Databricks Connect release notes

This page lists releases and maintenance updates issued for Databricks Connect.

The Databricks Runtime version of the cluster must be greater than or equal to the Databricks Connect version. Databricks recommends using the latest version to receive any bug fixes and security updates. For system environment information for Databricks Runtime releases, see Databricks Runtime release notes versions and compatibility.

Databricks Connect for Databricks Runtime 19

These Databricks Connect versions are compatible with Databricks Runtime 19.

Databricks Connect 19.0.0 (Python)

July 31, 2026

The Databricks Connect client is updated to support Databricks Runtime 19.

Databricks Connect 19.0.1 (Scala)

August 10, 2026

Minor fixes and internal improvements.

Databricks Connect 19.0.0 (Scala)

July 31, 2026

The Databricks Connect client is updated to support Databricks Runtime 19.

Databricks Connect for Databricks Runtime 18

These Databricks Connect versions are compatible with Databricks Runtime 18.

Databricks Connect 18.3.3 (Python)

July 22, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.2 (Python)

July 15, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.1 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.0 (Python)

June 16, 2026

The Databricks Connect client is updated to support Databricks Runtime 18 LTS.

Databricks Connect 18.3.3 (Scala)

July 22, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.2 (Scala)

July 15, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.1 (Scala)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.3.0 (Scala)

June 16, 2026

The Databricks Connect client is updated to support Databricks Runtime 18 LTS.

Databricks Connect for Databricks Runtime 18.2

Databricks Connect 18.2.4 (Python)

July 14, 2026

Minor fixes and internal improvements.

Databricks Connect 18.2.3 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.2.2 (Python)

June 16, 2026

Improved error messages for operations that are not supported on classic compute (standard access mode) or serverless compute (for example, RDDs and JVM-dependent attributes).

Databricks Connect 18.2.1 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 18.2.0 (Python)

May 6, 2026

The Databricks Connect client is updated to support Databricks Runtime 18.2.

• The databricks-sdk dependency lower bound has been raised from >=0.46.0 to >=0.93.0.
• Upper bounds have been added to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, zstandard<1, numpy<3, databricks-sdk<1.
• PyPy is no longer supported. It is not actively maintained and numpy, a Databricks Connect dependency, does not support it. See SPARK-56186.

Databricks Connect 18.2.1 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 18.2.0 (Scala)

May 12, 2026

The Databricks Connect client is updated to support Databricks Runtime 18.2.

Databricks Connect for Databricks Runtime 18.1

Databricks Connect 18.1.8 (Python)

July 31, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.7 (Python)

July 14, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.6 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.5 (Python)

June 16, 2026

Improved error messages for operations that are not supported on classic compute (standard access mode) or serverless compute (for example, RDDs and JVM-dependent attributes).

Databricks Connect 18.1.4 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.3 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, zstandard<1, numpy<3, databricks-sdk<1.

Databricks Connect 18.1.1 (Python)

March 3, 2026

Fixed an issue where Databricks Connect did not work in IPython shells.

Databricks Connect 18.1.0 (Python)

February 26, 2026

The Databricks Connect client is updated to support Databricks Runtime 18.1.

• Automatic management of UDF dependencies is now available in Public Preview. See Automatic management of UDF dependencies.
• UDFs with dependencies now support local packages, folders, and Python files. See Manage UDF dependencies.
• SPARK-54849 Upgraded the minimum required version of pyarrow from pyarrow>=11.0.0 to pyarrow>=18.0.0.

Databricks Connect 18.1.5 (Scala)

July 31, 2026

Fixed a thread-safety race condition in Arrow result deserialization that could intermittently fail with ScalaReflectionException: <none> is not a module when query results were deserialized concurrently. See SPARK-57371.

Databricks Connect 18.1.4 (Scala)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.3 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 18.1.2 (Scala)

May 28, 2026

SparkThrowable exceptions now propagate the message template, so client-side error formatting matches the server's templated error rendering.

Databricks Connect 18.1.1 (Scala)

April 28, 2026

The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.

Databricks Connect 18.1.0 (Scala)

February 26, 2026

The Databricks Connect client is updated to support Databricks Runtime 18.1.

Databricks Connect for Databricks Runtime 18.0

Databricks Connect 18.0.9 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.8 (Python)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.7 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.6 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, zstandard<1, numpy<3, databricks-sdk<1.

