them measured for contrast.</figcaption>
+						<figcaption>Five themes and four looks: how bright, and how it reads.</figcaption>
 					</figure>
 					<figure class="shot">
 						<img
@@ -669,11 +986,43 @@
 						/>
 						<figcaption>What has been done on the server, across every project.</figcaption>
 					</figure>
+					<figure class="shot">
+						<img
+							src="screenshots/delegated.webp"
+							alt="Waiting on others: tasks that have been handed to somebody else, grouped under the person who has them."
+							width="1400"
+							height="900"
+							loading="lazy"
+						/>
+						<figcaption>Waiting on others — grouped by who has it.</figcaption>
+					</figure>
+					<figure class="shot">
+						<img
+							src="screenshots/notes-source.webp"
+							alt="A note with the source view open: the same note as Markdown, with headings, a list, a quotation, a wiki link and a fenced code block."
+							width="1400"
+							height="900"
+							loading="lazy"
+						/>
+						<figcaption>The same note as Markdown. What is on disk is what you see.</figcaption>
+					</figure>
+					<figure class="shot">
+						<img
+							src="screenshots/settings.webp"
+							alt="The integrations settings page: connecting Gmail with an OAuth client, and below it the calendar feed address."
+							width="1400"
+							height="900"
+							loading="lazy"
+						/>
+						<figcaption>Integrations: your mailbox, your calendar feed, your address.</figcaption>
+					</figure>
 				</div>
 			</section>
 
-			<section>
+			<section id="features">
 				<h2>What it does</h2>
+
+				<p class="group-label">Doing the work</p>
 				<div class="grid">
 					<div class="cell">
 						<h3>Reads what you type</h3>
@@ -682,6 +1031,50 @@
 							one line — in Danish and English, mixed freely.
 						</p>
 					</div>
+					<div class="cell">
+						<h3>Nothing to remember</h3>
+						<p>
+							Start typing anywhere and you are writing a task, with the letter you
+							pressed already in the field. <code>⌘K</code> goes anywhere else.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Search that reads Danish</h3>
+						<p>
+							Type <code>gron</code> and find <code>grøn</code>. Unicode does not
+							consider <code>ø</code> an accented <code>o</code>, so a search that
+							leans on that alone is broken for the people this was built for.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Repeats, and says so in words</h3>
+						<p>
+							Recurring tasks are RRULE underneath and "every other Tuesday" on the
+							screen. Reminders are absolute or relative, and go out to the devices
+							you have allowed them on.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Notes, beside the work</h3>
+						<p>
+							Rich text to write in, Markdown on disk — and a source view to see it.
+							Images are pasted or dragged in. <code>#project</code> and
+							<code>[[another note]]</code> are links, and a task shows what has been
+							written about it. Archive what is done with it; the bin is for mistakes.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Waiting on others</h3>
+						<p>
+							What you have handed to somebody else, on its own page, grouped under
+							the person who has it. The thing a shared list stops being able to tell
+							you the moment it is shared.
+						</p>
+					</div>
+				</div>
+
+				<p class="group-label">Connected to what you already use</p>
+				<div class="grid">
 					<div class="cell">
 						<h3>Shared, properly</h3>
 						<p>
@@ -696,6 +1089,27 @@
 							here. Not a read-only feed pretending.
 						</p>
 					</div>
+					<div class="cell">
+						<h3>A calendar of its own</h3>
+						<p>
+							A month or a week of what is due, and a secret ICS address to subscribe
+							to from Apple Calendar, Google or Thunderbird. A Google calendar can be
+							laid over the top, read-only.
+						</p>
+						<p class="caveat">
+							The Google half is written to Google's documentation and has not yet
+							been run against Google. The feed and CalDAV have.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Your mail, as an inbox</h3>
+						<p>
+							Star a mail in Gmail, or flag one over IMAP in iCloud or Fastmail, and
+							it becomes a task. Forward one to your own address instead and the
+							subject is parsed like anything else you type. One way: unstarring it
+							later does nothing to the task you made.
+						</p>
+					</div>
 					<div class="cell">
 						<h3>Claude, connected</h3>
 						<p>
@@ -704,35 +1118,39 @@
 						</p>
 					</div>
 					<div class="cell">
-						<h3>Your mail, as an inbox</h3>
+						<h3>AI only if you ask for it</h3>
 						<p>
-							Star a mail in Gmail, or flag one in iCloud or Fastmail, and it becomes
-							a task. Forward one to your own address instead and the subject is
-							parsed like anything else you type. One way: unstarring it later does
-							nothing to the task you made.
+							Off by default, your own key, and your choice of Anthropic, OpenAI,
+							Google or something local. Nothing leaves the server until you have
+							filled that in, and everything works without it.
 						</p>
 					</div>
+				</div>
+
+				<p class="group-label">Yours, and checkable</p>
+				<div class="grid">
 					<div class="cell">
-						<h3>Nothing to remember</h3>
+						<h3>More than one way in</h3>
 						<p>
-							Start typing anywhere and you are writing a task, with the letter you
-							pressed already in the field. Sign in with a passkey and there is no
-							password to forget either.
-						</p>
-					</div>
-					<div class="cell">
-						<h3>Yours to leave</h3>
-						<p>
-							Import from Todoist, export back to it, or take everything as JSON. The
-							round trip is tested, not promised.
+							Passkeys, or a password hashed with Argon2id and a second factor with
+							recovery codes. Every device you are signed in on is listed, and any of
+							them can be signed out from here.
 						</p>
 					</div>
 					<div class="cell">
 						<h3>Keys stay behind</h3>
 						<p>
-							Mail tokens and mailbox passwords are encrypted in the database, and the
-							key lives in a file beside it — not inside it. So a backup you download,
-							or one that ends up somewhere it should not, does not open your mail.
+							Mail tokens, mailbox passwords, second-factor secrets and AI keys are
+							encrypted in the database, and the key lives in a file beside it — not
+							inside it. So a backup you download, or one that ends up somewhere it
+							should not, does not open your mail.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>Nobody signs themselves up</h3>
+						<p>
+							The first account is the administrator, and after that people arrive by
+							invitation. An address on the internet is not an open door.
 						</p>
 					</div>
 					<div class="cell">
@@ -743,50 +1161,157 @@
 							instance's record, not the person's.
 						</p>
 					</div>
+					<div class="cell">
+						<h3>Yours to leave</h3>
+						<p>
+							Import from Todoist, export back to it, take everything as JSON, or
+							hand the notes a zip of Markdown and get the same zip back out. The
+							round trip is tested, not promised.
+						</p>
+					</div>
+					<div class="cell">
+						<h3>It counts itself, out loud</h3>
+						<p>
+							Once a day an instance says it exists — a random id, the version, and
+							nothing else. No tasks, no names, no addresses. It is on to begin with,
+							the settings page prints the exact payload, and one switch stops it.
+						</p>
+					</div>
 				</div>
 			</section>
 
