Promote Eve persona and typography state

This commit is contained in:
Meta Vens 2026-06-16 18:00:56 +01:00
commit 83dee183f7
13 changed files with 113 additions and 35 deletions

View file

@ -136,7 +136,7 @@ The current iOS app consumes Mimir's `/eve/deck` WebSocket broker. That broker i
already shaped like the future API: provider manifests, retained dashboard
state, provider switching, commands, and typed node metadata. The VoidBot
provider proves that Eve can render an app-specific native cockpit from a
structured mesh snapshot: CTB rail, avatar images, selected Face/status pane,
structured mesh snapshot: CTB rail, avatar images, selected Persona/status pane,
state tree, and detail panel.
The Android proof under `android/` builds directly against the installed Android

View file

@ -293,8 +293,9 @@ The portable style document should look like data:
"colorBackgroundTop": "#01040b",
"colorPanel": "#0a2239",
"colorText": "#efe5c9",
"fontTitle": "Press Start 2P",
"fontBody": "VT323",
"fontDisplay": "\"Montserrat\", \"Zen Kaku Gothic New\", \"M PLUS 1\", sans-serif",
"fontTitle": "\"Montserrat\", \"Zen Kaku Gothic New\", \"M PLUS 1\", sans-serif",
"fontBody": "\"M PLUS 1\", \"Ubuntu Sans\", Ubuntu, sans-serif",
"borderWidthPx": 4,
"pixelArt": true,
"scanlineOverlay": true
@ -315,6 +316,17 @@ Rules should target component roles, variants, and state, not arbitrary global
text selectors. If a style cannot explain which component role or variant it
serves, it probably belongs in a runtime lowerer or nowhere.
Typography roles carry script-coverage obligations. GUI body text should use
`M PLUS 1` and `Ubuntu Sans`/Ubuntu for the clean contemporary reading voice.
GUI display text may use Montserrat for Latin, but Japanese display needs a
matching contemporary fallback such as `Zen Kaku Gothic New` before falling
back to a body face.
Pixel text is a TUI raster-font concern, not a GUI webfont substitution. A TUI
lowerer that claims Japanese pixel typography must expose fixed-cell raster
atlases for display, body, and small scales, with Latin and hiragana supported
and katakana supported inside each advertised cell size.
## What CultUI Takes From CSS
Keep:
@ -374,6 +386,9 @@ Each runtime needs one style lowerer:
Runtime lowerers must report capability gaps. If Android cannot yet render
Repixelizer's pixel-art image preview or cleanup canvas, the surface should say
that plainly instead of faking a control that cannot carry the command.
If a lowering target cannot load the declared Japanese-capable display fonts or
TUI raster atlases, it must report that typography capability gap rather than
silently rendering hiragana with tofu or an unrelated default face.
### Parity Unit Boundary

View file

@ -1,7 +1,7 @@
# Eve Face Memory
# Eve Persona Memory
This is the human-readable persistent memory surface for Eve's repo Face. The
typed `.voidbot/state/eve.cc` file remains the live Face-state authority; this
This is the human-readable persistent memory surface for Eve's repo Persona. The
typed `.voidbot/state/eve.cc` file remains the live Persona-state authority; this
document is the stable re-entry map that belongs in the repo body.
## Mission
@ -56,8 +56,8 @@ CultLib/CultMesh is Eve's substrate. If the substrate cannot expose field-level
reactivity, authority, prediction, and reconciliation, Eve's magic becomes a
gesture instead of a machine.
VoidBot and repo Faces are Eve's social field. Eve surfaces should let their
state become legible without flattening each Face into one central dashboard
VoidBot and repo Personas are Eve's social field. Eve surfaces should let their
state become legible without flattening each Persona into one central dashboard
voice.
## Current Doctrine

View file

@ -247,9 +247,29 @@ Minimum common tokens:
- `radiusPanel`
- `fontBody`
- `fontTitle`
- `fontDisplay`
- `pixelArt`
- `motion`
Typography tokens describe roles, not one runtime's installed fonts. `fontBody`
should favor `M PLUS 1` and `Ubuntu Sans`/Ubuntu for GUI lowerers. `fontDisplay`
and `fontTitle` may use Montserrat for Latin text, but must include a
Japanese-capable contemporary display fallback such as `Zen Kaku Gothic New`;
Ubuntu remains the body voice, not the flashy title answer.
TUI pixel typography is a separate lowering capability, not a GUI webfont
fallback. A TUI lowerer that claims Japanese pixel support must expose fixed
raster-cell fonts for each advertised scale, with Latin, hiragana, and katakana
in the same cell grid. The expected capability roles are:
- `tuiPixelDisplay`: large raster display labels, roughly 16px or larger.
- `tuiPixelBody`: readable dense raster text, roughly 12px.
- `tuiPixelSmall`: the smallest readable kana lane, roughly 10px.
If a TUI target lacks a Unicode raster atlas for one of those scales, it should
publish a capability gap instead of silently falling back to tofu, proportional
Japanese text, or a Latin-only console font.
Product surfaces may add scoped token groups such as:
- `streamPixels.characterPreview`