| 2026-09-24 |
Document gnauth host (192.168.1.167) and add inventory item
...
Host gnauth (192.168.1.167) was previously undocumented. Verified live via SSH during gnexus-auth update deploy: Ubuntu 22.04.5 VM on DL380 G6, Laravel + nginx + MySQL, queue worker systemd unit, deploy procedure, 2026-09-24 update b0f50ed→be4aea0 with SESSION_LIFETIME=43200.
|
Document poisoned-pending protection: validate-before-write, full rollback on failed apply, delete_pending_change tool, corrupt-JSON tolerance
...
Задача gntodo #3: задокументирована новая защита от отравленных pending-changes (validate-before-write, полный rollback, delete_pending_change, устойчивость к corrupt JSON). Коммит c77bcb1 на gnbook.
|
fix(server): prevent poisoned pending-changes (validate before write, full rollback, delete tool)
...
- create(): reject secret-pattern payloads and unsafe doc targets BEFORE writing the pending JSON; rejected requests leave no file
- apply(): on validation failure, remove the change's own pending JSON (full rollback, doc + inventory-item paths)
- PendingChangeRepository.delete(): pending/corrupt records deletable, applied records refused
- list()/get()/apply(): tolerate corrupt JSON files (status=corrupt) instead of crashing
- validation.py: reusable scan_text_for_secrets() (behavior of repo-wide scan unchanged)
- MCP: new delete_pending_change tool; update_doc cleans up its pending JSON on failure
- REST: DELETE /changes/{id} (404/409), POST /changes returns 400 on rejection
- tests: test_pending_changes_fix.py (9 cases); updated rollback expectation in test_pending_changes.py
|
| 2026-09-22 |
gntodo updated to dcb6976: MCP project management (12 tools) + detailing fallback, api-only rebuild
...
Update 2026-09-22 (42aada2 → dcb6976) documented: MCP project management tools (create/update/archive/restore_project, 12 total) and detailing fallback (dropped title details preserved). Backend-only rebuild (api image), web/db untouched, no migrations. Verified: api healthy, /api/health 200 local and public, MCP list_projects works.
|
tgbots-vm: дока хоста (192.168.1.254) + инвентарь, таймзона Europe/Kyiv установлена 2026-09-22
...
New host doc 10-systems/hosts/tgbots-vm.md (192.168.1.254, hostname tgbots, KVM VM, Ubuntu 22.04.5, three Telegram bot services) and inventory item hosts/tgbots-vm. Timezone Europe/Kyiv set 2026-09-22 (was Etc/UTC), bot services restarted. Disambiguation note added to 10-systems/hosts/tgbots.md (192.168.1.178, vmk, no bots).
|
tgbots.md: добавлена заметка «do not confuse» со ссылкой на новую доку tgbots-vm (192.168.1.254), где живут боты
...
Обнаружена путаница двух хостов: tgbots (192.168.1.178, vmk, без ботов) и tgbots-vm (192.168.1.254, hostname tgbots, все Telegram-боты). Добавлена заметка-разъяснение в обе доки.
|
gntranscribe: фикс применён — except TimedOut + PTB таймауты 30s, бэкап bot.py.bak-20260922, сервис перезапущен и верифицирован
...
Фикс применён на tgbots-vm 2026-09-22 (owner-approved): отдельный except TimedOut, PTB таймауты 30s, бэкап bot.py.bak-20260922, сервис перезапущен и верифицирован.
|
gntranscribe: задокументирован инцидент 2026-09-22 — «Timed out» был таймаутом Telegram API, не whisper; добавлено правило триажа
...
Инцидент 2026-09-22: ошибка «сервис транскрибации недоступен» была вызвана таймаутом Telegram Bot API (PTB TimedOut), а не whisper-server. Зафиксирован known failure mode и правило триажа по логам.
|
inventory: добавлен сервис gntodo (GNexus Tasks)
...
Created inventory service item gntodo (GNexus Tasks): host gntasks (192.168.1.179), upstream http://192.168.1.179:8135, public https://tasks.gnexus.space, TLS on internal-proxy-vps (LE cert to 2026-12-20), health /api/health, repo gnexus-tasks.git, MCP endpoint /mcp/ with per-user Bearer tokens. Applied pending change 20260922005535-abb1e790.
|
gntasks.md: MCP integration section (gntodo → navi rei)
...
