Skip to content

AI Assistant

The demo loads when visible.

Reset demo restores the original sample data and clears its conversation. To use your own prompts, follow the API-client setup guide.

Both examples use predefined requests without a backend or model key. To interpret your own wording, connect your API client as shown in the quick start and configure your backend.

Your application chooses which operations are available. Filtering and sorting are the starting point; reading totals, changing data and working with products such as Gantt require additional configuration. Supported data and column edits show a preview for the user to approve.

  1. A user enters a request in the grid’s prompt or chat.
  2. Your interpreter translates those words into a structured command, action or message. It can run on your backend using a model provider you choose.
  3. The assistant validates the result and uses the existing grid or product APIs. Data and column edits wait for approval.

AiAssistantPlugin supplies the grid integration. Provider SDKs and credentials stay on your backend. Row values are excluded from interpretation requests by default; see Control what the model receives for optional disclosures.

How the AI assistant works A user describes a task. The assistant discloses allowed metadata to your backend, which calls your model. The returned JSON is validated before existing grid APIs execute it. Writes pause for approval. User Your application Assistant Browser Your backend Policy + credentials Model Your chosen provider Grid APIs Browser / data owner Describe a task Allowed metadata + enabled actions Server policy + typed schema One complete JSON result Typed result (no execution) Execute through existing APIs; writes pause for approval Updated view or action result Validate response and current revision
Your backend holds provider credentials and enforces its policy. The browser validates the response before calling existing grid APIs. Raw rows are excluded by default.

Follow the first three guides for an initial integration, then open the topic you need:

GuideWhat you will learn
1. Connect your first assistantPrerequisites, plugin setup and your first filter/sort request.
2. Choose actions and allow editsEnable reads, layout changes and approved data or column edits.
3. Control what the model receivesChoose disclosed fields, row samples, memory and result evidence.
Answer questions about dataClarify requests and answer from permitted analytical results.
Connect custom actions and productsIntegrate business services, Pivot, Pivot Charts, Gantt, Scheduler or Kanban.
Work with server-side dataApply criteria through your remote loader and respect paging.
Customize chat, history and languageConfigure presentation, suggestions, prompt recall and translations.
Set up the AI backendRun the reference service and prepare an authenticated deployment.
Test your integrationVerify behavior without live model calls.