Smart Chat settings

Deprecated page

This page is retained temporarily and may contain outdated details. Use the current Smart Chat local and API model documentation for current behavior. Start with Getting started with Smart Chat API Extension for the shortest verified workflow.

Smart Chat Pro settings control thread instructions, send behavior, Lookup context, and the configured local or cloud model used for future responses.

Open settings

Open settings from the Smart Chat Pro row in the Smart Plugins Store or through the current Smart Environment settings surface.

A model marked Current is selected. It is not proof that credentials, account access, or a local server can answer.

Custom instructions

Choose Edit custom instructions to manage the default instructions used by new API Chat threads.

Keep stable rules short and testable, for example:

Use only attached context for source-backed claims.
Name the source notes used.
State what evidence is missing instead of guessing.

Thread-specific instructions can override the default for that thread.

Send behavior

The default send shortcut is Shift + Enter. Change it in Chat Settings when the current shortcut conflicts with multiline writing.

Streaming changes how partial output is displayed. It does not change the source set, model, or trust status of the completed response.

Lookup context

Lookup context settings control the candidate source count and result type used when Smart Chat retrieves likely notes or blocks.

Lower limits reduce prompt size and noise.
Higher limits increase recall and review cost.
Sources return whole notes.
Blocks return smaller sections when block preparation is available.

Lookup proposes candidates. Review and remove weak, stale, duplicated, or untrusted sources before sending.

Configure a chat model

Under Chat models, choose + New.
Select an enabled provider, such as a supported cloud provider or Ollama for a local model.
Enter the required API key or local host.
Choose Refresh Models when the provider needs to load its available model list.
Select Chat Model.
Save or close the editor.
If several models exist, choose the intended row under Default chat model.
Confirm it is marked Current.
Run Test or Test model.

Do not send note content when the test returns an authorization error such as 401.

Common failures

State	Smallest recovery
`MISSING MODEL`	Create or finish one model row, select it under Default chat model, confirm Current, and run Test.
`MISSING PROVIDER`	Repair the provider, credential, account access, or local server, then run Test again.
Send unavailable	Enter a prompt and confirm one usable model/provider pair.
Lookup context unavailable	Enter a sufficiently specific question and confirm Smart Environment retrieval readiness.
Generic answer	Improve the question and reviewed source set before adding more context.

Data boundary

A local model can run on the device when its local server is available.
A cloud chat provider receives the prompt and context sent with the request.
Smart Chat does not attach the active note or whole vault automatically.
Provider codeblocks and Pro API Chat have different persistence contracts; review Smart Chat before choosing a thread path.