Databricks Connect 18.0.5 (Python)

March 3, 2026

Fixed an issue where Databricks Connect did not work in IPython shells.

Databricks Connect 18.0.4 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.3 (Python)

February 12, 2026

• Fixed an issue that caused PySpark import to fail on Windows. (SPARK-54745)

Databricks Connect 18.0.2 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.1 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 18.0.0 (Python)

December 10, 2025

The Databricks Connect client is updated to support Databricks Runtime 18.0 (EoS).

• UDFs with dependencies now support pyspark.sql.streaming.DataStreamWriter.foreach and pyspark.sql.streaming.DataStreamWriter.foreachBatch. See Manage UDF dependencies.
• SPARK-50657 Upgraded the minimum required version of pyarrow from pyarrow>=10.0.0 to pyarrow>=11.0.0.

Databricks Connect 18.0.5 (Scala)

July 14, 2026

Fixed a thread-safety race condition in Arrow result deserialization that caused intermittent failures with the error ScalaReflectionException: <none> is not a module when query results were deserialized concurrently. See SPARK-57371.

Databricks Connect 18.0.4 (Scala)

June 24, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.3 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 18.0.2 (Scala)

May 28, 2026

SparkThrowable exceptions now propagate the message template, so client-side error formatting matches the server's templated error rendering.

Databricks Connect 18.0.1 (Scala)

April 28, 2026

The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.

Databricks Connect 18.0.0 (Scala)

December 11, 2025

The Databricks Connect client is updated to support Databricks Runtime 18.0 (EoS).

Databricks Connect for Databricks Runtime 17.3 LTS

Databricks Connect 17.3.12 (Python)

July 14, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.11 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.10 (Python)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.9 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.8 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, zstandard<1, numpy<3, databricks-sdk<1.

Databricks Connect 17.3.7 (Python)

March 3, 2026

Fixed an issue where Databricks Connect did not work in IPython shells.

Databricks Connect 17.3.6 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.5 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.4 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 17.3.3 (Python)

January 12, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.2 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 17.3.1 (Python)

November 5, 2025

The Py4J dependency requirement has been relaxed from py4j==0.10.9.9 to py4j>=0.10.9.7,<0.10.9.10 to ensure consistency with PySpark 3.5 and future releases. See SPARK-54015.

Databricks Connect 17.3.0 (Python)

October 28, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.3 LTS.

• Databricks Connect for Python support for serverless compute is now generally available (GA).
• Serverless sessions no longer expire after 10 minutes of inactivity.
• SPARK-53917 Large local relations (2GB+) are now supported.
• zstandard>=0.23.0 has been added as a dependency.

Databricks Connect 17.3.8 (Scala)

July 14, 2026

Fixed a thread-safety race condition in Arrow result deserialization that caused intermittent failures with an error ScalaReflectionException: <none> is not a module when query results were deserialized concurrently. See SPARK-57371.

Databricks Connect 17.3.7 (Scala)

June 24, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.6 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.5 (Scala)

April 23, 2026

The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.

Databricks Connect 17.3.4 (Scala)

March 10, 2026

Minor fixes and internal improvements.

Databricks Connect 17.3.2 (Scala)

January 6, 2026

Fixed a client-server syncing issue that could cause failures when executing UDFs on serverless compute.

Databricks Connect 17.3.1 (Scala)

December 6, 2025

When connecting to serverless compute, Databricks Connect now prints a warning instead of throwing an error if the Databricks Connect client version is lower than the serverless version. This ensures that Databricks Connect continues to work when the serverless environment is upgraded.

Databricks Connect 17.3.0 (Scala)

December 2, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.3 LTS.

• Serverless sessions no longer expire after 10 minutes of inactivity.
• SPARK-53917 Large local relations (2GB+) are now supported.

Databricks Connect for Databricks Runtime 17.2

Databricks Connect 17.2.10 (Python)

March 3, 2026

Minor fixes and internal improvements.

Databricks Connect 17.2.9 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 17.2.8 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 17.2.7 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 17.2.6 (Python)

January 12, 2026

Minor fixes and internal improvements.

Databricks Connect 17.2.5 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 17.2.4 (Python)

November 5, 2025

The Py4J dependency requirement has been relaxed from py4j==0.10.9.9 to py4j>=0.10.9.7,<0.10.9.10 to ensure consistency with PySpark 3.5 and future releases. See SPARK-54015.

