Explorations

Explorations is a conversational UI feature in Tinybird that lets you explore and interact with your data using natural language.

  • Use a notebook-style interface to explore your data with tables and charts.
  • Ask questions in natural language using the AI chat, which generates contextualized SQL for your queries.

Explorations are sandbox environments where you can test queries on your ingested data. For example, use Explorations to quickly query real-time production data, debug existing queries, or understand your Workspace resources.

To start using Explorations, go to Tinybird Cloud, log in to your account and select Explorations in the sidebar.

To create a new Exploration, submit your own prompt or select one of the built-in suggestions. To run a query using SQL instead of natural language (without using the LLM), select the Write a query option.

Explorations initial screen

How it works

Based on your prompt, a large language model (LLM) generates contextualized SQL with detailed explanations to help you understand the results. SQL queries are structured in nodes. Each Exploration can have one or more nodes. Nodes in Explorations are similar to nodes in Pipes. A node can query other nodes within the same Exploration. You can edit and rerun any SQL generated by the LLM in the SQL editor.

The chat history contains the entire conversation, and you can ask follow-up questions at any time.

Similar to other LLM-based interfaces, each Exploration maintains the context of previous messages. For unrelated questions or data, create a new Exploration for better results.

Fix errors

If the AI-generated query is incorrect, a Fix error option appears. Click it to ask the AI to fix the error.

Results as table or time series

You can quickly switch between table and time series visualization by clicking the Switch to time series/table button in the top-right of the SQL editor.

For time series visualization, the initial configuration is inferred from your query results if possible. You can further configure your time series.

  • The X axis should use date or datetime data.
  • The Y axis should use numeric data.

Example of an exploration with chat history on the left and generated SQL and table results on the right

Define your own workspace rules

You can fine-tune how the assistant responds to get faster and more relevant results. Inside an Exploration, click the workspace rules link below the chat prompt input to add your custom rules.

Limitations & Privacy

Explorations are available to all roles within a Workspace.

The Explorations AI chat is only available in Tinybird Cloud. Tinybird anonymously processes a limited sample of your data to deliver faster, more accurate results.

Currently, Explorations cannot be shared or exported. Each user can only access their own Explorations.

Next steps

Updated