part 39
Previews in the settings: a mockup drawn by the game's real renderer
Seven of the 24 appearance settings have no text at all. An analysis of 1 November 2026 proposes one pinned preview per card, as in the Lines configurator, and the mockup draws it with the game's real renderer. So far only the first step is in the game.
The settings window of Audamir has 24 items of appearance and controls, and the name of most of them does not say what changes when it is switched. The analysis of 1 November 2026 (document docs/SETTINGS-PREVIEW.md) summed it up like this:
the settings are already quite large, with many options; a preview of the visualisation will be needed, like the configurator of Rokgard-Lines has. Then it will be clear what a setting does.
Rokgard is the realm of the board games Colony and Lines; the article on realms says what a realm is. The analysis found that seven settings have no word in the catalogue, and that the game can already draw a picture for almost every one of them, because the board, the dice and the tokens are drawn by the real BoardRenderer. All twelve decisions were approved as recommended. Only the first step has so far got into the game: the panel on the Kastgard card in version 0.82.2.
What was measured: seven settings without a word
In the code (js/core/settings.js, version 0.81.2) the group look has 24 records. The ⚙ window divides them into three cards: All realms has 9 settings, Kastgard 8 and Your hero · settings 7.
The document counts how much text each setting has in the catalogue:
Without text remain Player colours, Your colour, Theme, Sound, Animation speed, Dice-count numbers and Avatar in the corner. The analysis worked with 23 records from state 0.80.1. The twenty-fourth setting, Avatar motion (avatarMotion), came in version 0.81.2 and has descriptions of both its values, so today seven of the 24 are fully explained.
The document concludes from this that names are not enough. It proposes a pair: a picture next to the setting, and 14 new descriptions where the game has no text for the value (decision D7 below).
What can be drawn
For each of the 24 records the analysis finds what changes and where, and divides them by the way the preview works:
| Kind | Count | What it is | Settings |
|---|---|---|---|
| Live | 10 | the picture changes together with the value | highlight of conquests, dots, numbers, chances on the borders, player colours, your colour, player markers, avatar in the corner, avatar motion, live picture |
| Demo ▶ | 7 | the button plays 1 – 2 seconds of the behaviour | attack confirmation, warning at the end of a turn, animation of opponents, animation speed; three music settings with a tone |
| Sample sentence | 3 | a sentence in which the value can be seen | avatar comments, opponents' comments, gender |
| No preview | 4 | nothing to draw | Language, Theme, Sound, Live music |
More important than the division is that the game already draws the picture almost always. The preview scene in the mockup is a board of eight hexagons given as data: the mask, the owners, the dice counts and two conquests. The board is made by Grid.fromMask('hex', …) and drawn by BoardRenderer, the same one the replay player uses (replayview.js, line 222). Frames, dots, numbers and player markers are flags the renderer already knows. The chances on the borders are computed by a small function from the rule of the game, which takes the draw rule as a parameter. The drawing of the board had been pulled out into a shared kit earlier, which is the subject of the article on the board kit.
Where the game has no picture, the analysis says so. Music: the button “Listen” would need a function that plays four bars of the chosen style, and there is no such function in js/. Decision D10 therefore postpones the music to step P5, and the volume may use a tone.
The only “deliberately bad state” is at the player colours: the switch “Colour vision: normal / red-green (simulated)” puts an SVG filter on the canvas, and red and green merge into one colour. It shows why there is a second palette for people with colour blindness, which a sentence would not say so easily.
Model: the Lines configurator, but the picture follows the row
The note on the first configurator of Lines from 31 October 2026 (DESIGN §96) said that the preview panel is good, but impractical for many settings. The wish was to keep it always in view, so that the picture is redrawn at once and the settings can be scrolled. From that came ConfigPane (js/ui/config-pane.js, 53 lines, DESIGN §96.8): from 820 px width two columns with the pinned preview on the right; below this width a compact preview at the top that can be folded away. On a phone of 375 × 812 the Lines preview takes 266 px, that is 33% of the height.