Databricks Connect 17.2.3 (Python)

October 28, 2025

Minor fixes and internal improvements.

Databricks Connect 17.2.2 (Python)

September 29, 2025

Minor fixes and internal improvements.

Databricks Connect 17.2.1 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 17.2.0 (Python)

September 15, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.2 (EoS).

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.

Databricks Connect 17.2.3 (Scala)

January 6, 2026

Fixed a client-server syncing issue that could cause failures when executing UDFs on serverless compute.

Databricks Connect 17.2.2 (Scala)

December 6, 2025

When connecting to serverless compute, Databricks Connect now prints a warning instead of throwing an error if the Databricks Connect client version is lower than the serverless version. This ensures that Databricks Connect continues to work when the serverless environment is upgraded.

Databricks Connect 17.2.1 (Scala)

September 16, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.2 (EoS).

Databricks Connect for Databricks Runtime 17.1

Databricks Connect 17.1.10 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 17.1.9 (Python)

January 12, 2026

Minor fixes and internal improvements.

Databricks Connect 17.1.8 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.7 (Python)

November 5, 2025

The Py4J dependency requirement has been relaxed from py4j==0.10.9.9 to py4j>=0.10.9.7,<0.10.9.10 to ensure consistency with PySpark 3.5 and future releases. See SPARK-54015.

Databricks Connect 17.1.6 (Python)

October 28, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.5 (Python)

September 29, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.4 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.3 (Python)

September 15, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.2 (Python)

September 2, 2025

Minor fixes and internal improvements.

Databricks Connect 17.1.1 (Python)

August 18, 2025

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.
• Minor fixes and internal improvements.

Databricks Connect 17.1.0 (Python)

August 4, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.1 (EoS).

Databricks Connect 17.1.4 (Scala)

January 6, 2026

Fixed a client-server syncing issue that could cause failures when executing UDFs on serverless compute.

Databricks Connect 17.1.3 (Scala)

December 6, 2025

When connecting to serverless compute, Databricks Connect now prints a warning instead of throwing an error if the Databricks Connect client version is lower than the serverless version. This ensures that Databricks Connect continues to work when the serverless environment is upgraded.

Databricks Connect 17.1.1 (Scala)

July 28, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.1 (EoS).

Databricks Connect for Databricks Runtime 17.0

Databricks Connect 17.0.10 (Python)

November 5, 2025

The Py4J dependency requirement has been relaxed from py4j==0.10.9.9 to py4j>=0.10.9.7,<0.10.9.10 to ensure consistency with PySpark 3.5 and future releases. See SPARK-54015.

Databricks Connect 17.0.9 (Python)

October 28, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.8 (Python)

September 29, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.7 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.6 (Python)

September 15, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.5 (Python)

September 2, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.4 (Python)

August 18, 2025

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.
• Minor fixes and internal improvements.

Databricks Connect 17.0.3 (Python)

August 4, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.2 (Python)

July 23, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.1 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.0 (Python)

June 12, 2025

The Databricks Connect client is updated to support Databricks Runtime 17.0 (EoS).

Databricks Connect 17.0.5 (Scala)

January 6, 2026

Fixed a client-server syncing issue that could cause failures when executing UDFs on serverless compute.

Databricks Connect 17.0.4 (Scala)

December 6, 2025

When connecting to serverless compute, Databricks Connect now prints a warning instead of throwing an error if the Databricks Connect client version is lower than the serverless version. This ensures that Databricks Connect continues to work when the serverless environment is upgraded.

Databricks Connect 17.0.2 (Scala)

August 8, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.1 (Scala)

July 24, 2025

Minor fixes and internal improvements.

Databricks Connect 17.0.0 (Scala)

June 12, 2025

Features and improvements:

• The Databricks Connect client is updated to support Databricks Runtime 17.0 (EoS).

  Note

  All versions of Databricks Connect 17 (Scala) are only compatible with Spark 4 and Scala 2.13.

  To use this client with sbt, update the build.sbt file with Scala version 2.13 and declare the dependency:

  libraryDependencies += "com.databricks" %% "databricks-connect" % "17.0.+"
  

• Databricks Connect for Scala now supports connecting to serverless compute. This feature is available in Beta. See Configure a connection to serverless compute.

