# Definitions

The VIA definition files built into the binary. `embed.go` embeds every JSON file
in this directory, so that `go install` delivers them: it copies a binary and
creates no data directory, and a board VIA does not carry cannot be fetched.

This is not the directory the tool reads at runtime. That is the per-user
`definitions/` directory the top-level README names; `keyboard fetch` writes
there, and a file placed there — or passed with `--definition` — takes precedence
over the one built in. To use or correct a definition, put it there, not here.

`keyboard fetch` does not replace a file that is already in the per-user
directory, because that file may be one you have edited and nothing on disk says
which it is. It refuses and names the file; `--force` overwrites. A board's file
is recognised by the `vendorId`/`productId` it carries, so the check follows a
renamed file too, and `--force` removes the file it replaced rather than leaving
two files for one board.

## What belongs here

A board's file, only when VIA's own collection does not carry that board, kept as
the vendor serves it. A file is matched to a board by the `vendorId`/`productId`
it carries, so only the `.json` extension matters, not the name — but it must be
that extension in lowercase, because both this directory's embed pattern and the
loader that reads it skip anything else. A file added here takes effect after a
rebuild.

## The files here

Every file here comes from the vendor's own [product wiki](https://wiki.wobkey.com/en/Products),
which publishes a VIA JSON per model and per connection mode. The names here are
the names the vendor serves, so a file can be compared against what the page
offers; only the extension was changed, because this directory's embed pattern
and the loader both match `.json` and nothing else.

| File | Board | Mode | From |
|------|-------|------|------|
| `impact80.json` | Impact 80 | — | [Driver & Firmware](https://wiki.wobkey.com/en/Products/PMOKEY-Impact-80/Driver-Firmware) |
| `Aurora80-RGB.json` | PMOKEY Aurora 80 | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/PMOKEY-Aurora-80/Driver-Firmware) |
| `WAVE75wiredonlyRGB.json` | PMOKEY Wave 75 | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/PMOKEY-Wave-75/Driver-Firmware) |
| `ZEN65-RGB.json` | ZEN 65 | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/ZEN-65/Driver-Firmware) |
| `pmo_p75_gaming_keyboard.json` | PMOKEY P75 | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/PMOKEY-P75/Driver-Firmware) |
| `pmo_p75_gaming_keyboard_2p4g.json` | PMOKEY P75 | 2.4G | the same page |
| `Rainy-75-RGB-USB-3.0.json` | Rainy 75 RGB | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/Rainy-75/Driver-Firmware) |
| `Rainy-75-RGB-2.4G-3.0.json` | Rainy 75 RGB | 2.4G | the same page |
| `Crush 80 RGB-USB.json` | Crush 80 RGB | wired | [Driver & Firmware](https://wiki.wobkey.com/en/Products/Crush-80/Driver-Firmware) |
| `Crush 80 RGB-2.4G.json` | Crush 80 RGB | 2.4G | the same page |

Those pages carry more files than are here: the Rainy 75 page also has Non-RGB
(Lite), DE/UK-ISO and JIS, and the Crush 80 page a Non-RGB pair. A board in one
of those modes has no definition at all — lookup matches `vendorId` *and*
`productId`, so no file of another mode stands in for it. Such a board is still
driven, through raw effect IDs, and `keyboard info` reports `"effects": false`
for it.

## Crush 80 and Rainy 75 share their USB IDs

The vendor issues the same pair of IDs to two different boards:

| | wired | 2.4G |
|---|---|---|
| Rainy 75 RGB | `0x320F/0x5055` | `0x320F/0x5088` |
| Crush 80 RGB | `0x320F/0x5055` | `0x320F/0x5088` |

Nothing in the USB descriptor tells them apart, so a definition lookup cannot
either. Both files for a pair are kept because the vendor publishes both, and
`FindDefinition` takes the first match in the order `LoadDefinitionsFS` returns,
which is sorted by path — `Crush 80 RGB-USB.json` sorts before
`Rainy-75-RGB-USB-3.0.json`, so a Rainy 75 is reported as a Crush 80. The two
catalogs happen to name the same 19 effects at the same IDs, so the effect names
are right either way and only the reported model name is wrong. Renaming one file
is all it takes to change which board wins.