-			<section>
+			<section id="recently">
+				<h2>Lately</h2>
+				<p class="muted">
+					A landing page can say a project is alive; a list of what landed says it
+					better. This is not a changelog — every release has notes, and they are on
+					<a href="https://github.com/kristianwind/verdande/releases">the releases page</a>.
+				</p>
+				<ul class="lately">
+					<li>
+						<span class="ver">0.34</span>
+						Sections can be dragged into the order the work actually happens in. The
+						sidebar keeps its head still while the projects scroll.
+					</li>
+					<li>
+						<span class="ver">0.33</span>
+						Notes grew an archive, selection of many at once, pictures you can paste
+						or drag in, a source view, and a copy button on code blocks. Search
+						stopped answering by date and started answering by relevance.
+					</li>
+					<li>
+						<span class="ver">0.32</span>
+						A calendar of its own as a menu item, with a read-only Google calendar
+						that can be laid over it.
+					</li>
+					<li>
+						<span class="ver">0.31</span>
+						Mailboxes over IMAP, so flagging a mail in iCloud or Fastmail makes a
+						task — not only Gmail.
+					</li>
+					<li>
+						<span class="ver">0.30</span>
+						An installation count that says exactly what it sends, and can be turned
+						off in one click.
+					</li>
+				</ul>
+			</section>
+
+			<section id="running">
 				<h2>Running it</h2>
 
