NodemancyDocsMapLibraryCards
Troubleshooting

Generation says it isn't configured

You are almost certainly in the metered lane. The free one is one row further down the same menu, and it is the one most accounts should be using.

First: check which lane you wanted

The Content section has two boxes, and they are grouped by cost rather than by what they do:

metered · costs units · needs a paid plan ✦ Generate the app calls a model for you ⟲ Weave the app finds connections for you free · always available ⇥ Import paste in what any AI wrote for you ↓ Lens skill the file you give that AI first ↓ Groups & relations

If what you wanted was “get an AI to write nodes for my lens”, that is Import plus Lens skill — the free pair. You bring a model you already pay for, the app never makes a call, and nothing is metered because nothing is spent.

This is not a consolation prize. The free lane gets your lens's full context into the model — its groups, its relations, the exact names of every node already there — which is what actually determines whether the output links up. Bring your own AI.

If you did mean the metered lane

Two things have to be true before a hosted call happens, and they are checked in this order:

  • Hosted AI is switched on for your account. It is off by default and nothing turns it on for you. Account menu → AI & Privacy.
  • Your plan includes units. The free plan has none — so metered here means unavailable rather than cheaper.

Matching the message

What it saysWhat it means
“Hosted generation isn't configured on this deployment”Nothing to do with your account. This deployment has no model key at all, so the metered lane is switched off for everybody on it. The free lane still works.
“Hosted AI is off for this account”The switch, and the one you can act on. AI & Privacy → turn it on, and choose a scope while you are there.
Something about your allowance or limitYou are out of units. Note that this message comes after the consent check, so if you are seeing it, hosted AI is already on — you are not one toggle away.
“temporarily unavailable. Nothing was charged”A ceiling above your account, not your account. Try later; nothing was spent.
“Sign in to generate”Your session expired. Sign in again.
When a call fails after it starts — the model returns nothing usable, or errors — the message says nothing was charged, and that is accurate rather than reassuring-sounding. Your allowance is only reduced by work that produced something.

Turning hosted AI on

Account menu → AI & Privacy. As well as the switch you pick a scope, which decides how much of your graph a model may read: session only, which is the default and sends nothing but what you type, or graph-aware, which lets it use your existing entries as context.

Whichever you choose, nodes you have marked restricted are never sent. That floor is in the code path every model call passes through, not in a settings row. What leaves this app has the detail.

Related