Settings > Models decides which of the models your DataMind server offers appear in the chat model picker. The page lists the models, grouped by vendor, and gives each one a switch: turn it off to take it out of the pickers. Nothing is hidden to begin with.
<name> in chat, so switching it off hides that model from the pickers.Unsaved changes: <count>, and reads No changes while nothing is staged.Saved.To change your mind before saving, choose Cancel: it restores the saved choice and drops your edits.
<vendor> models. Its header counts what is shown (3 of 5 shown), and while only some of the vendor's models are on, the switch shows a mixed state.One DataMind model always stays visible.1M or 200K; DataMind's own rows never show one, by design.Type in Search models. The search matches a model's name, id, vendor and description, and vendors with no match stay hidden while you search. When nothing matches, the page says No model matches your search.
Choose Refresh to read the current list from the provider. While it runs, the button reads Refreshing…, and Apply waits for it to finish.
Someone else changed these settings while this card was open.Some deployments store the settings document read-only. There the page shows The settings document is read-only in this deployment. Reading and searching still work, but the switches, Refresh and Apply are disabled.
Loading the provider directory failed with the reason, and offers Retry.No model source is configured for this app.No models are available from this provider right now.Not offered right now, so you can always switch it back on.DataMind Oda reaches its models through providers. A provider is one route: an address, a key and a set of limits. Those are set up with your app, not on this page; this page decides only which models the pickers offer. A provider change reaches the next request without a restart.
For the provider side, see Set up models and providers and Provider route settings reference. Other problems you meet in Oda are listed in Updates and troubleshooting.
Open Settings > Models, turn off the switch beside the model, and choose Apply. The model disappears from the pickers, while a chat that is already using it keeps working.
One DataMind model always stays visible, so its switch is locked and reads One DataMind model always stays visible. Turn another DataMind model on first if you want to swap which one stays.
It turns every model of that vendor on or off at once. Its header counts what is shown (3 of 5 shown), and while only some models are on, the switch shows a mixed state. On the DataMind card, switching everything off keeps its first model on.
Not offered right now?The model is hidden and the catalogue no longer reports it. It stays listed under its vendor so you can restore it, but the provider does not offer it right now.
The catalogue marks it as a recent addition. Nothing to do: it is already offered in the pickers unless you hide it.
The save did not go through, so Oda left your changes in place and shows the reason. A refusal over a concurrent change reads Someone else changed these settings while this card was open. Close the card and reopen it to continue.
The deployment stores the settings document read-only; the page says The settings document is read-only in this deployment. You can still read the lists and search them.
Two cases. No model source is configured for this app. means nothing feeds the page, so there is nothing to manage. No models are available from this provider right now. means the source answered with no models. If instead the page says Loading the provider directory failed, choose Retry.
Choose Refresh. The page reads the provider's current list, so the model appears without an app update.
That is the model's context window, shown for other vendors' models. DataMind's own rows never show one, by design.