Databricks Connect for Databricks Runtime 16.4 LTS

Databricks Connect 16.4.21 (Python)

August 11, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.20 (Python)

July 14, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.19 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.18 (Python)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.17 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.16 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, databricks-sdk<1.

Databricks Connect 16.4.15 (Python)

March 3, 2026

Fixed an issue where Databricks Connect did not work in IPython shells.

Databricks Connect 16.4.14 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.13 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.12 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 16.4.11 (Python)

January 12, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.10 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.9 (Python)

November 5, 2025

The Py4J dependency requirement has been relaxed from py4j==0.10.9.9 to py4j>=0.10.9.7,<0.10.9.10 to ensure consistency with PySpark 3.5 and future releases. See SPARK-54015.

Databricks Connect 16.4.8 (Python)

October 28, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.7 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.6 (Python)

September 15, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.5 (Python)

September 2, 2025

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.
• Minor fixes and internal improvements.

Databricks Connect 16.4.4 (Python)

August 18, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.3 (Python)

July 23, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.2 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.1 (Python)

June 23, 2025

New features and improvements:

• When using serverless compute, UDFs are executed using Serverless environment. See Python base environment.

Databricks Connect 16.4.0 (Python)

May 15, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.4 LTS.

New features and improvements:

• UDFs with dependencies are now supported. See Manage UDF dependencies.

Databricks Connect 16.4.10 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.9 (Scala)

May 28, 2026

Minor fixes and internal improvements.

Databricks Connect 16.4.8 (Scala)

April 28, 2026

The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.

Databricks Connect 16.4.6 (Scala)

August 7, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.5 (Scala)

August 6, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.4 (Scala)

August 6, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.3 (Scala)

July 25, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.2 (Scala)

June 12, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.1 (Scala)

June 2, 2025

Minor fixes and internal improvements.

Databricks Connect 16.4.0 (Scala)

May 15, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.4 LTS.

New features and improvements:

• DatabricksSession.builder.getOrCreate() is now supported on both dedicated and standard clusters. See Databricks workspace behavior.

Databricks Connect for Databricks Runtime 16.3

Databricks Connect 16.3.5 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 16.3.4 (Python)

June 12, 2025

Minor fixes and internal improvements.

Databricks Connect 16.3.3 (Python)

May 20, 2025

Minor fixes and internal improvements.

Databricks Connect 16.3.2 (Python)

April 24, 2025

Minor fixes and internal improvements.

Databricks Connect 16.3.0 (Python)

March 20, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.3 (EoS).

Databricks Connect 16.3.1 (Scala)

June 2, 2025

Minor fixes and internal improvements.

Databricks Connect 16.3.0 (Scala)

March 31, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.3 (EoS).

New features and improvements:

• Session validation is turned on by default. This will validate that the Databricks Connect version and the Scala version are compatible with the Databricks Runtime version. This can be bypassed by setting validateSession(false).

Databricks Connect for Databricks Runtime 16.2

Databricks Connect 16.2.6 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 16.2.5 (Python)

June 12, 2025

Minor fixes and internal improvements.

Databricks Connect 16.2.4 (Python)

May 20, 2025

Minor fixes and internal improvements.

Databricks Connect 16.2.3 (Python)

April 24, 2025

Minor fixes and internal improvements.

Databricks Connect 16.2.2 (Python)

March 31, 2025

Bug fixes:

• Fixed a bug that caused Databricks Connect to fail while closing the Spark session.

Databricks Connect 16.2.1 (Python)

March 20, 2025

New features and improvements:

• A redundant instantiation of a new Spark session no longer occurs when SparkSession.getActiveSession is called.

Bug fixes:

• Fixed a segfault issue whenever to_pandas was called on an empty PyArrow table.

Databricks Connect 16.2.0 (Python)

February 18, 2025

• The Databricks Connect client is updated to support Databricks Runtime 16.2 (EoS).
• Databricks Connect Python 16.2 can only be installed in Python 3.12 environment.

Databricks Connect 16.2.0 (Scala)

February 18, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.2 (EoS).

Databricks Connect for Databricks Runtime 16.1

Databricks Connect 16.1.6 (Python)

June 12, 2025

Minor fixes and internal improvements.

Databricks Connect 16.1.5 (Python)

May 20, 2025

Minor fixes and internal improvements.

Databricks Connect 16.1.4 (Python)

April 24, 2025

Minor fixes and internal improvements.

