ATX AnalyticsTable is a desktop SQL client and reporting workspace for Google BigQuery. You write the query, BigQuery runs it, and the result lands in a local DuckDB database inside your project — on your disk, not in another vendor's cloud. From there you filter, pivot, chart and schedule without paying for a new scan each time, and personal data is masked before anything leaves your machine.
The Google BigQuery connection form in ATX AnalyticsTable 1.7, filled in with the example values below. Secrets go into the encrypted vault, never into the project.
A service account that can run jobs (gcloud): gcloud iam service-accounts create atx-reporting --display-name="ATX AnalyticsTable reporting" gcloud projects add-iam-policy-binding example-analytics-prod \ --member="serviceAccount:atx-reporting@example-analytics-prod.iam.gserviceaccount.com" \ --role="roles/bigquery.jobUser"
Read access to one dataset only (BigQuery SQL): GRANT `roles/bigquery.dataViewer` ON SCHEMA `example-analytics-prod.marketing` TO "serviceAccount:atx-reporting@example-analytics-prod.iam.gserviceaccount.com";
Reporting query — partition filter first, e-mails hashed in BigQuery: SELECT event_date, country, TO_HEX(SHA256(LOWER(TRIM(user_email)))) AS user_key, COUNT(*) AS sessions FROM `example-analytics-prod.marketing.web_sessions` WHERE event_date BETWEEN DATE_SUB(CURRENT_DATE(), INTERVAL 30 DAY) AND CURRENT_DATE() GROUP BY event_date, country, user_key; — event_date is the partitioning column: the WHERE clause decides how much is scanned and billed. Save the result into the project's DuckDB and page over it there — once billed, not once per page.
Then reshape locally — a Run SQL step on the project's DuckDB: -- "input" is the table the previous pipeline step produced (the import above). -- This runs on your machine, in the project's DuckDB, not on the database. SELECT region, SUM(revenue) AS revenue, COUNT(DISTINCT customer_key) AS customers FROM input GROUP BY region ORDER BY revenue DESC; — Column names follow the query above; on the local engine the pipeline's current table is registered as input before the script runs.
BigQuery column-level access (policy tags) and dynamic data masking are evaluated for the identity that runs the job. BigQuery connections cannot carry a per-user identity yet — every query runs as the connection's service account — so per-person masking belongs in the project's data-access policy (layer 3).
BigQuery itself is a Google Cloud service, so the query needs a route to it. Everything after the import is local: the rows sit in the project's DuckDB on your disk, and pivots, charts, pipelines, PDF reports and masking over them run with no network at all. The licence validates offline, and ATX receives none of your data.
No. The app talks to the BigQuery API directly from your machine (or your own server installation). ATX receives none of your queries, results or credentials, and the software sends no telemetry.
It keeps raw addresses inside BigQuery, which is good. But an unsalted SHA-256 of an e-mail can be reversed by hashing a list of candidate addresses. Use it as a join key, and use the Mask personal data step (keyed with the project's random key) for pseudonyms that must hold.
Only what an administrator allows. The agent does not read from a saved connection by default, and anything sent to a model is pseudonymised first — names, e-mails, IBANs and IDs become stable tokens that are unmasked locally in the answer.
Filter on the partitioning column in the WHERE clause. For tables that require a partition filter, the app's preview filters to the newest partition with data on its own and says so.
This site needs JavaScript for the full experience. Every page above is reachable without it.