Bislang gab es keine vom Nutzer-Rollenmodell (creator/agentur/marke/
kanzlei) getrennte Betreiber-Rolle — jede Verwaltungsaufgabe (welche
Kanzlei darf im öffentlichen Verzeichnis stehen, wer sind unsere
Accounts) wäre nur per Hand in der Datenbank möglich gewesen. Admin
ist von Anfang an als fünfte app_user-Rolle im Datenmodell verankert,
nicht nachträglich aufgesetzt.
Migration 0004:
- app_user.role erlaubt zusätzlich 'admin' (kein Self-Service-Weg
dorthin — /register bietet die Rolle nicht an, erster Admin wird
einmalig per SQL angelegt, siehe CLAUDE.md).
- account.verified: Freigabe fürs kostenlose Kanzlei-Verzeichnis
(§ 49b Abs. 3 BRAO: reine Auflistung, kein Routing/keine Vermittlung).
- audit_log: append-only-Protokoll jeder Admin-Aktion (gleicher Trigger
wie finding/extraction/evidence_package).
Neue Routen:
- GET /admin, /admin/accounts, /admin/accounts/{id}: Accounts-Übersicht
und -Detail (Logins je Account), requireAdmin (404 statt 403 für
angemeldete Nicht-Admins, wie beim bestehenden Mandanten-404-Muster).
- POST /admin/accounts/{id}/verifizieren: Kanzlei-Freigabe umschalten,
schreibt einen Audit-Log-Eintrag.
- GET /admin/audit-log: Protokoll ansehen.
- GET /kanzleien: öffentliches Verzeichnis (kein Login), zeigt nur
Accounts, die sowohl verified sind als auch einen Nutzer der Rolle
"kanzlei" haben.
Volle Testsuite inkl. echter Postgres-Tests grün; End-to-End manuell
gegen einen laufenden Server verifiziert (Admin-Login, Verify-Toggle,
Erscheinen im öffentlichen Verzeichnis, Audit-Log-Eintrag, 404 für
Nicht-Admin-Zugriff).
Participant (the Verantwortungsmatrix — who briefed, who approved) is
not append-only like finding/extraction/evidence_package; getting a
role wrong and correcting it isn't rewriting evidence, so full CRUD is
legitimate here: Create/List/Get/Update/Delete. UpdateParticipant sets
approved_at the first time freigegeben flips to true and never moves it
again on subsequent no-op updates — it marks when approval first
happened, not "last touched".
ListSubmissionsForAccount is the query the upcoming archive overview
needs: every submission for a tenant plus a findings count and highest
severity, computed with the same anti-join ListCurrentFindings already
uses for "currently valid" findings.
Also fixed a real bug this surfaced: CreateFinding let a nil Sources
slice reach a NOT NULL TEXT[] column, which Postgres rejects with an
unhelpful constraint error instead of a clear message. It now normalizes
nil to an empty slice before inserting — matters for any future rule
that ships without a fundstelle entry.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
internal/auth is pure logic (bcrypt hashing, session token generation)
with no DB access — persistence for account/app_user/session lives in
internal/store like everything else, via migration 0003.
account is the tenant (Mandant); app_user is a login inside one account;
session is a real server-side row (not a signed stateless token) so
logout can actually end a session rather than the client just
forgetting a JWT. submission.account_id is NOT NULL — added directly
rather than the nullable-then-backfill dance, since no submission rows
exist anywhere yet (verified empty on the test server before writing
the migration). Added as migration 0003 (new file), not folded into an
earlier one, since 0001/0002 are already applied on the test server.
store.ErrNotFound lets callers distinguish "wrong email" / "unknown
session" from a genuine DB error — matters for login, where those two
cases should both fail closed but for different reasons.
Not yet wired into internal/web — that's the next commit. All of this
is tested against real Postgres (14 store tests green) but isn't
reachable from any HTTP handler yet.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Migration 0002 replaces finding.message with title/fix/sources (a
Postgres text[]). A finding needs to render into the dossier the way it
looked at the moment it was raised — referencing the current rules/*.yaml
by rule_id+version isn't safe once that file is edited for a later
version, since old wording isn't kept around as a separate live file.
Added as a new migration rather than editing 0001, since that's already
applied on the test server.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Create/Get for submission (including the one legitimate status
transition — submission is not append-only, unlike the other three),
Create/GetLatest for extraction and evidence_package, Create for
finding plus ListCurrentFindings which applies the anti-join documented
in the migration (a finding referenced by another row's `supersedes`
is not "current").
Tested against real Postgres, including that the append-only trigger
still rejects UPDATE on evidence_package via this new code path, and
that ListCurrentFindings actually hides a finding once a correction
supersedes it.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
internal/store connects via pgx and runs golang-migrate migrations
embedded in the binary (go:embed), so Deklarix stays a single binary
despite the move to Postgres. Schema covers the five MVP tables
(submission, asset, extraction, finding, evidence_package, participant).
extraction, finding and evidence_package are append-only by design: a
Postgres trigger rejects UPDATE/DELETE outright, since a corrigible
evidence archive isn't an evidence archive. Corrections to a finding are
new rows whose supersedes column points at the row they replace (set at
INSERT time on the new row, since the trigger blocks UPDATE on the old
one) — "currently valid" findings are the ones no other row supersedes.
scripts/test.sh now spins up a disposable Postgres container so the
store's integration tests (including the append-only guarantee) actually
run on every test.sh/release.sh invocation instead of silently skipping
for lack of DATABASE_URL.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
