| 1 | # 🤖 fichas |
| 2 | |
| 3 | The Oficina suite's card cabinet. One instance per team. |
| 4 | |
| 5 | A *ficha* is an index card; a *fichero* is the cabinet it sits in. |
| 6 | |
| 7 | **Nothing is built.** This repo is a design document and this README. |
| 8 | There is no module, no binary and no schema yet. |
| 9 | |
| 10 | ## What it is meant to be |
| 11 | |
| 12 | Typed records — *cards* — with append-only *entries* attached to them, and |
| 13 | references out to items in Docs, Sheets and Calendar. The first stack is a |
| 14 | CRM: a Customer card, a Call Log entry, and a summary on the card that is |
| 15 | derived from its entries and cites them. |
| 16 | |
| 17 | The ancestor is DabbleDB, not Microsoft Access. What is worth stealing is |
| 18 | schema-by-refinement, entity promotion and views-as-first-class-objects — |
| 19 | not the generic-database-builder positioning, whose best execution still |
| 20 | found no market. The part that could not have been built in 2008 is that |
| 21 | the model replaces HyperTalk: you reshape a card type in a sentence rather |
| 22 | than in a dialog. |
| 23 | |
| 24 | ## Why it is not part of memoria |
| 25 | |
| 26 | memoria is not an app of record and its decision 7 keeps it that way — it |
| 27 | serves nobody to ask, so it cannot own content that needs an owner, a |
| 28 | deletion story and a per-item visibility answer. Fichas owns its content |
| 29 | and answers for it; memoria indexes it, cites it and reaches into it like |
| 30 | any other sibling. |
| 31 | |
| 32 | Fichas would also be **the first app to serve the suite contract's server |
| 33 | half** — `suite/README.md` says that half "arrives with the first app that |
| 34 | serves one, and the fixtures get corrected by the second". memoria is the |
| 35 | waiting consumer. |
| 36 | |
| 37 | ## The design |
| 38 | |
| 39 | `docs/superpowers/specs/2026-09-19-fichas-design.md`. |
| 40 | |
| 41 | Read §1 first: half of its decisions table is marked **Proposed** and has |
| 42 | not been ruled on. A proposal read as a ruling is the failure this suite's |
| 43 | specs work hardest to prevent. |
| 44 | |
| 45 | Two things block slice 1, both in §10 — the seal-tier list (which card |
| 46 | fields leave the seal, and therefore what can be filtered, sorted and |
| 47 | grouped), and the reach credential ruling. |
| 48 | |
amadan