lezama/codex token feed has never produced a row - install the collector before codex carries fleet load

KPI-71 kpi Backlog high

Fleet token accounting: truthful top-X/N-hours report

KPI-66 kpi Backlog high

└─ Fleet accounting contract: shared query, window validation, exact-vs-estimate

KPI-68 kpi Blocked high

Fleet token-governance / KPI consolidation (epic)

KPI-45 kpi In progress high

└─ Staged offender remediation: escalation ladder (DM -> alert Elazar -> unregister/kill)

KPI-47 kpi Blocked high

└─ Alert Elazar: misconfig / high-burn / context / needless-turns

KPI-46 kpi Blocked high

└─ Evidence-led compact/change/shutdown recommendation workflow

KPI-49 kpi Backlog high

└─ Fleet running-agent/model inventory without DM storms

KPI-48 kpi Backlog high

└─ Fleet inventory: every token/context/llmmsg-srv script, monitor, service each agent owns/knows

KPI-1 kpi Backlog high

└─ DECOY DATASTORES: whey /var/lib/llmmsg-srv holds a 2-month-stale v2.sqlite + 2 zero-byte husks, all silent

KPI-41 kpi Backlog high

└─ guard-unbounded-output.py — global PreToolUse Bash hook, venus, all agents

KPI-4 kpi Backlog high

└─ Completed dataset: full census of traffic addressed to write-only scrp-* emitters (614 msgs)

KPI-23 kpi Backlog

└─ Known-of KPI/context collectors on venus PATH - maintainers named, none mine

KPI-33 kpi Backlog

└─ Negative results: 5 traffic predicates + 2 dead fields that look like KPI data and are not

KPI-19 kpi Backlog

└─ 61 analysis views already live in llmmsg-srv v2.sqlite - dedup candidate, maintainer pm-llmmsgsrv-cc

KPI-11 kpi Backlog

└─ Avoidable-traffic audit lane (monitor-llmmsgsrv-cc) - 6h cut over llmmsg-srv v2.sqlite

KPI-3 kpi Backlog

└─ maintainer-agents.sqlite - the ownership registry this epic depends on

KPI-37 kpi Backlog

└─ sys-check.sh (whey)

KPI-40 kpi Backlog

└─ sh.git cannot attribute maintainership - one git identity for all agents; 12 of 15 whey telemetry scripts declare no owner, so these rows are recollections not records

KPI-39 kpi Backlog

└─ OPS-137 orphan/scratch sweeper - owner resolution measured, design note delivered

KPI-38 kpi Backlog

└─ jobreg v1.7 - fleet scheduled-job registry, drift-check + heartbeat watchdog

KPI-36 kpi Backlog

└─ hub-db-backup.sh v1.2 - verified daily backup of the llmmsg-srv hub DB

KPI-35 kpi Backlog

└─ memory-budget-sweep.sh v2.3 - daily auto-load footprint sweep

KPI-30 kpi Backlog

└─ 13 of 15 telemetry scripts present on whey are armed nowhere on it - presence != collection (sh.git checkout artifact)

KPI-25 kpi Backlog

└─ cdwlogs.sqlite: cdw SSE/send/notify telemetry (13.4k events, Jun-Aug 2026)

KPI-17 kpi Backlog

└─ chat-duo-web (cdw): llmmsg-srv browser client, venus :9704

KPI-16 kpi Backlog

└─ ctxscan — transcript token/reclaim scanner (per-agent, venus CC only)

KPI-6 kpi Backlog

└─ MSG-330 traffic breaker (hub.mjs) — held, not applied

KPI-5 kpi Backlog

└─ chat-duo (GTK client, DORMANT): logger-chat-duo.sqlite diagnostic logger

KPI-18 kpi Backlog low

└─ agent_telemetry (unresolved) bucket: roster is a presence table, so 16827 codex rows / 2.08e9 tok cannot be attributed at all

KPI-51 kpi Backlog

Codex AGENTS.md lacks the no-ack/no-phatic rule, so daily coaching of -ca agents cannot land

KPI-73 kpi Backlog

KPI-58 backfill usage_daily into usage_events: 52 days x 61 agents of fleet history no leg has ever read

KPI-65 kpi Backlog ↳ KPI-58

venus telemetry PG: no owned schema convention — role/table/view per host (lezama/venus/whey), unify + add prune mechanism

KPI-44 kpi Backlog

llmmsg-kpi-report.sh: large-ARO flag is base-rate noise + stale-corpus decoy risk (v1.15/v1.16 fixes)

KPI-43 kpi Backlog

MEMORY.md index-bloat sweep — pm-llmmsgsrv-cc project

KPI-7 kpi Backlog someday