| 2026-10-05 |
hard-panel: обновление c11cd47..3867b9d — фикс maskable PWA-иконок; бэкап БД файловым копированием (sqlite3 в контейнере нет), флот 19/19, шары 4/4, сервисы 6/6, MCP initialize OK
...
Обновление панели c11cd47..3867b9d (фикс maskable PWA-иконок) выполнено navi rei 2026-10-05 08:25 по запросу владельца; все проверки зелёные.
|
hard-panel: добавлены 6 сервисов в health-мониторинг (все up, 200 OK)
...
Update 2026-10-05 02:40 (navi rei): через MCP service_create добавлены 6 сервисов в health-мониторинг панели — gntodo, gnexus-creds, gnexus-synapse, hard-panel, bugtrail, gnexus-book. Все health-эндпоинты предварительно проверены (200 OK), services_overview 6/6 up. Записывающие тулы вызваны прямым JSON-RPC на /mcp с персональным ключом владельца (в профиле server_admin экспонирована только read-группа).
|
hard-panel: обновление 40a8810..c11cd47 — PWA и MCP-тулы управления; бэкап БД на хост; флот 19/19
...
Обновление панели на .176 до c11cd47 (PWA + MCP-тулы управления), проверено живой верификацией: флот 19/19, MCP initialize OK, локальные compose-правки сохранены.
|
Update 2026-10-05 04:35: стек обновлён 0ba611c..2f6d8d4 (admin-сводка + мобильная адаптация), миграций нет, бэкап 28 KB, все health-проверки пройдены
...
Обновление стека 0ba611c..2f6d8d4 выполнено 2026-10-05 04:35 (navi rei): 2 коммита (admin-сводка, мобильная адаптация), миграций нет, бэкап preupdate-2026-10-05-0433.sql (28 KB), все проверки здоровья пройдены. Факты получены по SSH с хоста gnsynapse.
|
Synapse: MCP подключён к Navi rei, задокументирована Host-ловушка FastMCP
...
Update 2026-10-05 03:50: Navi rei подключила Synapse как MCP-сервер (mcp_servers.d/synapse.json, LAN endpoint 192.168.1.172:8013/mcp, Host: localhost:8013). Задокументирована ловушка FastMCP host-validation: 421 Invalid Host header на публичный https://synapse.gnexus.space/mcp (nginx передаёт $host) — фикс в Pending владельца. 35 тулов mcp__synapse__* зарегистрировано в Navi.
|
hard-panel: MCP-ключ владельца подключён к navi rei
...
navi rei: подключён персональный MCP-ключ владельца к navi rei — mcp_servers.d/hard-panel.json (streamable_http, LAN url http://192.168.1.176:8000/mcp, Host override localhost:8000 из-за валидации Host на стороне панели), профиль server_admin + read-группа (6 тулов). Проверено: initialize OK (GHard Monitor 1.30.0), tools/list OK, panel_overview — флот 19/19 online.
|
hard-panel: панель закрыта за gnexus-auth (OAuth PKCE), allowlist владельца, верификация ok
...
Панель закрыта за gnexus-auth: OAuth-клиент владельца (gnx_3Sst...), GHARD_AUTH_* в .env на .176 (allowlist goldmikki@gmail.com), контейнер пересоздан. Верификация: /auth/login → 302 на auth.gnexus.space с PKCE, API без сессии 401, webhook без подписи 401, SPA-редирект на логин подтверждён в бандле. Креды сохранены в gnexus-creds (секрет hard-panel gnexus-auth OAuth client).
|
Add hpanel.gnexus.space route to internal nginx proxy targets (GHard panel, TLS, /mcp blocked)
...
Публикация hpanel.gnexus.space: nginx-сайт на internal-proxy (.226) с TLS (LE, до 2027-01-02), апстрим http://192.168.1.176:8000, /mcp закрыт 403 на прокси. Проверено вживую: DNS, health, UI HTTP/2, 301.
|
| 2026-10-04 |
hard-panel: prod updated to 40a8810 (personal MCP keys), migrations applied, verification ok
...
