The Limerence handbook
A desktop app that reads your database and answers questions in plain language. What it does, where your data goes, how to teach it what your business means, and how to tell when it is wrong.
By the Limerence team · Living document
Where your data goes
What stays on your machine, what leaves for the model, and what stops a query from writing.
Question to SQL to answer
The loop, end to end, with the generated query visible at every step.
Start with a file on disk
Point it at a SQLite file and ask something real. Nothing to provision first.
Checking the answer
The short read of the generated SQL that catches most mistakes before you act on a number.
Before you install
What Limerence is, and what it isn't
A desktop app that reads your database and answers questions in plain language. The loop end to end, and the cases where a BI tool or a human analyst is still the better answer.
Where your data goes
What stays on your machine, what leaves for the model, and why a generated query cannot write to your database.
Installing
macOS and Windows installers, what the first launch asks for, and what the app does before you have connected anything.
Setting up
Connecting a database
What the app finds on its own, the credentials worth handing it, and how it learns and re-learns the shape of your schema.
Data sources
- PostgreSQLConnection profiles, TLS modes from disable through verify-full, search-path handling, and schema scoping.
- MySQL & MariaDBUser and grant setup, charset and collation gotchas, and the dialect differences SQL generation already accounts for.
- SQL ServerWindows versus SQL auth, schema versus database scoping, and what to expect against a read replica.
- SQLitePointing at a file on disk. The fastest way to try this on data you already have, with nothing to provision.
Connecting a model
Bringing your own key, which provider to start with, and what the app sends when it asks one a question.
Teaching it your business
A schema says orders and amount. It does not say which orders count as revenue. Spaces, and the definitions that outlive a conversation.
Working
Asking, and checking the answer
Question, plan, SQL, rows, answer. What each step exposes, the short read of the generated query that catches most mistakes, and how to turn a correction into knowledge.
Dashboards
Pinning a question so it re-runs against current data, and arranging the handful that are worth a standing look.
Reports
Longer-form output that carries the reasoning alongside the numbers, for the times a chart is not the deliverable.
Exporting
Export is how work leaves the app today. What each format carries, and what it deliberately leaves behind.
Over time
Automations
Questions on a schedule, where the results land, and what happens to a slot your machine was asleep for.
When your schema changes
The symptoms of a stale read, the safe re-sync order, and which knowledge you need to revisit after a rename.
Updates
How the app updates itself, what a release can change about generated SQL, and where to read what moved.
When something breaks
Model errors, refused connections, and empty results that should not be empty. What to check, in order, and what to send us.