DMS Document Schema

Document title

Optional description or summary

Reference to dms_document_types (custom document type)

Legacy document type string (deprecated, use documentTypeId)

Custom field values defined by the document type schema

Document lifecycle status

Linked CRM party

Linked contact person

Linked contract

Linked project

Linked employee profile

Assigned user for review

When true, document is published on the project customer portal (requires project.customerAccess feature releasedDocuments).

True if the document was uploaded by a customer via the public portal.

Customer party that uploaded this file (set when customerUploaded = true).

Owning org unit

Standort der eigenen Firma (location entity under the isSelfCompany party, same set as the print-target sites of ADR 0370). Purely organisational: it says which own office the document belongs to (Miet-/Versicherungsvertrag, Brandschutz, Netzdoku je Niederlassung). Absent/null = not assigned — no backfill required.

Manual + auto tags

Reference to dms_folders entity

Virtual folder path (e.g. /invoices/2026/)

User ID who has this document checked out

When the document was checked out

Auto-release lock after this time (default 24h)

Reference to files collection

MIME type of the document file

File size in bytes

S3 storage key

S3 key for thumbnail image

Number of pages (PDF)

Full text extracted from document for search

Text extraction pipeline status

OCR processing status

Which OCR method produced the text

OCR confidence score (0-100)

Reason why the last OCR attempt failed (exception message or diagnostic hint)

S3 key for the searchable PDF (original + invisible OCR text layer)

Why the invisible text layer could not be built, even though OCR itself succeeded. Without a text layer the PDF stays unselectable in the viewer, so this is a user-visible outcome and not just a log line. Absent on documents processed before this field existed — treat as "unknown", never as "no error".

Classification confidence score 0-1

How the document was classified. EINVOICE = ZUGFeRD/XRechnung Pre-Classifier matched the embedded SellerTradeParty/BuyerTradeParty against the tenant identity.

Diagnose-Spur der letzten Klassifikation: welche Regeln matchten, welche E-Invoice-Felder die Tenant-Identität trafen, oder bei MANUAL der vorherige Doktyp. Zeigt im UI „Warum wurde dieses Dokument so klassifiziert?".

Applied retention policy

When retention period expires

Zeitpunkt, an dem die Aufbewahrungsuhr gestartet ist (abgeleitet aus retentionStartTrigger des Doctypes).

Berechneter Ablauf der Aufbewahrungspflicht. Vor diesem Zeitpunkt blockiert der Soft-Delete-Endpoint und (bei retentionMode=compliance) auch S3 jede Löschung.

Effektiver Object-Lock-Modus dieses Dokuments (vom Doctype geerbt, beim Lock-Apply eingefroren).

S3-VersionId des PUT, an dem der Object-Lock haftet. Pflicht, weil Lock immer auf eine konkrete Version greift.

Sanity-Flag: true sobald PutObjectRetention erfolgreich gegen S3 gefeuert wurde. Daily-Integrity-Job verifiziert via GetObjectRetention.

Hex-codierter SHA-256-Hash des Originalinhalts beim Upload. Daily-Integrity-Job vergleicht gegen neu berechneten Hash zur Manipulationsprüfung (GoBD Rz. 110 Unveränderbarkeit).

Aufbewahrungs-Sperre (S3 PutObjectLegalHold). Wenn active=true, blockiert jede Löschung — auch nach Ablauf von retentionUntil. Verwendung: Steuerprüfung, Litigation Hold, Audit.

Previous versions of this document

Links to other entities

Incoming invoice specific data (Eingangsrechnung)

AI-generated document summary

AI-suggested tags

AI-suggested filing location

AI-suggested folder ID

Reference to potential duplicate document

Similarity score for duplicate detection (0-1)

Vector embedding generation status

When embeddings were last generated

Document expiry date (e.g. TÜV, certificates, warranties)

Days before expiry to send notification

Resolved Codemeta user ID matching the SharePoint uploader (via AAD object ID, UPN, or email). Null when no match found.

Storage backend currently holding the bytes.

Backend-specific item id (e.g. Graph driveItem id, S3 key, FS path).

Backend ETag at last sync — used for 3-way conflict detection.

Filing policy most recently applied to this document.

Backend-absolute path at the time of last filing (audit + quick display).

Pfad, an dem das Dokument beim Inbound-Sync ursprünglich lag (z.B. "/Kunden/ACME/Rechnungen/file.pdf"). Quelle der Pfad-Hoheit, wenn keine Filing-Policy mit overrideBackendPath greift.

Backend, das das Dokument ursprünglich geliefert hat (Inbound-Origin). Bleibt stabil, auch wenn ein späterer Move das aktive Backend ändert.

Ownership-Modell der Pfad-Hoheit (ADR 0391): "origin" = Backend-Original (Inbound-Sync), "user" = explizite Ordnerwahl beim Upload, "policy" = Filing-Policy-Template. Nur eine Policy mit overrideBackendPath=true darf "origin"/"user" überstimmen.

Filing lifecycle state (ADR 0051).

Conflict-locking state-machine (ADR 0052 §7.4).

Ingestion-Inbox quarantine flag (ADR 0051 §7.2.5).

Dokument stammt aus einem konfigurierten SharePoint-Posteingangsordner (ADR 0392). Unterdrückt das automatische filingStatus="filed" am Origin-Pfad, damit es ohne greifende Filing-Policy im Posteingang zur Triage landet. Fehlt auf allen Alt-Dokumenten → falsy → unverändertes Verhalten.

Sensitivity classification used by filing policies (§9.4).

When the post-ingestion pipeline acquired this lock. Null = idle.

BullMQ jobId of the running pipeline (used for explicit cancellation).

Current pipeline stage — surfaced to the user as a progress hint.

Lock TTL — past this point the lock is considered stale.

Set by POST /dms/documents/:id/unlock. The worker checks this between stages and bails on the next stage boundary.

Timestamp of the most recent successful outbound move/push.

BullMQ jobId of the last outbound-move attempt (used for diagnostics and idempotency).

Total outbound-move attempts (increments on every CAS-lock acquisition).

Per-document override of the document type's defaultEncryption. null = inherit from type.

Encryption metadata for ZK-protected files. Set on upload, never mutated; rotation requires re-upload.

Free-text notes