Databricks Connect 16.1.3 (Python)

March 31, 2025

Bug fixes:

• Fixed a bug that caused Databricks Connect to fail while closing the Spark session.

Databricks Connect 16.1.2 (Python)

March 20, 2025

New features and improvements:

• A redundant instantiation of a new Spark session no longer occurs when SparkSession.getActiveSession is called.

Bug fixes:

• Fixed a segfault issue whenever to_pandas was called on an empty PyArrow table.

Databricks Connect 16.1.1 (Python)

February 18, 2025

• zstd_compress/zstd_decompress/try_zstd_decompress can now be imported by wildcard import, i.e. from pyspark.sql.functions import *.
• Fixed namespace collisions when importing multiple databricks Python packages from PyPI.

Databricks Connect 16.1.0 (Python)

January 27, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.1 (EoS).

Databricks Connect 16.1.2 (Scala)

February 18, 2025

Minor fixes and internal improvements.

Databricks Connect 16.1.2 (Scala)

February 5, 2025

• Fixed an issue that caused Databricks Connect to use PAT authentication instead of OAuth authentication.

Databricks Connect 16.1.1 (Scala)

January 27, 2025

Minor fixes and internal improvements.

Databricks Connect 16.1.0 (Scala)

January 8, 2025

The Databricks Connect client is updated to support Databricks Runtime 16.1 (EoS).

Databricks Connect for Databricks Runtime 16.0

Databricks Connect 16.0.4 (Python)

May 20, 2025

Minor fixes and internal improvements.

Databricks Connect 16.0.3 (Python)

March 20, 2025

New features and improvements:

• A redundant instantiation of a new Spark session no longer occurs when SparkSession.getActiveSession is called.

Bug fixes:

• Fixed a segfault issue whenever to_pandas was called on an empty PyArrow table.

Databricks Connect 16.0.2 (Python)

February 18, 2025

• zstd_compress/zstd_decompress/try_zstd_decompress can now be imported by wildcard import, i.e. from pyspark.sql.functions import *.
• Fixed namespace collisions when importing multiple databricks Python packages from PyPI.

Databricks Connect 16.0.1 (Python)

January 27, 2025

• Error retry behavior has been improved.

Databricks Connect 16.0.0 (Python)

November 29, 2024

This release of Databricks Connect for Python supports Databricks Runtime 16.0 (EoS).

• Added DatabricksSession.builder.create() API which always creates a new Spark session. This API is disabled by default in notebooks. For more information, see Databricks workspace behavior.

Databricks Connect 16.0.4 (Scala)

February 18, 2025

Minor fixes and internal improvements.

Databricks Connect 16.0.3 (Scala)

February 5, 2025

• Fixed an issue that caused Databricks Connect to use PAT authentication instead of OAuth authentication.

Databricks Connect 16.0.2 (Scala)

January 27, 2025

Minor fixes and internal improvements.

Databricks Connect 16.0.1 (Scala)

January 8, 2025

• Error retry behavior has been improved.

Databricks Connect 16.0.0 (Scala)

December 3, 2024

The Databricks Connect client is updated to support Databricks Runtime 16.0 (EoS).

Databricks Connect for Databricks Runtime 15.4 LTS

Databricks Connect 15.4.25 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.24 (Python)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.23 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.22 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, databricks-sdk<1.

Databricks Connect 15.4.21 (Python)

March 3, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.20 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.19 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.18 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 15.4.17 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.16 (Python)

October 28, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.15 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.14 (Python)

September 15, 2025

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.
• Minor fixes and internal improvements.

Databricks Connect 15.4.13 (Python)

September 2, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.12 (Python)

July 23, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.11 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.10 (Python)

June 23, 2025

New features and improvements:

• When using serverless compute, UDFs are executed using Serverless environment. See Python base environment.

Databricks Connect 15.4.9 (Python)

May 20, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.8 (Python)

March 20, 2025

New features and improvements:

• A redundant instantiation of a new Spark session no longer occurs when SparkSession.getActiveSession is called.

Bug fixes:

• Fixed a segfault issue whenever to_pandas was called on an empty PyArrow table.

Databricks Connect 15.4.6 (Python)

February 18, 2025

• zstd_compress/zstd_decompress/try_zstd_decompress can now be imported by wildcard import, i.e. from pyspark.sql.functions import *.
• Fixed namespace collisions when importing multiple databricks Python packages from PyPI.

