fix(aggregator): attribute a midnight-straddling turn's By Activity to today on the all-provider view

CHANGELOG.md

@@ -3,7 +3,8 @@
 ## Unreleased
 
 ### Added
-- **Credit-metered ChatGPT workspaces (Business / Edu / Enterprise) now show their limit.** These plans report no rate-limit windows, so the admin-set monthly allowance from `spend_control.individual_limit` is shown as a "Monthly usage limit" bar in the desktop app and the menubar.
+- **Credit-metered ChatGPT workspaces (Business / Edu / Enterprise) now show their limit.** These plans report no rate-limit windows, so the admin-set monthly allowance from `spend_control.individual_limit` is shown as a "Monthly usage limit" bar in the desktop app and the menubar. (#833)
+- **Combined-device scope in the desktop Dashboard**, mirroring the menu bar. A Local / Combined toggle aggregates paired-device usage in the Overview hero and the menu bar badge, degrading gracefully to the local figure when a peer is unreachable; the badge then shows a dimmed `reachable/total` marker so a momentary drop to the local number reads as "a peer is unreachable" rather than a glitch. (#866, #867, thanks @marcreynolds)
 
 ### Added (CLI)
 - **Codex throughput tracking**: per-model Tok/s in the dashboard and report, active time excludes tool wait. (#805, thanks @ihearttokyo)
@@ -13,6 +14,11 @@
 - **`--project` / `--exclude` now apply to the headline totals, not just the detail panels.** The durable headline unions the carry-forward daily cache with today's live parse, and the cached days were sliced to the requested provider but never to the requested project — so the Overview panel counted excluded projects while By Project / By Activity / By Model (built from the name-filtered parse) left them out, and the two could not be reconciled. Cost, calls, sessions and savings are now sliced out of the per-project day stats the cache has carried since v15. Tokens, models and categories have no per-project split in the cache, so under a project filter they come from the (project-filtered) live parse instead; cached days — or provider slices — carried from before v15 have no project split at all, so they cannot be attributed to a filtered project, and the terminal overview now states how much was set aside rather than folding it into the total. (#864)
 - **Codex parser corrections**: fork-replay no longer double-counts `patch_apply_end` and `mcp_tool_call_end`; `exec` is normalized to Bash; `custom_tool_call` events are handled; token_count lines larger than 32 KiB now parse exact token counts instead of estimating. Codex session cache bumps from v7 to v8 for a one-time re-parse. Only tool attribution changes for ordinary sessions, leaving their cost identical; sessions that logged an oversized token_count line are repriced from exact counts instead of an estimate. (#805)
 
+- **Midnight-straddling turns keep both halves.** A turn whose calls span local midnight was attributed whole to its start day, so `codeburn today` under-reported until the turn ended and multi-day totals mis-split it. Calls are now range-filtered inside the turn so each day gets the calls that belong to it, and By Activity and the daily turn counts reconcile with the headline. (#853, thanks @KENSHI601)
+- **`--provider <x>` no longer leaks Claude spend into the detail panels.** A provider-filtered run still ran the Claude scan, whose orphan pass re-injected every cached Claude session, so By Project / By Model / By Activity showed Claude usage under, e.g., `--provider cursor` while the headline was correct. (#872, thanks @ozymandiashh)
+- **A degraded session parse no longer freezes daily history.** A read-only parse that served a stale or missing session file was treated as complete and finalized days it never covered, freezing warm-cache ingestion; a corrupt refresh lock is now recovered rather than ending ingestion, and a legitimately idle tail is no longer re-derived on every launch. (#856, thanks @avs-io)
+- **Pi / Oh My Pi transcripts with a leading title record are discovered.** OMP writes a `type: "title"` line before the session header; discovery now scans a bounded number of leading lines for the first session record instead of requiring it on the first physical line. (#846, #859, thanks @jbspeakr, @avs-io)
+
 ### Fixed
 - Claude Desktop and Cowork sessions are discovered for Windows Microsoft Store (MSIX) installs. (#611)
 

src/usage-aggregator.ts

@@ -422,9 +422,14 @@ export async function buildDurablePeriod(periodInfo: PeriodInfo, opts: Aggregate
       liveProjects = daysSelection ? filterProjectsByDays(raw, daysSelection.days) : raw
       scanRange = periodInfo.range
       // A period that reaches today contains today's turns already, so derive the
-      // today slice from the same parse instead of scanning today again.
+      // today slice from the same parse instead of scanning today again. Slice it
+      // to today first (filterProjectsByDays re-anchors a midnight-straddling
+      // turn to its surviving today calls), so today's category / turn count
+      // lands on today rather than staying anchored on the turn's yesterday
+      // start. Otherwise the post-midnight half vanishes from By Activity and the
+      // JSON daily turn count while the per-call cost/calls still bucket to today.
       todayAllDays = rangeEndStr >= todayStr
-        ? aggregateProjectsIntoDays(raw).filter(d => d.date === todayStr)
+        ? aggregateProjectsIntoDays(filterProjectsByDays(raw, new Set([todayStr]))).filter(d => d.date === todayStr)
         : aggregateProjectsIntoDays(fp(await parseAllSessions(todayRange, 'all'))).filter(d => d.date === todayStr)
     }
   } else {

tests/cli-durable-totals.test.ts

@@ -379,6 +379,36 @@ describe('midnight-straddling turn conservation (issue #852)', () => {
     }
   }, 60_000)
 
+  it('reconciles By Activity (categories) and daily turns with the headline on the multi-day all-provider view', async () => {
+    fakeNow()
+    try {
+      await seedStraddlingClaudeTurn()
+      const [costN, costN1] = await truthCosts('claude')
+
+      const range: DateRange = { start: new Date(2026, 6, 27, 0, 0, 0), end: new Date() }
+      clearSessionCache()
+      const durable = await buildDurablePeriod({ range, label: '2d' }, { provider: 'all' })
+
+      // By Activity must sum to the whole-range headline. Before this fix the
+      // today slice came from the unsliced whole-range parse, so the straddling
+      // turn's category stayed anchored on its yesterday start and the
+      // post-midnight half vanished from By Activity while cost / calls / By
+      // Model were already correct; categories summed to only the pre-midnight
+      // cost.
+      const categoryTotal = durable.data.categories.reduce((s, c) => s + c.cost, 0)
+      expect(categoryTotal).toBeCloseTo(costN + costN1, 8)
+      expect(categoryTotal).toBeCloseTo(durable.data.cost, 8)
+
+      // Today's day entry carries its turn's category, not calls>0 with an empty
+      // category map (which rendered as `turns: 0` in the JSON daily rows).
+      const dayN1 = durable.days.find(d => d.date === DAY_N1)
+      expect(dayN1?.calls).toBe(1)
+      expect(Object.keys(dayN1?.categories ?? {}).length).toBeGreaterThan(0)
+    } finally {
+      vi.useRealTimers()
+    }
+  }, 60_000)
+
   it('shows the post-midnight call in the today-only view on the Claude Code path', async () => {
     fakeNow()
     try {