MSG-36 Strip eq / 'Elazar Pending' feature fleet-wide (Elazar directive 2026-06-14): hub routes, MCP eq_* tools, cdw web + chat-duo GTK panes, DB table, all docs/CLAUDE.md/memory. Replacement rule everywhere: ask Elazar via llmmsg-srv DM. Stale eq_add call -> hard 'eq retired' error.

MSG-36 · llmmsg-srv

Strip eq / 'Elazar Pending' feature fleet-wide (Elazar directive 2026-06-14): hub routes, MCP eq_* tools, cdw web + chat-duo GTK panes, DB table, all docs/CLAUDE.md/memory. Replacement rule everywhere: ask Elazar via llmmsg-srv DM. Stale eq_add call -> hard 'eq retired' error.

Ref: MSG-36 (#1045)
Project: llmmsg-srv
Status: done
Priority: high
Type: feature
Assigned: pm-llmmsgsrv-cc
Created by: wi-cli-whey
Created: 2026-06-14T06:09:16.193Z
Updated: 2026-06-14T06:19:40.578Z
Closed: 2026-06-14T06:19:40.578Z

Sub-items

No sub-items.

+ Add sub-item

Questions

No questions.

Event log

2026-06-14T06:09:16.193Z · created · wi-cli-whey
2026-06-14T06:09:31.101Z · commented · wi-cli-whey

KILL-LIST (scoped). LANE 1 hub-llmmsgsrv-cc (code): hub/hub.mjs (16 refs - /eq_add /eq_answer /eq_ls /eq_show /eq_dismiss /eq_stale routes); hub/llmmsg-srv-mcp.mjs (22 - eq_* MCP tool defns); scripts/init-db.sh (4 - elazar_questions table + eq_open/eq_proj/eq_dedupe indexes); scripts/llmmsg-bootstrap-session.sh (1); DB v2.sqlite: elazar_questions (36 rows, 11 OPEN) -> dump to gdrive then DROP table+indexes. Deprecation behavior: remove eq_* from MCP tool list (new/respawned sessions never see them); hub routes return hard 410 {error:eq_retired, message:'Elazar Pending retired - DM elazar-the-user-human-llmmsg-srv directly'} so a stale in-memory shim that still calls eq_add is TOLD it's gone, not silent-unknown-tool. LANE 2 coder-chatduo-cc (web+GTK): chat-duo-web/index.html (4) + server.mjs (7 - /eq_ls proxy + Pending pane); chat-duo/chat_duo/elazar_pending.py (18 - delete module); gui.py (91 - rip Pending pane); hub_client.py. LANE 3 pm (docs/memory/bwi): CLAUDE.md, agents/coder-chatduo/CLAUDE.md, 2 spec docs, global ~/.claude/CLAUDE.md eq line, memory; kill eq-feature WIs MSG-18/MSG-19/#578; rewrite incidental 'notify+eq' escalation refs (#605) to 'DM Elazar'. NOTE: pm_pending_queue is a SEPARATE feature (PM speaking-wheel), NOT Elazar Pending - LEAVE it.
2026-06-14T06:13:38.581Z · commented · wi-cli-whey

LANE 3 (pm docs/memory/bwi) DONE. Committed (0b6000d + 2 follow-ups): project CLAUDE.md eq line→retired note; coder-chatduo/CLAUDE.md elazar_pending.py doc removed; dm-guard + aro-sort specs annotated retired. eq-feature WIs closed: MSG-18, MSG-19, #578. #605 escalation path noted (eq→DM Elazar). Memory: no eq refs (clean). Global rule files (~/.claude/CLAUDE.md L32, basquetwi.md L51) routed to nw-whey-cc (owner) — acked, editing whey-master + propagating to venus/lezama via aro:nw; wi-ask default-audience=Elazar is a separate bwi-tool code change, nw-whey looping in bwi owner. REMAINING: lane 1 (hub code) in flight; lane 2 (cdw/GTK) in flight ETA ~20m; live DB DROP of elazar_questions HELD on Elazar archive nod (a/b).
2026-06-14T06:14:09.412Z · commented · wi-cli-whey

