generated: '2026-08-02'
method: searched
source: https://docs.ocient.com/release-notes
docs: https://docs.ocient.com/release-notes
scheme:
  platform: major.minor (no patch component in published notes)
  drivers: semver
  dated: partial
  note: JDBC release notes carry ISO dates per version; platform release notes are
    versioned but undated.
streams:
- id: platform
  name: OcientAIQ Unified Data Platform
  url: https://docs.ocient.com/ocientaiq-unified-data-platform-release-notes
  current: '27.1'
  history_depth: 'versions published back to 12.0'
- id: jdbc
  name: JDBC driver
  url: https://docs.ocient.com/jdbc-release-notes
  current: 4.2.1
  distribution: https://mvnrepository.com/artifact/com.ocient/ocient-jdbc4
- id: pyocient
  name: pyocient Python driver
  url: https://docs.ocient.com/pyocient-release-notes
  current: 3.8.0
  distribution: https://pypi.org/project/pyocient/
entries:
- version: '27.1'
  stream: platform
  date: null
  breaking: true
  highlights:
  - Data Extract Tool gained Apache Parquet file extraction.
  - Added CREATE TRANSACTIONAL PIPELINE with an optional START FOREGROUND clause that
    blocks the SQL session until the pipeline completes or fails, rolling back all
    uncommitted data on failure.
  - JDBC bulk loading can now stage data through an S3-compatible object store by
    setting bulkLoadMode to s3, instead of SSH or SFTP to Loader Nodes.
  additions:
  - Transaction support in the database and in the JDBC and pyocient drivers, grouping
    statements into a single committed or discarded unit of work.
  - DROP TASK SQL statement to remove orphaned or completed distributed tasks.
  - New system catalog tables sys.pipeline_errors_historical,
    sys.pipeline_events_historical, sys.pipeline_files_historical,
    sys.pipeline_metrics_historical.
  - transaction_id and transaction_prefix_len columns on sys.queries and
    sys.completed_queries; transaction_id on sys.pipelines.
  - Information schema views pipeline_status_historical, pipelines_historical,
    transactional_pipeline_status.
  breaking_changes:
  - The array_to_string SQL function no longer adds quotes around values of VARCHAR
    arrays.
- version: '27.0'
  stream: platform
  date: null
  breaking: false
  highlights:
  - Added the ALTER PIPELINE SQL statement for schema evolution, allowing the SELECT
    clause of a data pipeline to be modified for columns, filters, and transformations
    without recreating the pipeline.
  additions:
  - Ensemble machine learning models — Bagging, Boosting, and Stacking.
- version: '26.1'
  stream: platform
  date: null
  breaking: false
  highlights:
  - Added the line_number metadata key for loading the line number of a file in data
    pipelines, plus new extract options for delimited and CSV formats.
  additions:
  - Regression version of the Decision Tree model.
  - featureSubsetStrategy and maxCellsToFetch options on the Random Forest model.
  - DRAIN PAGES SQL statement to manually migrate data from pages to segments.
  - NOINDEX option for time bucket clauses in CREATE TABLE.
  - ALTER SYSTEM SET DEFAULT STORAGESPACE, ALTER CLUSTER ADD STORAGESPACE, and
    ALTER CLUSTER REMOVE STORAGESPACE statements.
  - LOCATION and OPTIONS keywords on CREATE TASK.
- version: 4.2.1
  stream: jdbc
  date: '2026-06-26'
  breaking: false
  highlights:
  - Internal updates only.
- version: 4.2.0
  stream: jdbc
  date: '2026-06-16'
  breaking: false
  highlights:
  - Added transactional support — group multiple statements into a single transaction
    by disabling autocommit and ending with COMMIT or ROLLBACK.
- version: 4.1.0
  stream: jdbc
  date: '2026-04-01'
  breaking: true
  highlights:
  - The CANCEL and KILL commands now take the query UUID as a quoted string.
  additions:
  - New connection properties bulkLoadSshHostKeyVerification,
    bulkLoadSshKnownHostsPath, and maxRowsPerInsertBatch.
window: 'Recent window only — the live release notes are the source of truth.'