v4.1.0-beta.29

Add files via upload

v4.1.0-beta.28

Add files via upload

v4.0.0 - Hyperbolic Awesomeness, Details Panel, Tablet/Desktop view and more integrations!

[4.0.0]

A major release. The headline is a second way to see your network - the
Hyperbolic view, an explorable tree - and a details panel for every
device, available in both views. The classic Flat diagram is still there and
looks as it did. Also includes everything drafted for 3.3.1, which never
shipped on its own: MikroTik and OPNsense Auto-Discovery, and the Summary
badge colours moving into the Summary section.

Added

Auto-Discovery (from the unreleased 3.3.1)

• MikroTik - three integrations: the core MikroTik integration,
  MikroTik Router (mikrotik_router) and MikroTik Extended
  (mikrotik_extended). Each finds the device as a Router/Gateway and,
  because nothing tells a router from a wireless bridge, also offers it as an
  Access Point node, plus its clients. MikroTik Router and MikroTik Extended
  also fill in the Router's Real-time Download/Upload from the WAN
  interface's traffic sensors, and MikroTik Extended fills in the WAN IP
  too (its cloud public address, or the WAN interface's address).
• OPNsense - finds the Router/Gateway from its WAN interface status, with
  the WAN's Real-time Download/Upload (the interface's in and out
  kilobytes per second), and its clients. Its built-in speed test is now a
  speed test source ("OPNsense (built-in)"); its Unbound Blocklist
  switches (the single legacy toggle and any individually named DNSBL rules)
  appear as DNS Filtering candidates; and its VPN status sensors, which report
  "up", count as active.

Hyperbolic view

• View Mode: choose Flat (the original diagram) or Hyperbolic from
  two tiles at the top of the Layout and Discover pages.
• An explorable, zoomable tree of the whole network on a Poincare disk, with
  the Router at the centre. Drag to pan, pinch or use + / - to zoom, tap a
  node to centre it, tap it again to open it, and use the target button to
  return to the centre.
• Every device type is a node: Internet (with its quota ring), Router, LAN,
  Switches, Access Points, Servers and their VMs and containers, count
  circles, individual clients, monitored services, and an Unknown node for
  clients that can't be placed.
• Clients layout: attach clients to the access point or switch they are
  connected to, or collect them in groups.
• Node Sizes: one slider for all nodes, plus a slider per type.
• Tree Appearance: outline and background colours with their own
  transparency, an optional Glass style (frosted, with adjustable blur),
  and a transparent card background.
• Offline Nodes: colours for the "!" badge on an offline node and for the
  dotted line leading to it.
• Arrangement: Auto, Phone or Tablet. On a tablet the tree can be sized to
  fill the screen height exactly ("fit tree to screen height") or to a width of
  your choice.

Details panel (Hyperbolic view, and Flat view when switched on)

• Select any node to see its details: name and state, live throughput, IP
  addresses (WAN, secondary WAN, LAN, and a device's own), VPN, firewall and
  DNS-filtering state.
• Device Page button: opens the device in Settings > Devices, under the
  integration it comes from. A node with no device (or a user without access to
  Settings) gets More Info instead. Choose between the two with
  Header Button.
• Clients list: every client connected to the selected device, with a dot in
  its online or offline colour, online ones first. Tap one to jump to it.
• Client Info: for a single client - status and since when, last online,
  connected to, IP, MAC, hostname, SSID, VLAN, switch port, band, channel,
  signal, link speed, traffic, uptime, manufacturer, guest network, area, and
  any other attributes the tracker reports.
• Device statistics: CPU, RAM, temperature, traffic received and sent, PoE
  power, ports up, uptime, firmware and model, for any device that reports them
  (UniFi Device Info and similar).
• PoE Ports: a switch's PoE ports with the watts each is delivering,
  busiest first, from the switch's own attributes or from per-port PoE sensors.
• Servers: CPU, memory, disk, uptime, temperature, load and network
  readings from the sensors on the server's device, plus a Services list of
  its VMs and containers headed by a total, for example "11 of 12 online". The
  total comes from count entities you choose, from count sensors on the device,
  from the guests the integration lists, or from a Home Assistant label you
  put on the guests (works with any integration).
• History graphs: Throughput, Latency (ping and jitter), Connected clients,
  Response time, Resources (CPU, memory and disk for a server; CPU and memory
  for a VM or container) and a client's own Throughput. Each has its own
  timescale buttons (5m, 30m, 1h, 12h, 24h, 72h); the choice is remembered on
  that device. The graphs follow live readings as they arrive.
• Graph Colors: a colour for each kind of series - Download, Upload, Ping,
  Jitter, Connected Clients, Response Time, CPU, Memory and Disk.
• Show as Pop-up: show the panel as a sheet that slides up from the bottom
  instead of beside or below the diagram. On a tablet the summary then moves to
  the top and the diagram takes the whole card.
• Panel background and outline colours, and card shape and shadow that follow
  your theme (--border-radius, --box-shadow).

