Lumo by HorusBI

By HorusBI

Get New App Alerts

Open in ChatGPT

Open in Claude

Overview

Ask questions about your company's business data in Lumo, the BI platform by HorusBI, and get the numbers with a link to check them. Read-only. Lumo is the BI platform by HorusBI. This connector lets ChatGPT read the BI applications you already have access to in Lumo: tables, columns, curated business indicators (KPIs), ready-made measures and aggregated results. What ChatGPT can do with it: - Find where a business concept lives: indicators, measures and columns across your applications, each indicator with a ready-to-run query. - Describe an application: tables, columns, measures and how they relate. - Look up the values stored in the data, so a name you type (a customer, a product, a store) matches the exact stored value before filtering. - Run aggregated queries with filters, sorting and limits (up to 500 rows per call), with totals over all matching rows. Every query result carries the date of the data and a link to the same result as a report in Lumo, so you can check each number. The connector is read-only: it never creates, edits or deletes anything. It uses your own Lumo permissions, so you only see the applications, rows and columns you can already see when logged in. Your company administrator can block access per group or user. Requires a Lumo account. Sign-in uses OAuth in your browser.

Tools

get_app

ChatGPT

Describes one app: its tables (fact or dimension, date of the last load), the columns of each with ref, type, role, unit and how they enter a query (agg or bucket), the app expressions and indicators, and the relationships that let columns of different tables be combined in one query. Every column and expression is listed with its complete ref; these are the refs that query accepts for the app.

query

ChatGPT

Runs an aggregated query and returns the numbers: raw numbers, each column's unit, a summary over all matching rows, how many rows were cut, the filters and period applied, the date of the data and a link to the result. fields: { ref } groups by a text column, a date or a dimension; { ref, agg } measures, agg being sum, avg, min, max, count or count_distinct (numeric columns; count and count_distinct work on any column); a date groups by day, or by { ref, bucket } with bucket day, month, year or month_of_year; an app expression is already aggregated and goes as { ref } only. filters: { ref, op, value }, op being = != > < >= <= between, not between, in, not in, like, not like, is null, is not null; value is one item, or a list for in and between ([start, end], dates as YYYY-MM-DD); agg on a filter cuts on the aggregated value. An empty value means blank: = "" keeps the rows without value (null or empty text), != "" leaves them out; != with a filled value keeps the blank rows. A text filter with = or in is checked against the values of the column before the query runs. orderBy: { ref, dir } on a ref that is in fields, dir asc or desc. limit defaults to 10 rows (31 when a date is grouped) and goes up to 500; without orderBy the rows are an arbitrary sample, not a ranking. The refs it accepts are those of the app's schema, as get_app lists them. Columns of different tables combine when the app has a relationship between them. Percent columns come as fractions (0.25 = 25%); durations come in seconds or minutes and sizes in bytes, as the unit says.

search_catalog

ChatGPT

Finds where a concept lives in the user's data. With terms, returns the matching indicators, app expressions and columns of every app the user can see: an indicator comes with its ready query call, a column or expression with its ref and the field that goes in query. Without terms, returns the catalog: each app with its desk, whether it is a personal copy or was published, a sample of its columns (the date columns and the most used in dashboards) and its curated indicators. An indicator is a curated definition of a number; an expression is a ready-made, already aggregated measure of an app; a ref is the name of a column or expression in the app, written as the interface shows it. The catalog carries a sample of each app's columns; the complete schema (tables, columns, relationships) of an app is what get_app returns. It does not search the data itself; search_values does.

search_values

ChatGPT

Looks at the values that exist in the data. With ref: the distinct values of that text or date column, without counts, optionally narrowed by contains; every word of contains has to appear in the value, ignoring accent and case, as a whole word or a prefix ('rest aurora' matches 'Restaurante Aurora'). Without ref and with contains: looks for that text in the text columns of the app (up to 20 columns, dimension tables first, a few values per column, within 20 seconds) and lists the columns left out by the limit or the time. mode sample: example rows of a table, without technical columns such as ids and uuids.