Screens

22 screens described. A screen is a page of the application, with what it shows and what it allows.

ScreenRolesFirst lineLinks
About this wikiReaderThe page of trust: everything that lets a reader judge whether the site read is current. When it was built, from which repositories, at which version each was read, what the site leaves out and how a page is corrected. The footer of every page links to it, "See the sources and their versions".
Alphabetical indexGlossary owner, ReaderEvery word the documentation uses, with or without a definition, by initial letter. Letters without an entry stay in view, struck through and inert. Each word is listed with its type, the first line of its page and the number of pages citing it; a word without a definition is dotted and described by the passage that uses it most. Sorting follows the collation rules of the project locale, accents included, and the words without a definition are never set apart: they are the working list of a glossary owner.
API pageIntegrator, ReaderThe page of an interface whose contract the build imported. Same shell as the entity page, laid out at the grain an interface is worked at: its operations. The tree lists them under the interface, a table matches them to the contract and states the gaps between the contract and the notes, the contract stands in a block of its own, shown and never copied into the text, and the panel keeps five keys and lifts the operations to the top of the related pages. An API without an imported contract keeps the generic entity page.
CategoryReaderThe list of one category of a space: a folder of the repository at any depth, or a year or a month of a space whose every note is dated, where a card of the space page, a folder of the tree or a step of the breadcrumb leads. A category does not unfold in the tree, it opens as a page, the missing link between the space and the note.
Colour schemeReaderThe switch of the bar of every page that takes the site from its light palette to its dark one and back. The dark scheme is a second palette, not an inversion of the light one: its colours are measured on their own, and nothing a reader does in one scheme is lost in the other. The choice follows the system preference, stays remembered once made, and the switch stands in the same place in both schemes.
Decision pageReaderThe page of an architecture decision: short, dated, with its consequences. Same shell as the entity page: the tree of the space on the left, the title and the panel where a reader expects them. A reader comes to check two things, whether the decision holds and since when, and finds them at the head of the panel; the rest of the page says what the decision changed.
Document pageReaderThe page of an office document, a slide deck, a report, a spreadsheet or a PDF, alone or merged with its note. Same shell as the entity page: the tree of the space on the left, the title and the panel where a reader expects them. Where the entity page puts the note, the document itself stands, its text and its notes one tab away, the original one click away.
Edge casesReaderWhat shows when something is missing. A static site has no support desk: the page is the only recourse. It names the cause in plain words, says what stays reachable despite the lack, and offers at least two ways out; no partial lack hides the rest of the page, and the navigation is always kept. A notice is allowed on these states alone, nowhere else in the site.
Entity pageReaderThe page of a note, laid out as a wiki a corporate reader already knows: the text in the centre, what the tool computed in the right panel. One template serves every type; only the labels of the attributes and the order of neighbours change, and both come from the profile. A type may have a page of its own: a component a theme or the type's module provides for it is rendered instead of the generic template, which stays the fallback of every type.
Exploring the siteThe path a reader follows through the generated site, from a word to the notes that use it and back, without a server and without leaving the browser. Every step is a page of the site, so the process is filed among the screens it walks through.
Gallery indexTheme authorThe page set concordance gallery writes so that a theme is styled and checked without building a corpus: every slot of the site rendered with fixture view models, one page per slot and state, under an index that says who renders each slot.
HomeReaderTwo ways into a corpus: a word, or a space. The page opens on a question, "What are you looking for?", and a search field that answers it as the reader types, because every word used anywhere in the documentation has a page, defined or not. The most cited pages stand next to it as shortcuts. Then the spaces, the sources the site is fed by, each with its freshness from the git history, and the pages changed last. A space that has not moved past the staleness threshold is named in an alert.
Keyword pageGlossary owner, ReaderThe page of a word nobody defined. Same template as the entity page, without the markdown and the declared properties. A banner says that no note exists and how many passages were recorded. Three numbers only: occurrences, files, sources. The passages follow, grouped by page in corpus order, two in view per page, with their context. The words that accompany the expression are its neighbourhood map, drawn from its co-occurrences, and expressions with a similar form are offered as a lead, worded to assert nothing.
Meeting pageReaderThe page of a working session that exists as a note, a transcript and a deck, read as one page. Same shell as the entity page: the tree of the space on the left, the title and the panel where a reader expects them. Where the entity page puts the note, the three files stand as tabs; what the meeting produced, a decision written elsewhere, is named in a callout rather than mixed into the text.
Mentions panelReaderThe block of the entity page headed "Related pages", which answers one question: which pages evoke this entity, and did a person write a link or did the tool recognise the words. One entry per related page, ordered by number of passages, written links and recognised words counted alike; an entry is marked "Cited" when its page writes a link. A page that nothing links to, a keyword page, a meeting or a document, lists the pages its own passages evoke instead. The first twenty mentions are in the served HTML; the rest loads from a JSON fragment specific to the entity, never from a global index. The block offers a text filter and a type filter, and without JavaScript the served pages stay readable and their links work. The threshold of twenty is configurable.
Neighbourhood mapReaderThe small map of the entity page that shows the closest entities at one hop, all labelled, and never stands alone: a list next to it carries the same information in text. Deliberately modest, six nodes by default and twelve at most; beyond the cap the map draws the nodes kept and says how many neighbours the page has in total, and a page without any neighbour points at the mentions panel instead.
Pinned pagesReaderThe row of tabs under the bar of every page of the generated site, holding the pages a reader chose to keep: pinned from the header of a page, unpinned by hand, kept in the browser from page to page and from one visit to the next. Tabs, but chosen: nothing preloads, nothing closes by mistake, and the selection survives a reload. The exploration trail that once followed the pages visited is gone, because it had no rule to start over; the one breadcrumb left is the filing of the page.
SearchReaderFull-text search over the whole corpus, without a server. The index is generated at build and loaded in fragments as the user types. Facets on type, source, domain and application carry counts frozen at build; they combine, and a facet with no result is disabled rather than hidden. The query and the active filters live in the URL, so a search can be sent to a colleague and replays exactly.
Side panelReaderEach of the two columns beside the text of the entity page, the tree of the space on the left and the panel of what the tool computed on the right, and the handle on its edge that folds it. Folding gives the room to the wide content of the note, tables, figures, code and documents, while the prose keeps its measure; the browser remembers what a reader folded, page after page, and nothing of it enters the address.
SpaceReaderThe page of one source, where a row of the home or of the spaces page leads: one changes page, nothing unfolds. A space reads like a small wiki within the wiki: its own search, its own vocabulary, its own news.

20 screens of 22 — pagination by twenty. The Links column counts the related pages, which brings the most central screens of the journey to the top.