[Read on Dashport](https://dashport.ai/documentation/how-chat-works)

Asking ChatHow Chat works

Copy page

[

Dashport Docsv2.0](/documentation)

Getting Started

IntroductionCreating your account5-minute quick start

Connecting Data

Supported integrationsConnect a databaseColumn comments & AI contextSuperSheets as a sourceCredential securityRefresh intervals

Using SuperSheets

What is SuperSheets?Creating a sheetData entry & column typesFormulas & calculationsFormatting & selectionSort, filter & columnsRow operationsValidation & frozen panesImport & exportSharing & activity logKeyboard shortcutsUse as a data source

Asking Chat

How Chat worksWriting effective promptsExample conversationsBuilding charts via ChatLimits & best practices

Building Dashboards

Three ways to buildDrag-and-drop builderAsk Chat to buildSharing & permissions

Asking ChatHow Chat works

Copy page

## How Chat Works[](#ch-how)

When you ask Chat a question, here's what happens under the hood:

1

Intent classification

Chat classifies your message as a data query, a dashboard action, or a conversational question.

2

Schema lookup

Chat retrieves the schema of your connected data sources (table names, column names, types) to understand what data is available.

3

Query generation

For data questions, Chat generates a query (SQL or API call) against your live data source and executes it.

4

Response generation

The query results are formatted into a human-readable answer, optionally with a chart.

5

Dashboard action (if requested)

If you asked Chat to create a chart, it programmatically adds the widget to the specified Superdashboard.

Note

Chat has access to your data schema but not your raw data values — it generates queries that are run against your data in real time. Your data never leaves your environment to train AI models.

Previous

Asking Chat

Next

Writing effective prompts
