> ## Documentation Index
> Fetch the complete documentation index at: https://docs.popsink.com/llms.txt
> Use this file to discover all available pages before exploring further.

# v0.1.0-alpha.191 28/09/2026

***

**Description**: Backfill - a SQL Server source can load its history into BigQuery.

### Changes

* Backfill - a SQL Server source can load its history into BigQuery (#4475).
* Salesforce source - an SObject is synced on the cursor it has, and a failed read holds that object alone (#4515).
* Salesforce source - a connector takes 100 SObjects, capped on the whitelist it provisions from (#4510).
* Oracle source - the CDC adapter is a connector setting (#4508).
* DP - A MySQL binary column reaches a target as base64, never as a Python repr (#4478).
* DP - a declared primary key is checked against the columns the target writes (#4505).
* Snow target - the durable-offset read leaves the poll thread (#4507).
* BigQuery target - a Debezium date reaches a legacy INTEGER column, and a row that cannot be written stops the worker (#4506).
* DP - MARIADB\_SOURCE: connector type, mariadb-to-kafka strategy, flavour pre-flight (#4503).
* \[ARCH] The Snowflake target listings take a port instead of building their own adapter (#4517).
* \[FRONT] Part 1 of #4518: the pipeline stores read /pipelines/\* through the service (#4520).
* DP - A missing primary key says why, and what it costs (#4491).
* \[SEC] A pipeline write authorizes every connector it names (#4527).
* \[TESTS] Pin what the wizard's first subscriptions carry (#4529).
* DP - MariaDB's introspection says it reads the key, like MySQL's (#4534).
* BigQuery target - a batch is written, or the worker stops holding it (#4519).
* Release cut - data-plane 2026.09.28, chart 0.1.0-alpha.191 (#4525).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.