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

# AI Assistant

> Ask data questions in natural language with the ToeBox AI Assistant

The ToeBox AI Assistant is a conversational analytics tool built into the Analytics page. Ask questions about your data in plain English and get charts, tables, and insights in return.

## Using the AI Assistant

<Steps>
  <Step title="Open the Assistant">
    Navigate to **Analytics** and click the **AI Assistant** tab.
  </Step>

  <Step title="Ask a question">
    Type a question in natural language. For example:

    * "What were our top 10 products by revenue last month?"
    * "Show me daily order count for the past 30 days"
    * "Compare this quarter's sales to last quarter"
    * "What's the average order value by customer segment?"
  </Step>

  <Step title="Review the results">
    The assistant generates a SQL query, executes it against your connected data source, and displays the results as a chart or table. You can see:

    * The generated SQL query
    * A visualization of the results
    * The raw data in table format
  </Step>

  <Step title="Export to dashboard">
    If the chart is useful, click **Export to Dashboard** to add it directly to one of your dashboards. The chart, query, and configuration are all preserved.
  </Step>
</Steps>

## Model selection

The AI Assistant can use different models depending on your preference and available API keys:

| Provider        | Models                                                      |
| --------------- | ----------------------------------------------------------- |
| **Anthropic**   | Claude Sonnet 4.6, Claude Haiku 4.5, Claude Opus 4.6        |
| **OpenRouter**  | GPT-4o, GPT-4.1, Gemini 2.5 Flash/Pro, Llama 4, DeepSeek V3 |
| **AWS Bedrock** | Claude Sonnet 4.6, Claude Haiku 4.5                         |

Select your preferred model from the dropdown in the Assistant panel. If you have configured BYOK (Bring Your Own Key), your keys are used automatically.

<Tip>Claude Sonnet 4.6 is recommended for the AI Assistant. It provides the best balance of SQL generation quality and response speed.</Tip>

## How it works

1. Your natural language question is sent to the selected AI model
2. The model generates a SQL query based on your database schema
3. The query is executed against your connected data source (read-only)
4. Results are returned and automatically visualized as the most appropriate chart type
5. You can modify the chart type, formatting, or query as needed

## Tips for better results

* **Be specific** -- "Revenue by product category for Q1 2025" is better than "show me revenue"
* **Mention time ranges** -- The assistant handles dates well when you specify them
* **Reference table names** -- If you know your schema, mention specific tables for more accurate queries
* **Iterate** -- Follow up with refinements like "break that down by month" or "add a filter for orders over \$100"

<Note>The AI Assistant only executes read-only queries. It cannot modify your data.</Note>