cctally 1.89.0 → 1.89.2

package/CHANGELOG.md

@@ -5,6 +5,22 @@ based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ## [Unreleased]
 
+## [1.89.2] - 2026-07-31
+
+### Fixed
+- Internal (maintainer-only): the public repository's CI matrix no longer fails on a test that requires maintainer-local tooling to be present. v1.89.1 stopped the remote-test transport harness from being treated as mandatory on a public checkout, but a test in the shipped suite still asserted that the same harness appears in a planned full-suite run — so the public matrix moved from failing before any test ran to failing on that one test, again across all three Python versions. That assertion now applies only in the private checkout, keyed off the very marker the runner itself uses to decide the harness is required, so the public suite skips it and the private tree still asserts it in full. (#446)
+
+## [1.89.1] - 2026-07-31
+
+### Fixed
+- Cache Report no longer shows a healthy measured 0% on a day with no activity yet — the panel, spotlight, daily rows, sparkline and net-$ bars now mark an unmeasured day explicitly (#443).
+- Cache Report daily rows whose anomaly predicates could not be evaluated now render a neutral flag with an explanatory legend, instead of a green check (#443).
+- Cache Report single-source views now show a provider's degraded or stale status chip, matching what the all-sources view already showed (#443).
+- The all-sources Cache Report summary no longer reports an anomaly for data the panel beside it calls "building baseline" (#443).
+- An empty Cache Report source no longer renders KPIs and a healthy verdict alongside its empty label (#443).
+- A Cache Report that failed to build now says so, instead of showing an indefinite loading state (#443).
+- Internal (maintainer-only): the public repository's CI matrix no longer fails before running a single test. v1.89.0 started treating three harnesses as mandatory so that a lost executable bit could not silently drop one from a full-suite run, but one of them — the remote-test transport harness — is maintainer-local tooling that the public mirror deliberately never carries, so on every public checkout the guard refused immediately and took all three Python versions down with it. The requirement now applies only in the private checkout, where the file is supposed to exist, while the public matrix goes on running the shipped subset as designed. Nothing the guard was added to catch has been given up: in the private tree a harness that loses its executable bit — or disappears outright — still fails the run. (#446)
+
 ## [1.89.0] - 2026-07-31
 
 ### Fixed

package/bin/_cctally_dashboard_cache_report.py

@@ -135,6 +135,12 @@ class CacheReportDailyRow:
     net_usd: float
     anomaly_triggered: bool
     anomaly_reasons: tuple[str, ...]
+    # #443 S1. `anomaly_unevaluated` = the predicates the classifier declined
+    # to run for this row; `observed` is False only for the builder's
+    # synthetic today row. Both default so their absence reproduces pre-S1
+    # rendering exactly.
+    anomaly_unevaluated: tuple[str, ...] = ()
+    observed: bool = True
 
 
 @dataclass(frozen=True)
@@ -161,6 +167,9 @@ class CacheReportTodaySpotlight:
     anomaly_triggered: bool
     anomaly_reasons: tuple[str, ...]
     baseline_daily_row_count: int
+    # #443 S1 — see CacheReportDailyRow.
+    anomaly_unevaluated: tuple[str, ...] = ()
+    observed: bool = True
 
 
 def _cache_report_snapshot_to_dict(cr: "CacheReportSnapshot | None") -> "dict | None":
@@ -189,6 +198,8 @@ def _cache_report_snapshot_to_dict(cr: "CacheReportSnapshot | None") -> "dict |
             "anomaly_triggered": cr.today.anomaly_triggered,
             "anomaly_reasons": list(cr.today.anomaly_reasons),
             "baseline_daily_row_count": cr.today.baseline_daily_row_count,
+            "anomaly_unevaluated": list(cr.today.anomaly_unevaluated),
+            "observed": cr.today.observed,
         },
         "days": [
             {
@@ -203,6 +214,8 @@ def _cache_report_snapshot_to_dict(cr: "CacheReportSnapshot | None") -> "dict |
                 "net_usd": d.net_usd,
                 "anomaly_triggered": d.anomaly_triggered,
                 "anomaly_reasons": list(d.anomaly_reasons),
+                "anomaly_unevaluated": list(d.anomaly_unevaluated),
+                "observed": d.observed,
             }
             for d in cr.days
         ],
@@ -319,6 +332,12 @@ def _cache_report_empty(
 ):
     """The empty (no in-window entries) ``CacheReportSnapshot`` — factored so the
     warm and cold builder paths share one ``is_empty`` return (#272 §6)."""
+    crk = _cache_report_load_kernel()
+    # #443 S1 — the dataclass defaults would claim ``observed=True`` with an
+    # empty ``anomaly_unevaluated`` for a day that was definitionally never
+    # measured or classified. That is the fabricating default this session
+    # removes everywhere else, so it is stated explicitly here too even though
+    # the empty branch short-circuits before any of it renders.
     empty_today = CacheReportTodaySpotlight(
         date=today_iso,
         cache_hit_percent=0.0,
@@ -328,6 +347,8 @@ def _cache_report_empty(
         anomaly_triggered=False,
         anomaly_reasons=(),
         baseline_daily_row_count=0,
+        anomaly_unevaluated=tuple(crk.CACHE_ANOMALY_PREDICATES),
+        observed=False,
     )
     return CacheReportSnapshot(
         window_days=window_days,
@@ -555,6 +576,10 @@ def build_cache_report_snapshot(
             anomaly_triggered=False,
             anomaly_reasons=(),
             baseline_daily_row_count=baseline_daily_row_count,
+            # This row never reaches the classifier, so nothing was evaluated
+            # for it and nothing was measured behind it (#443 S1 F1/F2).
+            anomaly_unevaluated=tuple(crk.CACHE_ANOMALY_PREDICATES),
+            observed=False,
         )
     else:
         today_spotlight = CacheReportTodaySpotlight(
@@ -568,6 +593,8 @@ def build_cache_report_snapshot(
             anomaly_triggered=today_row.anomaly_triggered,
             anomaly_reasons=tuple(today_row.anomaly_reasons),
             baseline_daily_row_count=baseline_daily_row_count,
+            anomaly_unevaluated=tuple(today_row.anomaly_unevaluated),
+            observed=True,
         )
 
