| 1.0 | Basic SNMP monitoring, an initial dashboard, and email alerts. |
| 1.1 | LLDP topology detection, automatically distinguishing an uplink from an end-device port. |
| 1.2 | Device lookup by IP/MAC down to the physical port, and 48-hour traffic history for every interface. |
| 1.15.7 | Fix: when the monitoring process froze, the automatic watchdog sometimes could not see it and left it running, so monitoring stayed down. The watchdog now finds a frozen monitoring process even when its process search is slow, and restarts it. |
| 1.15.6 | Fix: Telegram messages that could not be sent during a short outage are kept and sent late instead of being lost. A switch's back-up message gives the time it came back and, when it rebooted, the reboot too. |
| 1.15.5 | Fix: after hardening the server, switch photos could not be uploaded or deleted in the Layout tab. The hardening script now keeps write access to the photos folder. |
| 1.15.4 | Fix: the server hardening script could stop halfway and leave files that only the site's account could open. It now keeps SYSTEM and administrator access, checks the result, and puts the previous permissions back by itself if anything fails. |
| 1.15.3 | New: the search box also finds ports by their description or port name. Clicking a result opens the switch page with the port highlighted. |
| 1.15.2 | Fix: a false "monitoring stopped" alert after a server restart. The external watchdog now waits 10 minutes after the server boots and never stops a monitoring process that started less than 5 minutes ago. |
| 1.15.1 | Fixes: the outage history now also shows switch reboots (switches that rebooted together in one row) and the periods when monitoring itself was not running. The version number is shown next to the logo. |
| 1.15.0 | The system now reports on itself: a red banner on every screen and a Telegram/e-mail alert when the monitoring process is not running, and a message when it resumes; an external watchdog that alerts and restarts a stuck process; a low disk space alert; a switch reboot alert; a "not checked since" marker on devices with stale data. Fixes: reboots detected after a long monitoring gap, the web site starts while the database is locked, and the database log file is kept small. |
| 1.14.0 | New: a "Help" page in the menu — a short user guide inside the system for every signed-in user: what the system does, how to sign in, what each screen is for, common tasks step by step, common problems and safety rules. The guide is updated together with every change in how the system is operated. |
| 1.13.2 | Fix: a single SNMP timeout still sent a DOWN alert, followed moments later by "back up". A DOWN alert from the SNMP poll now needs 3 failures in a row and is not sent at all while the device still answers ping. Devices that answer ping are still caught within about 40 seconds by the fast check. |
| 1.13.1 | Fixes: false DOWN alerts followed moments later by "back up" — the fast ping check now waits up to 2 seconds, tries up to 3 times, and asks the switch over SNMP before declaring an outage; a switch that answers is not declared down. A real outage is still caught in about 40 seconds. An outage left open with no "back up" message (a race between the fast ping and the SNMP round) is now closed automatically. |
| 1.13.0 | New: the port-changes log now records cable connect and disconnect ("Connected"/"Disconnected"), not only speed, VLAN and description changes; it is recorded at every poll (every 5 minutes). Fixes: switches were shown only by IP address in the Port Watchdog, dashboard, map, search and in alert e-mails and Telegram messages; the switch name is now shown. In the Port Watchdog a single failed read discarded the switch's known port state, so a disconnect at that moment was never reported; the state is now kept and the switch is flagged until reads succeed again. |
| 1.12.2 | Fixes: in the Hebrew interface MAC addresses, IPs and host names were shown scrambled in search results and in the CSV import (for example the parts of a MAC reordered); they are now always left to right. An outage caught by the fast ping now starts at the first missed ping instead of the confirmation, and the outage-history note explains the fast ping. Security: real network data (a sample device list and an SNMP community string) was removed from the product, and every example in the interface now uses generic addresses. |
| 1.12.1 | Release-process fix: after an update the web server could be left stopped ("503") while the deploy reported success. The deploy now waits for the service to stop, starts it again and checks that the live site answers with the new version before the release is published. The application itself is unchanged. |
| 1.12.0 | The "Open in IMC" link was removed from the alerts log, the recurring-alerts panel and the port-details panel in the Switch Layout tab. Nothing else changes: the system never read or stored IMC data. |
| 1.11.2 | Fix: a settings change could be lost without anyone noticing, because the save button sat at the bottom of a long page. The button now stays pinned to the bottom of the screen, "You have unsaved changes" appears as soon as something changes, and leaving the page with unsaved changes asks first. |
| 1.11.1 | Fix: the Telegram test message succeeded even when "Send alerts to Telegram" was off, so it looked set up while real alerts were never sent. With the box off, the test still sends the message but now shows a red warning that real alerts will not go out until the box is ticked and saved. |