Databricks Connect 15.4.5 (Python)

January 27, 2025

• Error retry behavior has been improved.

Databricks Connect 15.4.4 (Python)

November 29, 2024

• Fixed an issue where df.write.parquet did not correctly recognize the gzip compression option.

Databricks Connect 15.4.3 (Python)

October 31, 2024

Minor fixes and internal improvements.

Databricks Connect 15.4.2 (Python)

September 13, 2024

• Fixed a bug where error messages were not complete. Error messages now include the stack trace from Spark if it is relevant.

Databricks Connect 15.4.1 (Python)

September 2, 2024

• validateSession() and the databricks-connect test command now fail when the Databricks Connect version is incompatible with the Databricks serverless compute version. For information about Databricks Connect serverless support, see Configure a connection to serverless compute.

Databricks Connect 15.4.0 (Python)

August 27, 2024

This release of Databricks Connect for Python supports Databricks Runtime 15.4 LTS.

Databricks Connect 15.4.9 (Scala)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.8 (Scala)

May 28, 2026

Minor fixes and internal improvements.

Databricks Connect 15.4.7 (Scala)

April 28, 2026

The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.

Databricks Connect 15.4.4 (Scala)

February 18, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.3 (Scala)

February 5, 2025

• Fixed an issue that caused Databricks Connect to use PAT authentication instead of OAuth authentication.

Databricks Connect 15.4.2 (Scala)

January 27, 2025

Minor fixes and internal improvements.

Databricks Connect 15.4.1 (Scala)

January 8, 2025

• Error retry behavior has been improved.

Databricks Connect 15.4.0 (Scala)

September 3, 2024

This release of Databricks Connect for Scala supports Databricks Runtime 15.4 LTS.

Databricks Connect for Databricks Runtime 15.3

Databricks Connect 15.3.6 (Python)

January 8, 2025

• Error retry behavior has been improved.
• Fixed an issue where df.write.parquet did not correctly recognize the gzip compression option.

Databricks Connect 15.3.5 (Python)

October 31, 2024

Minor fixes and internal improvements.

Databricks Connect 15.3.4 (Python)

September 13, 2024

• Fixed a bug where error messages were not complete. Error messages now include the stack trace from Spark if it is relevant.

Databricks Connect 15.3.3 (Python)

September 2, 2024

• validateSession() and the databricks-connect test command now fail when the Databricks Connect version is incompatible with the Databricks serverless compute version. For information about Databricks Connect serverless support, see Configure a connection to serverless compute.

Databricks Connect 15.3.2 (Python)

August 27, 2024

• A warning now occurs when the Databricks Connect version is incompatible with the serverless version.
• Improved error message when the local Python version does not match the serverless version.

Databricks Connect 15.3.1 (Python)

July 18, 2024

• Serverless Spark Connect sessions expire after 10 minutes of inactivity. Databricks Connect now automatically closes expired sessions on the client side. Subsequent calls to DatabricksSession.builder.getOrCreate() will return a new active connect session.

Databricks Connect 15.3.0 (Python)

July 15, 2024

• Added a new function timestamp_add for adding the specified number of units to the given timestamp. See SPARK-48369.
• Added a new function timestamp_diff for getting the difference between the timestamps in the specified units by truncating the fraction. See SPARK-48272.
• Added a new function try_parse_json, which attempts to parse a JSON string and construct a Variant value. If the attempt fails, the result is null. See SPARK-47922.
• Changed the default value for spark.sql.legacy.timeParserPolicy from EXCEPTION to CORRECTED. Instead of raising an INCONSISTENT_BEHAVIOR_CROSS_VERSION error, CANNOT_PARSE_TIMESTAMP will be raised if ANSI mode is enabled. NULL will be returned if ANSI mode is disabled. See SPARK-47719.

Databricks Connect 15.3.1 (Scala)

September 3, 2024

Minor internal fixes and improvements.