Flat view with details

• Show Details Panel for the Flat view: the diagram sits beside (tablet) or
  above (phone) a summary card and the same details panel. Tap a circle to see
  its details and again to open it; the selected circle gets a ring in its own
  border colour.

Summary

• The summary is now a card of its own: optional Summary title, background
  and outline colours, and up to six items (it was five).
• With no details panel beside it (switched off, or shown as a pop-up), the
  summary's badges are centred across the card instead of left-aligned.

Clients

• Sub-Group Width: Even - every sub-group exactly the same width, one equal
  column per group across the whole Clients box.

IP addresses and discovery

• Managed switches and servers can show their IP address as a pill above
  their circle, with an IP Address Entity setting and badge colours. If none
  is set, the card finds the address itself: the device's own attributes, a
  sensor on the same device, or the host in the device's web address.
• A gateway's LAN IP and WAN IP are read from the right attributes (for
  example UniFi Device Info's lan_ip and wan_ip), so one sensor can supply
  both.
• UniFi Device Info discovery now pairs a gateway with its WAN sensor and
  carries each switch's address into the configuration.

Changed

• Summary badge colours moved to the Summary section (from the unreleased
  3.3.1). The Speed Test Download, Speed Test Upload, Real-time Download,
  Real-time Upload and Latency and Jitter badges each have a Badge Color and
  Badge Icon Color under Layout > Summary, shown for the items you have
  chosen, and they apply to the whole card. The Real-time badges now have
  their own pair (they used to share the Download and Upload colours). The
  Internet page no longer has Download Badge Background Color, Download
  Badge Icon Color, Upload Badge Background Color, Upload Badge Icon
  Color, Ping Badge Background Color, Ping Badge Icon Color, or
  Download Line and Upload Line. Colours you set in those fields
  (internet.colors.download_badge, download_badge_icon, upload_badge,
  upload_badge_icon, ping_badge, ping_badge_icon) are not carried over:
  the badges start from the defaults (green, pink and cyan, with
  theme-coloured icons) unless you set the new summary_*_color options.
• The Internet connection line follows the global Flow Line Color (from the
  unreleased 3.3.1), like every other line on the card. It used the Internet's
  own Download and Upload line colours.
• Summary Position offers Top and Bottom only. Left and Right had
  stopped doing anything in the Hyperbolic view and the Flat view with a details
  panel (both placed the summary as Bottom would). A Left or Right saved by an
  older version is still shown, marked as an older setting, and the classic Flat
  strip still honours it.
• Advanced is now Layout, rebuilt as collapsible sections: General,
  Graph, Summary, Details Panel, and Sizing and Animations. Show IP
  Addressing moved from General to Graph.
• The wording "Connected Devices" / "Devices" is now "Connected Clients" /
  "Clients" throughout. Configuration keys are unchanged, so existing
  configurations keep working.
• With no quota entities set, the Internet circle's ring now uses its Border
  Color (it used the Quota Remaining Color).
• A text state that happens to start with a number (an uptime such as
  "3d 4h") is no longer shown as that number.
• Tree circles are opaque, so a line never shows through a node.

Fixed

• Speed test discovery only considers sensor entities of the speed test
  integrations (from the unreleased 3.3.1), so the other entities some of them
  create are no longer offered as readings.

v3.3.0 - More integrations, more internet, more ways to display

[3.3.0]

A major release built around Monitoring: bring Uptime Kuma, Ping,
Gatus, and UptimeRobot straight into the diagram as their own badges
and circles, not just router-brand integrations. Also adds a
Secondary Internet entity for a backup or genuine second
connection, real-time Download/Upload summary items sourced from the
Router itself, a Card Size control (Compact, Fit to Width, or a
manual Scaled percentage) aimed at fitting large networks on a phone
screen, keyword search across every Auto-Discovery list, and
support for additional integrations.

