Merge branch 'main' of git.arirex.me:iris/ai-app

AGENTS.md

@@ -457,6 +457,21 @@ machine belongs in `~/.claude/TOOLCHAIN.md` (toolchain versions) or
   the screen opens. The reset is not optional: without it the window is
   spliced onto rows that are no longer adjacent to it, which reads as
   ordinary output.
+- **The five-hour window has no reset time between blocks, and that is not
+  a missing value.** The usage API anchors it to the block it started in --
+  measured 2026-08-31, the reset came back as exactly five hours after work
+  resumed, and the weekly windows in the same response carried the identical
+  microsecond, so both are computed from one `now()` at request time. When
+  no block is running there is nothing to reset and `resets_at` is `null`;
+  the same response shows other idle windows with the same shape. The weekly
+  ones always have a reset because a week is always running, which is why
+  "the others seem fine".
+  So `resets_at` absent means **not running**, and only a timestamp that
+  arrives and cannot be parsed is unknown. The app collapsed both into one
+  null and the session bar said "reset time unknown" for a machine behaving
+  perfectly -- while the usage dialog, reading the same field, quietly drew
+  nothing. `WindowEnd` in `ResetCountdown.kt` is now the one rule both go
+  through.
 - **A transcript page used to cost the whole transcript.** `read_window`
   read and parsed every line and then kept the last `limit` of them, so the
   work was the size of the conversation rather than the size of the answer:
@@ -477,9 +492,24 @@ machine belongs in `~/.claude/TOOLCHAIN.md` (toolchain versions) or
   rows has to measure a screen rather than name a number: the history
   cushion was eight rows, which on a tool-heavy transcript is less than one
   screenful, and the reader hit the end of what was loaded on every swipe
-  and stood there for a round trip. That was "scrolling is laggy" -- not a
+  and stood there for a round trip. It is `HISTORY_SCREENS` viewports now,
-  slow frame. It is `HISTORY_SCREENS` viewports now, counted from what is
+  counted from what is actually on screen. Measured at the server, which is
-  actually on screen.
+  the one number here that does not depend on how the emulator renders:
+  against a 24,000-event transcript, ten swipes asked for ten pages before
+  and three after.
+- **What the transcript screen costs to scroll, for whoever measures it
+  next.** Taken 2026-08-30 on the GPU emulator (`emu up` provides one; a
+  frame number from the software rasteriser means nothing -- see
+  `~/.claude/MACHINE.md`), against a real imported transcript with the debug
+  server at `--delay 120`. Settled and flinging fast, both into fresh
+  history and back through rows already drawn: **5.2-5.9% janky frames, 99th
+  percentile 29-32ms, 0-2 slow UI-thread frames.** The stock Settings app on
+  the same device is 3.3% and 38ms, so this is at the platform floor and
+  what is left is the emulator rather than the app. The number that is *not*
+  at the floor is the first few seconds after opening a session, where every
+  row on the way is being composed for the first time; that is inherent to a
+  lazy list and it is why a measurement taken before the screen settles
+  reads three times worse. **Settle first, then reset `gfxinfo`.**
 - **Only `fetchTranscript` was off the main thread; the fold was not.**
   `foldEvent` returns a new list per event, so a page is that many copies of
   a growing list -- fine at 80 events and about 300,000 element copies at

app/androidApp/src/main/kotlin/com/example/aiapp/ResetCountdown.kt

@@ -20,18 +20,42 @@ fun formatSpan(until: Duration): String =
     }
 
 /**
- * Time from [now] until [resetsAt], or null when that timestamp cannot be read.
+ * What is known about when a usage window ends.
  *
- * Null rather than a zero duration, because a string this app failed to parse is not a window that
+ * Three answers rather than a nullable duration, because two of them shared `null` and they are not
- * has just run out: a caller given zero for both would tell the reader to refresh on the strength
+ * the same thing at all. A window the server sent no reset time for is one that is **not running**:
- * of something nobody measured.
+ * the five-hour window is anchored to the block it started in, so between sessions there is nothing
+ * counting down and the API says so by omitting the field -- measured against a live response on
+ * 2026-08-31, where the five-hour window's reset was exactly five hours after the moment work
+ * resumed. A timestamp that did arrive and could not be read is the genuinely unknown case, and it
+ * is the only one worth those words.
+ *
+ * Collapsing them put "reset time unknown" on the session bar for a machine behaving perfectly, on
+ * the one row somebody reads before starting something big -- and the usage dialog, looking at the
+ * same field, quietly drew nothing. Two rules for one missing value; this is the rule.
+ */
+sealed class WindowEnd {
+    /** No reset time was sent, so nothing is running in this window. Not a failure to find out. */
+    data object NotRunning : WindowEnd()
+
+    /** A timestamp arrived and could not be read. The one case that is actually unknown. */
+    data object Unreadable : WindowEnd()
+
+    /** How long is left. Negative once the window is past, which each caller words for itself. */
+    data class Ends(val until: Duration) : WindowEnd()
+}
+
+/**
+ * [resetsAt] as the server sent it -- absent, unreadable, or a moment -- against [now].
  *
  * [now] is a parameter rather than read here so a caller can drive it from state and have the
  * countdown recompute on its own schedule.
  */
-fun remainingUntil(resetsAt: String, now: OffsetDateTime): Duration? =
+fun windowEnd(resetsAt: String?, now: OffsetDateTime): WindowEnd {
-    try {
+    if (resetsAt == null) return WindowEnd.NotRunning
-        Duration.between(now, OffsetDateTime.parse(resetsAt))
+    return try {
+        WindowEnd.Ends(Duration.between(now, OffsetDateTime.parse(resetsAt)))
     } catch (_: Exception) {
-        null
+        WindowEnd.Unreadable
     }
+}

