v4.2.0

🎉 New Features

• Added support for ApplicationField.creator().
• Added support for CredentialClient.creator() and CredentialClient.purge_by_guid().
• Added support for optional parameters (host and port) in Connection.creator().

⛑️ Breaking Changes

• Changed the default setting for tracking IndexSearch logs (save_search_log=False). Previously, it was set to True, which was causing frequent Out of Memory (OOM) issues in Metastore pods.

🥗 QOL Improvements

• Generated the latest typedef models.
• Refactored OpenLineageClient.create_connection() to use the new CredentialClient.creator().

Full Changelog: 4.1.0...4.2.0

v4.1.0

🎉 New Features

• Added support for the following workflow packages:

  Connectors

  • OracleCrawler

  Utilities

  • LineageBuilder
  • LineageGenerator
  • APITokenConnectionAdmin

• Extended the WorkflowClient.run() method to accept raw workflow JSON strings.

🐞 Bug Fixes

• Updated the maximum API token Time-to-Expire (TTE) to 5 years. Previously, the value was set to 13 years (409,968,000 seconds), which was reverted to 5 years due to an integer overflow issue in Keycloak. For more details, see Keycloak Issue #19671.

🥗 QOL Improvements

• Added support for Python 3.10, 3.11, 3.12 and 3.13.
• Increased the default read timeout for AtlanClient to 900 seconds (15 minutes).

Full Changelog: 4.0.2...4.1.0

v4.0.2

🎉 New Features

• Added the ability to the custom package logger to log unhandled exceptions.
• Added support for creating OpenLineage connections (OpenLineageClient.create_connection()).

🐞 Bug Fixes

• Fixed handling of deleted AtlanTagNames in Purpose asset deserialization. Previously, a ValueError was raised.

🥗 QOL Improvements

• Added an "Installing for development" section to the README.md.
• Removed pyatlan-codeql.yaml in favor of using the organization-level CodeQL workflow to avoid configuration conflicts and failures.

Full Changelog: 4.0.1...4.0.2

v4.0.1

🎉 New Features

• Added a new connector type (CUSTOM).
• Added support for the DatabricksCrawler and DatabricksMiner workflow packages.
• Added the creator() method for the following assets:
  • Custom
  • Dataverse
  • TablePartition

🐞 Bug Fixes

• Fixed _user_id handling issue in pyatlan.pkg.utils.get_client().

🥗 QOL Improvements

• Enhanced test coverage with additional unit tests for the append_terms, replace_terms, and remove_terms methods.

❌ Deprecated Changes

• table_name and table_qualified_name in Column.creator() now emit a DeprecationWarning. Users are advised to use parent_name and parent_qualified_name instead.

Full Changelog: 4.0.0...4.0.1

v3.1.3

🐞 Bug Fixes

• Fixed _user_id handling issue in pyatlan.pkg.utils.get_client().

v4.0.0

🎉 New Features

• Added support for the MongoDBCrawler workflow package.
• Added creator() and updater() methods for the Procedure asset.

🐞 Bug Fixes

• Fixed WorkflowClient.find_by_type() method to use regular expressions.
• Fixed AttributeDef.cardinality and AttributeDef.type_name handling for the new multi_value_select attribute option.
• Fixed issues with the AssetClient.append_terms(), remove_terms(), and replace_terms() methods:
  • These methods now use SaveSemantic.APPEND/REPLACE/REMOVE, which is more optimized and faster than the previous approach that required full asset retrieval with relationships for updates.
• Fixed S3Object.create_with_prefix() and creator() to ensure the s3_bucket_name field is correctly set.

🥗 QOL Improvements

• Added an OSV vulnerability-scan workflow job to GitHub Actions.

⛑️ Breaking Changes

• The default behavior of AssetClient.get_by_guid() and AssetClient.get_by_qualified_name() has changed:

  • By default, these methods now ignore relationships during asset retrieval (ignore_relationships: bool = True). Previously, this was set to False.
  • This change is intentional, aiming to prevent users from retrieving more information than necessary, thereby reducing the overall runtime of requests.
  • Users can now use the attributes and related_attributes optional parameters to specify the exact details required for each search result. This ensures precise and efficient searches. See: Advanced Examples - Read.

  # In this example, we are retrieving the "userDescription" attribute
  # for both the glossary and its terms (related attribute).
  # You can also retrieve other attributes as illustrated below:
  
  glossary = client.asset.get_by_guid(
        guid="b4113341-251b-4adc-81fb-2420501c30e6",
        asset_type=AtlasGlossary,
        min_ext_info=False,
        ignore_relationships=True,
        attributes=[AtlasGlossary.USER_DESCRIPTION, AtlasGlossary.TERMS],
        related_attributes=[AtlasGlossaryTerm.USER_DESCRIPTION]
    )
  
  glossary = client.asset.get_by_qualified_name(
      asset_type=AtlasGlossary,
      qualified_name="pXkf3RUvsIOIG8xnn0W3O",
      min_ext_info=False,
      ignore_relationships=True,
      attributes=[AtlasGlossary.USER_DESCRIPTION, AtlasGlossary.TERMS],
      related_attributes=[AtlasGlossaryTerm.USER_DESCRIPTION]
  )

Full Changelog: 3.1.2...4.0.0

v3.1.2

🎉 New features

• Enabled the use of the Retry-After header for handling rate-limit retries.
• Added support for OpenTelemetry logging in custom packages.
• Added creator() methods for Anaplan assets.
• Added creator() methods for Insights assets (Collection, Folder, Query).
• Added support for the following new connector types:
  • ANAPLAN
  • AWS_ECS
  • AWS_BATCH
  • AWS_LAMBDA
  • AWS_SAGEMAKER
  • DATAVERSE

🥗 QOL improvements

• Generated the latest typedef models.
• Upgraded jinja2 from 3.1.4 to 3.1.5 to address a security vulnerability.
• Fixed pagination assertions in AuditSearch unit tests and integration tests.