-				<h3 class="sub">As a Rune in Yggdrasil Panel</h3>
-				<p class="muted">
-					This is what it was built for. <strong>Runes → Carve a rune</strong>, upload
-					<code>rune/verdande.yaml</code>, create a server from it, set the public address
-					and your mail server, and start it. No port forwarding — give it a subdomain and
-					the panel routes it through your Cloudflare Tunnel.
-				</p>
-
-				<h3 class="sub">Or with Docker, anywhere</h3>
-				<pre><code>docker run -d --name verdande \
+				<div class="panels">
+					<div class="panel">
+						<h3>As a Rune in Yggdrasil Panel</h3>
+						<p>
+							This is what it was built for. <strong>Runes → Carve a rune</strong>,
+							upload <code>rune/verdande.yaml</code>, create a server from it, set the
+							public address and your mail server, and start it. No port forwarding —
+							give it a subdomain and the panel routes it through your Cloudflare
+							Tunnel.
+						</p>
+					</div>
+					<div class="panel">
+						<h3>Or with Docker, anywhere</h3>
+						<p>Open the address afterwards and create the first account. That one is the administrator.</p>
+						<pre><code>docker run -d --name verdande \
   -p 8080:8080 \
   -v verdande-data:/data \
   -e VERDANDE_BASE_URL=https://todo.example.dk \
   ghcr.io/kristianwind/verdande:latest</code></pre>
-				<p class="muted" style="margin-top: 1rem">
-					Open the address and create the first account. That one is the administrator.
-				</p>
+					</div>
+				</div>
+
 				<ul class="plain">
 					<li>No database to provision — SQLite is inside the binary.</li>
 					<li>Distroless image: no shell, no package manager, one static file.</li>
 					<li>amd64 and arm64, so a Raspberry Pi is fine.</li>
 					<li>Nightly backups, fourteen kept, written with <code>VACUUM INTO</code>.</li>
+					<li>Twenty-four migrations, run on start. Upgrading is pulling the new image.</li>
 				</ul>
 			</section>
 
-			<section>
+			<section id="limits">
 				<h2>What it will not do</h2>
-				<p class="muted">
-					No karma or points. No team workspaces. No native apps. No mail integration
-					beyond Gmail and the forwarding address. No languages beyond Danish and English.
-					And no sync protocol of its own — CalDAV and the API <em>are</em> the sync.
-				</p>
+				<ul class="nogo">
+					<li>No karma, streaks or points.</li>
+					<li>No team workspaces — projects and the people in them are the whole model.</li>
+					<li>No native apps. It is a PWA, and it installs on a phone.</li>
+					<li>No Outlook or Exchange: Microsoft has closed IMAP to passwords and wants its own flow.</li>
+					<li>No languages beyond Danish and English.</li>
+					<li>No sync protocol of its own — CalDAV and the API <em>are</em> the sync.</li>
+					<li>No hosted version. There is nowhere to sign up, and that is the point.</li>
+				</ul>
+			</section>
+
+			<section id="risk">
+				<h2>At your own risk</h2>
+				<div class="risk">
+					<p>
+						verdande is <strong>MIT licensed and provided as is, without warranty of
+						any kind</strong>. That sentence is in every MIT licence and it is usually
+						skipped, so here it is in the open: if you run this, you run it at your
+						own risk, and nobody is on the hook if it loses your work.
+					</p>
+					<p>
+						It is one person's project, at version 0.x, and 0.x means the shape can
+						still change. <strong>You host it, so the data and the backups are
+						yours.</strong> It writes a backup every night and keeps fourteen — and a
+						backup nobody has ever restored is not yet a backup. Restore one before
+						you need to.
+					</p>
+					<p>
+						Security is worked on rather than claimed. There have been two full
+						reviews of the code, every finding is written down and closed, and what is
+						protected and what is not is spelled out in
+						<a href="https://github.com/kristianwind/verdande/blob/main/SECURITY.md">SECURITY.md</a>
+						— including the parts that are a deliberate trade-off. <strong>Nobody
+						outside the project has audited it.</strong> The last review found two ways
+						to reach something you should not and closed both; a third review will find
+						more, which is the point of having one.
+					</p>
+					<p>
+						It is built to sit behind something — a tunnel, a reverse proxy, a VPN —
+						put there by somebody who knows what they exposed. Found a hole? Mail it
+						rather than opening an issue; the address is in SECURITY.md and you get an
+						answer within a week.
+					</p>
+				</div>
 			</section>
 
 			<footer>
+				<!-- The Norn belongs at the end. It was the third paragraph of the hero,
+				     where it stood between somebody arriving and the buttons they came
+				     for — and it is a closing thought, not an opening one. -->
+				<p class="norn">
+					Verdande is one of the three Norns — the one who spins <em>that which is
+					becoming</em>. Not what has been, and not what is fated to be: the thing you are
+					in the middle of. Which is what a to-do list is.
+				</p>
 				<p>
 					MIT licensed ·
 					<a href="https://github.com/kristianwind/verdande">github.com/kristianwind/verdande</a>
+					· <a href="https://kristianwind.github.io/verdande/">Documentation</a>
 				</p>
 				<p>
 					Built to run as a Rune in