wiki audit 2026-11-21: session-scoped capture re-diagnosis (124 Pulse rows, ~57 uncaptured turns, gateway-restart hypothesis NOT proven); NEW page — first content_policy_blocked digest failures (gmail 08eb86f02d37 + freshrss ecc1fc13655a); watchdog streak 115->210 (single incident row); wiki-audit credit outage RESOLVED (two nights lost, 11-20 delivery suppressed); ledger corrections (4 unclosed health-monitor rows, not 5; audit incident closed 2026-11-20T01:19:13; 7 unclosed cron_incidents); skills-index regenerated 299 skills/67 cats; freshness 5 stale/132 fresh/3 no-date; Quartz rebuilt+deployed (879 files)

This commit is contained in:
Tony0410
2026-11-21 00:40:00 +08:00
parent f603c29d17
commit ca0ff5aa80
11 changed files with 484 additions and 31 deletions
+27
View File
@@ -202,3 +202,30 @@ Primary agent memory store for Hermes.
rather than fail. The **underlying cold start was not optimised** — only the budget was raised.
- **Capture and coverage unchanged** from the 10-18 read (working 760/786 = 96.7 %, episodic
1050/1056 = 99.4 %; nothing indicates a reindex is due). No monitor reds in the window.
## Re-read 2026-11-21 (audit window 2026-11-20; live probes 2026-11-21 00:0x–00:5x AWST)
**The capture fault is *session-scoped*, and its boundary is exact — the 11-20 "conversational silence"
reading is withdrawn.**
- **One session has stopped capturing entirely.** The desktop DM session `20261031_190813_95fbde` owns
**124 Pulse/automation rows** in `working_memory` (84 under the bare id, 40 under
`hermes_agent:main:local:dm:`), spanning **2026-10-31 11:09:16 → 2026-11-13 05:10:35 UTC (= 13:10:35 AWST)**
and **nothing after** — while the session itself kept producing user-role turns. Verified against
`state.db`: last captured turn **13:06:05 AWST**, first uncaptured **13:21:00 AWST**, **~57 turns** since.
The **11-13 13:14 gateway restart** sits inside that 15-minute boundary and is the leading hypothesis (a
watermark left at-or-past the tip). **NOT PROVEN** — the oldest surviving `agent.log` begins
**2026-11-13 19:03:34**, so the window has rotated out of the log horizon.
- **Capture still works for *new* sessions.** The single write after that boundary —
`2026-11-18 11:25:03 UTC = 19:25:03 AWST` — came from `hermes_20261118_191808_5c9e08`, a session created by a
genuine prompt that evening. The pipeline is not globally dead; it fails for the long-lived session.
- **Global silence now 52 h 35 m** (11-18 19:25:03 AWST → this read); exactly **one** write in eight days.
`working_memory` **1,131** / `episodic_memory` **1,250** — both unchanged since the 11-20 read.
- **The health monitor was already saying this.** `ea79a2107e36` fails with
`STALE MEMORY PIPELINE: capture missed: user message`, `failure_streak = 4`; incident
`ea79a2_ace0f263e18a` first seen 2026-11-21 00:03:12, alerted 00:03:17, unclosed.
- **Consolidation still 0:** 199 of the last 200 cycles consolidated nothing.
- **Conflicts still 59** against a threshold of 50; the watchdog has now been red for **210 consecutive
15-minute slots** (11-18 20:01:15 → 11-21 00:31:16 (00:16:15 at the first read)).
→ [[incidents/2026-11-14-mnemosyne-capture-stall]], [[incidents/2026-11-18-mnemosyne-watchdog-conflict-threshold-storm]]