> ## Documentation Index
> Fetch the complete documentation index at: https://docs.rowbase.com/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Assistant

> Use AI to transform, clean, and understand your data

## Overview

Rowbase includes a powerful AI assistant that helps you work with data through natural language. Chat with your data, get intelligent suggestions, and use AI-powered operations.

## AI Chat

Talk to your data using natural language. The AI assistant understands your datasets and can help you:

* Explore and understand your data
* Create transformations using plain English
* Debug data quality issues
* Get recommendations for cleaning and organizing

### Starting a Conversation

1. Open a dataset
2. Click the **Chat** icon
3. Ask a question or describe what you want to do

### Example Prompts

**Understanding data:**

* "What columns does this dataset have?"
* "Show me rows where status is empty"
* "What's the distribution of values in the category column?"

**Creating transformations:**

* "Convert all email addresses to lowercase"
* "Parse the date column from MM/DD/YYYY format"
* "Create a new column that combines first and last name"

**Data quality:**

* "Find duplicate rows based on email"
* "Which rows have invalid phone numbers?"
* "Flag rows where quantity is negative"

## AI-Suggested Operations

When you import data, the AI analyzes it and suggests operations to clean and transform it.

### How Suggestions Work

1. **Import data** - Upload a CSV or paste data
2. **Analysis** - AI analyzes the data structure and quality
3. **Suggestions** - You'll see suggested operations at the top of the pipeline
4. **Accept or dismiss** - Review each suggestion and accept or dismiss

### Types of Suggestions

* **Header detection** - Rename columns from detected headers
* **Type parsing** - Convert text to numbers, dates, or booleans
* **Normalization** - Standardize formats (snake\_case, trim whitespace)
* **Quality fixes** - Handle empty values, standardize case

<Tip>
  You don't have to accept every suggestion. Review each one and accept only the ones that make sense for your use case.
</Tip>

## AI Operations

Beyond chat and suggestions, Rowbase includes AI-powered operations you can add to your pipeline.

### Use AI

Generate content based on other columns using AI.

**Use cases:**

* Categorize products from descriptions
* Extract entities from text
* Generate summaries
* Translate content
* Enrich data with derived fields

```
Input: Product description
Prompt: "Categorize this product into: Electronics, Clothing, Home, or Other"
Output: AI-generated category
```

### Detect Headers

When data has messy or missing headers, AI can detect the actual column names from the data.

```
Before: _UNNAMED_0, _UNNAMED_1, _UNNAMED_2
After: customer_name, email, signup_date
```

## Best Practices

<AccordionGroup>
  <Accordion title="Be specific in prompts" icon="comment">
    The more specific your request, the better the result. Instead of "fix the dates", try "parse dates from DD/MM/YYYY format to ISO format".
  </Accordion>

  <Accordion title="Review suggestions carefully" icon="eye">
    AI suggestions are helpful but not always perfect. Always review the preview before accepting.
  </Accordion>

  <Accordion title="Use chat for exploration" icon="magnifying-glass">
    Chat is great for understanding your data before deciding on transformations. Ask questions first, then build your pipeline.
  </Accordion>

  <Accordion title="Combine AI with manual operations" icon="wand-magic-sparkles">
    AI is powerful for complex tasks, but sometimes a simple filter or rename is all you need. Use the right tool for the job.
  </Accordion>
</AccordionGroup>