#1053/#1045 Hub127 deployment profile — source preparation and activation
OBSERVED_AT=2026-10-12 11:01+08

- Production read-only snapshot: localhost9200 /health200, secured, Hub126,
  sessions330; listener BunPID3381400, cwd existing private Hub directory.
  Deployed profile still selects runtime-v109-preview126, mode700;
  original hub.env mode600. No credential contents read or reported.
- Software127 published from exact main31b0dbdae3192f471c8812574c118795658120f0,
  release38193173842 SUCCESS. Anonymous metadata exact version/SRI and unchanged
  dependency set verified; tarball bytes/hash match preceding publication proof.
- Narrow change: profile selects sibling v110/127; exact dependency lock updated
  without transitive churn; existing launcher fixture expects127 with126 rollback.
  Side-thread explicit0 behavior, base launcher, PM2, ports and secret source unchanged.
- Git diff126 source42fcf114... to127 source31b0dbda lists only Provider/secret
  handlers and contracts; no DB/schema/adapter file changes. No actual production
  DB backup was uploaded to Actions, no model or secret request sent.
- Existing Actions Hub deployment profile Docker job will verify launch path
  and explicit rollback. It uses fake Bun/ss and synthetic private env; not a
  real Hub migration or production activation proof. No local tests executed.
- Current process/configuration/data unchanged. Source PR, CI acceptance,
  normal main merge, private installation, backup and actual activation must
  remain distinct. Do not install branch launchers into production.

## Activation increment — 2026-10-12 11:34+08

- Source-preparation snapshot above remains historical, not the current state.
  PR2641 head479d79f05003b1cc898aae4d3f94dc244f764677 passed10 mandatory
  backend checks and existing Actions Docker launcher job114645584763 /
  run38198227403. Normal main squash3f956831eb949636b97ef964d5c4bd3ce9d97426
  merged03:08:25Z. Synthetic launcher check is NOT a real-data migration test.
- Installed exact package/lock into private sibling runtime-v110-preview127;
  npm ci --ignore-scripts, no global installation or install scripts. Public
  software remains main31b0dbdae3192f471c8812574c118795658120f0, SHA256
  ef9ed47cff0a1d6b65ef100dbf684f4e2c430e802ee75809bdb4db3c62fc6e54.
- Owner-private backup directory backups/scoped1053-p127.v4sV7M contains
  previous profile/base and consistent VACUUM snapshots (mode600). Latest
  cutover snapshot:747225088B, integrity ok,330 sessions/78778 tasks. No
  credential/database contents uploaded to Git or Actions. Data restore is
  operator-owned; these snapshots are not an off-machine disaster-recovery drill.
- Atomically deployed the exact Gitmain profile (mode700), base launcher and
  original hub.env unchanged. Only pm2 restart commhub-hub, no --update-env;
  no Daemon/business node/account/session change. Activated03:19:35.611Z
  (11:19:35.611+08), BunPID3381400 ->3775153, original PM2 parent1613.
- Actual process selects runtime-v110-preview127. Repeated /health200:
  version0.9.0-preview.127, secured,330 sessions; restartcount84->85 stable.
  Around11:25:125 SSE connections,116 recent5min heartbeats (before/after),
  114 identities with heartbeats after activation. Natural persisted events:
  delivered19/started19/replied21/failed1; independent task.ack0. The failed
  task is unclassified, zero ACK events are NOT an ACK PASS. No synthetic
  production traffic or model inference sent to inflate acceptance.
- Rollback: restore owner-private before126 profile with mode700 and restart
  only commhub-hub; retain old runtime-v109-preview126 and current live DB.
  Do not rewind live sessions/Provider state. Six recovery questions and the
  unchanged key source are in deploy/hub/README.md; off-machine master-key
  recovery inventory remains NOT_COVERED, not a claim of full recoverability.
- A diagnostic attempt to read entire1.16GB stdout failed in the reader with
  RangeError; Hub stayed healthy. Bounded last1MiB stdout contains this startup
  marker with zero scoped-vault-missing/panic matches. Stderr tail has no startup
  marker: scoped counts UNKNOWN, not zero. No log deletion/prune performed.
- No local tests/Docker executed. Existing Actions core/release gates reused;
  local integrity/process/heartbeat reads above are deployment observations.
  Daemon-vdev fresh11:32 heartbeat shows runtime136/original identity/idle0,
  but TMHR task1110d499 ended failed without full upgrade receipt. CLI/PID/SSE/
  Provider/config/rollback acceptance remains pending; do not close1053/1045
  from heartbeat alone. Client249 selected platforms publicly delivered
  separately; mirror and target activation are not inherited from build success.

## Daemon receipt increment — 2026-10-12 11:48+08

- TMHR read-only task20529e90-ecfe-4ad5-ada5-efec069e5b77 replied03:48:03Z:
  daemon-vdev actual CLI170/runtime136 remain running; original node identity,
  workdir/config hash unchanged, old private installation and0600 backup kept.
  Owner reports fresh idle/inflight0/queue0 and successful redacted Provider
  readback. Independent Hub03:46 read confirms136 and uptime902sec.
- Original upgrade task1110d499 remains FAILED due11:30 receipt deadline;
  this later audit does not rewrite that history. No additional install/restart/
  inference performed by TMHR. Live PID numbers are not included in its reply.
  SSE evidence is established Hub connection plus fresh heartbeat, not a
  separately executed protocol-level regression or real model entitlement.
- Original1053 delivery5/5 and1045 delivery3/3 can close on formal publication
  and target installation evidence plus the existing Actions core. No added
  production-secret/inference/SSE test is required to expand their last item.
  Generic key receipt still means next full node startup, not current TUI env.
- Local daemon-dev remains133 at PID3420628/PPID1, cwdcomm-ou, directory mismatch
  warning, no original-name tmux/CLI wrapper. Its current config SHA256
  7207b7cb129919513b5641cc9718dca7fa4fda178666ea3d57558f9d0a42a84b differs
  from historical4cf8ee...; preserve it. Asked the current workspace Owner to
  confirm launch/maintenance ownership before touching the live process.
- Installed inactive private pair provider906-cli170-runtime136, mode700,
  generated package/lock600, npm --save-exact --ignore-scripts,250 packages.
  Original public CLI/runtime tarball hashes above verified; mainf86012a is
  still an ancestor. No global package/config/env/identity change. Inactive
  preparation is NOT activation or a newly executed local test. Old167/133
  and current live process retained while ownership is being coordinated.
- Startup/upgrade/rollback authority remains published main CLI daemon start,
  docs-site/docs/deploy/daemon.md and existing fleet runbook. No new custom
  launcher/port/secret source/boot topology. Exact transitive private-lock
  archival and off-host data/key-backup inventory remain recovery gaps; this
  is not an empty-machine recovery drill. No private data/token added to Git.