The analysis wants to use the panel as it is, with one difference. In Lines all controls change one picture (the board), while in the settings 24 controls make about eight kinds of picture. Therefore:
- each card has one place for a preview and a description under it;
- the scene follows the row the player touched (click, Tab, the mouse over the row, a change of value); rows that share a scene (the four board settings of Kastgard) do not swap the picture, they only change what it shows;
- the card opens with the scene of its first row, so the place is never empty;
- the scene never writes a setting, it only reads the configuration;
- rows without a scene (Language, Theme, Sound, Live music) have only a description; the ⚙ window widens from 600 to 1,040 px when the card has a preview.
The pinned panel was chosen. On a phone it takes 195 px (24% of 812) with the four board settings and 257 px (32%) with a demo or on the hero screen, which is a little less than the 266 px of Lines. The numbers are from measuring the mockup at a real 375 px.
Architecture: the preview is data and the core does not know the realm
The analysis rests on the rules of DESIGN §111, by which the core does not name a realm and the data travels, the code does not. The realm writes into its manifest which scene belongs to which setting, and gives the API a hook that makes the scene:
// realm manifest, next to `pages`
previews: { changes: 'board', dicePips: 'board', showNumbers: 'board', hints: 'board', ... }
// API hook, next to `page(id, host, ctx)`
preview(scene, host, ctx) -> { update(cfg, key), frame?(now), destroy() }
The platform has its own table for its keys (animation speed, the three music settings, gender) and for the four without a preview. The scenes are those of every running realm that registered them: the player colours get two scenes in a build with Kastgard and Rokgard, and one in a build with a single realm. A test reads the source of the core files, to find no realm name in them. A second test says that **every key of the group look has a decision** (a scene, or an explicit “none”), so a 25th setting cannot come in without an explanation, just as lookCards does not let in a setting without a card.
| Where | What | Lines (estimate) |
|---|---|---|
js/ui/settings-preview.js (new) | panel in the card, the “follow the row” logic, description, composition; platform scenes | 220 |
js/ui/preview-board.js (new) | board host: canvas, renderer, loop, resizing | 70 |
js/realms/kastgard/ui/previews.js (new) | board scene (eight cells as data), chances | 90 |
js/realms/rokgard/ui/previews.js (new) | scene with tokens | 40 |
existing files and style.css | register, windows, CSS | 60 + 60 |
The four new files have 420 lines together, and 29 text keys in nine languages will be added, that is 261 strings: 3 for the panel, 14 descriptions and 12 for the demos. The first step needs seven keys (63 strings). All the numbers are an estimate from the mockup, not a measurement of the game.
The description appears under the picture only where the game has no text for the value, for example with the hints turned off, where the game says “No chances are shown on the borders.”. One of the 14 descriptions belonged to the provisional Avatar motion; today that setting has its own descriptions for both values, so 13 remain.
The mockup

docs/mockups/settings-preview.html is a proof, not a picture. It does not imitate the board, the dice and the tokens, the game's real renderer draws them. The controls change data and the renderer draws it. Of the 233 kilobytes of the file, 144 kilobytes are the game's code, copied without change from state 0.80.1. The bar of the mockup switches the window, the place of the preview, the phone width, the theme, the language and the build.
According to the document, the mockup was checked in the browser pane at 375 and at 1,100 px, in the dark and the light theme, in English and in Slovak:
| Check | Result according to the document and its notes |
|---|---|
| walk through the states at 375 px (language, theme, build, place, window, each row) | 720 states, 0 errors, 0 horizontal overflows |
| state with the ▶ Show button pressed | 92 states without an error |
| the real Tab key | moves the row and the preview |
| console | clean |
| repetition after adding the 24th setting | 576 states without an error (notes on the mockup; the document stayed at 720) |
The counts 720 and 576 cannot be reconciled, because the walk-through script was not saved. Both come from the notes on the mockup, not from a repetition in this article.
The mockup itself says what is real and what is an example. Real: the renderer, the dice, the tokens, the frames, the chances, the texts from the catalogues and the sentence with the forms of gender. Examples: the number 91% in the attack-confirmation sentence, the base times of the demos (900 and 220 ms are an estimate, the coefficients 1.7 / 1 / 0.45 / 0 are from the game), the tone behind the Listen button (not a music style), the live picture of dots, and the labels LIVE, DEMO and LINE next to the rows.
