Five things the first cut of accounts-from-content left open.
The first person and the responsible people are set up side by side,
and both try again a minute apart: a seed that waited twenty minutes on
Kanidm used to hold everyone else up, and a responsible person who
failed waited for the next restart.
The responsible are gone over again whenever the content changes.
watch_for_reload publishes portal.content.reloaded once new content is
in place, and seed.rs listens.
Being responsible ends when the content stops saying so: someone no
page names any more leaves RESPONSIBLE_GROUP in the same pass, if
portal is who put them there (Invited.already_member). The account
stays.
An invite keeps itself current. portal_invites has an expired state;
signing in closes a person's invites, a sweep moves the ones whose link
ran out, and a desk's move from expired back to open makes a fresh link
and mails it. Invites made at start are on file with the desks'.
A suggested edit is a case in the built-in portal_edit_suggestions
bucket, the person who made it is mailed at each move, and the card
lists what they sent and where it stands.
Found on the way, against Kanidm 1.11.2: a link's end comes as seconds
since the epoch, which portal did not read, so invite mails said
"until" and nothing; and a mail search answered with a 5xx was taken
for a refusal.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GT37Z1Xtfd9pUuQtMTg6Yt
From an empty Kanidm to everyone at their own desk: what in the content
decides who needs an account, the four things Kanidm needs for portal,
the first person, what happens at start, how everyone after comes in
from the desks, what the people who ask the questions can do, and what
each failure looks like.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GT37Z1Xtfd9pUuQtMTg6Yt
Every person a page names as responsible gets a Kanidm account at start
if they have none, and a mail saying they are listed as responsible for
asking that question, with the link to sign in; RESPONSIBLE_GROUP puts
them in one group. SEED_ADMIN_EMAIL invites the first person into the
most privileged group - the one whose desk may invite into the most -
once. Signed in, a responsible person may suggest a change to their
own pages; checked on the server, kept, published, and mailed to the
site's inbox.
Fixes an invite of someone new on a Kanidm with only the onboarding
account: the person is created with their address in one call, since
idm_people_on_boarding may create but not modify (it answered 403).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GT37Z1Xtfd9pUuQtMTg6Yt
An arrow fixed at the left edge, half-way up the screen, and the same
link in words at the foot of the page. It goes one level down: the
nearest page below the current path that the visitor may open, found by
taking the path apart a segment at a time, and the front door when none
of those is a page (/status/oppslag -> /, /review/abc -> /review). The
front door itself has none. Labelled with where it goes: 'Tilbake til
starten', or the target page's question in quotation marks.
Content stops needing a 'back' alternative on every page, and a visitor
who went up a branch whose context did not apply is never stranded.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GT37Z1Xtfd9pUuQtMTg6Yt
A MemoryStore meant every restart signed out every desk on the site
being deployed. Bearable while a desk is three people, and not the
kind of thing that gets better on its own.
They live in NATS KV now: the JetStream already under the events, the
projections, the content reload and the access questions. No new
service, no new store, and the same bucket a second instance of a site
would read if there ever is one. `create` draws a fresh id rather than
writing over a session already at that key, the bucket's `max_age`
collects what nobody comes back for, and a record's own expiry is what
answers a load, whatever the bucket still holds.
Proved the only way it can be: a store opened twice against one
JetStream, which is what two runs of the binary are.
Also fixed, while there: the integration test asserted a record's
projected state was one that `aggregates::transition` never writes -
it appends events, the desk handler updates the projection - so it
failed whenever it was actually run. It now asserts what it says it
does, that a mailed reply leaves the state alone. test.yml has been
red since 0.3.47 while publish.yml stayed green, which is how v0.4.0
went out over it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two halves of the same evening.
Uploads were one-way. A `type: file` field put an object in the bucket
and wrote its key into the record, and nothing ever read it back - a
desk showed `0aba5734-...-drawing.pdf` as a string with no link, so a
builder could receive drawings and never open one.
`GET /attachment/{bucket}/{record}/{field}` streams it. The key is not
the credential and is not taken from the caller: it is read out of the
record named, so a file is reachable only through a record somebody is
allowed to see, and a key that leaks off a desk is a link to nothing.
And there was no rate limiting anywhere. A burst of 30 per caller IP,
refilled one every two seconds, outside the session layer so anyone
over it is turned away before a handler is entered. Well above a
person filling in a form - a page load fetches several resources at
once - and well below being a nuisance. PORTAL_RATE_BURST and
PORTAL_RATE_SECONDS retune it; either 0 turns it off for an instance
behind something that already does this.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
/upload asked nothing. It took a file, capped it at 20 MB, and wrote
it to the bucket for whoever sent it - no session, no page, no policy.