| 1.11.0 | Choose which alerts are sent by email: the email settings have a list with the same types as Telegram (device down, device back up, path outage, traffic load, CPU, memory and port load), so you can, for example, get only outages by email instead of every busy port. Email and Telegram are chosen separately, and anything not ticked is still recorded in the alerts log. Email can now also carry "device back up" and path outages, and quiet hours now apply to every alert email. |
| 1.10.0 | Fix: a short outage (under about 5 minutes) could go undetected entirely, because the regular polling only checks each device every few minutes. An independent, fast ping check now pings every device every 20 seconds; after two consecutive missed pings (about 40 seconds) the device is marked down and the same alert (email, Telegram, outage history) fires as before, while the regular polling keeps running unchanged. A device that blocks ICMP (for example a firewall) is never marked down by this check. |
| 1.9.0 | An "Open in IMC" link on a busy-port alert (in the alerts log and the recurring-alerts panel) and on the port-details panel in the Switch Layout tab: opens the organization's HPE IMC traffic-analysis (NTA) screen in a new tab, to check who is generating the load without leaving the switch's own sFlow collector. Link only — no IMC data is collected or stored. (Removed in 1.12.0.) |
| 1.8.0 | A Switch Layout tab on every switch page: the front of the switch with each port coloured (status, speed, VLAN, load or errors), a panel for every stack member, and a click on a port shows the devices connected to it with their Inventory class. Below each member the rear panel shows the state of the power supplies and fans. Read-only. Power supply reading is also fixed on the HPE 5130 (an empty slot is no longer shown as a fault) and the Aruba CX (a real power supply fault is shown). |
| 1.7.0 | Session length per user: on the Users screen you choose 8 hours (default), 24 hours, 7 days, 30 days or 1 year for each user, so a control-room user is not logged out of the Dashboard after 8 hours. It applies from the next sign-in and is written to the audit log. The sidebar also shows the real logo now instead of a simplified redrawing. |
| 1.6.1 | Telegram alerts fix: the test buttons ("Send test message" and "Find Chat ID") now save the settings on screen before running, and the reason for a failure is shown instead of a generic error. |
| 1.6.0 | Telegram alerts alongside email: in Settings you choose which kinds of alerts go to Telegram (device down, recovery, path outage, traffic load, CPU, memory and overloaded port), with a delay for port alerts so a port that fills up for a moment and clears never reaches your phone. Email keeps receiving everything, and there are buttons to find the Chat ID and to send a test message. Temperature colours now follow the limits the switch itself reports (instead of a fixed 45 and 60), so healthy switches with hot chip sensors are no longer shown as hot. |
| 1.5.0 | Three new Dashboard windows. Outage history: when each device went down, when it came back and how long it was down (24 hours, 7 or 30 days); an outage in which every device drops together is shown as a single path-outage event. An uptime log for every device: how long it has been running, when it booted, 7- and 30-day availability, and a click on a row opens its log of reboots and outages. And the hottest switches by temperature. Temperature reading is also fixed: it worked for 37 of 99 devices and now works for 91 (Aruba CX reports in thousandths of a degree and HPE 5130 was read from the wrong column), and dialogs opened in English are no longer laid out right to left. History is recorded from the day of installation. |
| 1.4.4 | Security hardening, part two: sign-in, sessions and the audit log. Logging out, deleting a user, changing a password or role, and resetting 2FA now end open sessions, and the role is read from the database on every request. Failed sign-ins are counted per user and per address, 2FA codes have their own limit and cannot be reused, and the client address is written to the audit log, which now also records the actions that were missing (users, devices, alert thresholds, report schedules, map and license). Replacing 2FA needs the current code, AD groups are matched exactly, and local passwords follow a policy (12+ characters, also in the installer wizard). The mail server's TLS certificate is now verified (configurable in Settings), the Uptime report, which never worked, is fixed, and libraries with known vulnerabilities are updated. |
| 1.4.3 | Security hardening from an OWASP Top 10 review: the server's source code can no longer be downloaded without logging in. SNMP community strings and SNMPv3 credentials are no longer returned to any user by the API or the CSV export, and editing a device leaves them untouched unless a new value is typed. The update installer only installs an update signed by the vendor. Added security headers, input checks for scans, imports and map uploads, and neutralised Excel formulas in CSV files. The Audit page is for administrators only, and clearing the log is itself logged. Also added a server-hardening script for the file permissions. |
| 1.4.2 | Closing alerts: on the Alerts page you can tick several alerts (Shift+click selects a range of rows) and close them, or close every open alert at once, and each closure is written to the audit log. Traffic, CPU and memory alerts close by themselves when the value drops back under the threshold, and alerts of a deleted device close 14 days after it was deleted. An alert that keeps coming back over the week is listed in a "Needs review" panel at the top of the page, and the Dashboard's Open Alerts card now shows the real total instead of counting only the 10 latest alerts. |