Added

• Monitoring - a new main menu, and a new way of drawing a service's
  status directly on the diagram rather than only on the device that
  runs it. Auto-discovery covers four integrations, each confirmed
  against home-assistant/core's own test suite where one exists:
  Uptime Kuma, Gatus, and UptimeRobot (all three checked
  directly against their real entity/snapshot test fixtures), and
  Ping (from its documented round_trip_time_avg attribute). Each
  discovered or hand-added service gets its own entity, name override,
  icon, and border/icon colors (with separate colors while it's down),
  plus a Type of Auto, External, or Internal:
  • External services draw in a new dashed box above Internet,
    joined to it by dotted lines - one per circle when a Secondary
    Internet is configured.
  • Internal services can be assigned to Clients (their own
    "Monitored" sub-group) or to any device or client already
    configured anywhere on the card, not just Router and Servers -
    Internet, LAN, the Core Switch, any Switch or Access Point at any
    depth, any Server, or any individual Client. A device with several
    services shows the worst state among them.
  • Auto resolves live, from the monitor's own data: its tags,
    then the address it checks (a public address is External; a
    private range, single-word host, or .local/.lan name is
    Internal). Gatus exposes neither, so Auto falls back to checking
    its name for a keyword like "External" or "Internal" (Gatus puts
    its group name into the entity name); UptimeRobot, with nothing to
    go on, defaults to External, since a cloud checker can only be
    watching something reachable from the internet. The discovery list
    and the service editor both flag a service Auto can't classify.
  • An optional response-time badge, per service (Show Response
    Time Badge), at the bottom of its own circle (not on a device
    badge) - from its Response Time entity, or a status entity's own
    round_trip_time_avg attribute - colored globally by the same
    single-color-or-threshold choice PoE badges use.
  • A down badge flashes, following the card's own Enable Animations
    switch like every other animation on the card.
  • Global settings: badge colors and icons by state (up/down/pending/
    maintenance), badge location, the icon shown when several services
    share one device badge, how an unavailable entity is treated
    (Pending by default, or Down), Box Border Color/Radius for the
    External box, and Show Monitored Names - independent of the
    Clients page's own Show Names, so either can be on without the
    other.
  • Both the auto-discovery list and the list of services already
    added can be filtered by integration, each showing how many of
    each it found; filtering the already-added list also turns "Delete
    All" into "Delete Shown".
• Real-time Download / Upload summary items, sourced from two new
  entity pickers on the Router page (Download Speed Entity,
  Upload Speed Entity). With no Router configured, they fall back
  to the Primary Access Point's own Download/Upload entities instead;
  a Router that is configured but has no speed entity for a direction
  shows nothing for that direction rather than borrowing the Primary
  AP's. Router discovery for FRITZ!Box, pfSense, ASUSWRT,
  and AsusRouter (already supported for the router entity itself)
  now also fills these in automatically where each integration exposes
  live throughput: FRITZ!Box's and pfSense's own WAN rate sensors
  (pfSense from the default gateway's interface specifically, both
  directions from the same WAN), and ASUSWRT's and AsusRouter's
  "Download/Upload speed" sensors - both confirmed against
  home-assistant/core's own test fixtures, including which sensors are
  enabled by default and which are only created for interfaces
  selected in AsusRouter's own options. ASUSWRT's speed sensors carry
  a caution in the discovery list, since they're reported to read 0 or
  identical values on some routers. They're now the default Summary
  Item 1 and 2, ahead of the plain Speedtest-sourced Download/Upload.
• Secondary WAN IP Address, a new Router-page field shown only once
  a Secondary Internet exists, displayed under the secondary circle the
  same way the primary's WAN Address already is. AsusRouter fills it in
  automatically - confirmed against its own official change log
  (wan_ip / wan_ip_secondary sensors, added in integration 0.24.0
  alongside its secondary-WAN support). Checked as well against UniFi
  and TP-Link (Omada, Deco, Router): neither exposes a distinct
  secondary WAN address in any integration this card already supports,
  so there's nothing to auto-fill there yet.
• Card Size (Advanced → Layout) - one choice replacing the size
  sliders as the way to fit more on a small screen:
  • Normal - no change; the size sliders below work as set.
  • Compact - every circle, badge, and icon shrinks to its
    slider's own minimum, the column gap tightens to its minimum, and
    every name on the card - client, monitored-service, and the label
    inside a Router/Switch/AP/LAN/Server circle alike - hides, all
    without touching the size sliders themselves, so switching back to
    Normal is instant.
  • Fit to Width - automatically zooms the whole diagram down just
    enough to fit the card's width, recalculating whenever the card
    resizes or the network changes, down to a 40% floor.
  • Scaled - a manual zoom percentage (40-100%), remembered on its
    own even when switching to another Card Size and back.
  • A separate Horizontal Scroll toggle applies to every Card Size
    except Fit to Width (already sized not to need it) - lets a
    diagram wider than the card pan sideways, in both directions, to
    reach content that overflows on either side, instead of it being
    clipped by the card's own edge, which it previously always was,
    silently, with no way to see the rest.
  • A separate Hide Names toggle (Advanced → Layout) is a global
    override independent of Card Size or either page's own Show Names:
    on, it hides every name on the card at once, everywhere Compact
    does, without switching the rest of Compact's sizing on too.
• Secondary Internet: a second internet_secondary entity draws
  alongside the primary, with the same fields (bandwidth, quota,
  colors) and its own dashed circle outline, for a backup line or a
  genuine second connection - always the same Circle and Icon Size as
  the primary, so the one pair of sliders governs both. Edited from
  inside the Internet page itself (Primary/Secondary tabs) rather than
  a separate menu entry, and Speedtest discovery adds or removes it
  with the same plain buttons as the primary, acting on whichever tab
  is open. A Mode choice - Active/Active or Active/Standby,
  defaulting to Active/Standby - decides whether the secondary is drawn
  solid and animated (carrying its own traffic) or dashed and still (a
  backup, showing no flow at all while idle); in Active/Standby, it
  switches to solid and starts animating automatically the moment the
  primary actually goes down and the secondary is the one carrying
  traffic, then back to dashed and still once the primary recovers.
  Every line in the assembly - both drop lines, the T-bar, and the stem
  into whatever sits below it - uses the one global Flow Line Color
  rather than the Internet's own Download/Upload colors, at the same
  tighter spacing a typical Access Point's own dual line uses, and the
  T-bar reaches the true outer edge of a paired download/upload line
  rather than stopping short of it. Where the Primary AP itself sits
  directly under Internet with no Router or Switch between them, its
  line uses the normal line color and its own Download/Upload entities
  to decide a single or double line, rather than Internet's.
• Keyword search, alongside the integration filter, on every
  Auto-Discovery list on the card - Speedtest, ISP, pfSense, AdGuard,
  Routers, Main Switches, Nodes, Clients, Containers, and Monitoring -
  matching a candidate by its name or its entity id. Where a list
  already has an integration filter, search narrows within it rather
  than replacing it.
• ASUSWRT Auto-Discovery - experimental (the built-in asuswrt
  integration, confirmed against home-assistant/core's own source; not
  the same as the community AsusRouter integration already supported,
  and the two can be installed side by side). Router: uses the
  integration's own "Devices connected" sensor, which is enabled by
  default, tied to the router's own device, and goes unavailable
  whenever the router can't be polled - so it's a genuine router-status
  entity, and also fills in the Router's LAN Connected Devices count.
  Only if that sensor has been disabled does it fall back to a client
  tracker as a best-effort proxy. Every configured ASUSWRT router is
  offered as its own candidate. Clients: every tracked device - by
  default only devices that report a name, unless "Track unknown
  devices" is turned on in the integration itself. Nothing is offe...

