56 lines
3.2 KiB
Markdown
56 lines
3.2 KiB
Markdown
# Run summary: title spacing, section gap, and the server's title cap
|
|
|
|
Item: the run summary's title was reported glued to the top of the pane, glued to
|
|
the artifacts section, and "cut off". Android-only change (`android/`).
|
|
|
|
## What changed
|
|
|
|
- `android/app/src/main/java/net/awan/pipelines/ui/screens/RunSummaryTab.kt`: the
|
|
summary now leads with the run's identity block (title, status row, facts) and
|
|
the artifacts collapsible section follows it; the identity block carries the
|
|
screen's standard top spacing above it and the standard stack step below it.
|
|
The title keeps `maxLines = RUN_VIEW_TITLE_LINES` (`Int.MAX_VALUE`) and
|
|
`TextOverflow.Clip`.
|
|
- `android/app/src/main/java/net/awan/pipelines/ui/theme/Dimens.kt`: two named
|
|
tokens, `SUMMARY_TOP_SPACING` (= `SPACING_L`) and `SUMMARY_SECTION_GAP`
|
|
(= `SPACING_S`); no new number, no raw `dp` at the call site.
|
|
- `android/app/src/test/java/net/awan/pipelines/ui/RunTitlesTest.kt`: the
|
|
run-identity/title-line test gained the source-scan cases pinning the title's
|
|
budget/`Clip`, the identity block leading the artifacts section, its two
|
|
spacing tokens' call sites, and their values.
|
|
|
|
## Confirmed runtime limitation: a run title is capped at 80 runes by the server
|
|
|
|
The "cut off" title is not app-side clipping. The app renders the whole string it
|
|
is given: `runLabel` (`ui/RunIdentity.kt`) does not shorten a title and the
|
|
summary's title line takes `RUN_VIEW_TITLE_LINES = Int.MAX_VALUE` with
|
|
`TextOverflow.Clip`. The string itself is bounded before it reaches the app:
|
|
|
|
- `internal/summarise/summarise.go` defines `TitleCap = 80` runes and applies it
|
|
through `capRunes` in `Title` (model-derived), `DeriveTitle` (prompt fallback)
|
|
and `ExplicitTitle` (`--title`/rename), so every stored run title is already
|
|
capped when the server persists it.
|
|
- Evidence (18 Oct 2026): `pipelines inspect 347` returns
|
|
`"title": "Two chat UX gaps: the working indicator does not distinguish thinking from waiti"`
|
|
— exactly 80 runes, ending mid-word — while that run's input prompt's first line
|
|
is 82 runes (`… from waiting`). A `uiautomator` dump of the run summary shows
|
|
the same 80-rune string, i.e. the app holds exactly what the server stored.
|
|
|
|
So a title can end mid-word by design of the server's cap, and no Android change
|
|
can render the missing characters. Widening or raising `TitleCap` is a
|
|
Pipelines-server change (Go), outside this item's Android-only scope and not
|
|
something this item deploys.
|
|
|
|
## Documentation now stale (for `docs-fold`)
|
|
|
|
- `android/README.md`, section "Run title readability: whole in the run view, two
|
|
lines in the list (17 Oct 2026)" (~line 6278): says the run view "guarantees the
|
|
operator can read it whole". It should also state that the stored title is
|
|
itself capped at 80 runes by `internal/summarise/summarise.go` (`TitleCap`), so
|
|
"whole" means the complete stored title, which can end mid-word; only the app's
|
|
own renderer is guaranteed not to shorten it.
|
|
- `android/README.md`, section "Collapsible artifacts gallery in the run summary
|
|
(16 Oct 2026)" (~line 4130-4136): "The order is unchanged: the artifacts section
|
|
still leads the summary and the facts item follows it" is now wrong — the
|
|
summary leads with the identity block and the artifacts section follows.
|