Skip to content

Changelog

All notable changes to TwinCommit. Versions follow semantic versioning.

  • twincommit check now covers every new source and target.
  • MCP server tools check_source and tail_events.
  • Databricks target: soft_deletes option and liquid clustering on primary keys.
  • Kafka target: protobuf format.
  • Default batch.flush_interval for lakehouse targets is now taken from merge_interval.
  • Oracle: SCN ordering across RAC threads after an instance restart.
  • MCP server (@twincommit/mcp) with eight tools, read-only by default.
  • control block with token authentication and allow_writes.
  • Snowflake target via Snowpipe Streaming.
  • SQL Server: reading log backups when the active log has been truncated.
  • Apache Hive target (ACID and Iceberg modes).
  • Apache Spark Structured Streaming connector.
  • Leader election in the Helm chart.
  • Oracle and SQL Server sources.
  • Databricks and Kafka targets, including a Debezium-compatible format.
  • state_schema default renamed to twincommit.
  • First general availability release.
  • Db2 LUW and PostgreSQL sources; PostgreSQL and SQL Server targets.
  • check, ddl, snapshot, run, status, verify and logdump commands.
  • Prometheus /metrics and JSON /status endpoints.
  • Exactly-once, commit-ordered apply with in-transaction checkpoints.