v3.2.0 - Endless Possibilities + discovery for Netgear, Fast.com and Mobile App (all experimental)

[3.2.0]

Adds unlimited-depth tiers (a fed Switch, and now an Access Point too,
can feed further Switches/Access Points/Servers to any depth), renames
Homelab to Server everywhere it's shown, per-server Security targeting,
Netgear and Fast.com Auto-Discovery, a real expandable tree for the
Nodes editor page with reordering and continuous numbering at every
tier, a "Core Switch" rename for the top-level Switch menu, and a
confirmed fix for bandwidth connectors drawing a phantom second strand
whenever only one of Download/Upload was configured.

Added

• Unlimited tiers. A fed Switch can carry its own access_points,
  fed_switches, and homelabs, and those can carry theirs, to any
  depth - not just the one level of Fed Switches supported previously.
  Every extra level adds one row to the diagram between the Access
  Point row and Connected Devices. A Switch/AP feeding several children
  is centred over every column beneath it, with its own stem and mini
  bus-line; offline state cascades down the whole chain; PoE totals sum
  Access Points, Servers, and Switches at every depth.
• An Access Point can be a parent too. Straight off the bus or
  nested under a Switch, an Access Point can now feed its own
  fed_switches, homelabs, and further access_points (a wireless
  mesh) - everywhere a Switch could feed something, an Access Point now
  can as well. An Access Point that feeds anything also gets its own
  dotted line down to the Clients box, the same as a Primary AP in
  Tiered layout, with its Connected Devices circle hanging off that
  line if it has one.
• A Server can hang off anything, not only sit at the top level -
  under any Switch or Access Point, at any depth, with its own
  Containers/VMs, PoE, and Connected Devices intact. (See the Homelab →
  Server rename below.)
• Per-server Security targeting. VPN, Firewall, DNS Filtering, and
  Reverse Proxy are still configured once and pointed at a device type
  (Server among them), but each Server can now choose which of those
  items it actually shows:

  homelabs:
    - entity: sensor.server_a
      security: { dns: true, reverse_proxy: true }
    - entity: sensor.server_b
      security: { vpn: true }