Added "MCP integration into navi swarm (2026-09-22, rei)" section to 10-systems/hosts/gntasks.md: per-user MCP credential reference (gnexus-creds), live-verified endpoint (initialize 200, 8 tools), rei config mcp_servers.d/gntodo.json + server_admin profile update. Also removed 4 stale pending-change files from 90-maintenance/pending-changes/ (3 contained a raw credential from failed attempts, 1 targeted a non-markdown file).
|
| 2026-09-21 |
Webhook secret ротирован (whsec_kgL...), подписанный тест пройден: новый 200, старый 400
...
Webhook secret ротирован Женей 2026-09-21 (whsec_DlRa... → whsec_kgL...): вписан в .env, api пересоздан, подписанный HMAC-тест пройден (новый 200, старый 400). Факт из чата с владельцем.
|
gntodo обновлён до ac8aefb: мультитенантность, вебхуки SSO, per-user MCP-токены; GAUTH_WEBHOOK_SECRET настроен, бэкап снят
...
Обновление gntodo до ac8aefb (мультитенантность ТЗ 0.34, вебхуки ТЗ 3.16, per-user MCP-токены) выполнено и верифицировано 2026-09-21; GAUTH_WEBHOOK_SECRET настроен; бэкап перед миграцией в /srv/gntodo-backups/.
|
SSO настроен: client_id/secret в .env, api пересоздан, креды в gnexus-creds (ca8836f7)
...
SSO-клиент gntodo зарегистрирован Женей в auth.gnexus.space, client_id/secret вписаны в .env, api пересоздан и healthy. Обнаружено при настройке SSO 2026-09-21.
|
Добавлена секция gntodo (GNexus Tasks): стек на :8135, SSO pending, Ollama .130/qwen3.5:2b
...
На хост .179 задеплоен сервис gntodo (GNexus Tasks): docker compose стек на порту 8135, публичный URL tasks.gnexus.space через прокси .226. Обнаружено при деплое 2026-09-21.
|
Добавлен сайт tasks.gnexus.space (upstream 192.168.1.179:8135) и отдельный certbot-сертификат (до 2026-12-20)
...
Деплой gntodo: новый сайт tasks.gnexus.space (upstream 192.168.1.179:8135) и отдельный LE-сертификат на internal-proxy-vps. Обнаружено при деплое GNexus Tasks 2026-09-21.
|
Add host gntasks (192.168.1.179) with navi instance ayame
...
Deployed navi instance ayame (profile navi_code) on new VPS 192.168.1.179 (gntasks): docker compose plugin, navi-1 deploy branch e276b1e, install.sh, swarm key from hive, HIVE_URL, announce_ok verified, peer list online. Credentials in gnexus-creds secret e4e28fe1.
|
| 2026-09-20 |
Document ogg/opus rejection and WAV conversion pipeline in GNtranscribe
...
Live debugging 2026-09-20: whisper-server on bserv-host rejects ogg/opus with HTTP 400 "Invalid request" (verified: ogg→400, wav→exact transcript). GNtranscribe bot now converts all incoming audio to 16 kHz mono WAV via ffmpeg (installed on tg_bots VM, 4.4.2). Download from Telegram CDN moved to aiohttp (PTB default 5s read timeout too short). Whisper calls retried up to 3 attempts.
|
Add GNtranscribe telegram bot (doc + inventory + host doc update)
...
GNtranscribe (@gntranscribe_bot) deployed 2026-09-20 on VM tg_bots (192.168.1.254) by navi rei: /home/ubuntu/bots/GNtranscribe, systemd gntranscribe.service (enabled, Restart=always), python-telegram-bot 21.10. Forwards voice/audio/video to whisper-server on bserv-host (192.168.1.130:8008), asyncio queue max 20, single worker, optional chat_id whitelist via ALLOWED_CHAT_IDS, commands /start /help /id /status. Verified: getMe ok, direct whisper transcription test passed, two service restarts passed. Bot token stored in gnexus-creds secret "GNtranscribe token" (881b7fa9-1775-4058-b256-04aa5737e62d).
|
| 2026-09-19 |
Update gitbucket VM inventory: runs navi instance pilar
...
gitbucket VM now runs navi instance pilar (deployed 2026-09-19 by rei) in addition to GitBucket.
|
Add hosts inventory item for VM gitbucket (GitBucket + navi pilar)
...
navi instance pilar (profile navi_code) deployed on VM gitbucket 2026-09-19 by rei. Host doc + hosts inventory item created.
|
Create host doc for VM gitbucket: GitBucket + navi instance pilar (deployed 2026-09-19)
...
