Use the Ask AI drawer

DeltaVault’s Ask AI drawer is where you talk to the AI directly. This guide covers opening the drawer, attaching context to a message, finding a past chat, session titles, and the working indicator you see while a turn runs.

Open the drawer

Open Ask AI and the drawer opens to a composer-first home: a greeting, a message box, and a set of suggested skills you can run. Type a message into the box and send it to start a chat.

A small blue dot on one of the suggested skills means your organization added it or customized it: hover the dot and it reads Added by your organization. A skill with no dot is one of DeltaVault’s built-in skills, unchanged by your organization.

Starting a chat leaves the home view behind: the greeting and suggested skills step aside for your conversation. The plus button in the drawer header returns you to the home view: it carries no visible label, but its name, New chat, shows on hover.

The docked, resizable pane

Ask AI docks to the right of the page instead of covering it. The page content, including a canvas, reflows to sit beside the pane rather than dimming behind it, so you keep working while the pane is open. Drag the pane’s left edge to widen it (double-click the edge to reset to the default width); the width is remembered in this browser. The pane is non-modal: clicking the page behind it does not close it, and pressing Escape does not either. Close it from the X in the pane header when you are done.

Attach context to a message

Next to the message box, the + control lets you attach context before you send. You can attach:

• Current selection. A snapshot of whatever is selected on the page at that moment, for example a table, a business entity, or a workspace.
• Knowledge document. Pick one from a picker of your organization’s or project’s knowledge documents.

Each attachment shows as a chip above the message box. You can remove a chip any time before you send; once you send the message, what you attached travels into that chat turn as context for the assistant.

Find a past chat

Select the history control in the drawer header to open a history view of your recent chats. Chats are grouped by Today, Yesterday, This week, and Older, and a search box lets you find one by name. An attention dot on the control appears when one of your chats or runs is still working or waiting on you.

Open any chat from the history view to resume it in place, right where you left off.

The history view also lists your recent skill runs, not just chats. Open a run that is still working to watch its progress, or a finished one to reread its result, the same way you reopen a chat.

A count badge on the Ask AI button shows how many of your runs and chats are still in flight, so you can tell at a glance when something is still working.

The history view is for finding and resuming your recent activity. To see where your AI credits are going, open See your AI usage.

Session titles

Each chat is named automatically from its first exchange, so your history reads as a list of topics rather than timestamps. A chat you started before this feature shipped reads Untitled chat.

Watch a turn run

While a turn is running, a line above the message box tells you what the assistant is doing right then: Thinking… while it reasons, or Running followed by the name of the tool it is currently using, ending in an ellipsis.

When the assistant uses one or more tools during a turn, those steps collapse into a summary reading Used N actions once the turn finishes. Select the summary to expand it and see each individual step.

Watch a skill reason and propose

When you run one of the suggested skills, the panel switches to a live view of the run instead of a bare “Streaming” message. You see the run’s progress as it happens, and where the skill allows it, a readable summary of the model’s reasoning as it forms.

The suggestion skills that propose business domains, business entities, relationships, or entity enrichment reason about your catalog before they propose. While the run works, the panel shows that reasoning as a live, expanding summary, so you can follow what the model is weighing. Once the proposal is ready, the reasoning collapses into a Show AI reasoning disclosure you can open at any time to reread it.