  A Server with no security key shows every item pointed at it,
  exactly as before - additive, not a breaking change.
• Netgear Auto-Discovery - experimental. Covers the official
  netgear integration (Orbi, Nighthawk, and other pynetgear-supported
  routers), confirmed against home-assistant/core's own source. Offers
  a Router candidate via one of its connected devices' trackers as a
  reachability proxy - the same technique OpenWrt (LuCI) already uses,
  since Netgear has no dedicated router-status entity, and its
  one enabled-by-default entity (the firmware Update entity) reports
  "off" when simply up to date - which this card's offline check
  would otherwise misread as offline. Every attached device is offered
  as a Client; Netgear's own integration never distinguishes a
  satellite, switch, or AP from an ordinary phone or laptop, so nothing
  from it is ever offered as a Switch/AP node. Marked experimental
  because of the client-proxy reachability check and less real-world
  mileage than the router integrations already supported.
• Fast.com added to Speedtest Auto-Discovery (fastdotcom platform,
  confirmed against home-assistant/core's own source) - a single
  Download-only sensor, no Ping/Jitter/Upload/ISP, since that's the
  only figure fast.com measures. Its single_config_entry manifest
  flag means at most one Fast.com service will ever be found.
• Home Assistant Mobile App Auto-Discovery (mobile_app platform,
  confirmed against home-assistant/core's own source) - every device
  running the official Companion app gets one device_tracker entity,
  offered on the Clients page. Clients-only: unlike every router-brand
  integration above, there's no Router/Node concept here at all - a
  phone or tablet is always a personal device, never network hardware.
• Editor: the Nodes page is a real expandable tree.
  • Every Switch or Access Point row has a chevron that expands it in
    place, with a live online/offline dot and a count of everything
    below it.
  • "+ Access Point / + Switch / + Server" sit directly inside the same
    card as the row they add to, ahead of its children, instead of at
    the bottom of a long flat list.
  • "Add AP / Add Switch / Add Server" are three separate buttons at
    the bottom of the page, each creating the right node type directly,
    instead of one "Add Node" button that always defaulted to an
    Access Point.
  • Move up/down now works at every tier, not just the top level - an
    Access Point, fed Switch, or Server reorders among its own siblings
    of the same kind under the same parent (Access Points, Switches,
    and Servers are always drawn as three separate blocks, so that's
    the only reordering that's meaningful).
  • Fallback labels ("Switch 3", "AP 4", ...) now share one running
    count per kind across the whole tree, instead of restarting at 1
    under every new parent.
  • A discovered candidate's "Add as" dropdown now lists every eligible
    Switch/Access Point in the tree, with a breadcrumb ("D-Link ›
    Workshop PoE"), not only top-level Nodes - this also lets a
    discovered Proxmox Server attach under an existing node instead of
    only ever becoming a new one.
  • Deleting a Switch/AP with devices below it asks for confirmation
    first. Only one Access Point can be Primary, wherever it sits in
    the tree.

Changed

• "Homelab" is now "Server" everywhere it's shown - buttons,
  headers, tooltips, dropdown options, and the diagram's own fallback
  label text. Display-only: the YAML keys (homelab, homelabs, the
  homelab Security target value, ...) are unchanged, so no existing
  config needs to change.
• The top-level "Switch" menu is now "Core Switch" - its entry in
  the menu list and its own page header only. Every Node-level "Switch"
  (Fed Switches, the node-type dropdown, "Main Switch" as a Security
  target, tree labels, ...) is unchanged, since that's a distinct,
  still-called-Switch concept from the one Core element between
  Router/Gateway and Nodes.
• A Switch's mini bus-line is now drawn as one exact centre-to-centre
  stretch per child, instead of a single line inset by a fixed amount
  from the group's edges - this replaces the previous inset-based
  adjustment, which could still overshoot the last drop by a few pixels
  in a mixed AP/Fed-Switch row.
• Four long inline descriptions became hover tooltips (the same small
  (i) icon already used elsewhere), instead of a paragraph sitting
  underneath the control: Clients page - Grouping, Show Client Names;
  Advanced page - Show IP Addressing, Enable Animations.

Known limitations

• A Switch that feeds other devices doesn't show its own Connected
  Devices circle - there's no free row for it. An Access Point in the
  same position does (see "Added" above).
• A Primary AP anywhere but straight off the bus (inside a Switch's
  group, or hanging off another Access Point) still renders Flat, not
  Tiered.
• Each Security item (VPN, Firewall, DNS, Reverse Proxy) still has one
  entity; Servers can each choose whether to show it, not point it at a
  different entity per server.
• Auto-Discovery still never knows which switch physically feeds which
  switch/AP - no supported integration exposes that uplink topology to
  Home Assistant - so a deep, discovered tree still has to be assembled
  by hand via the attach-to-existing dropdown or the Node editor.
  Netgear's own integration goes a step further and can't tell a
  satellite, switch, or AP apart from an ordinary client at all, so its
  devices only ever show up as Clients.
• The Nodes tree's expanded/collapsed state lives only in the open
  editor session; reordering or deleting inside a Switch collapses
  whatever was expanded beneath it.

v3.1.7 - zip file fix, client names, device area groupings etc

[3.1.7]

