Unify the wireless (airOS) device view onto the shared statCards grid — signal
(colored + meter), signal côté borne, CCQ, débit lien TX/RX, capacité airMAX,
SSID, fréquence, uptime, gestion, point d'accès (link) — same glanceable cards as
fiber, replacing the old chip row.
"See what's happening" — the three things F shows, now in OPS:
- actual traffic: live eth0 sparkline (already present)
- DHCP leases = the devices behind the CPE: ops_airos.php mode=leases SSHes the CPE
(port 2194) -> cat /tmp/dhcpd.leases -> {mac, ip, host, vendor(OUI), remaining};
hub airosLeases + GET /collab/airos-leases; UI renders an on-demand "Appareils du
client (DHCP)" table with refresh.
- wifi/wired clients = those leases.
Verified: leases for Bryson CPE -> Cisco-Linksys 172.31.1.19 (OUI vendor resolved),
through the hub too. Build clean, leak 0; deployed.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fiber twin of the wireless-MAC issue: when an ONT is replaced, F's `fibre` table
holds the current serial + OLT coords, but OPS keeps the old serial → dostuff
queries a serial no longer on the OLT → shows "offline" for an online customer
(C-LPB4: OPS had TPLGC4160688/ontid 0; F fibre id 13890 = TPLG55684520/ontid 2).
- hub: fibreById() reads F `fibre` by the SE's legacy_fibre_id (hub reaches
gestionclient); reconcileFibreOnu() aligns serial_number/gpon_serial/olt_ip/
slot/port/ontid; fibreLive({equipment}) reconciles THEN dostuffs with the current
serial. GET /collab/fibre-live, POST /collab/fibre-sync, and syncFibreOnus()
(park-wide, batched) folded into the daily 03h tick alongside wireless.
- ops: the fiber hero + "Do Stuff en direct" now call /collab/fibre-live (auto
reconcile on load), reflect the corrected serial locally, and show "ONU
synchronisée depuis F (remplacement)".
Verified: fibreLive(EQP-0000011366) auto-corrected TPLGC4160688 → TPLG55684520
(logged), and dostuff on the current serial = online (Rx -14.39 / Tx -13.78, bon,
uptime 1h17). Build clean, leak 0; deployed index.d652a898.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fast page-load status (no slow status.cgi): ops_airos.php gains a `mode=status`
path that returns RADIUS-only state (online, since, current MAC, AP IP) in ~ms.
Hub airosStatus() + GET /collab/airos-status; EquipmentDetail loads it on mount for
wireless → "● En ligne depuis 9h · IP …" badge (with MAC auto-sync on this call too).
Daily park-wide reconcile (03h ET): ops_airos.php `bulk=1` returns every wireless
service's current radacct MAC in one call (2 queries + merge); hub syncWirelessMacs()
indexes Service Equipment by mgmt IP and updates only changed MACs. Scheduled via a
/30-min tick that fires once in the 03h ET window (WIRELESS_MAC_SYNC=off to disable),
started lazily on first request (not at require → no timer in CLI/tests). Manual
trigger: POST /collab/airos-mac-sync.
Fixed erp.update success detection (returns {ok:false}, never throws) in autoSyncMac
+ syncWirelessMacs — counts are now accurate. Verified: reconcile bulk 351 / checked
333 / updated / errs=18 = pre-existing broken service_location links (flagged
separately, non-blocking). Deployed hub+F+SPA (index.5444d59b); leak 0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
MAC updates now flow into OPS automatically. On any signal read (wireless panel
OR the assistant's check_signal), the hub's airosSignal resolves the Service
Equipment record and, if the RADIUS-authoritative MAC (radacct.callingstationid,
auto-updated on WPA2 re-auth) differs from the stored mac_address, it updates
Service Equipment server-side — idempotent, fill/correct, no click.
UI shows a green "MAC synchronisée automatiquement depuis RADIUS (était …)" note
and reflects it in the Identity field; the manual button remains only as a fallback
if the auto-update can't run. Manual serial/MAC edit (InlineField) still available.