     # Daily rows — newest first, capped at ``window_days``.
@@ -615,6 +642,8 @@ def build_cache_report_snapshot(
                 net_usd=0.0,
                 anomaly_triggered=False,
                 anomaly_reasons=(),
+                anomaly_unevaluated=tuple(crk.CACHE_ANOMALY_PREDICATES),
+                observed=False,
             )
         )
     days_newest_first.extend(
@@ -630,6 +659,8 @@ def build_cache_report_snapshot(
             net_usd=r.net_usd,
             anomaly_triggered=r.anomaly_triggered,
             anomaly_reasons=tuple(r.anomaly_reasons),
+            anomaly_unevaluated=tuple(r.anomaly_unevaluated),
+            observed=True,
         )
         for r in raw_days_newest_first
     )

package/bin/_lib_cache_report.py

@@ -119,6 +119,11 @@ CACHE_REPORT_MIN_BASELINE_SESSIONS = 10
 # dashboard/web/src/types/envelope.ts:71 — keeps the two surfaces in
 # lockstep so a typo on either side fails type-check.
 CacheAnomalyReason = Literal["net_negative", "cache_drop"]
+# Every predicate _classify_anomalies can run, in reason-append order. The
+# TypeScript mirror is CACHE_ANOMALY_PREDICATES in cacheReportVerdict.ts.
+CACHE_ANOMALY_PREDICATES: tuple[CacheAnomalyReason, ...] = (
+    "net_negative", "cache_drop",
+)
 
 
 @dataclass
@@ -185,6 +190,10 @@ class CacheRow:
     # Anomaly (populated by _classify_anomalies)
     anomaly_triggered: bool = False
     anomaly_reasons: list[CacheAnomalyReason] = field(default_factory=list)
+    # Predicates the classifier did NOT run for this row. Empty means every
+    # predicate was evaluated; a non-empty list is what lets the dashboard
+    # distinguish "evaluated, clean" from "never evaluated" (issue #443 F2).
+    anomaly_unevaluated: list[CacheAnomalyReason] = field(default_factory=list)
 
     @property
     def total_tokens(self) -> int:
@@ -767,6 +776,7 @@ def _classify_anomalies(
         for row in rows:
             row.anomaly_triggered = False
             row.anomaly_reasons = []
+            row.anomaly_unevaluated = list(CACHE_ANOMALY_PREDICATES)
         return
     if not rows:
         return
@@ -782,24 +792,31 @@ def _classify_anomalies(
 
     for i, row in enumerate(rows):
         reasons: list[CacheAnomalyReason] = []
+        unevaluated: list[CacheAnomalyReason] = []
 
         # Trigger 1: net_negative (no baseline needed; cache-activity guard).
         if row.cache_creation_tokens + row.cache_read_tokens > 0:
             if row.net_usd < 0:
                 reasons.append("net_negative")
+        else:
+            unevaluated.append("net_negative")
 
         # Trigger 2: cache_drop (requires baseline).
         anchor = anchors[i]
+        median = None
         if anchor is not None:
             median = _compute_baseline_median(
                 rows, anchor=anchor,
                 window_days=window_days, min_samples=min_baseline,
                 exclude_row=row, is_session_mode=is_session_mode,
             )
-            if median is not None and (median - row.cache_hit_percent) >= threshold_pp:
-                reasons.append("cache_drop")
+        if median is None:
+            unevaluated.append("cache_drop")
+        elif (median - row.cache_hit_percent) >= threshold_pp:
+            reasons.append("cache_drop")
 
         row.anomaly_reasons = reasons
+        row.anomaly_unevaluated = unevaluated
         row.anomaly_triggered = bool(reasons)