docs: plan editorial presentation redesign
This commit is contained in:
@@ -0,0 +1,34 @@
|
||||
# Adopt Tailwind Additively For Source-Owned UI
|
||||
|
||||
Status: accepted
|
||||
|
||||
The web workspace will adopt Tailwind and shadcn-compatible conventions for
|
||||
new source-owned chat and presentation primitives because established AI-chat
|
||||
components provide better interaction and accessibility than maintaining a
|
||||
custom message surface. Existing console CSS will not be rewritten wholesale;
|
||||
old styles migrate only when their components are replaced.
|
||||
|
||||
## Considered Options
|
||||
|
||||
**Keep all styling hand-written.**
|
||||
This avoids another styling dependency, but repeats low-level work for chat,
|
||||
forms, focus states, and responsive composition. The existing presentation
|
||||
already demonstrates the cost of one-off component styling.
|
||||
|
||||
**Rewrite the whole console with Tailwind and a component suite.**
|
||||
This creates one styling mechanism eventually, but turns presentation work into
|
||||
a risky console migration. It also lets framework defaults replace deliberate
|
||||
product decisions before those decisions are ready.
|
||||
|
||||
**Adopt a complete AI application shell.**
|
||||
This supplies familiar chat interactions quickly, but also imports model,
|
||||
account, attachment, and transport assumptions that do not match the existing
|
||||
`AgentDriver` and workflow RPC boundaries.
|
||||
|
||||
## Consequences
|
||||
|
||||
Tailwind enters through the official Vite integration without Preflight, so it
|
||||
does not reset existing `/console` CSS. New source-owned primitives may follow
|
||||
shadcn-compatible structure and copy narrowly selected open-source components,
|
||||
while existing surfaces migrate only when replaced. The project retains
|
||||
ownership of component code, accessibility, tokens, and domain behavior.
|
||||
@@ -0,0 +1,33 @@
|
||||
# Use An Editorial Canvas Around Shared Product Surfaces
|
||||
|
||||
Status: accepted
|
||||
|
||||
The defense presentation will keep one warm editorial identity and temporarily
|
||||
expand real shared console components within a persistent presentation frame.
|
||||
It will not switch the entire deck into a dark product theme or maintain
|
||||
presentation-only copies of product UI, because those approaches break
|
||||
narrative continuity and make the demo less credible.
|
||||
|
||||
## Considered Options
|
||||
|
||||
**Switch between light slides and a dark application stage.**
|
||||
This makes product moments visually distinct, but repeated whole-stage theme
|
||||
changes feel like separate applications and make transitions harder to follow.
|
||||
|
||||
**Build presentation-only product replicas.**
|
||||
This gives complete control over appearance, but creates evidence that is
|
||||
visually persuasive while no longer being the actual product surface. Fixes and
|
||||
interaction behavior would also diverge between `/console` and `/present`.
|
||||
|
||||
**Keep the current dashboard shell as the presentation.**
|
||||
This preserves working behavior, but persistent rails, controls, labels, and
|
||||
card grids compete with the argument and consume the limited `1280x720` frame.
|
||||
|
||||
## Consequences
|
||||
|
||||
The audience view uses one fixed, scalable `1280x720` Editorial Canvas. Shared
|
||||
console graphs, receipts, evidence, and approval forms may expand within it;
|
||||
presentation code owns framing and choreography rather than alternate product
|
||||
implementations. Locally dark terminal, code, or evidence insets remain valid
|
||||
when they communicate content type, but the main sequence does not switch its
|
||||
global theme.
|
||||
Reference in New Issue
Block a user