AI Assistant
Asking the assistant
The AI Assistant is a chat workspace where you type in plain language and get a written answer grounded in your own ThreatWeaver data. It can also take action for you — but only after you confirm each step. This page covers everything on the screen and the workflows that get real work done.
What it's for
- Ask questions about your environment and read a written answer, instead of building a query or a chart yourself.
- Start a task by describing it — the assistant proposes the action and waits for your go-ahead before anything changes.
- Keep a history of your conversations so you can return to, rename, or export them later.
The screens in it
When you open the assistant you see two areas side by side:
- The conversation panel (right) — a welcome screen and suggestions when you're starting fresh, or your running chat once you begin.
- The history sidebar (left) — your past conversations, with search and a button to start a new one.
The welcome screen
Before you send anything, the panel shows a "What can I help with?" heading and a row of suggestion buttons — one-click starting points. These are personalized to your workspace, so what you see can change over time; Build a dashboard, Schedule a scan, Generate a report, Review tickets, and Critical CVEs are illustrative examples, not a fixed list. Click any suggestion to send it as your first message, or just start typing your own.
Every control explained
The message box
The message box sits at the bottom of the conversation.
| Control | What it does |
|---|---|
| Message field | Type your question or instruction here. Press Enter to send, or Shift + Enter to add a new line. |
| Send button | Sends your message. It's disabled until you've typed something and while the assistant is still working. |
| Character counter | Appears as you approach the length limit for a single message and turns red at the maximum. |
| "BETA" label | A reminder that the assistant is an evolving feature. |
| "Verify important actions before executing" | A standing note to review anything the assistant proposes before you confirm it. |
While the assistant is working, an animated indicator and a short status label (for example Thinking…) show that a reply is on the way.
The conversation
Your messages appear on the right; the assistant's answers appear on the left with an assistant icon. Answers are formatted text — headings, lists, and links render properly. Hover over any assistant answer to reveal a Copy control that copies that answer to your clipboard.
On a long conversation, a Load earlier messages button appears above your messages, inside this panel — it loads older messages within the conversation you currently have open. This is separate from the history sidebar's own Load more, which pages through your list of past conversations (see below).
Each message shows a timestamp. Some user messages also get a small topic tag (for example Dashboard, Scan, Report, Ticket, or Analysis) that reflects what the message is about.
Confirming an action
When you ask the assistant to do something rather than just answer, it doesn't act straight away. Instead it shows an action card headed Action required that describes exactly what it's about to do, with two buttons:
| Control | What it does |
|---|---|
| Confirm | Carries out the proposed action. The card changes to Executing…, then to a green Completed state with a short result message. |
| Cancel | Discards the proposed action — but also clears your entire current conversation, not just that pending step. All messages in the panel disappear and you start fresh. Nothing changes in your ThreatWeaver environment itself. |
| Open | Appears on a completed action when there's somewhere to go — for example the new dashboard or report — and takes you straight there. |
The assistant never carries out an action on its own. Anything that creates or changes something in ThreatWeaver waits behind a Confirm button that describes what will happen first.
Clicking Cancel doesn't just drop the pending action — it wipes every message in the conversation you're viewing. If you want to keep what's been said so far, use Export first (see the header, below).
Answering a follow-up with options
Sometimes the assistant needs a detail before it can continue — a time range, a system type, a source. When it does, it may show a choices card with selectable answers instead of asking you to type:
| Control | What it does |
|---|---|
| Option rows | Click an option to choose it. For a single-choice question your click submits the answer immediately. |
| Select all that apply | For multi-choice questions, tick each option you want, then click Submit. A counter shows how many you've selected. |
| Other… | Opens a small field so you can type a custom answer if none of the options fit. |
| Skip | Available on optional questions — moves on without answering. |
The history sidebar
The sidebar keeps every conversation so you can pick up where you left off.
| Control | What it does |
|---|---|
| New chat | Starts a fresh conversation and clears the panel. |
| Search | Filters your conversations by title. Press Ctrl + K (or Cmd + K) to jump straight to the search box. |
| Conversation list | Click any past conversation to reopen it, with its messages restored — including the outcome of any action you confirmed. |
| Load more | Below the list, loads the next page of older conversations. This is separate from the conversation panel's own Load earlier messages, which loads older messages within the chat you have open (see above). |
| Select conversations | Switches the sidebar into multi-select mode so you can act on several conversations at once. |
Hover a conversation, or open its "..." menu, for per-conversation controls:
| Control | What it does |
|---|---|
| Pin / Unpin | Keeps a conversation pinned to the top of the list. |
| Rename | Renames that conversation directly from the sidebar — separate from renaming the conversation you currently have open, which you do from the header (below). |
| Archive | Removes the conversation from your active list without deleting it. |
| Delete | Permanently removes that one conversation, after you confirm. |
In multi-select mode, use Select all to select every visible conversation, then Archive or Delete to act on all of them together.
At the bottom of the sidebar, Clear all history permanently deletes every conversation in your history — not just the one you're viewing — by removing each one. This cannot be undone, and you'll be asked to confirm before anything is deleted.
The header
The header runs across the top of the conversation panel.
| Control | What it does |
|---|---|
| Conversation title | Click the title to rename the current conversation. Press Enter to save or Esc to cancel. |
| Export | Copies the whole conversation to your clipboard as formatted text you can paste into a document or ticket. |
| "active" badge | Indicates the assistant is ready. |
Resuming an unfinished task
If you leave in the middle of a multi-step task, the welcome screen shows a "You have an active … session" banner the next time you return:
| Control | What it does |
|---|---|
| Continue | Picks the task back up where you left off. |
| Dismiss | Clears the unfinished task so you can start fresh. |
Safety and limits, in plain terms
- Answers are AI-generated. They're grounded in your data, but they can be incomplete or wrong — verify anything important before you rely on it.
- Actions always need your confirmation. The assistant proposes; you decide.
- The assistant works through an AI provider your organization sets up. If no provider is configured yet, or if a proposed provider isn't approved by your organization's policy, the assistant will say so and point you to an administrator. See Admin.
- There are usage limits. If your organization reaches its daily AI limit, or if too many requests come in at once, the assistant will tell you to try again later rather than fail silently.
- If the assistant is briefly unavailable, you'll see a short message with a Dismiss button; try again in a moment.
Common workflows
Workflow: ask a question and act on the answer
-
On the welcome screen, click a suggestion — for example Critical CVEs — or type your own question and press Enter.
Step 1 — sending a question from the message box -
Read the assistant's answer. Ask a follow-up in the same box to refine it — the assistant remembers the conversation so far.
Step 2 — the assistant's written answer, with a follow-up question -
When you're ready to act, ask for it — for example "Build a dashboard for these." The assistant shows an Action required card describing what it will do. Click Confirm.
Step 3 — the action card with Confirm and Cancel -
The card turns to Completed. Click Open to go straight to the result.
Step 4 — the completed action with an Open button
Workflow: answer a follow-up with options
- Ask for something that needs a detail — for example "Generate a report."
- When the assistant shows a choices card, click an option (or tick several and click Submit). Use Other… to type your own answer if needed.
- The assistant continues with your answer and proposes the next step.
Workflow: revisit and rename a past conversation
- In the sidebar, use Search (or Ctrl + K / Cmd + K) to find an earlier conversation, then click it to reopen it.
- Click the conversation title in the header, type a clearer name, and press Enter.
- Click Export to copy the conversation as text for a report or ticket.