Using the AI Assistant

The AI Assistant is a built-in conversational helper that can answer questions about your project and perform actions directly — such as creating geometry, assigning materials, configuring physics, running simulations, and analyzing results.

Prerequisites

The AI Assistant requires:

Third-party data sharing enabled for your organization. An organization administrator can enable this under Organization settings. The assistant uses third-party large language model services to process questions and data. See the FAQ for instructions.
AI Assistant scope assigned to the user account.

If either requirement is not met, the assistant button is hidden or the panel shows a message explaining how to enable it.

Opening the assistant

The AI Assistant button is located in the top bar of the project view, on the right side. Select it to open a resizable side panel.

The panel can be resized by dragging its left edge (200–600 px). The panel state and width persist across page loads within the same browser session.

When the panel is closed and a response is still generating, the button shows a dot indicator.

Chat interface

Type a message in the input field at the bottom of the panel and press Enter to send, or select Send message. Press Shift+Enter to insert a new line.

While the assistant is generating a response, a Stop generating button appears. Selecting it stops the current response.

Responses are streamed in real time. Code blocks include syntax highlighting and a copy button on hover.

Starting a conversation

When the panel opens with no active session, a welcome message reads: “How can I help you today?” with the subtitle “Ask me anything about your simulations, analysis, or design.”

Effort level

Before sending the first message, select an effort level from the dropdown at the bottom of the panel. Three levels are available:

Level	Use when
Low	Quick questions, simple lookups
Medium	General simulation setup, multi-step tasks
High	Complex analysis, geometry creation, tasks requiring deep reasoning

The effort level is locked after the first message in a session. To change it, start a new session.

Attaching content

The assistant accepts several types of attachments via next to the input field:

Screenshot — Attach a screenshot of the current model view. When enabled, a screenshot is included automatically with each message. Toggle this on or off from the menu.
Image — Upload an image file or paste an image from the clipboard (Ctrl+V / Cmd+V ).
File — Upload a file (PDF, CSV, text, Office documents, and other common formats).

Files can also be dragged and dropped directly onto the assistant panel. Attached items appear as removable chips above the input field.

Voice input

Hold Ctrl+Space to record a voice message. Release to send. The recording is transcribed to text automatically.

What the assistant can do

The assistant has access to project-aware tools organized by simulation domain. It can read your project state and, with your approval, make changes.

Domain	Example actions
Geometry	Create, modify, and inspect geometry elements; view the entity tree
Materials	Assign materials to regions; edit material properties
Physics	Add physics definitions; configure interactions and boundary conditions
Mesh	Create meshes; adjust mesh settings and refinements
Simulations	Start simulations; monitor job status; read simulation logs
Visualization	Display field results on the model
Plotting	Create result plots from simulation output values
Expressions	Work with variables, functions, and expressions
Variable overrides	Create and edit variable overrides for parametric sweeps
Regions	Manage region definitions and selectors
Documentation	Search the Allsolve documentation for relevant information

The assistant is multilingual — write your question in any language and it responds in the same language. Names and descriptions created by tool calls also use the language of your message.

Tool calls

When the assistant determines it needs to perform an action (for example, creating a geometry element or starting a simulation), it displays the proposed action as a tool call. You can:

Select Accept to let the assistant execute the action.
Select Reject to prevent the action.

The tool call section is expandable, showing details of what the assistant intends to do.

Auto-accept

Under Settings, turn on Auto accept tool calls to let the assistant execute actions without waiting for manual approval each time.

Thinking

For complex questions, the assistant may show an expandable Thinking section above its response, indicating its reasoning process. Once complete, it displays how long the thinking took (for example, “Thought for 12s”). Select the section to expand or collapse it.

Session management

Each chat session is tied to the currently open project. Sessions from other projects appear as read-only.

Session tabs and history

Active sessions appear as tabs at the top of the panel. Session titles are generated automatically based on the conversation content.
Select New chat to start a fresh session.
Select Chat history to browse previous sessions, grouped by Today and Previous. Use the search field to filter sessions by title.
Sessions can be deleted from the history list or via the More options menu (three-dot icon).

Copying a conversation

Select More options → Copy chat to clipboard to copy the full conversation as text.

Feedback

After each assistant response, thumbs up and thumbs down icons appear on hover. Select one to send feedback. A “Feedback sent” label replaces the feedback buttons once submitted.

Token usage

A token usage indicator at the bottom of the panel shows how much of the available context window has been used in the current session. Hover over it for details including prompt tokens, cached tokens, and cache write tokens.

As the conversation grows, the context usage increases. Starting a new session resets the token count.