Update 2026-10-04 23:55: prod updated b6080d0..40a8810 (4 commits, 16 files +834/-19) — personal MCP keys: /mcp-keys page (issue/revoke mcp_* tokens, plaintext shown once, sha256 hash + hint in DB), require_mcp guard (personal mcp_* Bearer OR static GHARD_ADMIN_TOKEN super-token OR gnexus-auth cookie session), idempotent startup migrations (_try_alter users.system_role snapshot + users.blocked from auth webhooks; CREATE TABLE IF NOT EXISTS mcp_tokens). docker-compose.yml untouched upstream — stash pop clean, local edits preserved. DB backup before update: /data/hard-panel.db.bak-sqlite-20261004-2155-premcp (1.6 GB, sqlite3 backup API incl. WAL). Verification: health ok, fleet 19/19 online (transient offline after restart recovered in ~90s), agents ingesting 200 OK, MCP initialize OK, /mcp-keys 200, synapse reporter disabled as before.
|

hard-panel: prod updated to b6080d0 (auth, ui-kit 1.0, health-checker, Synapse integration), verification ok
...
Update 2026-10-04 21:30: prod updated edb6cff..b6080d0 (19 commits, 44 files +3897/-251, MCP serverInfo 1.30.0). New: gnexus-auth OAuth PKCE, gnexus-ui-kit 1.0 Verdant theme, services health-checker (per-service pages, handbook spec §5 self-report parsing), events journal, Gnexus Synapse integration (stage 4), i18n en/uk/ru + Settings, MCP prompt agent_guide + tools services_overview/service_incidents/events_recent. Upstream compose now forwards all documented GHARD_ env vars. Local compose conflict resolved manually (strict token form + 4 NAS volumes preserved). DB backups taken before update (file copy + sqlite3 backup API, 1.6 GB each, in volume /data). Schema extended at startup (oauth_states, sessions, users, services, service_samples, events). Verification: health ok, fleet 19/19 online, agents ingesting 200 OK, MCP initialize OK, /mcp unauth 406, SPA 200 with i18n, synapse reporter correctly disabled. Known expected: /auth/login 500 (GHARD_AUTH_BASE_URL unset) — gnexus-auth binding is the next stage.
|
gn-synapse: prod updated to 0ba611c (i18n + user_prefs/blocked), migration applied, verification ok
...
Update 2026-10-04 18:40: prod updated f13a251..0ba611c (3 commits: i18n locale override in Settings, handbook sync, ecosystem conventions). Diffstat 40 files +2281/-452 (i18n ru/en/uk, all views). Migration 20261004_0009_user_prefs applied (a4f8d2c17b90, table user_prefs: user_id PK, locale override, blocked flag from gnexus-auth webhooks gating owner MCP tokens). Pre-update pg_dump taken per runbook (backup-synapse-2026-10-04-1505-preupdate.sql, 26 KB). .env unchanged. Full verification passed (api healthy, worker processing tasks, healthz/readyz ok, SPA 200, /mcp 401 unauth, public healthz ok).
|
Record handbook local clone on rei and push path via pilar
...
Handbook cloned on navi rei host at /home/ubuntu/workspace/gnexus-handbook (HEAD 1a4a888). Push from rei requires a GitBucket token; pushes currently go via navi instance pilar on the gitbucket VM. Structure updated: 10-platform now includes health, i18n, mcp.
|
Update handbook doc after 20-deployment tone revision (commit 64362fc)
...
Handbook updated to commit 64362fc: 20-deployment rewritten as informational reference for developing agents (hosting, domains, backups describe how things are arranged, not prescriptions); README updated accordingly. Doc updated with latest SHA and structure note.
|
Document gnexus-handbook policy repository and cross-link from GitBucket
...
New application doc for the gnexus-handbook policy/conventions repository (created 2026-10-04, initial commit e10b4c3, 14 files, verified via GitBucket API). Cross-reference added to the Gnexus core group in the GitBucket doc.
|
gn-synapse: prod updated to f13a251 (personal MCP keys), migration mcp_tokens applied, verification ok
...
