Understand this database
Every table has a story. I want to know who's involved.
Open a SQL connection and click Understand in the sidebar (or Tools › Understand this schema). Lily sends the schema of the current schema to your AI tool and turns it into a guide: what the database is about, which areas it has, and what each table is for, even when the names are cryptic and there are no foreign keys.
It works on PostgreSQL, MySQL, SQLite, SQL Server, Trino and ClickHouse. MongoDB and DynamoDB have a smaller version, Understand collection and Understand table, on their engine pages.
What it reads
Table names, column names and types, primary keys, declared foreign keys and the foreign keys Lily guesses from names. No rows. For each table it sends up to 16 columns and counts the rest. See What the AI sees.
The guide
The result opens as a tour called Meet the cast:
- the domain and a summary of the database;
- areas, groups of tables that belong together;
- Who's who: every table with a friendly name, its role and how important it is (core, supporting or lookup);
- each entity's detail, with its relationships, Open table and Query ideas.
Copy puts the summary on the clipboard as Markdown.
Query ideas
Query ideas on an entity asks for three to five everyday queries, like recent orders for a customer. Each one has to be a read with at least one :parameter, and Lily checks it with EXPLAIN (never runs it) before showing it. Pick one to open it in the editor, fill in the parameters and run it.
Find where something lives
The search field at the top takes a question like where do we store refunds? Lily shortlists tables locally, asks the AI to rank them and checks every answer against the real schema. Click a result to open the table. Only names, types and roles are sent.
Big schemas
Up to 25 tables are analyzed in one request. Bigger schemas go in batches of 18, saved as they finish, then summarized. If the analysis stops (time limit, an error, you closed it), the guide shows how far it got; close and reopen the sheet to continue from there. Reanalyze starts over.
Keeping it current
The analysis is saved per connection and schema, so it opens instantly next time. When the schema changes, a banner counts new, removed and altered tables, and Update analyzes only those. For small schemas, or when most of it changed, Update redoes everything.
Where it helps
Once a schema is understood, Lily adds that business context to the other AI features: Ask, generated SQL, index advice, Investigate, migrations, query ideas and fake data. That is how customers finds the table called tb_usr.
Opening Understand also runs the sensitive-data scan for that schema.