Fixed

• Client sub-group boxes rendered enormous whenever Show Client
  Names was on - root-caused via direct browser DevTools inspection
  after an initial layout-based fix didn't resolve it. The actual
  cause was a JavaScript string-concatenation bug in 3.1.6:
  individual_devices_group_padding arrives as the string "6"
  (config values commonly do), and the extra bottom padding added for
  Show Client Names was computed as pad + 18 - since pad was a
  string, + concatenated rather than added, producing "618" (as in
  618px) instead of 24. Every group box independently carried this
  same broken padding, which is why it affected every box uniformly
  regardless of its own content. pad is now coerced to a real Number
  at its source, not just at the one call site that happened to do
  arithmetic on it, so the same class of bug can't resurface through
  some other future use of the same variable.
• Invalid <rect> attribute errors on every Clients box render
  ("Invalid value for <rect> attribute width="calc(100% - 2px)"",
  visible in the browser console) - calc() is only valid for SVG
  geometry properties when set through CSS (the style attribute or a
  stylesheet), not as a plain XML attribute value, which is how width
  and height were set on the dashed-border rect drawn behind the
  Clients box and each of its sub-groups. Moved both into the style
  attribute alongside rx/ry, which were already handled correctly
  this way. A real, confirmed bug, but a cosmetic one on its own - the
  SVG's own outer size was already correct via CSS regardless, so this
  didn't contribute to the height issue above.
• Sub-group boxes no longer stretch to match whichever sibling is
  tallest (align-items: flex-start instead of stretch on their
  shared flex container). This turned out not to be the cause of the
  height issue above, but is a real fix in its own right: without it,
  one heavily-populated group wrapping onto multiple rows could still
  impose its height on every sibling regardless of their own content.
• The Internet circle's outline now defaults to the Billing
  Remaining color when no billing entities are configured at all,
  instead of falling back to the Internet circle's general theme
  color. Previously, once billing entities were set, the plain circle
  border went transparent in favor of the SVG progress ring drawn on
  top of it (colored by Billing Remaining/Billing Progress) - but
  without billing configured, the border fell back to whatever generic
  color the Internet circle otherwise uses, which read as unrelated
  to billing styling entirely. This distinguishes "billing configured
  but a sensor value is temporarily unavailable" (still falls back to
  the general circle color, unchanged, so the border doesn't flicker
  colors as a sensor briefly goes unavailable) from "billing was never
  configured for this Internet entity" (now uses Billing Remaining
  color specifically).
• The editor is no longer a separate lazily-loaded file -

v3.1.2 - Switch only fix

[3.1.2]

Fixed

• A bus made entirely of Switch nodes rendered crushed toward the
  bottom whenever no Access Point existed anywhere on the diagram -
  the grid row reserved for a Fed Switch's own circle and connector was
  sized by whether any Access Point existed (hasAnyAp), while the
  content actually placed into that row was correctly gated by whether
  an Access Point or a Fed Switch existed (hasApRow). An
  Access-Point-free, Switch/Fed-Switch-only bus had real content
  rendered into a row deliberately sized to 0px, squashing every Fed
  Switch and the Clients box beneath it upward into an overlapping
  mess. Row height now uses the same condition that gates the content.

v3.1.1 - Editor bug fix

Fixed

• Router page's WAN Address Entity picker rendered as broken raw
  text (".label=WAN Address Entity @value-changed=..." visible
  directly in the editor) instead of an entity picker - introduced
  earlier in this same cycle by the info-tooltip conversion above,
  which accidentally dropped the picker's own opening

v3.1.0 - Discover all the things! Unifi, Omada, Synology, and AsusRouter autodiscovery + Performance Enhancements

[3.1.0]

Adds Auto-Discovery for UniFi Network and TP-Link Omada across Router,
Switch, Nodes, and Clients, catches up this changelog on the
Synology SRM and AsusRouter (AiMesh) discovery already in the codebase
but never previously documented, adds the ability to attach a
discovered Access Point or Switch under an already-configured Node
instead of always creating a new one, Client grouping by Home Assistant
Area, an optional "Show Client Names" label, and a substantial editor
and rendering performance pass aimed at the Companion app.

Added

• Synology SRM and AsusRouter (AiMesh) Auto-Discovery, present in
  the codebase but never previously documented in this changelog:
  Synology SRM for Router/Gateway and Clients - like LuCI, it has no
  dedicated router-status entity of its own, so its candidate uses one
  of its client-tracker entities as a reachability proxy too, though
  best-effort only, since this integration's older polling model
  doesn't guarantee it goes unavailable the moment the router drops the
  way LuCI's does. AsusRouter for Router/Gateway (a real connectivity
  sensor on its main AiMesh unit, as reliable as TP-Link's), its
  satellite units as Access Point Nodes (defaulting to Primary AP when
  added, same as a TP-Link Deco satellite), and its Clients.
• UniFi Network and TP-Link Omada Auto-Discovery for Router/Gateway,
  Switches, Access Points, and Clients, confirmed directly against each
  integration's own source. UniFi has no dedicated "device type"
  attribute, so classification goes by matching the HA device
  registry's model field against known Ubiquiti product-family
  prefixes; Omada exposes its own device type directly on the device
  tracker's type attribute, so its classification is exact. Neither
  integration exposes uplink/topology data to Home Assistant, so
  results are flat lists, same as every other integration this card
  supports - grouping a switch under the node it actually feeds is
  still up to you (see the next item). Client grouping, Guest Network
  badge detection, and wired/SSID/VLAN sub-grouping were all extended
  to recognize Omada's attribute names (guest, switch_mac, ssid,
  vlan_id) and AsusRouter's guest attribute alongside the existing
  TP-Link Deco/UniFi ones.
• Attach a discovered Access Point or Switch under an existing Node
  instead of always creating a new one - a per-candidate dropdown on
  the Nodes discovery list sends it straight into an already-configured
  Node's Access Points or Fed Switches. Only Nodes that already have
  their own Switch are offered, since Access Points/Fed Switches only
  ever hang off a Node's Switch.
• Auto-Discovery on the Switch page: finds devices UniFi or Omada
  have already classified as switches - the same classification used
  for Fed Switch and Nodes discovery, not a separate check of its own.
  No other integration this card supports models a distinct switch
  role, so nothing else is scanned here.
• Group Clients by Device Area - a new "Device Area" option
  alongside SSID/VLAN/Connected AP, resolving each client's assigned
  Home Assistant Area (falling back to its device's Area if the entity
  itself has none), independent of any integration.
• Show Client Names toggle (Clients → Layout) - displays each
  client's name in small text directly below its circle. Off by
  default, since it noticeably increases visual density on a Clients
  box with a lot of devices; the full name remains available on
  tap/hover either way.