| 1.4.1 | A duration for every alert threshold: the threshold window has a new "Must stay above the threshold for (minutes)" field, so a specific switch can be set to alert only after it has stayed above the threshold for, say, 10 minutes. The default is 5 minutes, as before. The thresholds table gains a Duration column and an Edit button. Also fixed: saving an alert threshold for a single port failed silently; several ports on one device can now each have their own threshold, with its own duration. |
| 1.4.0 | Physical switch count on the Dashboard: a new card next to "Total Devices" shows how many real switches are on the network, not just how many IP addresses are monitored. One address is often a stack of 2 to 10 switches; the number is read from the equipment itself (ENTITY-MIB) and refreshed hourly. Also, HPE switches that Inventory classified as "Computers" are now classified as network infrastructure: a device whose IP address is monitored by the system is infrastructure. |
| 1.3.12 | Custom categories in Inventory: an admin can create a category of their own (e.g. "Radiology & Cardiology") straight from a VLAN's dropdown, pick an icon, and rename or delete it later. The new category gets its own summary card and works in the filter and the table like a built-in one. Each VLAN also gains an "Incl. identified" checkbox that makes the VLAN's category apply to every device on it, not only the ones that were still unknown. |
| 1.3.11 | Added a read-only database schema tool (check-schema.js) that lists the tables and shows each table's columns and row count. It finds the database the same way the server does, so it works on both standard and IIS installs. |
| 1.3.10 | VLAN names in Inventory: a new VLANs tab lists every VLAN the system detects on the network, read directly from the equipment (ARP tables and switch port VLANs), with its device count, how many of those are still unknown, and how many ports and switches carry it. An admin can name each VLAN (e.g. 220 = Cameras) and give it a category, and unknown devices on that VLAN are then classified with it. The device table gains a VLAN column and a VLAN filter. The diagnostics report also shows the poller's Scheduled Task on IIS installs. |
| 1.3.9 | Inventory fixes: the page showed only "Error" and never loaded because of a wrong column name in the query, and the query itself timed out against 970K records. Every ARP-learned endpoint was also missing from the page — 5,266 records including all of the wireless subnets — which is why searching an address in one of those subnets returned nothing. Missing IP addresses are now recovered by cross-referencing MAC addresses (coverage 64%→94%), Aruba access points that were being counted as computers are classified correctly (0→571), and reverse-DNS name resolution plus search over switch port descriptions were added. |
| 1.3.8 | Added an Endpoint Inventory page — classifies all active MACs by OUI (vendor) into categories: Computers, Printers, APs/WiFi, Cameras, Medical Devices, Network Infrastructure, VMs, and Unknown. Includes summary cards, a searchable/filterable paginated table, and Hebrew + English UI. |
| 1.3.7 | Added a "Collect Diagnostics" tool (Start Menu shortcut) that gathers a safe-to-share report — version, service status, recent logs, database schema/counts only (never row data), and .env key names only — for remote troubleshooting. Also fixed a second deploy pipeline gap where the installer's own scripts folder could silently go stale. |
| 1.3.6 | Fixed a deploy pipeline gap that had silently blocked every backend/db code change from reaching the live server since v1.3.4 — this is why the audit log and main search stopped working. Also fixed endpoint IP/hostname not showing on some ports, false "fan failed" hardware alarms on HPE 5130 switches, and added a VLAN column to the ports table. |
| 1.3.5 | Added hostname resolution (reverse DNS) for endpoints — every IP shown for a connected device now also shows its computer name, when available via DNS, in the port endpoints table and both search views. |
| 1.3.4 | Fixed the Audit page still showing Hebrew in English mode — log messages are now rebuilt in the active language instead of showing raw Hebrew text stored in the database. Fixed the website's "latest" download link, which was pointing at a file that was never actually uploaded. |
| 1.3.3 | Full English translation coverage — every page (Dashboard, Devices, Alerts, Trends, Audit, Topology, Map, Tools, Reports, Admin, License, Login) is now fully translated, not just the sidebar. Alert messages are now rebuilt in the active language instead of showing Hebrew text stored in the database. |
| 1.3.2 | Full RTL/LTR layout switch — switching to English moves the sidebar to the left and flips the entire UI to LTR. Fixed version chip to appear inside the sidebar instead of floating over buttons. |
| 1.3.1 | Fixed SNMP poller not running since Aug 13 (no Repetition Interval on Scheduled Task) — traffic charts and alerts now update again. Full UI translation coverage including all 4 license banners. Fixed sidebar scrolling with page content. |
| 1.3.0 | Standalone Port Watchdog page in the main menu — real-time port change monitoring with beep and browser notification. Fixed Hebrew/English language toggle in the sidebar. |
TK Comms Sentinel