Software de gerenciamento de documentos construído em torno de versões

Every document in DocumentMS is a record with a history rather than a file with a name. Uploading a revision creates a new version and leaves the previous one intact and retrievable, so "final_v3_revised_FINAL.docx" stops being a filing convention. The current version is marked unambiguously; superseded versions remain readable but are clearly labelled as superseded, which is the distinction quality and safety auditors test for.

Version history is unlimited. That sounds like a storage decision, but it is really a legal one: once you start pruning history you have to be able to justify what was pruned and when, and the cost of keeping it is almost always lower than the cost of explaining its absence.

Documents are organised by category and folder, which most teams can map from their existing structure on day one. That deliberate familiarity is why we did not go metadata-only: a taxonomy nobody recognises gets worked around, and a worked-around system stops being the record.

Alongside the digital hierarchy, document locations track where a physical original lives — building, room, cabinet, box — including its history as it moves. Boxes get relocated during office moves and the paper trail is usually the first casualty; recording location as data rather than as a label on a shelf means an original can still be produced years later.

Checking a document out locks it for editing and tells everyone else who holds it. When it is checked back in, the edit becomes a new version with an author and a timestamp. The alternative — everyone editing a copy and reconciling later — is how organisations end up with two documents that both look authoritative.

Check-out state is visible in the repository listing, not buried in a properties dialog, because the value of a lock is that people can see it before they start work.

Document numbers are generated by the system according to rules you configure per document type, which removes the two failure modes of manual numbering: duplicates and gaps. Numbering that is consistent is what makes a document referenceable in a procedure, a contract or an audit finding.

Watermarking applies a visible marker to previews and downloads — draft status, confidentiality level, recipient identity — so a document that leaves the system still carries its context. It is not a security control on its own, and we do not present it as one; it is a way to make the status of a printed page obvious to whoever is holding it.

Every uploaded file is hashed, and a matching hash flags the upload as a duplicate of an existing document regardless of what it has been renamed to. This matters more than it sounds: duplicated documents are the main reason storage bills grow without anyone deciding to grow them, and they are how a superseded version gets circulated as current.

Detection is a prompt, not a block. There are legitimate reasons to hold the same bytes in two places — a contract filed under both parties, for instance — so the system tells you and lets you decide, then records the decision.

Deleting a document moves it to a recycle bin visible only to account administrators. Ordinary users cannot see or restore the contents of the bin, which is what stops it becoming a shadow repository, and administrators can restore a document with its version history and audit trail intact.

Permanent deletion is a separate, logged action. It exists because retention schedules and erasure requests sometimes require genuine destruction, and a system that cannot actually delete anything fails a different set of obligations from one that deletes too easily.

Documents arrive by upload, by drag and drop, from a scanner, from a watched mailbox that files an email and its attachments into a specific folder, or generated from a template. URLs can be stored and embedded alongside documents, so a reference to a standard hosted elsewhere sits in the same folder as the procedure that cites it.

Whatever the route in, the same things happen on arrival: the file is numbered, hashed against what is already there, OCR-processed if it needs it, and subjected to the retention rule of its document type.