Verified: corrupted EQP-0000100001 mac to 00:00:00:00:00:00 → one airosSignal read
self-healed it to RADIUS 24:5A:4C:30:CE:AA (logged). SPA deployed; leak-check 0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
RADIUS MAC (user's ask — auto-follows equipment swaps): ops_airos.php resolves the
service's radius_user (device→service) and reads the latest radacct session →
callingstationid = current CPE MAC (auto-updated on WPA2-RADIUS re-auth) and
nasipaddress = the AP IP (reliable — replaces the flaky radio-MAC guess). Verified
on Bryson: RADIUS shows the current MAC and a *different* MAC 2 days prior (proving
the swap was auto-tracked). EquipmentDetail flags a RADIUS≠stored MAC mismatch with
a 1-click "Mettre à jour" (updateDoc) — manual serial/MAC edit (InlineField) stays.
The RADIUS DB isn't reachable from the hub (no GRANT) → done via the F bridge.
Live client-ethernet throughput (F parity): ops_airos returns eth0 rx/tx byte
counters + ts; EquipmentDetail's wireless panel gets a "Débit en direct" toggle that
polls every ~6s (self-rescheduling, no overlap) and draws ▼download/▲upload
sparklines (eth0 tx=download, rx=upload). Verified live: ↓8.0/↑1.9 kbps on Bryson.
Also: AP link now uses the reliable nasipaddress; signal tiles keep interval coloring.
Deployed (hub + F bridge + SPA index.27375132); leak-check 0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Backfill (scripts/backfill-wireless-equip.js): mirror F wireless CPEs
(airos_ac/airosm/cambium) into OPS Service Equipment, resolving customer +
service_location via device→service→delivery_id→Service Location(legacy_delivery_id),
dedup by MAC-or-IP, serial=MAC when F sn is null, undo manifest. Park was already
~99% synced — created the 2 genuinely-missing (incl. James A. Bryson) + flagged
devices without service/MAC. Verified: Bryson EQP-0000100001 → live LiteBeam 5AC
signal -74/-72, airMAX 93/67.
Signal interval coloring (F parity): wifiSigColor (airOS: >=-60 green … <-78 red)
and fibreRxColor (Rx -8..-25 green, -25..-28 orange, else red) color the live
signal tiles (wireless + fibre). Assistant analysis (check_signal advice) kept.
Device + AP links in service details: CPE management GUI link (Interface →
https://ip) in the wireless panel; AP link when its mgmt IP resolves (ops_airos.php
resolves the AP by radio MAC in F's device table, best-effort → apIp/ap_ip), else
shows the AP MAC. Hub normalizeAiros passes apIp through.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The hub can't reach the CPE network, so wireless signal now flows through a
new read-only F endpoint that CAN (F sits on that network + already has the
airOS code + device creds):
- services/legacy-bridge/ops_airos.php — token-gated (X-Ops-Token, same secret
as ops_reassign.php), SSRF-safe (only curls a mgmt IP present in F's `device`
table). Resolves device by serial/mac/ip, logs into the CPE (airOS >=8.5
/api/auth or <8.5 /login.cgi), reads status.cgi, returns normalized signal /
signal-AP / CCQ / TX-RX rate / airMAX capacity / model / uptime. v8 gets full
data; v6 falls back to top-level wireless.signal; CCQ clamped to 0-100.
- hub: airosSignal now calls the F bridge FIRST (OPS_LEGACY_URL host +
/app/data/ops_legacy.token), n8n webhook demoted to post-migration fallback.
Shared normalizeAiros() for both paths.
Verified end-to-end (hub -> F -> CPE): NanoBeam 5AC signal -53 / airMAX
140/143 Mbps; v6 NanoStation signal -47; offline CPE -> clean {ok:false}.
Auth reject + SSRF guard verified. docs updated (F bridge primary).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Last F Do Stuff gap was wireless CPEs (signal, signal-AP, CCQ, TX/RX rate,
airMAX capacity). Empirically confirmed the hub cannot reach the CPE mgmt
network (all probes timed out) — so, exactly like the fiber TP-Link path,
this goes through n8n.