Changed

• Icon rendering switched from <ha-icon> to <ha-svg-icon> for
  every fixed/default icon on the card (badges, node circles, the
  offline "X" marker) - skips the async icon-metadata lookup ha-icon
  performs per instance, worthwhile since a populated topology can put
  40-80 icons on screen at once. Any icon overridden with a custom
  string still renders via <ha-icon> as before, so customizability is
  unaffected.
• The card and editor are now minified as part of the release
  build (Rollup + Terser) rather than shipping full readable source -
  previously only the bundled Lit library was minified; the card's own
  several thousand lines of source shipped completely untouched,
  roughly doubling the JavaScript the browser had to parse.
• Discover page section order: Router/Gateway now comes before
  Security · DNS Filtering, which in turn comes before Nodes.
• Clients page reorganized: the "Labels" section is gone - "Show
  Client Names" now lives under a renamed "Layout" section (previously
  "Box & Connector"). Sub-sections now run Clients → Auto-Discovery →
  Grouping → Guest Network → Layout.
• Guest Network badge default icon changed from
  mdi:account-question to mdi:account.
• Long explanatory paragraphs under section headings throughout the
  editor (the Auto-Discovery blocks especially) are now a small info
  icon carrying the same text as a hover tooltip, instead of
  always-visible text pushing that section's actual controls further
  down the page.

v3.0.1 - IMPORTANT: Now locally bundled, and enhanced auto discovery!

[3.0.1]

A major release built around Auto-Discovery: scan a supported
integration directly from the editor to populate Router, Nodes,
Clients, Internet, Security, and Homelab, instead of hand-picking
every entity yourself. Also introduces IP Address badges and renames
Individual Devices to Clients.

Added

• Auto-Discovery framework across most editor pages, all following
  the same pattern: scan, review a list of what was found, select the
  ones you want, add. Nothing is ever applied without an explicit
  choice, and re-selecting a different candidate than the one already
  in place prompts for confirmation before overwriting anything.
• Router/Gateway discovery: TP-Link Router, TP-Link Deco (its
  master unit), and OpenWrt via the LuCI integration (which has no
  dedicated router-status entity of its own, so its candidate uses one
  of its client trackers as a reachability proxy - a legitimate signal
  since HA marks every entity on that integration unavailable together
  the moment the router can't be reached). All candidates are listed
  together so you choose between them - e.g. a TP-Link Router acting as
  gateway with a Deco mesh riding behind it - rather than one being
  silently preferred. TP-Link Router discovery also finds its WAN/LAN
  IPv4 sensors and its "Total wired clients" sensor (wired into LAN
  Connected Devices) alongside the router pick itself.
• Nodes discovery: one "Scan for Nodes" button covers TP-Link Deco
  (satellite units, added as Access Points with Backhaul Type/Speed,
  Connected Devices, Download/Upload, and IP Address all wired up
  automatically from each unit's own entities) and Proxmox VE
  (physical/cluster nodes, added as Homelab Server nodes with their
  VMs/Containers attached as badges automatically). Results filter by
  Integration and by Type (Access Point / Switch / Homelab - Switch
  reserved for a future scanner). A Deco master unit is included as a
  selectable candidate here too, in case you'd rather add it as a Node
  than use it as Router, and defaults to Primary AP when added.
• Clients discovery (formerly "Individual Devices" - see Changed):
  TP-Link Deco, TP-Link Router, and OpenWrt (LuCI) client trackers,
  filterable by integration, with Select All and pagination (10 per
  page). Each added Client can be manually named, defaulting to its
  entity's Friendly Name.
• Internet discovery: Speedtest.net, Ookla Speedtest, Cloudflare
  Speed Test, and LibreSpeed for Ping, Jitter, Download, and Upload
  (plus ISP name as a one-time default for the Name Override); Aussie
  Broadband, Starlink, and Start.ca for Billing Total, Billing
  Remaining, Total Downloaded, Total Uploaded, and (Starlink) the
  Internet connectivity entity itself. Every service found on every
  matching integration is listed separately - including multiple
  services on one account, such as several Aussie Broadband lines - so
  you pick the right one instead of one being assumed.
• Security discovery: AdGuard Home's "queries blocked" sensor for
  the DNS Filtering entity, listing every instance found (e.g. a
  primary and a secondary DNS server).
• Homelab discovery: "Scan for Container/VM" on a Homelab node
  combines Portainer (Docker containers) and Proxmox VE (VMs and LXC
  containers) into one list.
• IP Address badges: optional WAN/LAN address badges on the Router
  (WAN centered below Internet, LAN centered below Router) and an
  address badge on every Access Point (above the circle for a
  non-Primary AP and for a Flat-layout Primary AP; below the circle for
  a Tiered-layout Primary AP), plus a WAN badge below Internet when the
  Primary AP itself is acting as the router. Gated by a new "Show IP
  Addressing" toggle (Advanced → Layout, off by default), sized to
  match PoE badges, with independent background/text color pickers on
  both the Router and each Access Point, and a new IP Address Badge
  Opacity slider (0-100%, default 100%, Advanced → Layout).
• Delete All button for Clients, with a confirmation prompt.
• Remove action on every already-applied auto-discovered candidate
  (Router, Security/AdGuard, Clients, Nodes, Container/VM) - replaces
  the old static "Selected"/"Already added" label with an actionable
  button, styled in the secondary text color, that clears exactly the
  fields that candidate contributed and makes it selectable again after
  a confirmation prompt.

