Snowflake Python APIs release notes for 2026¶

This article contains the release notes for the Snowflake Python APIs, including the following when applicable:

Version 1.13.1 (Aug 12, 2026)¶

Code Bundle API:
- Added typed BundleSpec and CodeBundleSpecification models for code bundles and code bundle executions.
- Added iter() support for code bundle executions.
Dynamic Table API:
- Added initialization_warehouse, frozen_where, backfill_from, default_ddl_collation, log_level, row_timestamp, cluster_by_raw, and start_at fields to DynamicTable.
- Added DataMetricSchedule, DataMetricCronSchedule, DataMetricMinutesSchedule, and DataMetricTriggerOnChangesSchedule models for data metric scheduling.
- Extended refresh_mode with ADAPTIVE and CUSTOM_INCREMENTAL values.

Task and DAG APIs:
- Added condition parameter to DAG to support setting a WHEN condition on the root task.
- Added the recursive argument to TaskResource.fetch_task_dependents.
- Added support for defining task schedules and target completion intervals at a granularity of seconds, including intervals under a minute.
- Added overlap_policy on Task and on the DAG root task.
- Added success_integration and execute_as_user on Task and DAGTask.
Code Bundle API:
- Added support for managing code bundles through root.databases[<db>].schemas[<schema>].code_bundles (create, iterate, fetch, add a version, execute, and drop).
- Added support for code bundle executions through root.code_bundle_execution (execute a code bundle from a stage location, and fetch the status of or cancel an execution).
- Added the _SNOWFLAKE_SKIP_ASYNC_EXEC_POLLING environment variable. When set to true, a request submitted with asyncExec=true that receives a 202 Accepted response returns that accepted response (carrying the job id) immediately, instead of polling the execution to completion. Transient-error retries, for example, 429/503/504, are unaffected.

copy.copy, copy.deepcopy, and pickling of IntervalSchedule and TargetCompletionInterval now work; previously they would raise an exception.
Security improvements - for more information, see CVE-2025-24793 (https://www.cve.org/CVERecord?id=CVE-2026-19594):
- Improved validation of path parameters; values consisting solely of dot-segments (. or ..) now raise a clear ValueError.
- Improved encoding of URL query parameters; resource names containing special characters, such as &, =, or #, are now percent-encoded correctly, preventing failures on rename, swap, and clone operations for resources whose names include such characters.
- Fixed DagRun._repr_html_() to correctly escape special characters in property values.
- Improved validation of single-quoted string literals used in stage paths; strings with unescaped embedded quotes are no longer incorrectly treated as valid.
- Improved sanitization of stage path values; backslash sequences immediately preceding a single quote are now doubled.

Proxy configuration through environment variables now honors both the uppercase HTTPS_PROXY and the lowercase https_proxy variable.
The CA bundle used for TLS verification now honors the SSL_CERT_FILE environment variable (both uppercase SSL_CERT_FILE and lowercase ssl_cert_file), allowing connections through corporate proxies with a custom trust store.

Added support for setting (set_tags), unsetting (unset_tags), and fetching tag assignments (get_tags). Tagging support for specific resources is introduced in the following Snowflake server releases:
- 10.3: Alert, database, database role, dynamic table, event table, image repository, network policy, notebook, password policy, pipe, procedure, role, schema, stream, table, task, user, user-defined function, view, warehouse.
- 10.4: API integration, catalog integration, compute pool, function, iceberg table, notification integration, Streamlit.

The DAGTask object type now accepts custom objects with a to_sql() method as task definitions.
The UserDefinedFunction object type now supports executing scalar UDFs using the execute method.

Creating, fetching, and listing stored procedures that use a staged handler (where the body property is empty) no longer raises a ValidationError.
Pydantic deprecation warnings related to class-based config and the update_forward_refs, parse_obj, and _iter methods no longer occur.