It was the one server entry point that did not ask the question every
other one asks, and on a site whose forms are public that makes the
route public too: a free disk for anyone who finds the host.
Harmless while no instance had storage configured, which is why it sat
there. kasse now has a garage wired to klingenberg's portal, so it is
not harmless any more.
An upload now carries the question and the alternative the file is
being attached to. The page has to exist, the alternative has to be
open and to actually declare a `type: file` field, and the policy has
to allow Submit on that page for whoever is asking - checked after the
bytes are read and before any of them are written.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The Suspense fallback on a question is the whole served page until
the question resolves - no header, no wordmark - so the word
"loading…" was the first and only thing a visitor or a crawler read
about the site. True, and not about them. It holds its space and
stays quiet now; `aria-busy` says what is happening to a reader that
would otherwise find an empty document. Same for the two resource
slots inside a page, which keep a line's worth of room.
Also gone: the `gitea_org_repos` resource kind, which no site ever
named. `gitea_starred` and `gitea_releases` stay.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Five things a building authority needed and portal could not say,
found by building one and replaying its cases through the engine.
The one that mattered: a submitter could only ever act on a record
nobody had touched. `access.rs` pinned a self_transition's from-state
to `open`, though the runtime had always passed the real state and
asked. So the applicant written to for a missing drawing could not
send it, and nobody could withdraw once anything had happened.
`self_transition` now takes `from: [state, ...]` and compiles one rule
per state; omitted it still means the initial state alone, which is
what every unsubscribe link already meant.
Then the four beside it. A record can belong to another record
(`type: record`, `of:`), with `where: { field, equals }` on a listing
as the other half, so an application's page lists the objections that
name it. A public view can be narrowed to the states worth showing,
so a register that is public by law is public about what was decided
and not about what was withdrawn. A state can grant a group, so an
approval that means membership needs no second step. And a state can
be moved by a clock (`after: { days, to }`, src/deadlines.rs) - held
to the same declared edge as a desk button, firing the same mail,
publishing the same event.
Also: the README is a README again, and everything that was crowding
it is in docs/ - architecture, content reference, state machines,
access, operations.
95 tests, up from 82.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Whatever reads the mailbox publishes on portal.mail.received; portal
appends the note after checking the record exists in a bucket this
site has and the sender is the address the record itself holds. A
note never moves a case: a mailbox is the one door anyone may knock
on, so it cannot drive a state machine even if someone forges their
way past the checks. The desks show the conversation beside the case.
Tested: the address parser, every refusal (a stranger, no address to
compare, an empty reply, the same reply twice, an unknown contract),
and the whole path through a real JetStream - the note lands, the
state does not move, and no event is written.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`type: location` puts a pin on a map and stores {lat, lon, zoom}; a
record that holds one renders as a still, on a desk and in any
resource card. `place:` on a question is the counterpart to `event:`:
event makes a page an occasion with a time, place makes it somewhere,
shows the map under the question, and opens any location field on the
page there instead of on the globe.
Every tile and every still is fetched by portal and served from this
origin (src/maps.rs, MapLibre vendored in public/), so a page with a
map on it still talks to nobody but us: no third-party request, no
consent banner, and the token never leaves the server. MAPBOX_TOKEN
unset turns the whole thing off, the way GARAGE_* does for uploads.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
`mail: { to, subject, body }` on a state in aggregates.yaml sends when a
record enters that state, `to: submitter` or one address, templates
over the record's fields plus {site}, {chain}, {email}. An invite mails
its one-time link built in, reworded by `mail.invite` in site.yaml,
which also carries the sender (`mail.from`, `mail.reply_to`); without
it a site sends nothing and says so in the log. Portal renders the
text and an HTML alternative and publishes one message per state
change on portal.mail.send in the WORMHOLE stream, id <bucket>:<record>
:<state> so a retried transition sends once; uhhm/gdo holds the
durable consumer and talks SMTP. The lint requires a required email
field on every form recording into a bucket that mails the submitter.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
people: the Kanidm person lookup reads the body (200 + null is a
missing person in Kanidm 1.11).
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The existence check went by status alone, so every invitee looked like
an existing account: creation was skipped and the group step failed.
Read the body; `null` means nobody there.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Three releases went out with no section because the edit that moves
Unreleased under a version heading was a no-op when no Unreleased
section existed, and nothing checked. Now the publish job fails when
the tag has no section.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
Who may do what to what was five checks in five files: `qualifies` on
a page, `requires_group`/`public` on a resource, a desk's transition
allow-list, a self_transition's email check, the automation token.
