You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
With our v26.1 release, we will extend the AI-powered capabilities of both the DevExtreme [DataGrid](https://js.devexpress.com/Documentation/Guide/UI_Components/DataGrid/Overview/) and [TreeList](https://js.devexpress.com/Documentation/Guide/UI_Components/TreeList/Overview/). We plan to introduce a DevExtreme Chat-based AI assistant that will allow you to use natural language to interact with DataGrid/TreeList. The DataGrid/TreeList AI assistant will be able to complete the following actions:
2
-
* Sort, filter, search, and group data.
3
-
* Resize, reorder, fix/unfix, and show/hide columns.
4
-
* Change pagination settings.
5
-
<!--split-->
1
+
The AI Assistant for the DevExtreme [DataGrid](/Documentation/Guide/UI_Components/DataGrid/Overview/) allows you to interact with the component using natural language. The Chat also supports [speech-to-text input](/Documentation/ApiReference/UI_Components/dxChat/Configuration/#speechToTextEnabled), ideal for hands-free interactions or entering longer prompts.
2
+
3
+
You can use/configure the following DataGrid features in the AI Assistant Chat prompts:
6
4
7
-
This feature is particularly well suited for the following usage scenarios:
5
+
-[Filtering and Searching](/Documentation/Guide/UI_Components/DataGrid/Filtering_and_Searching/)
-[Column Fixing](/Documentation/Guide/UI_Components/DataGrid/Columns/Column_Fixing/), [Resizing](/Documentation/Guide/UI_Components/DataGrid/Columns/Column_Sizing/), and [Reordering](/Documentation/Guide/UI_Components/DataGrid/Columns/Column_Reordering/)
13
+
14
+
The AI Assistant feature is also available for the DevExtreme [TreeList](/Documentation/ApiReference/UI_Components/dxTreeList/Configuration/#aiAssistant) component.
15
+
<!--split-->
8
16
9
-
***Large Datasets**: Use the power of AI to browse large data sets. Apply complex filter values with natural language.
17
+
In this demo, the AI Assistant is enabled for a DataGrid that displays mock sales data with over 1500 records.
AI services used for this demo have been rate and data limited. As such, you may experience performance-related delays when exploring the capabilities of the DataGrid AI Assistant.
13
22
23
+
When connected to your own AI model/service without rate and data limits, the AI Assistant will perform seamlessly, without artificial delays. Note that DevExtreme does not offer an AI REST API and does not ship any built-in LLMs/SLMs.
14
24
15
-
***Accessibility-first Applications**: Allow users to interact with your application in a manner that is most appropriate for each specific situation: while at their desk, or on the go (when their hands are busy and they require voice commands to interact with your app).
25
+
[/note]
16
26
27
+
To enable the AI Assistant, configure the [aiIntegration](/Documentation/ApiReference/UI_Components/dxDataGrid/Configuration/#aiIntegration) or **aiAssistant**.[aiIntegration](/Documentation/ApiReference/UI_Components/dxDataGrid/Configuration/aiAssistant/#aiIntegration) object and set the **aiAssistant**.[enabled](/Documentation/ApiReference/UI_Components/dxDataGrid/Configuration/aiAssistant/#enabled) property to `true`. Once activated, the DataGrid adds a predefined item (*"aiAssistantButton"*) to the toolbar. This button opens the AI Assistant Chat in a draggable pop-up window.
17
28
18
-
***Replicable Configurations**: Apply the same prompt to recreate complex grid configurations across sessions. Reduce time spent on tedious layout adjustments.
29
+
This demo also configures [suggestions](/Documentation/ApiReference/UI_Components/dxChat/Configuration/#suggestions) for the AI Assistant Chat. These buttons allow you to interact with the assistant in one click using predefined prompts. For additional information about suggestions, refer to the following demo: [DevExtreme Chat - Prompt Suggestions](/Demos/WidgetsGallery/Demo/Chat/PromptSuggestions/).
// TODO: Implement system prompt when grid operations are ready
46
-
system: 'You are a helpful AI assistant for a data grid component. The user sends a natural language request describing an operation on the grid (e.g., sorting, filtering, grouping). You receive the user\'s message, a context object describing the current grid state, and a JSON schema describing the available commands and their arguments. Your task is to interpret the user\'s request and return a JSON object with one field: "actions" — an array of command objects, each with "name" (the command name) and "args" (an object of arguments matching the schema). Output must be a valid JSON string, directly parsable by JSON.parse. Do not include any markdown, formatting, or extra text — only the raw JSON object.',
45
+
system: 'You are a helpful AI assistant for a data grid component. The user sends a natural language request describing an operation on the grid (e.g., sorting, filtering, grouping). You receive the user\'s message, a context object describing the current grid state, and a JSON schema describing the available commands and their arguments. Your task is to interpret the user\'s request and return a JSON object with one field: "actions" — an array of command objects, each with "name" (the command name) and "args" (an object of arguments matching the schema). Output must be a valid JSON string, directly parsable by JSON.parse. Do not include any markdown, formatting, or extra text — only the raw JSON object.\n\nCRITICAL RULE FOR OPTIONAL ARGUMENTS: If an optional argument is not used, the field MUST be ENTIRELY ABSENT from the JSON object — the key must not appear at all. NEVER emit an optional field with value null, empty string "", empty array [], or any placeholder. This rule overrides any instinct to "complete" the object — omitted IS the value.',
0 commit comments