Skip to main content
The Text-to-Chart feature in ClickHouse Observability allows you to create visualizations by describing what you want to see in plain text. Rather than manually selecting metrics, filters, and group-by fields, you can type a prompt such as “Error counts by service over the last 3 days” and the AI Assistant will generate the corresponding chart automatically. This feature uses a large language model (LLM) to convert your text prompt into a query, then builds the visualization in the Chart Explorer.

​
Prerequisites

A team admin must enable Generative AI in Team Settings under Configure Generative AI. See Enabling Generative AI for instructions. AI models and providers are managed automatically by the platform.

​
Using Text-to-Chart

2

Select a data source

Expand AI Assistant at the top of Chart Explorer if it is collapsed. Choose the data source you want to visualize from the selector beside the prompt field — for example, Logs or Traces.
3

Enter a text prompt

In the AI Assistant prompt field, type a natural language description of the chart you want to create. For example:
  • Error counts by service over the last 3 days
  • Latency breakdown by endpoint
  • Count of events over time grouped by severity
Click Generate to create the chart from your prompt. Review the generated configuration in the builder below. You can adjust the chart type, filters, and grouping, then click Run to update the visualization.

​
Example prompts

The following prompts demonstrate common use cases when working with observability data: Prompts can reference any column or attribute available in your configured data sources. Include specific column names, filters, and time ranges to help the AI generate the chart you want.

​
Limitations

  • Only logs and traces are supported as data sources. Prometheus metrics are not yet supported.
  • Chart accuracy depends on the clarity of the prompt and the structure of the underlying data. If a generated chart does not match expectations, try rephrasing the prompt or specifying column names explicitly.

​
Further reading