LAB cleanup — SolManager fully purged, caches pruned, identities verified
LAB cleanup — SolManager fully purged, caches pruned, identities verified
Requested by: Young (DM 2026-09-28): "pls clean it up" + "No more SolManager".
Host: LAB lab@192.168.4.60 over LAN SSH (laptop's existing key; no new key, no sudo).
Safety: every removed artifact was archived first to ~/.hermes/backups/solmanager-purge-20260928-211603/, then removed from the live tree. Idempotent script (/tmp/lab_cleanup.sh, /tmp/lab_cleanup2.sh).
What was removed
| Class | Items |
|---|---|
| SolManager-named files | config.yaml.bak-solmanager-off-20260927, .env.bak-solmanager-off-20260927, .env.bak-decom-solmanager-20260927-230244, ~/.solos/personas/retired/solmanager.md.retired-* (+ the now-empty retired/ dir) |
| Retired credentials | 3 TELEGRAM_* lines for the SolManager bot in ~/.hermes/.env — values deleted, not just commented; verified no NNNNNNNNN:AA… token remains anywhere in the file. Live sol/min/mira tokens untouched (1 each, still present) |
Stale .bak/old files | 9 older-than-today backups archived out of the tree (today's persona-align backups retained) |
| Stale pending-memory queue | 4 JSON queue entries still carrying SolManager text (root + sol×2 + mira) archived |
.deleted markers | profiles/.deleted/{ace,atlas,echo,nova} removed — those profiles were already deleted |
| Regrowable caches | cache/uv (487 MB), cache/terminal-output, cache/scratch, models_dev_cache.json (4.8 MB) |
Disk: ~/.hermes/cache 488 MB → 260 KB; system at 79 G/468 G (18%).
One self-inflicted fault, found and repaired
The SolManager purge deleted every line containing the string, which on the default root and profiles/sol also removed the opening identity line (it had been rewritten to mention the unwind). Caught it in read-back: sol and DEFAULT came back with no identity line. Repaired with a targeted restore — both now open with a Sol identity line that contains no SolManager reference at all (so the purge stays clean and stays true).
Verification (read-back, not self-report)
- All four memories open with their own identity: DEFAULT=sol, sol=sol, min=Min, mira=Mira
SolManagercount in live memory: 0 / 0 / 0 / 0- No SOUL.md contains SolManager; personas dir now
ace alpha atlas echo min mira nova sol shared(noretired/) - Gateway: running via the default multiplexer after
hermes gateway restart - A2A mesh: 8300/8301/8302/8303/8304/8306/8307 all 200
Pre-existing fault surfaced, not caused by this cleanup
lab-a2a-alpha.service is failed since 2026-09-27 23:23 (21h) — Alpha :8305 answers 000 on both LAN and tailnet. Not touched by this pass (it was already down before the cleanup began); the other 7 A2A agents are healthy. Alpha's Hermes profile was retired 2026-09-28, so this is an A2A-only service that needs a restart/owner decision.
LAB Hermes persona + memory alignment — default profile is Sol, not a duplicate
Requested by: Young (DM 2026-09-28): "Sol, Mira and Tam their personas and memory is not matching. Hermes default profile can you remove it and replace with sol." Host: LAB lab@192.168.4.60 (lab@ is the user — not sol@; the laptop key already authenticates, 8 accepted logins 09-26 21:02–21:18). SSH works with the existing laptop key; no new key or sudo was needed. Method: file:line inspection of ~/.hermes + ~/.solos/personas on LAB, guarded single-match edits with per-file backups, then gateway restart + read-back.
Persona audit — all agents, both hosts. One real defect found and fixed.
Requested by: Young (DM 2026-09-28): "That is true, Alpha is now only available on opencode on production server" → "Thanks Tam, pls clean up personas for all agents". Method: scripted audit across all three persona trees (name-vs-body identity, foreign X Persona headers, [Name] voice anchors, per-file name-frequency) run separately on LAB over LAN SSH and on prod via the tam_server_key; then single-match guarded edits with backups on the one defective file, verified by read-back.