what you can do

EEG Lab

EEG Lab — what you can do

/admin/d/eeg · pillar devices

Read a Neurosity crown, train binary intent grids, and decode a trained intention into a trigger that runs anything.

Fetch this, never remember it

These change per install and per release. Call for them - a list written here would be a future lie.

What	Call	Why
whether a crown is connected, and whether it is simulated	`GET /api/eeg/status`	ALWAYS READ THIS FIRST. `simulated: true` means the bands are generated, not measured — anything trained from them is trained on a sine wave. `configured: false` means no credential exists yet, which is the ordinary first-run state and not an error.
the grids and their quadrants	`GET /api/eeg/grids`	each quadrant is one binary axis. `settings.scoreThreshold`, `minMatches` and `cooldownMs` are what decide whether a decode runs anything — they are the answer to "it recognises me but nothing happens".
what a quadrant has been trained on	`GET /api/eeg/grids/:gridId/samples`	a quadrant with samples on only one pole will match almost anything. The counts per labelRole are the honest measure of how much training exists.
the trained models	`GET /api/eeg/grids/:gridId/models`	keyed by quadrant. A quadrant absent from this map has never been trained, which reads on screen as "resting" and is a setup problem rather than a signal.
what a thought is bound to run	`GET /api/eeg/bindings`	each is "when I think X, do Y", stored as an ordinary rule. `allowSimulated: true` on any of them means a SIMULATED crown can fire it, which is fine for a demo and wrong for a door.
bind a thought to a function	`POST /api/eeg/bindings`	{intent, run, props} is the whole body. The intent must be spelled as it was TRAINED — a typo compiles to a condition that can never be true, so the rule arms green and never fires. Read the vocabulary first.
the poles this grid can actually produce	`GET /api/eeg/grids/:gridId/vocabulary`	the closed set a binding may name, each marked with whether its axis is trained. Bind against this rather than against what the operator called it in conversation.
turn a decode into a sentence, or into a chosen action	`POST /api/eeg/grids/:gridId/interpret`	deterministic by default; `useModel` adds a language model and `choose` makes it pick an action id. It NEVER runs what it picks. `usedModel: false` in the reply means you are reading the deterministic phrase — check `note` for why.
one decode of the newest frame	`GET /api/eeg/grids/:gridId/decode`	returns every quadrant's score, the command they compose, and whether it would fire. 409 means no frame has arrived — connect a crown first.
the live frame feed	`GET /api/eeg/stream`	SSE, about four frames a second, each carrying band power, the vendor metrics and a raw preview. It holds a crown session open: the server closes the session when the last watcher leaves, so do not hold this open to poll.

Tabs

Every one is a real URL. A tab with its own section below carries a real job.

Tab	Path	Group	What it is for
`crown`	-	The headset	Connect a Neurosity crown, watch signal quality, and see the live bands.
`signal`	-	The headset	Per-electrode contact and the raw trace — where to look when nothing decodes.
`grids`	-	The lab	The binary intent grids: each quadrant is one axis, drawn as a 26x26 sheet.
`training`	-	The lab	Hold an intention, capture it as a pole, and train the quadrant on what you captured.
`models`	-	The lab	What each quadrant has learned, how much it was trained on, and how to retrain it.
`bind`	-	The output	When I think YES, do this. One pole, one action, no language model in the way.
`listen`	-	The output	Live decoding: the moving cell, the command it spells, and whether it would fire.
`apps`	-	The output	Watch a metric and fire a trigger. No code is written and none is run.

API

Mounted prefixes that serve this domain: /api/eeg

The full route table is leumas-capabilities/reference/api-routes.md.