Decisions and plan
On 1 November 2026 all twelve decisions were approved as recommended, one by one. They concern the pinned panel on the card (D1–D3), the four settings that have no scene (D4), the bad state only for the palette (D5), the demos ▶ (D6), the 14 new descriptions (D7), the phone (D8), a scene for every realm in the hero window (D9), the music at P5 (D10), ownership (D11) and the 1,040 px window (D12).
The plan has six steps, each merged separately:
| Step | What | Estimate |
|---|---|---|
| P1 | Kastgard card: register, panel, board host, scene for conquests, dots, numbers and chances; 7 texts | 1.5 – 2 d |
| P2 | Your hero: colours, markers, avatar and its motion; Rokgard tokens | 1.5 d |
| P3 | demos ▶: attack confirmation, end-of-turn warning, opponents' animation, speed | 1.5 d |
| P4 | sample sentences: avatar comments, opponents' comments, gender | 1 d |
| P5 | music and live picture (optional; a new function for four bars) | 1 – 1.5 d |
| P6 | freeze: test of all keys, check at 375 and 1,100 px, chapter in DESIGN | 0.5 d |
The sum is 6 – 6.5 days without P5 and 7 – 8 days with it. Steps P1 to P4 touch the same four files and catalogues, so one session does them at a time.
Where it stands today and what was not measured
From 1 November 2026 (version 0.82.2) the first step P1 is in the game: the Kastgard card in the settings window has a pinned panel with a board of eight hexes, drawn by the real BoardRenderer, and the scene follows the row the player is on. Before the merge came checks by tests and a test in the browser. Nothing else from the previews is in the game yet: the hero settings, the demos, the sample sentences and the music (P2 to P5) had not been built at that date. Added on 3 November 2026: what has been built since is in the forthcoming article no. 45; this text stays as the state of 1 November.
The analysis itself lists what it did not measure. While building P1 three things were measured (DESIGN §128.3); the numbers were not recalculated in this article:
| What the analysis did not measure | Answer from step P1 |
|---|---|
Does the real ConfigPane hold the pinning inside the window scroll? | Yes, unchanged: the preview stays at top: 0 (one column) and top: 12px (two), while the rows scroll. The CSS of the mockup is not needed. |
Does StacksBoard.view(game) take a minimal game? | Yes: a grid, cells, players, rules and a round are enough. It gives real chances (24%, 45%, 91%); the hand-assembled view of the mockup disappeared. |
| How much does the canvas with the loop cost in the window? | 19 – 22 frames per second at 0.09 ms each, that is 0.17% of the main thread, while the conquest frame pulses; on a hidden card 0 frames in three seconds. |
Two things, on the other hand, did not show up in the mockup and only the real code showed them. The 900 px limit gave the panel 819 px, one under 820, and therefore the wide window switches on from 920 px. And the name of the setting in the header of the compact panel broke into syllables in Russian; this was fixed by a separate line for the name.
What could not be verified and what remains
- Comprehensibility. It was not measured whether the previews make the settings clearer; the starting point is the number of texts in the catalogue.
- A phone and a screen reader. The figure of 19 – 22 frames is from a desktop; the canvas with the loop was not measured on a phone, and the behaviour with a screen reader was not checked.
- Languages. The lengths of texts in German and Russian were measured only for the Russian name in step P1.
- A scene from territories. The default maps of Kastgard are territories of the type
blob; a scene with territories is not drawn. - Contradictions in the documents. Chapter 30 of
docs/ROADMAP.mdstill has “12 decisions wait for the user” in its heading, although §6 already records the decisions. The mockup uses the valuesrotatingandstaticfor Avatar motion, the codeturningandstill. - Remaining are P2 to P6 (hero settings, demos, sample sentences, music, freeze).
Sources and notes
- Analysis:
docs/SETTINGS-PREVIEW.md(1 Nov 2026), summarised also indocs/ROADMAP.md, chapter 30. The rule “the core does not name a realm” is in the same analysis and indocs/HOME-FINAL.md, both referring to DESIGN §111. - Numbers about the settings:
js/core/settings.js(version 0.81.2),js/i18n/en.js; counts of 1 Nov 2026. - Mockup and step P1:
docs/mockups/settings-preview.html, DESIGN §128. - Related articles: From one game, more games: realms, rulesets and a refactor that must not change anything, Two games on one board: the kit behind Colony and Lines.