Explore MotherDuck tables with AI
Find the relevant table in an unfamiliar database. Inspect its fields and a limited sample to understand what one row represents before counting or joining data.
Explore MotherDuck tables with AI when the obvious table name still leaves a question unanswered. An orders table might hold one row per purchase, one per item or one per status change. Counting those rows gives three different answers to a manager asking how many orders were completed.
The motherduck-explore guide starts with authorised catalogues, field types and available descriptions. It moves to a small sample or a focused profile when that helps resolve a specific uncertainty. An empty dispatch date, for instance, may be exactly what you expect for an order awaiting collection.
What it does
- Locate the database and table relevant to a question
- Read column types and existing descriptions
- Establish the business event represented by a row
- Inspect missing values in a field that matters
- Check whether a proposed key is unique enough for a join
Install and use
- Use a compatible agent with authorised read access to the target MotherDuck database. Identify the database and fields the agent may sample, especially if customer information is present.
- Ask, “Find where completed orders are stored. Explain what one row represents and avoid displaying customer names or addresses.”
- Compare the explanation with a familiar order. If its identifier appears more than once, determine why before counting orders or joining the table to another source.
FAQ
Source
Made by MotherDuck. Original skill: motherduck-explore, published on GitHub under the MIT licence.
View on GitHub