They are now compiled from content on every load into one Cedar
policy set (src/access.rs), and every server entry point asks it: open
and submit a page, read a resource, move a record, fire a self
transition, read a bucket as the automation service, extend an
invite. Each decision names the rules that granted it; refusals and
every mutation are published on portal.access.decided, and
portal.access.<site>.may answers the same question for other apps
over request-reply (proven live against uhhm's content). The policy
validates against a fixed schema, so a content shape the compiler
cannot express fails instead of granting nothing. Identity is still
Kanidm's; the browser keeps is_qualified as a rendering hint only.
`invite: {group}` on an alternative of a gated page onboards a person:
src/people.rs creates the account, records the email, adds the group
and mints a one-time credential link through Kanidm's REST API with a
service-account token (KANIDM_API_TOKEN, in idm_people_on_boarding).
The result is recorded with the answer into the runtime's
portal_invites bucket and announced on portal.people.invited. Validated:
an invite needs a gated page and required name and email fields.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
iris is versioned as portal. On a tag push this job clones uhhm/iris,
re-pins its portal dependency to the tag, sets its version to match,
builds and tests it, writes the changelog line, and pushes the commit
and tag - which iris's own CI publishes. The tarball ships the binary
just built. A portal change to something the lint reads fails the
release at this step. Needs the IRIS_PUSH_USER variable and
IRIS_PUSH_TOKEN secret on this repo.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
question_lint, needs_replay, the needs module and the local-checkout
loaders now live in uhhm/iris, which depends on this crate as a
library pinned to a release tag. The tarball ships the matching iris
binary (IRIS_RELEASE in publish.yml) in place of question_lint, so a
site that pins nothing extra keeps a lint at current/iris. Portal
loses about 1,100 lines and keeps only what runs at request time.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
The 0.3.34 rule listed every followup "further down" from the current
page, so the front door offered the thank-you pages before anything
was submitted, and a followup one typed in offered the ones beyond
it. The chain half was weaker than it looked too: it matched the
question the chain answered, which is a front door anyway.
Now a followup is offered only when the visitor has earned it: their
verified chain answered the alternative that leads there (the chain
index records the alternative from here on; older nodes fall back to
every recording alternative of the answered question, and nothing
beyond it), or it hangs off an earned page by a link that records
nothing. A page behind an answer they have not given is never offered.
The walk ends at any front door, so "back to the start" never fans out.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
Forty releases carried only a version bump in their commit, so what
changed lived in the commits between them and in the content repos'
pin bumps. The changelog is reconstructed from that history, newest
first, with an Unreleased section on top and a table of which version
each of the four sites runs. The publish workflow now posts the tag's
section as the Gitea release notes, and the README says to move the
Unreleased lines under a version heading before cutting.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
Until now nothing outside a live site exercised the NATS side: the
aggregate engine was tested by in-memory replay only. This test
submits an answer, moves it from a desk through the engine, checks
that an undeclared move and an unknown record are refused, and reads
the record back from its own event log. It runs when
PORTAL_TEST_NATS_URL is set; CI starts a throwaway nats:2.14.6-alpine
with JetStream for it and tears it down. Without the variable the
test passes as skipped, so `cargo test` stays green offline.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
content.rs was 2,373 lines doing five jobs. It is now a directory:
mod.rs keeps the types and the resolution rules (router, refs,
reachability, build_questions) and the tests; gitea.rs loads from
Gitea and watches for reload, through crate::http; validate.rs holds
the checks a repo must pass; markdown.rs the inline markdown;
handlers.rs the Axum routes. Each submodule is re-exported from
content, so no caller changes. The loader module is server-only and
gated as such. No behaviour change; 65 unit tests unchanged and green.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
Every outbound fetch - Gitea content at boot and on reload, the raw
asset proxy, the repo-info endpoint, the content-declared url resource
- built its own reqwest client with no timeout and read whole bodies
with .text(). `crate::http` is now the one client (5 s connect, 20 s
per request, 5 redirects) with capped body readers: 2 MB for a url
resource, 8 MB for a Gitea reply, 32 MB for a proxied asset; one byte
over is an error, never a truncated document.
The upload handler read a whole field into memory before comparing it
to the 20 MB cap, and axum's 2 MB default body limit had made that cap
unreachable anyway. It now reads chunk by chunk and refuses the moment
the cap is crossed; the /upload route raises the body limit to match.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01L4jrCgLiKKHAEFuZUJjckH
