Changelog

Added

• Real-time CDC support for PostgreSQL sources. See Real-time Streaming Guide.
• New webhook destination type for pushing events to HTTP endpoints.
• Pipeline-level environment variable overrides in acme.yml.

Changed

• Improved error messages for connection failures. Errors now include the connector name and a suggested fix.
• The Scheduler API now returns next_run_at in ISO 8601 format.

Fixed

• Fixed a race condition when running multiple pipelines targeting the same BigQuery table.
• Resolved memory leak in long-running streaming pipelines (>48h).

Added

• OAuth 2.0 authentication for API access.
• Support for Parquet file sources.
• acme test command for pipeline validation.

Changed

• Default batch size increased from 1,000 to 5,000 rows.
• Connector health checks now run every 30 seconds (was 60).

Fixed

• Fixed incorrect row counts in pipeline run summaries.
• S3 connector now handles eventual consistency properly.

Added

• Custom transform functions in Python.
• Pipeline dependency graphs with depends_on.
• Slack notification channel for pipeline failures.

Fixed

• CSV parser now handles quoted newlines correctly.
• Fixed timezone handling in MySQL datetime columns.

Added

• Initial support for streaming connectors (Kafka, Redis Streams).
• Dark mode for the web dashboard.
• Export pipeline runs as CSV.