- hub: ticket-collab.airosSignal({serial|ip}) + GET /collab/airos-signal.
Resolves mgmt IP/login from Service Equipment, calls n8n webhook
get_signal_airos (mirror of F device_ajax/airos_ac_ajax.php: airOS auth +
status.cgi parse), normalizes signal/signalAp/ccq/tx-rx rate/airMAX
capacity/model/version/uptime. Degrades gracefully: 404 -> registered:false.
- ops: EquipmentDetail shows a "Do Stuff en direct (sans-fil)" panel for
wireless equipment (isWireless heuristic), rendering the same tiles + feeding
"Copier pour le ticket". Shows an "en attente" note until the webhook exists.
Verified: hub live returns {ok:false,registered:false} for a real CPE (SN-772);
SPA build clean, leak-check 0 secrets, deployed (index.c4ccc86f.js). The one
remaining hop is creating the n8n get_signal_airos workflow (spec in docs).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- capacityByDay compte les jobs assignés SANS heure (untimed_used_h au prorata AM/PM/Soir)
- /roster/assign-job : garde 409 si le ticket F est déjà à un autre vrai tech (force pour écraser) ; SPA = dialog confirm
- mirrorAssign : start_time depuis F due_time (am/pm) + fix due_date (cur.→r.) + backfill fill-only dans watchLegacy (97% des tickets F = 'day' → le spread capacité est le fix porteur)
- pushAssignments : garde anti-clobber (conflits F skippés, force=1) + compteur conflicts
- ticket-collab : courriels de notification → deep link OPS /#/tickets?open= (fini le desk) ; TicketsPage lit ?open= ; Dashboard ouvre le ticket précis
- DetailModal + SupplierInvoices : lien desk gaté can('manage_settings')
- NOUVEAU JobMediaModule (photos tech géo-vérifiées + journal terrain) monté dans IssueDetail onsite + volet Planif ; hub /roster/job-media + /field/photo-file signé ; JobThread poll silencieux 45s
- legacy-sync emailList : strip caractères invisibles (soft hyphen U+00AD…) — fixait un 417 InvalidEmailAddress en boucle (compte 15988)
Déployé prod 2026-07-18 (hub restart + SPA /opt/ops-app), sondes live OK.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Planificateur « Suggérer » : 4 stratégies (smart / meilleurs d'abord / équilibré /
juste ce qu'il faut), compétences+niveaux par tech (édition inline), niveau requis
par compétence + par job, carte des tournées (1 couleur/tech, domicile→arrêts,
sélecteur de jour), fenêtre de dispatch auj.+demain (dates sélectionnées), règle
week-end + placeholder « en attente du quart », clustering + lasso + filtre-date
sur la carte, accès rapide « À assigner » (badge).
Boîte : liste /conversations allégée (45 Mo → ~1 Mo, 17×) + messages chargés à
l'ouverture. Rapports : cache SWR sur revenue-explorer (22×). Session : keep-alive
+ timeout fetch global + authFetch durci → fin des rechargements manuels.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Comms: sélecteur d'expéditeur « De: » (défaut groupe Support TARGO) via resolveSendFrom + alias vérifiés
- Notifs: prefs de feeds PAR utilisateur (/conversations/notif-prefs) + cloche à bascules ; boot tooltip-ux (clic prioritaire + anti-empilement)
- Courriel: invitation à évaluer = modèle Unlayer éditable (transactional-rating-invite-*) ; test-send via Gmail + expansion {{rating}} ; logo TARGO auto-hébergé sur le magasin d'actifs du hub
- Planif: bloc « sans déplacement » (damier, début de quart, alerte si pas de quart), quart éditable dans l'éditeur de jour, icônes de compétence en vue jour (TV pour télé), clic cellule → éditeur, clic gauche lane → liste / clic droit → menu quart, lien ↗ ticket par job
- Tickets: défaut « Non fermés » + correction du filtre « Mes tickets » (owner)
- Inbox: poll Gmail 1 min + rafraîchir à la demande (poll-now)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
