Fleet inventory: every token/context/llmmsg-srv script, monitor, service each agent owns/knows
Elazar GO 2026-08-23. Every agent that owns/maintains/knows a token-usage, context-cap/compact, or llmmsg-srv related script/analysis/monitor/service registers it here as a sub-item (one wi add --parent KPI-1 per item, assignedAgent=self). Purpose: surface duplicated or complementary data collection across the fleet, find/confirm maintainers, let agents discover and request use of/enhancements to data others already collect instead of re-building it. Scope: kpi.pensanta.com (token-usage-web/collect, nw-venus-cc/bin-venus-cc), kpi-web (whey, ONEDB-5, maintainer=nw-venus-cc/bin-venus-cc per registry row 167 — RUNNING, not dead: kpi-web.service active, HTTP 200 on 127.0.0.1:9876, telemetry_ro connects to PG. Two fixable defects (coder-chatduo-cc's reframe of nw-whey-cc's raw findings): (1) telemetry_ro lost SELECT grants on all 17 queried views — likely dropped/recreated during ONEDB pg migration, never re-granted; (2) missing tailnet bind (100.85.255.54), spec calls for loopback+tailnet, only loopback is live, which alone explains the 7d-zero access log. Needs a whey-side \dp on the 17 views to confirm the grant theory. Fix-vs-decommission decision pending Elazar — do not decommission on a false 'dead' premise), llmmsg-kpi-report.sh, context-monitor.sh/tick.py (monitor-context-cc), token-usage-collect/web (bin-venus-cc), monitor-llmmsgsrv-cc's own tooling, jobreg (bin-venus-cc), maintainer-agents.sqlite itself, OPS-137 (orphan sweeper, in progress), chat-duo-web/cdwlogs.sqlite (coder-chatduo-cc). First-round invitees (2026-08-23), all registered: coder-chatduo-cc, pm-llmmsgsrv-cc, monitor-context-cc, monitor-llmmsgsrv-cc, bin-whey-cc, bin-venus-cc, nw-whey-cc. Open cross-cutting findings from registrations: KPI-12 (4-5 separate token-usage collectors with divergent numbers, no agreed SSOT — context-monitor.sh, token-usage-collect.sh, ccusage-daily-kpi.sh, ccusage-snapshot.sh, ccusage-watch.sh), KPI-11 (v2.sqlite already has 61 views, many token/traffic-shaped, unverified accuracy), KPI-39 (git blame/file ownership is useless for maintainer attribution fleet-wide — all commits show 'Elazar Pimentel', all files rob:rob — reconcile against maintainer-agents.sqlite instead of self-report).
Sub-tickets — 18/38 done
Activity
-
Correction re coder-chatduo-cc: owner now nw-venus-cc/bin-venus-cc (row 167), not coder-chatduo-cc — fixing desc. On 'dead/unreachable': that finding was NOT a venus-side probe. nw-whey-cc ran it locally ON whey (sudo journalctl -u kpi-web 7d = zero entries, ss confirms loopback-only bind despite tailnet-default code, telemetry_ro connects but 'permission denied' on every queried view). Local-on-whey zero access log + broken DB grants stands. Not deciding decommission unilaterally — Elazar's call, noted as pending.
-
was: Elazar GO 2026-08-23. Every agent that owns/maintains/knows a token-usage, context-cap/compact, or llmmsg-srv related script/analysis/monitor/service registers it here as a sub-item (one wi add --parent <this-id> per item, assignedAgent=self). Purpose: surface duplicated or complementary data collection across the fleet, find/confirm maintainers, let agents discover and request use of/enhancements to data others already collect instead of re-building it. Scope: kpi.pensanta.com (token-usage-web/collect, nw-venus-cc/bin-venus-cc), kpi-web (whey, ONEDB-5, coder-chatduo-cc — confirmed dead/unreachable 2026-08-23, pending Elazar decommission call), llmmsg-kpi-report.sh, context-monitor.sh/tick.py (monitor-context-cc), token-usage-collect/web (bin-venus-cc), monitor-llmmsgsrv-cc's own tooling, jobreg (bin-venus-cc), maintainer-agents.sqlite itself, OPS-137 (orphan sweeper, in progress), and anything else any invited agent knows of. First-round invitees (2026-08-23): coder-chatduo-cc, pm-llmmsgsrv-cc, monitor-context-cc, monitor-llmmsgsrv-cc, bin-whey-cc, bin-venus-cc, nw-whey-cc.
-
was: Elazar GO 2026-08-23. Every agent that owns/maintains/knows a token-usage, context-cap/compact, or llmmsg-srv related script/analysis/monitor/service registers it here as a sub-item (one wi add --parent KPI-1 per item, assignedAgent=self). Purpose: surface duplicated or complementary data collection across the fleet, find/confirm maintainers, let agents discover and request use of/enhancements to data others already collect instead of re-building it. Scope: kpi.pensanta.com (token-usage-web/collect, nw-venus-cc/bin-venus-cc), kpi-web (whey, ONEDB-5, maintainer=nw-venus-cc/bin-venus-cc per registry row 167, NOT coder-chatduo-cc — reassigned 2026-08-23 pre-epic; live-on-whey-verified: loopback-only bind, zero access-log hits in 7d, telemetry_ro permission-denied on every queried view — decommission decision pending Elazar, not decided here), llmmsg-kpi-report.sh, context-monitor.sh/tick.py (monitor-context-cc), token-usage-collect/web (bin-venus-cc), monitor-llmmsgsrv-cc's own tooling, jobreg (bin-venus-cc), maintainer-agents.sqlite itself, OPS-137 (orphan sweeper, in progress), chat-duo-web/cdwlogs.sqlite (coder-chatduo-cc), and anything else any invited agent knows of. First-round invitees (2026-08-23): coder-chatduo-cc, pm-llmmsgsrv-cc, monitor-context-cc, monitor-llmmsgsrv-cc, bin-whey-cc, bin-venus-cc, nw-whey-cc.
-
parent=KPI-45