Describe Your Nodes
Every node on the canvas, every record in Environment and the workspace itself can carry two things about it: a Description, one short line that says what it is for, and AI notes, a list of short notes, one thought each. People read them, and every AI that works in the workspace reads them first.
What they are for
A node's name says what it is. Its description says what it is for, and its notes say what an AI should know before it works with it:
- Description: "Barcode scanners on packing line 3; one message per scan."
- AI note: "Line 3 stops for cleaning 2-3am, so gaps then are normal."
- AI note: "Device 14 reads a duplicate on restart."
So when someone asks "why is there a gap at 2am?", the AI already knows, instead of guessing from the node's name. Descriptions and notes guide AIs. They never grant a right: nothing in OpenX decides who may do what from them.
Write them on a node
- Select the node on the canvas. The Inspector shows Description and AI notes at the top.
- Type the description. The count under the box shows how much of its one short line you have used.
- Press Add note for each note and type it. Each note is added, changed or removed by itself with the bin button beside it. A note left empty is never kept.
- Save the workspace as usual.
Clearing the description, or removing the last note, leaves the node with none. A warm query needs a description, so its box is outlined red while it is empty.
A data store's containers have the same two boxes: open the database's Manage containers, expand a container, and save them with the container's other settings.
The workspace's own notes
The workspace has a description and instructions of its own in Workspace > Workspace Settings. Its AI notes sit under the description there: short things every AI should know about the whole workspace, such as "Stores close at 9pm, so no rows after that are normal." Add, change or remove them, then press Save Changes.
Write them in an Environment window
Every record under Environment has the same two boxes, in its own window:
- Provenance: in the form, saved with Save configuration.
- IoT: press Edit settings; they are saved with Save & Provision.
- Identity providers: open a directory and press Edit these details; they are saved with the provider.
- Communications: press Edit settings on an account; they are saved with Save account. A Microsoft 365 account has no such button, so its boxes have a Save of their own.
- Landing zone: once it is provisioned, the window has a form with the two boxes and a Save of its own.
- Data residency sits in the same landing zone form: the regions it may use, and whether a resource outside them is reported or refused. It is not a note: only a person sets it, it saves with Save and update Azure, and an AI cannot propose it. See Manage Private CALZ.
- Cloud connections, DNS zones and Source control: press the small edit icon on the connected cloud, the domain's tab or the connection, type, then press Save.
A Save of their own writes only these two fields. It never renews a connection, provisions or deploys anything.
When an AI proposes one
An AI can suggest a description or a note when it learns what a node, a record or the workspace is for. It never writes one itself: its suggestion is a proposal, and it shows in Proposals like any other change, named Describe and the node's name, with each field now and as proposed. Each note shows by itself, as added, changed or removed, so you read exactly what accepting does. Accept it and the node's Inspector shows it; reject it and nothing changes.
This works for every node, every Environment record (a cloud connection, the landing zone, IoT, DNS zones and the rest) and the workspace itself.
What every AI reads
- Thinky carries the whole workspace, so it reads every description and note as it works.
- An outside AI connected over MCP reads them with
get_workspace, and on each record's row ofretrieve_inventory. A record with none carries none. - Each kind's guide tells an AI what the two fields are for and that it may propose them.
Related
- Proposals: how a proposed change is shown, accepted and rejected.
- Working with Thinky: asking Thinky about your workspace.