Databricks Connect 15.3.0 (Scala)

July 22, 2024

Features and improvements:

• Supported Dataset.observe for getting stats of Columns of a DataFrame. See SPARK-47545.
• Added a new function timestamp_add for adding the specified number of units to the given timestamp. See SPARK-48369.
• Added a new function timestamp_diff for getting the difference between the timestamps in the specified units by truncating the fraction. See SPARK-48272.
• Added a new function try_mod for computing the remainder of the division of two numbers, which results in null on division by zero. See SPARK-48954.
• Added a new function try_parse_json, which attempts to parse a JSON string and construct a Variant value. If the attempt fails, the result is null. See SPARK-47922.
• Changed the default value for spark.sql.legacy.timeParserPolicy from EXCEPTION to CORRECTED. Instead of raising an INCONSISTENT_BEHAVIOR_CROSS_VERSION error, CANNOT_PARSE_TIMESTAMP will be raised if ANSI mode is enabled. NULL will be returned if ANSI mode is disabled. See SPARK-47719.

Databricks Connect for Databricks Runtime 15.2

Databricks Connect 15.2.5 (Python)

September 26, 2024

Minor fixes and internal improvements.

Databricks Connect 15.2.4 (Python)

September 13, 2024

• Fixed a bug where error messages were not complete. Error messages now include the stack trace from Spark if it is relevant.

Databricks Connect 15.2.3 (Python)

September 2, 2024

• validateSession() and the databricks-connect test command now fail when the Databricks Connect version is incompatible with the Databricks serverless compute version. For information about Databricks Connect serverless support, see Configure a connection to serverless compute.

Databricks Connect 15.2.2 (Python)

August 27, 2024

• A warning now occurs when the Databricks Connect version is incompatible with the serverless version.
• Improved error message when the local Python version does not match the serverless version.

Databricks Connect 15.2.1 (Python)

July 18, 2024

• Serverless spark connect sessions expire after 10 minutes of inactivity. Databricks Connect now automatically closes expired sessions on the client side. Subsequent calls to DatabricksSession.builder.getOrCreate() will return a new active connect session.

Databricks Connect 15.2.0 (Python)

July 15, 2024

• The split function now supports column type. See SPARK-47845.
• Added new functions is_variant_null, variant_get, try_variant_get, schema_of_variant, and schema_of_variant_agg to improve the usability of the Variant type. See SPARK-47890.
• Added new functions collate and collation to access collation capabilities. See SPARK-47395.
• Extensions are now allowed to log extended information in the explain plan. See SPARK-47289.

Databricks Connect 15.2.1 (Scala)

September 3, 2024

Minor internal fixes and improvements.

Databricks Connect 15.2.0 (Scala)

July 22, 2024

• The split function now supports column type. See SPARK-47845.
• Added new functions is_variant_null, variant_get, try_variant_get, schema_of_variant, and schema_of_variant_agg to improve the usability of the Variant type. See SPARK-47890.
• Added new functions collate and collation to access collation capabilities. See SPARK-47395.
• Extensions are now allowed to log extended information in the explain plan. See SPARK-47289.

Databricks Connect for Databricks Runtime 15.1

Databricks Connect 15.1.3 (Python)

September 13, 2024

• Fixed a bug where error messages were not complete. Error messages now include the stack trace from Spark if it is relevant.

Databricks Connect 15.1.2 (Python)

September 2, 2024

• Improved error message when the local Python version does not match the Python version of the serverless release.

Databricks Connect 15.1.1 (Python)

July 18, 2024

• Serverless spark connect sessions expire after 10 minutes of inactivity. Databricks Connect now automatically closes expired sessions on the client side. Subsequent calls to DatabricksSession.builder.getOrCreate() will return a new active connect session.
• Supported Python versions are now limited to 3.10 and 3.11, because the distutils dependency has been removed in Python 3.12.

Databricks Connect 15.1.0 (Python)

July 15, 2024

Features and improvements:

• Databricks Connect for Python now supports connecting to serverless compute. This feature is available in Public Preview. See Configure a connection to serverless compute.
• Added a new function try_cast, which attempts to cast to a different data type. If the attempt fails, the result is null. See SPARK-47669.
• Added a new function monthname that returns the three-letter abbreviated month name from a given date or timestamp. See SPARK-46515.

Bug fixes:

• Session stop() API is idempotent now and will not fail if the session is already closed by the server. See SPARK-48810.
• Supported numpy versions are now limited to 1.15<=numpy<2.0 because PySpark 3.5.0 is incompatible with numpy 2.0.

Databricks Connect 15.1.1 (Scala)

September 3, 2024

Minor internal fixes and improvements.

