Analog Way

Everything here speaks AWJ — the JSON-over-WebSocket interface the device's own web application uses, on TCP 10606. It is a leaf-oriented protocol: you can read and write a property if you already know its path, and you cannot enumerate anything, because a container answers with an empty object. That normally means a control application is limited to whatever paths somebody managed to observe.

The way past it is the vendor's own bundle. The Web RCS ships as an unminified webpack build that keeps its original TypeScript module paths and its generator's headers, so the entire object model — every collection, every property with its type, range, default, read-only flag and enum members — is recoverable from the file the switcher already serves you. Take the model from there, then read the leaves over AWJ. That is why the parameter catalogue in the control surface has real limits rather than guessed ones: 67 layer parameters and 19 screen controls, generated from the switcher rather than typed from a PDF.

It also explains a trap that runs through all four. Subscribing is a write — the Subscriptions list is a `replace` on a path like anything else — so a tool that means to watch a live frame and not touch it cannot get there by declining to call `set()`. The read-only paths here are built with no reachable write path at all, and poll instead.

More of this than is usual, and it corrected things rather than confirming them. The VPU map has read a real Aquilon C end to end in two configurations, and the second corrected two things the first had made look settled. The control surface's whole chain — a fader movement through to an opacity write, a TAKE, and the preset flip that follows it — has been exercised on a real Aquilon C, with the frame captured first and all 87 values verified restored afterwards. The preset-letter rule was settled by firing a take and re-reading rather than by reasoning about it.

What has not happened is just as specific. No physical control surface has ever been plugged into the surface host — the four controller maps come from published documentation, and a MIDI implementation chart is exactly the kind of document that is quietly wrong, which is why there is a bring-up mode. The VPU map has never seen a Link setup, a capacity other than 4K or 5K, a combined VPU, Optimized mode or Cut & Fill; its repository says which recorded capture would settle each one, and the guide for recording it assumes nothing.

There is no Aquilon here any more, so from here on the recorded captures are the ground truth. A capture from your frame is the single most useful thing anyone can send.

LivePremier Plus has no hosted demo and cannot have one. The interface you would be looking at belongs to Analog Way and is served by your switcher: the proxy carries it and contains no part of it, so without a device there is genuinely nothing to show. Run it from the container, a desktop build or a checkout on the same network as the frame.

The demo that does exist — Mynah's hosted site — is a demonstration of the command syntax and nothing more. A page served over HTTPS cannot open a connection to a switcher on your network, so no hosted copy of any of this will ever drive hardware, however much it looks like it might.

Analog Way, LivePremier, Aquilon, Midra, LiveCore, NeXtage, Ascender and Pulse2 are their vendor's names, and none of this is endorsed by, supported by or connected to them. Product names are used only to describe what the software talks to.

The protocol work was done against device behaviour and against files the device itself serves, not against a specification anybody published. Treat it as accurate for the frames and firmware it has been tried on and no further — each project's own page says exactly which those are.