v0.1.16 (Early Access)

New: the prospect decision card

Deciding whether to sign a prospect now opens a rebuilt two-column card. Who they are and what they pay sit at the top, what they want is on the left, and whether you can serve it is on the right.

• Every service a prospect wants is spelled out. A database, object storage, functions or Kubernetes customer used to show only a bare tag with no numbers. The card now lists every service they need with its real demand, VMs and otherwise, so you know what you are taking on before you sign.
• Hidden requirements are shown up front. Some customers need more than a site certification. They may need a working SIEM for audit logging, an air-gapped facility, more than one availability zone, or an isolated segment for their tier. These used to turn you away at sign-time with no warning. The card now lists them, split into hard requirements you must clear and softer recommendations.
• Isolate a customer as you accept them. The card can place a new customer on a network segment at the moment you sign. Keep the default (the shared network, or a tier rule that already covers them), add them to an existing segment, or create a dedicated one, optionally reserving hardware for it. Choosing an existing segment uses a searchable list, so it stays usable when you have many.
• Capacity reflects where the customer will actually run. When you isolate a customer onto dedicated hardware, the "can you serve it" read measures that segment's reserved hosts instead of the whole fleet, and updates as you change the isolation choice. You can see whether the dedicated hardware has room before you commit.

Network map

• The network map reads clearly at any size. The map used to shrink the whole graph to fit the panel, so on a larger network the labels turned to an unreadable blur. It now keeps text at a legible size and spreads the graph out to use the space, grouping your gear by region and availability zone so a multi-zone network fans out top to bottom instead of collapsing into a thin strip. When the map is bigger than the panel you can drag to pan around it.
• One label per link, not a pile. Where several matching cables run between the same two groups of gear (a redundant pair of uplinks, for example), the map used to stamp a full speed label on every one, so they stacked into an unreadable clump. Those now share a single "x N" label that reads their combined traffic, and the live flow still animates along each individual cable.
• New: see the map as your actual racks. Turn on the Racks view and the map redraws by where your equipment physically lives. Each rack becomes a labelled cabinet with its devices stacked in rack order, the carrier feeds sit inside the rack they terminate in, and the cables route between the cabinets, so you can read "what is plugged into what, and where it is" at a glance. Turn it off to return to the network-flow view.
• Racks show the name you gave them. The map now labels each rack with your own name for it instead of a generic "rack 5".
• Speed labels point to their cable. On a busy map it was hard to tell which speed reading belonged to which cable. Each label now draws a short line back to its cable, colour-matched to the link speed, and sits in the open space near the cable instead of crowding the gear.

Fixes

• Breakout cables no longer scramble their route. When you ran a QSFP breakout leg (a QSFP port cabled to an SFP port) and threaded it through cable managers, wall hooks or a rack roof anchor, the committed cable could order its waypoints wrong, crossing over itself into an "X" and looking like it skipped an anchor, even though the preview was correct. The route is now stored in a fixed direction, so the finished cable matches the preview no matter which end you started from. Re-lay any breakout cable that already looks crossed to clean it up.
• QSFP DAC cables are slimmer. The direct-attach copper cables that fan out of QSFP ports were drawn far too thick, so bundles splayed wide and sat badly in cable managers. Every QSFP DAC speed now uses one consistent, slimmer jacket that packs neatly into a manager rail.
• QSFP DAC connectors show their boot. On rack-mounted switches the strain-relief boot, where the cable meets the QSFP module, was not drawn, so the cable looked disconnected from the module. The full connector now renders on every rack device, matching what you see when you inspect a rack up close.
• Device front ports can no longer be reached from behind. On switches and other rack devices, standing on the rear power side let you interact with the front-face network ports straight through the chassis. Front ports are now reachable only from the front, and power inlets only from the rear.
• Renaming a rack renames only that rack. Editing a rack's name used to leave every rack showing the last name you typed, and the change never actually saved. Rack names now save correctly, apply only to the rack you edited, and commit when you press Enter as well as when you click Rename.
• Twinax cables stop reporting a rack as full when it isn't. Threading direct-attach copper cables through the same cable-management finger near the top of a rack could hit a false "24 cables, full" wall while the rest of the rack sat empty. Those cables now spill into the rack's full cable run instead of erroring.
• A customer's Functions stay on their dedicated hardware. If you isolated a customer onto dedicated hosts, their Functions were still landing on shared machines. Functions now respect isolation the same way VMs and every other service do, so an isolated customer's Functions run only on the hardware you reserved.
• The shop updates when you relocate. After moving to a new site, the build catalog kept showing the old site's products until you saved and reloaded. It now refreshes as soon as you arrive.
• A firewall's LAN port cannot be cabled to the gateway. The gateway is the internet-facing side, so running a firewall's LAN port into it was never valid, but the game accepted the cable and then would not let you remove it. That link is now rejected when you try to lay it, and every cable-rejection message reads in plain language that tells you what to do instead, rather than internal port names.
• Device panels show CPU that is actually in use. The in-world device panel showed reserved CPU rather than live load, so a lightly loaded host could look busier than it was. It now shows real CPU utilisation. Memory and storage still show what is reserved.
• The Accept button reads correctly. It only says "Accept anyway" when the fit is genuinely oversubscribed or over capacity, and it now follows your isolation choice: route a customer onto a dedicated segment that is too small and it arms the warning against that segment's hardware instead of the whole fleet. A comfortable fit, or a customer with no VM load, reads as a plain "Accept".
• The prospect card closes with the panel. Leaving the ops panel, or switching to another page, with a prospect card open no longer left it stuck open when you returned.

Notes

• No save or network protocol version change. Existing saves load normally.