The ideas behind Dédalo
Part of Dédalo in plain language · Previous: Ready for AI — on your terms
Written for curators, archivists, documentalists and humanities researchers. No technical background assumed. Each idea ends with a pointer into the technical manual, for when you want the full version — or want to hand it to whoever will install the system.
Software for cultural heritage usually fails in one of two ways. Either it is a spreadsheet with a nicer face — fast to start, impossible to keep coherent after three years — or it is a rigid database designed by someone who has never catalogued anything, where adding a field is a project. Dédalo is built on a handful of ideas that try to avoid both.
A record, a section, a field
Three words cover most of what you do.
- A record is one thing you catalogue: an object, an interview, a photograph, a person, a place, a bibliographic reference.
- A section holds all the records of the same kind. Coins is a section. Oral testimonies is a section. People is a section. If you come from databases: a section is a table; if you come from archives: it is a series of homogeneous units of description.
- A component is one field on a record — a title, a date, an extent, a photograph, a link to another record.
Components have kinds, and the kind is what makes a field intelligent. A date component understands imperfect dates. A geographic component understands coordinates and shapes. A media component understands that a file has a master, derivatives and a thumbnail. A relation component understands that its content is a pointer to another record, not a piece of text.
More: Sections · Components · Glossary
The shape of your catalogue is data
This is the idea Dédalo is built around, and the one worth understanding even if you read nothing else.
In most systems, the shape of the catalogue — which record types exist, which fields they have, in what order, with which vocabularies, in which languages — lives in the program. Changing it means changing the program: a developer, a release, a budget, a delay.
In Dédalo the shape of the catalogue is itself data, held in a structure Dédalo calls the ontology and edited inside Dédalo like anything else. Adding a field, creating a whole new kind of record, reordering a form, making a field repeatable or translatable, deciding that this field is a controlled vocabulary pointing at that thesaurus — these are acts of documentation, not acts of programming.
Why this matters for an institution
Your descriptive model can follow your discipline instead of the software's assumptions. It can also evolve: a project that starts with a modest inventory and grows into a full scientific catalogue does not have to be rebuilt, and a standard that changes — a new controlled vocabulary, a new aggregator profile — is a change you make yourself.
The trade is honest: this power lives with whoever designs your model, so the design deserves the same care as a cataloguing manual. Dédalo does not stop you from designing a bad model; it stops the software from imposing one.
Words that mean something: controlled vocabularies
A thesaurus in Dédalo is a tree of terms — places, personal names, subjects, materials, techniques, typologies — and the terms in it are ordinary records. That last point is the important one: a term can be described, dated, related, translated and cited exactly like an object, because it is a record.
So a place in your toponymic thesaurus is not the string "Vall de Cabó" repeated across nine hundred records. It is one record, with its own description, its own variants and its own history, that nine hundred records point at.
Links, not copies
When a record refers to another record, Dédalo stores a pointer, never a copy of its name. The consequences are the ones every documentalist wants:
- Correct once. Fix a name, a date, a spelling in the term record, and every record that refers to it is correct — immediately, everywhere.
- Ask backwards. Because the link is real, you can ask which testimonies mention this person? or what came from this excavation? without having planned that question in advance.
- No silent divergence. Two records cannot end up holding two slightly different spellings of the same place, because they hold no spelling at all.
This is the single largest difference between a catalogue and a spreadsheet, and it is what makes a heritage collection still usable after twenty years.
Many languages, two different kinds
Two things get translated in a heritage system, and confusing them causes trouble.
- The data — a title, a description, an abstract can exist in as many languages as your institution works in, each stored separately on the same record. Some values, by contrast, are the same in every language: an inventory number, a coordinate, a year.
- The interface — the buttons, menus and field labels a cataloguer sees.
A researcher can read the catalogue in one language while the data itself is served in another, and a translation workflow can run over the data plane without ever touching the program.
More: Internationalization · Language tools
Nothing is silently lost
Every change to a record is kept: what it was, who changed it, and when. You can read the history of a record and restore an earlier state.
For a scientific catalogue this is not a comfort feature. It is what makes an attribution defensible, a correction traceable and a disputed record reconstructible — and what lets you hand a collection to a colleague without losing the reasoning that produced it.
More: Time machine
Who may see what
Access is not "editor or reader". Permission is expressed per section and per field, at three levels: none, read, read-and-write. So a volunteer can catalogue the physical description of an object while never seeing the donor's name; an external researcher can read a collection they may not touch; a project can be visible to one team and invisible to the rest of the institution.
The same permissions govern every route into the data — the staff application, the public service and any AI assistant. See Three doors into the same collection.
Files are handled like heritage, not like attachments
An uploaded photograph, recording or video becomes a master that is never altered, plus the derivatives and thumbnails the interface needs. Recordings can be transcribed and indexed so that a passage — with its timecode — becomes a citable unit; long texts can carry page references.
Large files are served directly by the web server rather than carried through the program, which is why a multi-gigabyte video is an ordinary case.
More: Media pipeline · Transcription
Publishing is a decision, and it is reversible
Dédalo does not have a "public" switch for the institution. It has a decision per record, and a publication step that copies the marked records outwards into a separate database and into files. Un-marking a record removes it on the next run.
That is what lets an institution catalogue frankly — with the donor's conditions, the internal notes and the unresolved attributions — and still publish confidently.
Measuring what you produce
Dédalo carries a data-quality score, Raspa, that rates a catalogue from 0 to 10 across progressive levels: structured, ontologically modelled, traceable, translatable, openly processable. It is a way to argue about quality with evidence rather than impressions — useful in a funding application, and useful internally when deciding what to improve next.
A short translation table
The manual and your developers will use these words. Here is what they mean in yours.
| They say | You would say |
|---|---|
| section | a kind of record — a series, a register, a table |
| record | one catalogued item, in one section |
| component | a field on the form |
| ontology | the definition of your catalogue: which record types and fields exist, and how they behave |
| tipo | the internal identifier of one piece of that definition |
| locator | a pointer from one record to another |
| relation | a link between two records, one-way or mutual |
| thesaurus / hierarchy | a controlled vocabulary, as a tree of terms |
| descriptor | the preferred term of a concept (as against a non-preferred variant) |
| diffusion | publishing: producing the public copy |
| matrix | the database table where all records are stored |
| time machine | the history of a record, and the ability to restore it |
| API | a door another program may knock on |
| RAG / semantic search | finding records by meaning rather than by exact words |
| MCP | the agreed way an AI assistant may use Dédalo |
The full definitions, with their technical detail, are in the Glossary.
Where to go next
- Dédalo in plain language — back to the four pictures.
- Introduction and core hub — the technical manual's starting point, if you want to keep reading.
- Tools user guide — what your team will actually click on, tool by tool.
- The v7 rewrite — why the engine was rebuilt, written for developers and humanities readers alike.