A report built in a workbook keeps its logic where version control cannot see it: inside a binary .xlsx, in macros, in a query pasted into a hidden sheet. The usual result is a folder of report_final_v7.xlsx files and no answer to the question an auditor actually asks — who changed this figure, when, and to what. Git-native BI keeps the report's definition as small, readable text files in the project folder, so the same Git your data-engineering team already uses records every change as a diff that can be reviewed, explained and undone.
An .xlsx file is a zipped bundle of XML. Git can store it, but a one-cell change shows up as a changed binary, so nobody reviews it — and two people editing the same workbook cannot merge their work. Teams fall back on file names, e-mail attachments and shared drives, and the history of a production report lives in people's memories.
The fix is not a better naming convention. It is to separate what a report is (its queries, steps, schedule, filters and export rules) from the data it produces, and to keep the first part as text.
In ATX AnalyticsTable a project is a folder, and almost everything you build in it ends up there as JSON or SQL that a person can read:
Keep each project in a Git repository, with history and diffs for every change.
The data does not. Initialising a repository writes a managed block into .gitignore that excludes the project database (*.duckdb) and its write-ahead and journal files. That is not tidiness: the database is held open and locked while the application runs, so without those lines the first commit would fail. The block only ever adds missing lines — your own entries are never removed or reordered.
Credentials do not either. Connection secrets live in the encrypted vault, and the Git token itself is stored there and injected as an HTTP authorisation header — never written into a remote URL, a command line or a log file.
Analysts do not need a terminal. A Git panel in the sidebar initialises a repository on the project folder in one click (branch main), lists changes in plain language — new, edited, deleted, renamed, conflict — and commits everything with a message. Push and pull work against GitHub, GitLab or Bitbucket on the current branch, with a token validated in Settings before you rely on it.
The schedules file has its own version history with a coloured diff per revision, and any file can be restored to an earlier revision. Restore writes the old version back into the working tree for you to review and commit; it never rewrites history.
Because the definition is text, a change to a production report is a reviewable diff: a new WHERE clause in a .sql file, a changed export target in a pipeline step, a schedule moved from 06:00 to 05:30. Push the branch and review it as a pull request (merge request) in your Git host, with the same approval rules you use for code — branching and merging are done there, with the tools your team already has.
A saved query names its connection, and resolves it by id and then by name, because connection ids are per machine while the folder travels. A query whose connection the machine does not have refuses to open rather than silently running somewhere else.
Connection definitions export as a JSON bundle — encrypted under a passphrase when it has to carry secrets — so a standard set of connections is rolled out across a team rather than retyped. Together with a shared repository, every analyst opens the same report definition against the same named connections.
Git answers what changed in the report's definition. The audit trail answers who did what on the installation. ATX AnalyticsTable keeps an append-only, hash-chained audit log on every installation, and Git push, pull and restore are recorded in it alongside query runs, exports and credential access — so a reviewer can line the two up.
BI where the report's definition — queries, transformation steps, schedules, filters and export rules — is stored as text files in a repository, so every change is a commit that can be diffed, reviewed and restored, instead of living inside a binary workbook or a vendor's server.
You can commit .xlsx files, but Git sees them as binaries: no readable diff and no merge. Moving the report's logic into text — SQL files and JSON pipeline, schedule and export definitions — is what makes the history useful. ATX AnalyticsTable keeps the familiar grid and formulas while storing the definition as text, and still exports to Excel.
Settings has provider entries for GitHub, GitLab and Bitbucket — API base URL, repository and a token that is validated before you rely on it and kept in the encrypted vault. Self-hosted instances of those hosts are configured through the API base URL.
No. The managed .gitignore excludes the project database, and secrets stay in the vault. What is committed is the definition of the report, not its results.
This site needs JavaScript for the full experience. Every page above is reachable without it.