AI Assistant Settings

Everything on the Settings → AI Assistant tab — which model the assistant panels use, how they open and look, and how much of the conversation they remember.

Overview

The AI Assistant tab controls the embedded assistant panels that appear on screens across TotalApp. All of its settings are global: a change applies instantly to every panel, and there is no per-screen override. Open Settings → AI Assistant to find them.

Looking for how the panel works?

This page only describes the settings. For opening the panel, selected rows, supported screens and saving reports, see AI Assistant Panel.

Settings at a glance

Setting	Options	Default
AI Assistant model	Default, or a specific hosted model	Default
Panel Width	Narrow (288 px), Default (320 px), Wide (384 px)	Default (320 px)
Auto-open on screen visit	On / Off	Off
Typing indicator	On / Off	On
Show suggestion chips	On / Off	On
Auto-expand with Knowledge Graph	On / Off	On
Chat font size	Small, Medium	Small
Accent color source	Match App Mode, Custom	Match App Mode
Conversation memory (turns)	1 – 30	10

Model

AI Assistant model decides which model answers inside the panels. Leave it on Default and the assistants follow the AI engine you chose in Settings → Agentic. Pick a specific model from the list to use it for assistants only.

The list contains the hosted models from your providers (Modelion and OpenRouter). A model your account cannot use is shown but disabled.
The choice applies only when the AI engine is Cloud LLM (Hosted API). With any other engine the panel shows a note and keeps using that engine.
Test connection checks that the hosted gateway is reachable and shows the result next to the button.

Panel behaviour

Panel Width — Narrow (288 px), Default (320 px) or Wide (384 px). The panel is a fixed overlay on the right edge; while it is open, the screen reserves matching space so nothing is hidden behind it.
Auto-open on screen visit — opens the panel automatically when you reach a screen that has an assistant. Off by default; you then open it with the tab on the right edge.
Typing indicator — shows an animated spinner while the AI is generating a response. Turn it off and the panel simply waits until the answer arrives.
Show suggestion chips — on an empty chat, shows example question chips; clicking one sends it immediately. Turn it off for a clean empty state.
Auto-expand with Knowledge Graph — when a panel lets you pick a single My Knowledge source, up to 3 strongly connected sources from your Knowledge Graph are added automatically. They appear as a separate, removable list and are never silently merged.

Chat appearance

Chat font size — Small (11 px, the default) fits more text, especially at the Narrow width; Medium (12 px) is easier to read in long conversations.
Accent color source — Match App Mode (default) tints each panel with the colour of the module you are in. Choose Custom to use one fixed colour everywhere and pick it from the colour swatches that appear.

Context & memory

Conversation memory (turns) — how many past messages are sent as context with each new message. Drag the slider between 1 and 30 (default 10). Higher values give coherent multi-turn conversations but make requests slower and use more of your quota; lower values suit one-off analytical questions.
Live data injection — an information note: before each message the assistant can fetch live data from a connected service (for example Shopify products), so answers reflect the current state.

Frequently Asked Questions

Are AI Assistant settings per-screen or global?

Global. Changing any setting on this tab affects every assistant panel across TotalApp immediately. There is no per-screen override.

Does a higher conversation memory cost more or slow down responses?

Yes. The memory value controls how many previous messages travel with every request. A higher value (up to 30) gives the AI more conversational memory but increases the size of each request, which can slow responses and consume more of your usage quota. For mostly one-off questions, keep it low.

I picked an AI Assistant model, but nothing changed.

The model setting applies only when the AI engine is Cloud LLM (Hosted API). With Ollama or an in-browser engine the assistants keep using that engine.

Where do I start a new chat or reopen an old one?

In the assistant panel itself, not in Settings: use New Chat and the History list described in AI Assistant Panel.