ATX AnalyticsTable turns a Snowflake query into a local dataset you can keep working on with the network off. The warehouse does the heavy lifting once; the rows are saved into the project's DuckDB on your disk; reports, pivots, charts and scheduled PDFs run from there. Personal data is masked on the way in, on your machine.
The Snowflake connection form in ATX AnalyticsTable 1.7, filled in with the example values below. Secrets go into the encrypted vault, never into the project.
Create the key pair (on your machine): openssl genrsa 2048 | openssl pkcs8 -topk8 -v2 aes256 -out rsa_key.p8 openssl rsa -in rsa_key.p8 -pubout -out rsa_key.pub — Paste the TEXT of rsa_key.p8 into “Private key (PEM)”, not a file path. An old “BEGIN RSA PRIVATE KEY” file is refused with the conversion command.
Least-privilege reporting role (in Snowflake): CREATE ROLE IF NOT EXISTS REPORTING_RO; GRANT USAGE ON WAREHOUSE WH_REPORTING_XS TO ROLE REPORTING_RO; GRANT USAGE ON DATABASE SALES TO ROLE REPORTING_RO; GRANT USAGE ON SCHEMA SALES.PUBLIC TO ROLE REPORTING_RO; GRANT SELECT ON ALL TABLES IN SCHEMA SALES.PUBLIC TO ROLE REPORTING_RO; CREATE USER IF NOT EXISTS SVC_ATX_REPORTING TYPE = SERVICE DEFAULT_ROLE = REPORTING_RO; ALTER USER SVC_ATX_REPORTING SET RSA_PUBLIC_KEY = 'MIIBIjANBgkqh...'; -- body of rsa_key.pub GRANT ROLE REPORTING_RO TO USER SVC_ATX_REPORTING;
Reporting query — runs in Snowflake, e-mail addresses never leave it: SELECT o.order_date, o.region, SHA2(LOWER(TRIM(c.email)), 256) AS customer_key, COUNT(*) AS orders, SUM(o.amount) AS revenue FROM SALES.PUBLIC.ORDERS o JOIN SALES.PUBLIC.CUSTOMERS c ON c.customer_id = o.customer_id WHERE o.order_date >= DATEADD(day, -90, CURRENT_DATE()) GROUP BY o.order_date, o.region, customer_key ORDER BY o.order_date; — Every query carries the query tag, so this one shows up in QUERY_HISTORY under team=finance;app=atx.
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.
Snowflake's own Dynamic Data Masking and row access policies are evaluated for the role and user that run the query. With a key pair every analyst shares SVC_ATX_REPORTING, so they all see what that service user sees. To let Snowflake mask per person, sign in to AnalyticsTable with Entra ID, turn on on-behalf-of and use “Microsoft Entra ID (External OAuth)” with a token scope — the only Snowflake method that runs each query as the signed-in person.
Pulling fresh rows needs a route to Snowflake. Once imported, the data lives in the project's DuckDB on your disk: opening the project, filtering, pivoting, charting, running local pipeline steps and producing PDFs work with no network. The licence validates offline, and no ATX service is involved at any point.
The analysis is. Refreshing from Snowflake needs the network; everything you do with the imported rows — reports, pivots, charts, pipelines over local data, PDFs — runs on your machine without one.
Snowflake's masking and row access policies apply to the role and user that run the query. With a key pair everyone shares the service user; with Entra ID External OAuth and on-behalf-of, each query runs as the signed-in person and Snowflake masks per person.
In the project's DuckDB database on your disk (or on your server, for the server edition). Credentials sit in an encrypted vault, never in the project file.
Every query carries the connection's query tag, so QUERY_HISTORY shows what the reporting connection spent, and reporting from the local copy means a refresh is the only thing that costs credits.
This site needs JavaScript for the full experience. Every page above is reachable without it.