The AI Assistant is a conversational side panel that allows the user to create and interact with the dashboard using natural language. A system of specialised AI agents collaborate to query data, plan layouts, create widgets, and configure them based on the user's request.
The AI Assistant is an experimental feature. Behaviour will be slightly different depending on the LLM being used, and results can vary across queries.
Try It Out Copy Link
The example below shows an empty dashboard loaded with sales data from a fictional retail company. This includes sales, orders, and product information, which can be compared across different regions, segments, and product categories. Try one of the prompts below to see the AI in action:
CopyAdd bar charts showing the best performing stores and regions in terms of sales.
CopyAdd a table listing customers, region, segment, net sales and number of orders.
CopyBuild a line chart showing daily sales over time.
CopyCreate a pie chart showing sales by customer segment for the last calendar year.
What Can the AI Do? Copy Link
- Build complete dashboards from a single prompt - layout, widgets, data bindings, and formatting.
- Answer data questions by querying your data sources and summarising results.
- Add, move, and configure individual widgets - chart types, filters, titles, and more.
- Apply and manage filters at the dashboard or widget level.
- Plan iteratively - create, review, and update structured plans before executing changes.
- Work across multiple data sources using relationships and joins.
Quick Setup Copy Link
Getting the AI assistant running requires three things:
1. Register the AI Module Copy Link
import { AgStudioAiModule, AgStudioModuleRegistry } from 'ag-studio';
AgStudioModuleRegistry.registerModules([AgStudioAiModule]);
The AI module requires an AG Studio Pro with AI licence.
2. Provide an Adapter Copy Link
AG Studio does not bundle a connection to any LLM. You must implement the AgAiAssistant interface, which translates between Studio's request format and your AI provider. See Building an Adapter for a complete guide.
import type { AgAiAssistant } from 'ag-studio';
const adapter: AgAiAssistant = {
executeTurn: (request) => { /* your implementation */ },
};
3. Pass the Adapter to Studio Copy Link
const studioProperties = {
ai: adapter,
// other studio properties ...
}The AI panel appears automatically. See Configuration for panel visibility, state persistence, and more.
Key Concepts Copy Link
A quick reference for the terms used across the AI documentation.
| Term | Definition |
|---|---|
| Assistant | The conversational side panel that users interact with. It manages threads, orchestrates profiles, and presents results. |
| Thread | A persistent conversation container. Users can create multiple threads - e.g. one for sales analysis, another for inventory. |
| Adapter | Your implementation of AgAiAssistant - the bridge between AG Studio and your LLM provider. See Building an Adapter. |
| Profile | A specialised AI persona with its own instructions, tools, and delegation permissions. See Agentic Experience. |
| Tool | An action the AI can perform on the dashboard - viewing schemas, executing queries, adding widgets, and more. |
For a deeper look at sessions, exchanges, turns, and delegation, see the Agentic Experience page.
Next Steps Copy Link
- Building an Adapter - Implement the
AgAiAssistantinterface to connect to your LLM.