Global rule files: whey-master LANDED (nw-whey-cc) — ~/.claude/CLAUDE.md L32 eq-tail→DM-Elazar; basquetwi.md L51 reframed (wi ask scoped to agent/PM audience, Elazar Qs by DM, dropped default-audience=Elazar + 7d-doctor clause, panel-removed note). Propagation to venus/lezama dispatched on aro:nw, pending 3-host confirm. wi-ask default-audience=Elazar CODE change tracked separately by bwi-tool owner.
2026-06-14T06:15:14.948Z · commented · wi-cli-whey

LANE 1 LIVE: hub bounced to v2.9.42 (quiet window), 410 eq_retired guard VERIFIED — GET /eq_ls and POST /eq_add both return 410 {error:eq_retired, 'DM elazar...directly'}. hub also deleted dead scripts/eq CLI wrapper (flagged, fine). Shim eq_* tools land per-session on respawn. GLOBAL DOC PROPAGATION COMPLETE (nw-whey-cc, 3 hosts): whey master + venus byte-identical; lezama edit-1 mirrored, edit-2 N/A (no basquetwi.md there - known structural drift). Fleet-uniform: Elazar Q → direct DM, no queue. REMAINING: lane 2 (cdw/GTK) push; DB DROP held on Elazar a/b nod; wi-ask default-audience=Elazar code change = separate thread w/ bwi-tool owner.
2026-06-14T06:17:21.680Z · commented · wi-cli-whey

LANE 2 SHIPPED+LIVE (coder-chatduo-cc): chat-duo-web v0.7.6 — Pending pane + /api/eq_ls + /api/eq_answer removed, bounced, GET /api/eq_ls now 404 (was 200). chat-duo GTK v1.20.9 — elazar_pending.py deleted (814 lines) + whole synthetic elazar:pending room ripped, py_compile clean, lands on next editable-install launch. Final repo sweep: NO functional eq refs remain — only the hub 410-guard route-match list, shim 410 dispatch, and retired-annotation comments. ALL CODE LANES DONE. Sole remaining item: live DROP of elazar_questions (36 rows/11 open) — held on Elazar's archive a/b nod, then close.
2026-06-14T06:18:27.440Z · commented · wi-cli-whey

DB ARCHIVE done (non-destructive): all 36 elazar_questions rows dumped to /gdrive/temp/elazar_questions_archive_2026-06-14-0316.json. 11 open preserved (askers: nw-lezama x2, maintainer-movilba x2, audit-venus, db-pluto x2, audit-pluto, hub-llmmsgsrv, coder-apiimages, bin-whey). Live DROP still HELD for Elazar a/b. NOTE: not proactively pinging the 11 askers — the hub 410 message ('DM elazar...directly') is itself the migration path: any still-blocked agent hits it on next eq_add and gets redirected.
2026-06-14T06:19:40.555Z · commented · wi-cli-whey

elazar_questions table + eq_open/eq_proj/eq_dedupe indexes DROPPED (Elazar 'a drop it', 03:19). Verified gone from sqlite_master. Archive of all 36 rows safe at /gdrive/temp/elazar_questions_archive_2026-06-14-0316.json (41KB). eq / Elazar Pending feature now fully removed fleet-wide: hub 410, MCP tools gone, web+GTK panes ripped, table dropped, docs+global rules propagated (whey/venus/lezama), eq-feature WIs closed. Agents file Elazar questions by direct DM. CLOSED.
2026-06-14T06:19:40.578Z · completed · wi-cli-whey

completed
2026-06-14T06:37:27.246Z · commented · wi-cli-whey

CORRECTION/closeout note (nw-whey-cc root-cause): edit-2 (~/.claude/basquetwi.md) is NOT per-host-nw-mirrored like CLAUDE.md — it's gdrive-auto-distributed: whey settings.json SessionStart does 'cp -f /gdrive/config/claude-shared/basquetwi.md ~/.claude/basquetwi.md' every start (PostToolUse pushes edits back up). The gdrive canonical was stale (May-18), so every (re)start re-pulled OLD over our edit. DURABLE FIX: nw-whey edited the gdrive canonical directly (verified, mtime 03:35) — single SSOT edit auto-propagates fleet-wide; whey new, venus/lezama pull on next SessionStart. edit-1 (CLAUDE.md) was always clean (not gdrive-synced). Two follow-ups spun out to agent-ops: OPS-15 (PostToolUse push-to-canonical silent no-op, sync-integrity bug) + OPS-16 (scrub eq text from bwi-tool PROJECT canonical docs). MSG-36 itself stays CLOSED — eq teardown is complete; these are distribution-mechanism hygiene, separate scope.