Databricks Connect 15.1.0 (Scala)

July 22, 2024

Features and improvements:

• Added a new function try_cast, which attempts to cast to a different data type. If the attempt fails, the result is null. See SPARK-47669.
• Added a new function monthname that returns the three-letter abbreviated month name from a given date or timestamp. See SPARK-46515.

Bug fixes:

• Session stop() API is idempotent now and will not fail if the session is already closed by the server. See SPARK-48810.
• The bloomFilter function now throws an AnalysisException for invalid input. See SPARK-46787.

Databricks Connect for Databricks Runtime 14.3 LTS

Databricks Connect 14.3.22 (Python)

June 23, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.21 (Python)

June 16, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.20 (Python)

May 27, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.19 (Python)

April 30, 2026

• The client now automatically retries transient errors that occur when an intermediary proxy returns a non-gRPC response (for example, HTTP 404 from a brief ngrok tunnel drop). The retry attempts to resume the in-progress operation without re-executing it, so non-idempotent queries are not duplicated. Previously these transient routing failures surfaced as hard failures.
• Dependencies are now pinned to major versions to prevent unexpected breaking changes from upstream package updates: pyarrow<25, grpcio<2, grpcio-status<2, googleapis-common-protos<2, databricks-sdk<1.

Databricks Connect 14.3.18 (Python)

March 3, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.17 (Python)

February 26, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.16 (Python)

February 10, 2026

Minor fixes and internal improvements.

Databricks Connect 14.3.15 (Python)

January 27, 2026

Supported pandas versions are now limited to 1.0.5<=pandas<3 to prevent the pandas 3.0.0 release from breaking pyspark.pandas workloads.

Databricks Connect 14.3.14 (Python)

December 8, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.13 (Python)

September 22, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.12 (Python)

September 15, 2025

• Removed RetriesExceeded error from the Spark Connect Python client error handling logic. The client now prints a warning and throws the original exception after retries are exhausted. See SPARK-53307.
• Minor fixes and internal improvements.

Databricks Connect 14.3.11 (Python)

September 2, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.10 (Python)

July 11, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.9 (Python)

June 23, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.8 (Python)

March 20, 2025

• Minor fixes and internal improvements

Databricks Connect 14.3.7 (Python)

January 27, 2025

Minor fixes and internal improvements.

Databricks Connect 14.3.6 (Python)

January 8, 2025

• Error retry behavior has been improved.
• Fixed an issue where df.write.parquet did not correctly recognize the gzip compression option.

Databricks Connect 14.3.5 (Python)

November 29, 2024

Minor fixes and internal improvements.

Databricks Connect 14.3.4 (Python)

October 29, 2024

Minor fixes and internal improvements.

Databricks Connect 14.3.3 (Python)

September 11, 2024

• Calling spark.stop() on an already stopped or expired session will no longer fail.
• Fixed a bug where error messages were not complete. Error messages now include the stack trace from Spark if it is relevant.

Databricks Connect 14.3.2 (Python)

May 8, 2024

• Automatically re-try plan execution in some cases of SESSION_NOT_FOUND error from the service.
• Subsequent calls to SparkSession.builder.getOrCreate() after DatabricksSession.builder.getOrCreate() will return the active default connect session.

Databricks Connect 14.3.1 (Python)

February 22, 2024

Minor fixes and internal improvements.

Databricks Connect 14.3.0 (Python)

February 1, 2024

Features and improvements:

• Databricks Connect now ships with the databricks-connect binary. Users can now validate their Python environment and the connection to the Azure Databricks cluster by running: databricks-connect test.

Bug fixes:

• Limit the number of retries when Spark Connect requests fails to the service.
• spark.table() now accepts Delta table time travel syntax.
• Aggregate functions run on top of df.melt() now work. Seehttps://issues.apache.org/jira/browse/SPARK-46484.
• DataFrame joins using queries that contain common table expressions (CTE) now work. Seehttps://issues.apache.org/jira/browse/SPARK-46337.
• df.withColumnsRenamed() now retains original dictionary ordering. Seehttps://issues.apache.org/jira/browse/SPARK-46261.
• df.withColumn() now accepts an argument of Column type. Seehttps://issues.apache.org/jira/browse/SPARK-46028.
• The behavior of df.column() in Spark Connect is now more aligned with the behavior in classic Apache Spark. Seehttps://issues.apache.org/jira/browse/SPARK-45509.
• Referencing the join column in