Full Changelog: 3.1.1...3.1.2

v3.1.1

🐞 Bug fixes

• Fixed a ValidationError caused by the private field (_user_id) in the AtlanClient constructor.

Full Changelog: 3.1.0...3.1.1

v3.1.0

🎉 New features

• Added support for automatic token refresh and retrying API requests upon receiving a 401 (Unauthorized) response.

  To enable this feature, the following constants must be configured:

  1. Environment variables:

  For regenerating the bearer access token:

  • CLIENT_ID (string)
  • CLIENT_SECRET (string)

  2. Update AtlanClient field:

  • _user_id (string, default: None): The unique identifier (GUID) of the user that the client impersonates.

  Example:

  client = AtlanClient()
  client._user_id = "962c8f78-98a7-908f-9ec2-9e5b7ee7a09f"

⛑️ Breaking changes

• Introduced a new pagination approach in SearchLogClient.search() called search log bulk search (disabled by default). The SDK switches to this search operation automatically if results exceed a predefined threshold (e.g: 10,000 results). Alternatively, users can enable bulk search explicitly by setting bulk=True in SearchLogClient.search(). This breaking change affects searches that return more than 10,000 results — either the results will now be sorted differently or an error will be thrown.

• SearchLogClient.search() method will now raise an InvalidRequestError exception in the following scenarios:

  • when bulk search is enabled (bulk=True) and any user-specified sorting options are included in the search request.
  • when bulk search is disabled (bulk=False), the number of results exceeds the predefined threshold (e.g: 10,000 assets), and any user-specified sorting options are found in the search request.

  This is because the bulk search approach for search logs ignores user-specified sorting and instead reorders results by the createdAt timestamps of log entries to efficiently handle large volumes of search logs.

Full Changelog: 3.0.0...3.1.0

v3.0.0

🎉 New features

• Added a new connector type (BIGID).
• Added support for the following options in Batch operations:
  • update_only: bool (default: False)
  • track: bool (default: False)
  • case_insensitive: bool (default: False)
  • table_view_agnostic: bool (default: False)
  • creation_handling: AssetCreationHandling (default: AssetCreationHandling.FULL)
• Added default timeouts (read, connect) to AtlanClient:
  • Total default retries: int: 5
  • AtlanClient.connect_timeout: float (default: 30.0 seconds)
  • AtlanClient.read_timeout: float (default: 120.0 seconds)
• Added support for a new parent_type (SnowflakeDynamicTable) in Column.creator().
• Added exposure for source-specific custom attributes (e.g: Asset.custom_attributes).
• Added handling for error_cause and backend_error_id in ErrorInfo.

🐞 Bug fixes

• Fixed the generator to correctly handle the naming of the Asset.DOMAIN_GUIDS keyword search field.
• Fixed an issue where search pages (IndexSearchResults and AuditSearchResults) could overrun when the total results are just under the _MASS_EXTRACT_THRESHOLD.
• Fixed an issue with timestamp paging returning incomplete results when searching with a small page size (e.g: 2) and assets with the same creation time.

🥗 QOL improvements

• Generated the latest typedef models.
• Added the @init_guid decorator to the updater() method of assets to ensure that GUIDs are properly initialized and resolved in batch operations.
• Removed type_name validation from Table, View, and Materialised View to make them configurable when running Batch operations with table_view_agnostic=True.
• Removed deprecated AssetClient.get_lineage() integration tests.
• Updated integration test asset constants to align with the new tenant setup.

❌ Deprecated features

• Removed the deprecated AssetClient.get_lineage() method, which is slower and will no longer receive enhancements. Use the AssetClient.get_lineage_list() method instead.

⛑️ Breaking changes

• Batch now accepts AtlanClient as the first parameter, replacing the previous use of AssetClient.

• Introduced a new pagination approach in AuditClient.search() called audit bulk search (disabled by default). The SDK switches to this search operation automatically if results exceed a predefined threshold (e.g: 10,000 results). Alternatively, users can enable bulk search explicitly by setting bulk=True in AuditClient.search(). This breaking change affects searches that return more than 10,000 results — either the results will now be sorted differently or an error will be thrown.

• AuditClient.search() method will now raise an InvalidRequestError exception in the following scenarios:

  • when bulk search is enabled (bulk=True) and any user-specified sorting options are included in the search request.
  • when bulk search is disabled (bulk=False), the number of results exceeds the predefined threshold (e.g: 10,000 assets), and any user-specified sorting options are found in the search request.

  This is because the audit bulk search approach ignores user-specified sorting and instead reorders results based on the creation timestamps of assets to handle large volumes of assets efficiently.

Full Changelog: 2.7.0...3.0.0