Update 2026-10-04 15:00: prod updated e00539e..f13a251 (personal MCP keys feature + LoginView fix upstream). Local LoginView fix dropped (upstream 2b86b4b identical, diff f7c45c0..f8ff947). Migration 20261004_0008_mcp_tokens applied (c9a2d6e71f43, table mcp_tokens). .env unchanged (MCP_TOKEN semantics: optional superadmin backdoor, /mcp always mounted, personal mcp_ keys via SPA /mcp-keys). Post-update backup taken. Full verification passed (api healthy, worker processing tasks, healthz/readyz ok, SPA 200, /mcp 401 unauth, /mcp-keys 200, public healthz ok). Pending item about pushing LoginView fix closed by upstream.
|
hard-panel: fleet +gnsynapse (id 46, 19/19 online); gnsynapse host doc: stack deployed + monitoring section
...
Added gnsynapse (192.168.1.172, panel id 46) to hard-panel fleet table (19/19 online), updated fleet date/counters, added python3.10-venv pitfall for fresh cloud-init VMs, deployment history entry 2026-10-04. Host doc gnsynapse: Synapse stack marked deployed (was Planned), added Monitoring section (hard-monitor agent, /opt/hard-monitor, systemd unit, panel id 46). Verified via live panel API and host systemd status.
|
Synapse SSO configured: OAuth client creds applied, redirect verified
...
SSO configured 2026-10-04: owner registered OAuth client synapse in gnexus-auth; GAUTH_CLIENT_ID/SECRET/WEBHOOK_SECRET written to .env on gnsynapse and to gnexus-creds secret gn-synapse-prod-secrets (encrypted). api+worker recreated, api healthy. Verified: /auth/login → 307 to auth.gnexus.space/oauth/authorize with real client_id, PKCE S256, scope openid email profile roles permissions; authorize endpoint → 302 to /login (client registered). Pending list shrunk: telegram deferred to #29, push optional, LoginView fix push still open.
|
Synapse minor update deployed; upstream LoginView bug fixed locally
...
Minor update 82d9181..e00539e deployed on gnsynapse 2026-10-04 13:37 by navi rei. Upstream bug: e00539e deleted gnexus-mark.svg but left the reference in LoginView.vue:19 — Vite build failed; fixed locally (sed to synapse-mark.svg), stack rebuilt and verified (api healthy, healthz/readyz ok, new asset 200, old asset SPA-fallback). Fix not pushed upstream — owner action required. DB backup taken before update.
|
Deploy Gnexus Synapse: service doc, route update, host inventory update
...
Deployed 2026-10-04 by navi rei: docker compose stack (api+worker+postgres+redis) on gnsynapse 192.168.1.172:8013, public entry https://synapse.gnexus.space via nginx on internal-proxy-vps (.226) with fresh LE cert. Full pipeline smoke verified: event accepted 202, delivery delivered. Server-side call-home to gnexus-auth routed via LAN extra_hosts (override file). Secrets in gnexus-creds (gn-synapse-prod-secrets).
|
Add gnsynapse host (192.168.1.172) for Gnexus Synapse project
...
Host verified live over SSH 2026-10-04 by navi rei: Ubuntu 22.04.5, 2 vCPU/2GB RAM/15GB disk, Docker enabled, static IP via cloud-init netplan (dhcp4=no), timezone set to Europe/Kyiv. IP chosen from verified free range 192.168.1.170-180.
|
| 2026-10-03 |
pfSense: correct C6 stock VPN facts — C6 v2 supports OpenVPN server per official TP-Link FAQ
...
Correction: official TP-Link FAQ 1544 lists Archer C6 v2 (and A6 v2/v3) as supporting OpenVPN server in stock firmware. Previous claim 'stock C6 has NO VPN server' was wrong for v2. Updated candidate approaches: OpenVPN server on his C6 + pfSense client as option 1.
|
pfSense: friend router = Archer C6 (no VPN on stock, OpenWrt possible); SSH-proxy approach chosen; WG readiness verified
...
