Skip to content
Not publicly available.This describes the Loupestack API before launch. Endpoints, limits and fields can still change.

Glossary

Who may call an operation: logged in, any member, or owners and admins (Authentication).

The workspace every /v1/ call acts on. Each session has one; switch it with POST /api/auth/organization/set-active.

How a generated UI’s queries reach the API: the document asks the page that shows it, and the page calls the API (Embedding generated UIs).

A data source a workspace queries in place: a PostgreSQL database, or a BigQuery project. Its alias is the name used in SQL, for example crm in crm.public.orders. Its password or key is stored encrypted and never returned.

A series of questions and answers (or chat messages) of one user in one workspace. Only its author can read it.

A dashboard written by the language model from a prompt: an HTML document plus the saved queries it may run. Each change makes a new version; older versions keep working.

The workspace’s own storage, with the tables that sync jobs copied in. Its tables are named lake.<table>.

A connector queried in place, at the moment of the question. A BigQuery connector is live only when created with live: true; every read is then billed by Google.

A named measure in a table description, with a definition in words and a formula such as sum(amount). The formula is a hint for the model; it is never run.

The roles of a workspace membership. Owners and admins manage connectors, descriptions and sync jobs; members use them.

A checked, stored SELECT with declared parameters, run by name with parameter values. Clients see its name and parameters, never its SQL. Generated UIs create them.

How long calls answer: a stream of named events (Streaming responses).

Anything a workspace can query: the lake, or a live connector. GET /v1/sources lists their tables and columns.

A sync job copies a database table or an uploaded file into the lake on a schedule; each run of it is a sync run, with its status and row count.

What a table, its columns and its metrics mean, written by owners and admins. The model reads them when it answers questions and builds UIs.

The unit of tenancy: one team or company, with its own sources, lake, descriptions, conversations and UIs. Data never crosses from one workspace to another.