New navi instance pilar deployed on VM gitbucket (192.168.1.156) on 2026-09-19; host doc created per inventory creation workflow (docs field must point to an existing doc).
|
Fix radio repo last-commit year typo (2026-02-23 -> 2025-02-23)
...
Fix typo introduced in previous commit: radio repo last-commit year was changed to 2026 by mistake; restoring 2025-02-23 from the 2026-05-10 API snapshot.
|
Add VM disk layout and free-space observation for gitbucket VM (2026-09-19)
...
Disk-space check requested by owner; discovered VM disk path, partition layout and free space via read-only loop mount on hypervisor. No guest agent and no direct SSH into the VM.
|
| 2026-09-18 |
Document BugTrail service deployment: host doc + inventory (host, domain, service)
...
BugTrail deployed 2026-09-18 on host bugtrail (192.168.1.177): docker compose stack (postgres+server+caddy) on port 80, public https://bugtrail.gnexus.space via internal nginx proxy with separate Let's Encrypt certificate (CN=bugtrail.gnexus.space, expires 2026-12-17). DNS record already existed. Verified end-to-end: healthz ok, panel HTTP/2 200, git.gnexus.space unaffected.
|
Add bugtrail.gnexus.space proxy mapping with dedicated LE cert; refresh internal-proxy-vps doc and nginx proxy inventory
...
Issued dedicated Let's Encrypt certificate for bugtrail.gnexus.space on the internal proxy VPS (192.168.1.226), added HTTPS :443 server block to the bugtrail nginx site (proxy to 192.168.1.177:80), verified healthz/panel/git.gnexus.space. Documented the new mapping plus previously undocumented enabled sites (creds, fuckjira, transcribe-demo, vmk-marginator).
|
| 2026-09-17 |
Document host mine (192.168.1.160) — navi instance Kael (ex-Miner)
...
Host mine (192.168.1.160, user ubuntu, Ubuntu 22.04.5, Xeon X5650 12C, 15 GiB RAM) documented from a read-only self-report by the Kael agent via swarm peer protocol. Navi instance Kael (ex-Miner, renamed 2026-09-17, instance_id 347a599e-a98f-410f-98c3-ea21dd90b7a4) on port 8099, repo /home/ubuntu/navi-1 (commit 1fb0f2a), Postgres 127.0.0.1:5432, docker bridge br-161f7117df03. Owner hypothesis about former addresses 192.168.1.249/.229 checked: no traces in docs, inventory, secrets, or the 2026-05-10 network scan.
|
Update bugtrail inventory: navi-1 tsukiko instance, ports 8099/5432
...
Inventory hosts/bugtrail updated after navi-1 (tsukiko) deployment: navi.service active, ports 8099 (navi-server) and 5432 (postgres) listening, purpose set to navi-agent-instance. Doc 10-systems/hosts/bugtrail.md already committed in 9814867.
|
Document navi-1 (tsukiko) deployment on bugtrail: repo path, service, ports, hive, instance id
...
Deployed navi-1 instance tsukiko on bugtrail (2026-09-17): repo, service, ports, hive announce verified via SSH.
|
Add new LAN VPS bugtrail (192.168.1.177): host doc + hosts inventory item
...
New LAN VPS bugtrail@192.168.1.177 onboarded: read-only SSH recon verified login + sudo (same password), Ubuntu 22.04.5 LTS KVM VM (2 vCPU, 3.8 GiB RAM, 31G LVM), docker+containerd running with no containers, listening ports 22/53/43225, secondary IP 192.168.1.198 on enp1s0 (purpose unknown). Credentials stored in gnexus-creds secret d654f5aa-412b-4fe0-87ad-89cc640375bb. Purpose not yet specified by owner.
|
| 2026-09-16 |
Document navi.service (Navi Proxy, port 8099) on ovpn_reserv
...
Verified via SSH on 2026-09-16: VM ovpn_reserv (192.168.1.226) runs navi.service (systemd, user ubuntu, WorkingDirectory /home/ubuntu/navi-1, ExecStart .venv/bin/navi-server) listening on 0.0.0.0:8099 with /health endpoint. Config at /home/ubuntu/navi-1/.env (OLLAMA_API_KEY, OLLAMA_HOST=https://ollama.com). No ollama_backends.json exists. SSH access as user ubuntu (root denied). No secrets stored.
|