Owner confirmed friend's router model: TP-Link Archer C6. Verified via OpenWrt/TechInfoDepot: stock firmware has no VPN server (port forwarding only), OpenWrt supported on all hardware versions. Chosen approach documented: SSH port-forward proxy first, WireGuard as upgrade path. pfSense WireGuard readiness verified live (wg binary, if_wg.ko module, DHCP WAN).
|
Version 8a4814a: crypto migration done (salted KDF + AAD), alembic 0002, extension 0.2.4, new env vars (salt, CORS), update count ×7
...
Prod update 2026-10-03 10:21-10:30 UTC to 8a4814a: 5 commits with crypto changes (salted scrypt master KDF, AAD-bound ciphertexts, hardened endpoints). Required: new GNEXUS_CREDS_MASTER_KEY_SALT env var + explicit CORS origins (production validator), alembic 0002_notes_widen, crypto migration script (not in image — docker cp needed): 2 keys rewrapped, 130 ciphertexts rebound, idempotency verified. Extension 0.2.4 deployed. App crash-looped twice during setup (missing salt, missing CORS) — documented.
|
| 2026-10-02 |
Version 53acfb4 (Bearer fails closed); revocation semantics updated; extension rebuild recipe via node container; update count ×6
...
Prod update 2026-10-03 00:30 (21:29 UTC) to 53acfb4: Bearer auth fails closed (invalid/revoked token → 401, no session-cookie fallthrough) — resolves the 'revoked token still works' finding from 2026-10-02. Chrome extension rebuilt host-side in node:22 container (byte-identical to release, CRC-verified), stack rebuilt, DB untouched. Doc updated: revocation semantics corrected, extension rebuild recipe documented.
|
Add token revocation semantics: revocation immediate, extension popup cache TTL 5min can look alive, session-cookie fallthrough for same-browser requests
...
Owner reported a revoked token (657738c3459fe543, Chromium PC) still working. Verified: revocation is effective (revoked_at set, last_used_at frozen, audit shows access.failed bearer:gcr_657738c3 after revocation, empirical Bearer request → 401). Apparent 'still works' = extension popup cache (TTL 5 min) and/or session-cookie fallthrough in actor_from_request for same-browser requests. Documented semantics.
|
Version 287b7a0 (session death → sign-in without reload); update procedure count ×5
...
Prod update 2026-10-02 21:01 UTC to 287b7a0 (frontend-only: session death → sign-in without reload). No migrations, DB untouched, extension unchanged. Verified: app healthy, public health ok, front 200.
|
Version 27f95b4, extension 0.2.2 + pack v24; add Extension deployment section (zips committed in repo, no host rebuild); update procedure note about migrations
...
Prod update 2026-10-02 20:30 UTC to 27f95b4 (webhook fixes, revoked tokens UI, extension 0.2.2 + pack v24). DB untouched (no migrations). Verified: app healthy, public health ok, extension zips in container. Learned: extension zips are committed to repo and COPYed into image — no host-side rebuild needed (no node on host).
|
Add Backups section: daily pg_dump + .env snapshot, 10-day retention, root crontab 04:00 UTC; note extension secrets use source=domain and REST channel
...
Owner requested daily backups (10-day retention) after a false data-loss alarm. Backup script created, tested (EXIT=0, 64KB dump + .env snapshot), scheduled in root crontab at 04:00 UTC. Also documented: extension secrets use source=site domain and REST channel.
|
Add users section (2 accounts, all secrets owned by goldmikki), correct secret counts to 71 (69+2, 20/49 allow_mcp)
...
Live DB check 2026-10-02: secrets.user_id join shows all 71 secrets owned by goldmikki@gmail.com; second account lnikita2o1o@gmail.com (Popelous, SSO 2026-09-18) owns 0 secrets. Corrected counts: 69 actual + 2 archived, 20/49 allow_mcp split.
|
Add audit semantics section: snapshots persist after delete/rename, archived secrets stay in DB, audit IP is the internal proxy
...
Audit reconciliation 2026-10-02: audit keeps title snapshots after deletion/rename, archived secrets stay in DB, audit IP is the reverse proxy (192.168.1.226). Verified against live DB on gnexus-creds-host.
|