dorfteich/docs/architecture
Claude Fable 5 0bc36aa58c
Some checks failed
CI / Lint, typecheck, test (pull_request) Successful in 5m1s
CI / Build container images (pull_request) Successful in 2m48s
CI / Auth e2e pack (pull_request) Failing after 3m12s
CI / Import/export fidelity gate (pull_request) Has been skipped
#194: orphan-file sweep, drop the unused Attachment.deletedAt
Nightly sweep with two directions: attachments still unclaimed (pageId
null) after a 24 h grace period - claimed by no collab persist, page
upload, or import - are reclaimed (row, file, quota released); files on
the uploads volume without a database row (drift after a crashed
upload) are removed once older than the grace period. The grace period
protects the paste-then-insert window.

Deliberate deviation from the issue's content-reference idea, documented
in schema comment and operations.md: claimed attachments whose page
content no longer embeds them are NOT auto-deleted. The page attachments
panel lists claimed files as user-managed objects (inserting into the
document is optional there), so 'not embedded' is not 'unused' - an
auto-delete would destroy panel assets. Humans clean those up in the
panel or the pond file manager, which flags orphans already.

Attachment.deletedAt is removed by migration - deletion is hard
everywhere (sweep, purge, manual), there is no soft-delete state; the
never-true deletedAt:null filters in files/export queries went with it.

Refs #194

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0168Ph5uBmHm8X28CSVpbpnJ
2026-07-30 13:19:58 +02:00
..
adr #190: configurable session lifetime with a server-side idle timeout 2026-07-30 11:11:15 +02:00
data-model.md Add plugin storage, install API, and directory watcher (#71) 2026-07-10 16:55:26 +02:00
deployment.md Re-align the stage tables after the path change 2026-07-14 12:53:11 +02:00
operations.md #194: orphan-file sweep, drop the unused Attachment.deletedAt 2026-07-30 13:19:58 +02:00
permissions.md Scaffold pnpm monorepo with lint, format, and test tooling 2026-07-04 19:06:27 +02:00
plugin-architecture.md #155: ChordPro-Plugin — Akkordblätter/Leadsheets als Block 2026-07-20 02:11:36 +02:00
README.md docs: ADR 0017 — Barrierefreiheit als Standard-Anforderung 2026-07-21 17:17:26 +02:00
realtime-collaboration.md Scaffold pnpm monorepo with lint, format, and test tooling 2026-07-04 19:06:27 +02:00
roadmap.md Add M13 to the roadmap 2026-07-14 16:37:24 +02:00
security.md #191: feeds.enabled instance switch, feed-token log masking 2026-07-30 11:34:43 +02:00

Dorfteich — Architecture Overview

This directory is the authoritative architecture documentation. Every implementation story references the documents here; when a story and this documentation disagree, clarify before coding.

System context

Dorfteich is shipped as a set of Docker containers behind a reverse proxy. One deployment = one instance (e.g. dorfteich.online, or a self-hosted installation).

flowchart LR
    subgraph clients [Clients]
        B[Browser SPA<br/>React + TipTap + Yjs]
    end
    subgraph instance [Dorfteich instance - Docker Compose]
        RP[Reverse proxy]
        WEB[web<br/>static SPA assets]
        API[api<br/>NestJS REST]
        COLLAB[collab<br/>Hocuspocus WebSocket]
        PG[(PostgreSQL)]
        PAN[pandoc-server<br/>doc conversion]
        GOT[Gotenberg<br/>HTML to PDF]
        VOL[/uploads + plugins volume/]
    end
    B -- HTTPS --> RP
    RP --> WEB
    RP -- /api --> API
    RP -- /collab WebSocket --> COLLAB
    API --> PG
    COLLAB --> PG
    API --> PAN
    API --> GOT
    API --> VOL
    COLLAB -. permission checks .-> API
  • web serves the single-page application (static assets).
  • api owns all business logic: auth, ponds, pages, labels, permissions, quotas, import/export, plugin management, admin functions.
  • collab synchronizes CRDT documents (page content) and awareness (cursors) over WebSocket and persists document state to PostgreSQL. It authenticates clients with short-lived tokens issued by api.
  • pandoc-server and Gotenberg are internal-only conversion sidecars (Word/OpenOffice import/export, PDF export).
  • All page content lives in PostgreSQL; binary uploads (images, attachments) and installed plugins live on a Docker volume.

Documents

Architecture Decision Records (adr/)

ADR Decision
0001 TypeScript everywhere, pnpm monorepo
0002 PostgreSQL as the only database
0003 Yjs CRDT + Hocuspocus for real-time and offline collaboration
0004 TipTap (ProseMirror) as the WYSIWYG editor
0005 React + Vite single-page application
0006 NestJS + Prisma for the API server
0007 Cookie sessions, Argon2id, OIDC-ready identity model
0008 Sandboxed iframe plugins with a message-based API
0009 Pandoc + Gotenberg sidecars for import/export
0010 PostgreSQL full-text search behind a search interface
0011 Filesystem volume for uploads, DB-tracked quotas
0012 i18next with German and English from the start
0013 Page version history via Yjs snapshots, soft-delete trash
0014 CI/CD with Gitea Actions, staged promotion
0015 Nightly pg_dump + uploads sync, 30-day retention, off-host mirror
0016 Self-hosted Google Fonts, per-pond font configuration
0017 Accessibility (WCAG 2.1 AA) as a default requirement

Concept documents

Document Contents
data-model.md Entities and relations: users, ponds, pages, labels, grants, quotas, plugins
permissions.md Role model and the "most specific setting wins" resolution algorithm
realtime-collaboration.md CRDT document lifecycle, cursor sync, offline behavior, versioning hooks
plugin-architecture.md Plugin manifest, packaging, sandbox runtime, extension points, admin flows
deployment.md Compose stacks for Dev/Test/Int/Prod, environments, promotion pipeline
operations.md Monitoring, logging, backup/restore, update strategy for self-hosters
security.md Threat-driven security concept: authn/authz, sandboxing, uploads, secrets
roadmap.md Epics and milestones; the order stories are implemented in

Terminology

German (product vision) English (code, docs, issues)
Teich pond
Seite page
Teich-Admin Pond Admin
Bearbeitende Editor
Lesende Reader
Öffentlichkeit Public

Conventions for implementers

  • Language: English for code, comments, commit messages, and issues.
  • Write clear, human-readable code; document the "why", not the "what".
  • UI strings never appear hard-coded — always through i18n resources (see ADR 0012), with German and English translations added in the same change.
  • Every story lists the ADRs it depends on; read them before starting.
  • Every API route declares its access rule explicitly (a permission decorator, @Public(), or the Site-Admin guard — issue #52); permission checks run only through the shared resolution (permissions.md), never ad hoc. 404/403 policy: a denied read answers 404 so the existence of ponds and pages is not revealed; a denied write on something the user may read answers 403. Trash views require write capability and answer 404 on denial (ADR 0013).