Changed

• "Individual Devices" renamed to "Clients" throughout the editor
  and menus. The underlying individual_devices config key is
  unchanged, so existing configs and dashboards keep working.
• Discovery buttons renamed from "Use this" to "Select", for
  consistency across every discovery list.
• Primary AP layout now defaults to Flat (previously Tiered) for
  new cards; existing configs with an explicit layout are unaffected.
• Primary AP's own IP Address badge always renders above its
  circle, regardless of Tiered/Flat layout.
• Primary AP's backhaul is always treated as wired, regardless of
  any Backhaul Type entity - it's the mesh's own connection point to
  the router, not a hop backhauling through another AP.
• "PoE Badge Size" and "PoE Badge Font Size" renamed to "IP/PoE
  Badge Size" and "IP/PoE Font Size", reflecting that IP Address
  badges now share the same sizing.
• The Router-to-Primary-AP connector (Tiered layout) is 50% taller
  whenever Show IP Addressing is on, since the Router's LAN badge and
  the Primary AP's own badge sit at opposite ends of that segment and
  would otherwise overlap.
• IP Address badges are now plain text with no icon (previously
  included a small IP-network icon).
• The card now ships with lit bundled directly into the file,
  rather than importing it from a CDN at runtime.

Fixed

• TP-Link Router and Deco auto-discovery found nothing. The
  original classification relied on the Home Assistant device
  registry's model field containing a hint like "deco" or "router" -
  but real TP-Link Router/Deco devices report bare model numbers (e.g.
  "BE25") with no such hint, so it never matched. Router and Node
  discovery now use attributes confirmed directly against real device
  data instead: TP-Link Deco's device_type ("deco" vs "client") and
  master (true/false) attributes, and, for TP-Link Router, splitting
  purely by entity domain (device_tracker = clients, everything else
  on that device = router candidates, explicitly excluding buttons so a
  reboot button can never be mistaken for a status entity).
• TP-Link Router used the wrong entities. "Total Clients" was being
  used for both the Router status entity and the LAN Connected Devices
  entity; now uses "Connection Type" for Router and "Total wired
  clients" for LAN Connected Devices.
• Backhaul Type/Speed and node metrics used the wrong entity
  suffixes for TP-Link Deco. Entity IDs are slugified from the
  sensor's display name, not its internal key, so the real suffixes are
  backhaul_type / backhaul_speed / _down / _up /
  connected_clients - not connection_type / download_speed /
  upload_speed / client_count as originally assumed. Discovery also
  now skips these sensors when they're disabled (several forks ship
  them disabled by default) and falls back to reading connection_type
  as an attribute directly on the unit's own tracker entity, which is
  always live.
• "Select All" on a filtered Clients or Nodes list added devices from
  every integration, not just the filtered one. Select All now only
  affects entities in the currently visible (filtered) list, and
  preserves selections made under a different filter instead of
  discarding them.
• Wired-connection detection missed some TP-Link Deco setups. Now
  checks both the connection and connection_type attribute names,
  fixing incorrect "wireless" grouping/backhaul display where only the
  latter was present.