app/androidApp/src/main/kotlin/com/example/aiapp/SessionScreen.kt

@@ -88,13 +88,15 @@ private val LOADING_SPINNER = 48.dp
  * row is anything from one line to a page and a fixed count is therefore a distance only by
  * accident. Eight rows was the number, and on a tool-heavy transcript eight rows is less than one
  * screen: the reader reached the end of what was loaded on *every* swipe and waited a round trip
- * standing there. That is what "scrolling is laggy" turned out to be -- not a slow frame, but the
+ * standing there, which is a list running out of transcript rather than a slow frame.
- * list running out of transcript, which the emulator showed as a swipe that moved nothing for 689ms
- * and then jumped.
  *
  * Three, so a fling lands on rows that are already there and the page after them is on its way. The
  * cost of being generous is a page fetched that nobody reads; the cost of being mean is a list that
  * stops under a finger, and those are not the same size.
+ *
+ * Counted at the server rather than inferred from the screen, which is the only measurement here
+ * that does not depend on how the emulator renders: against a 24,000-event transcript at `--delay
+ * 120`, ten swipes asked for **ten** pages before this and **three** after.
  */
 private const val HISTORY_SCREENS = 3
 

app/androidApp/src/main/kotlin/com/example/aiapp/SessionUsageBar.kt

@@ -185,19 +185,22 @@ private fun UsageNote(text: String) {
  * The percentage on its own does not answer the question it gets asked, which is whether to start
  * something now; 80% with twenty minutes to go and 80% with four hours to go are opposite answers.
  *
- * The window's end has its own missing case, kept apart from the rest: a snapshot can arrive with
+ * The window's end has two missing cases and they are worded differently on purpose; see
- * no reset time, and saying "refresh soon" there would put a recommendation on the screen that
+ * [WindowEnd]. A window that is not running gets the percentage and nothing else, because there is
- * nothing measured. The percentage is still known, so it is still shown.
+ * no countdown to report and inventing one would be the same fault as inventing the number.
  */
 private fun fiveHourLabel(window: UsageWindow, now: OffsetDateTime): String {
     val percent = "${window.percent.toInt()}%"
-    val until = window.resetsAt?.let { remainingUntil(it, now) }
+    return when (val end = windowEnd(window.resetsAt, now)) {
-    return when {
+        // Between blocks the five-hour window has no reset time, and saying so is a fact about
-        until == null -> "$percent · reset time unknown"
+        // nothing: there is no window to run out. The percentage is the whole answer.
-        // Under a minute, including past the end: the number would round to "0m left", which reads
+        WindowEnd.NotRunning -> percent
-        // as a measurement rather than as the window having run out.
+        WindowEnd.Unreadable -> "$percent · reset time unreadable"
-        until < Duration.ofMinutes(1) -> "$percent · refresh soon"
+        is WindowEnd.Ends ->
-        else -> "$percent · ${formatSpan(until)} left"
+            // Under a minute, including past the end: the number would round to "0m left", which
+            // reads as a measurement rather than as the window having run out.
+            if (end.until < Duration.ofMinutes(1)) "$percent · refresh soon"
+            else "$percent · ${formatSpan(end.until)} left"
     }
 }
 

app/androidApp/src/main/kotlin/com/example/aiapp/UsageDialog.kt

@@ -204,10 +204,10 @@ private fun WindowBar(window: UsageWindow) {
             color = quotaColor(window.percent),
             modifier = Modifier.fillMaxWidth(),
         )
-        window.resetsAt?.let {
+        resetLine(window)?.let {
             Spacer(Modifier.height(2.dp))
             Text(
-                "resets ${formatReset(it)}",
+                it,
                 style = MaterialTheme.typography.bodySmall,
                 color = MaterialTheme.colorScheme.onSurfaceVariant,
             )
@@ -215,8 +215,19 @@ private fun WindowBar(window: UsageWindow) {
     }
 }
 
-/** "in 3h 12m" -- close enough for deciding whether to start a big task. */
+/**
-private fun formatReset(resetsAt: String): String {
+ * "resets in 3h 12m" -- close enough for deciding whether to start a big task -- or nothing.
-    val until = remainingUntil(resetsAt, OffsetDateTime.now()) ?: return "at $resetsAt"
+ *
-    return if (until.isNegative) "soon" else "in ${formatSpan(until)}"
+ * Null for a window that is not running, which is the case this row has always drawn as nothing and
-}
+ * is right to: there is no end to report. What it used to get wrong is the other missing case, a
+ * timestamp that arrived and could not be read: that was printed raw, so a parse failure appeared
+ * as an ISO string in the middle of a sentence written for a person. Both cases are named in
+ * [WindowEnd], and the session bar words them the same way.
+ */
+private fun resetLine(window: UsageWindow): String? =
+    when (val end = windowEnd(window.resetsAt, OffsetDateTime.now())) {
+        WindowEnd.NotRunning -> null
+        WindowEnd.Unreadable -> "reset time unreadable"
+        is WindowEnd.Ends ->
+            if (end.until.isNegative) "resets soon" else "resets in ${formatSpan(end.until)}"
+    }