snowflake-ml-python 2.2.0__tar.gz → 2.4.0__tar.gz

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/CHANGELOG.md

@@ -1,6 +1,62 @@
 # Release History
 
-## 2.2.0
+## 2.4.0
+
+### New Features
+
+* Generic: Support Python 3.14.
+
+* Registry: `FeatureSpec(dtype=DataType.OBJECT)` now accepts schema-free Python dictionaries as a
+  model input or output (SQL `OBJECT`). Callers can pass a dict to `ModelVersion.run`, an object
+  literal in SQL, or a JSON object over REST, without stringifying to JSON. Use `shape=(-1,)` for
+  a list of dictionaries (`ARRAY(OBJECT)`). Closed structs with a fixed field list remain
+  `FeatureGroupSpec`.
+
+* Experiment Tracking: Added `snowflake.ml.experiment.callback.catboost.SnowflakeCatboostCallback`, a
+  CatBoost training callback that logs metrics and, optionally, parameters to a Snowflake experiment.
+  Pass `params=model.get_params()` to log parameters. Model logging is handled separately after
+  training via ``ExperimentTracking.log_model``.
+
+* Experiment Tracking: `ExperimentTracking` accepts an optional `warnings_as_errors` parameter. When True,
+  conditions that are normally reported as warnings are raised instead.
+
+### Bug Fixes
+
+### Behavior Changes
+
+* Experiment Tracking: Re-constructing the `ExperimentTracking` singleton now logs an informational
+  message instead of emitting a `UserWarning`.
+
+* Registry: `log_model` no longer raises when `options={'relax_version': True}` is used with pip
+  requirements or Snowpark Container Services-only targets. The option is ignored, a warning is
+  emitted, and exact pip version specifications are kept.
+
+### Deprecations
+
+## 2.3.0 (2026-09-28)
+
+### New Features
+
+* Experiment Tracking: `ExperimentTracking` accepts an optional `model_registry` parameter. When provided,
+  `log_model` stores models in the registry's database and schema instead of the experiment's, so experiments
+  and models can live in separate schemas.
+
+### Bug Fixes
+
+* Feature Store: Offline reads of a feature view with several list aggregations or several secondary-key
+  aggregation windows used to return incomplete results in some scenarios. The bug has been fixed
+  and the existing feature views pick up the fix on their next read, with no re-registration or backfill required.
+
+### Behavior Changes
+
+* Feature Store: `update_feature_view()` now rejects a `target_lag` change on a feature view whose online store
+  is `POSTGRES`, instead of reporting success for a change that is not applied. Set `target_lag` when online
+  storage is first enabled, or delete and re-register the feature view to change it. Enabling or disabling
+  online storage, and `target_lag` changes on hybrid-table online stores, are unaffected.
+
+### Deprecations
+
+## 2.2.0 (2026-09-21)
 
 ### New Features
 
@@ -14,8 +70,17 @@
 
 ### Bug Fixes
 
+* Registry: `ModelVersion.load()` now compares the model owner to `CURRENT_ROLE()` (the role that
+  executes SQL) instead of `Session.get_current_role()`. Owner's-rights Streamlit apps and
+  `EXECUTE AS OWNER` procedures can load when statements run as the owner even if the caller's
+  primary role is not the owner.
+
 ### Behavior Changes
 
+* Dependencies: `scipy` is no longer a required dependency of `snowflake-ml-python`. It is still
+  installed transitively with `pip install snowflake-ml-python[scikit-learn]`. Direct `scipy` usage
+  in `snowflake.ml.modeling.preprocessing` already requires that extra.
+
 ### Deprecations
 
 ## 2.1.0 (2026-09-14)
@@ -37,21 +102,12 @@
 
 ### Bug Fixes
 
-* Registry: `ModelVersion.load()` now compares the model owner to `CURRENT_ROLE()` (the role that
-  executes SQL) instead of `Session.get_current_role()`. Owner's-rights Streamlit apps and
-  `EXECUTE AS OWNER` procedures can load when statements run as the owner even if the caller's
-  primary role is not the owner.
-
 * Registry: Fixed logging of MLflow models created with `mlflow.sklearn.save_model()`. The model is
   now re-logged from the scikit-learn estimator using the serialization format it was saved with,
   instead of from MLflow's PyFunc wrapper with the version-dependent default format.
 
 ### Behavior Changes
 
-* Dependencies: `scipy` is no longer a required dependency of `snowflake-ml-python`. It is still
-  installed transitively with `pip install snowflake-ml-python[scikit-learn]`. Direct `scipy` usage
-  in `snowflake.ml.modeling.preprocessing` already requires that extra.
-
 ### Deprecations
 
 ## 2.0.0 (2026-09-09)

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: snowflake-ml-python
-Version: 2.2.0
+Version: 2.4.0
 Summary: The machine learning client library that is used for interacting with Snowflake to build machine learning solutions.
 Author-email: "Snowflake, Inc" <support@snowflake.com>
 License: 
@@ -223,13 +223,15 @@ Classifier: Operating System :: OS Independent
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
 Classifier: Topic :: Database
 Classifier: Topic :: Software Development
 Classifier: Topic :: Software Development :: Libraries
 Classifier: Topic :: Software Development :: Libraries :: Application Frameworks
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
 Classifier: Topic :: Scientific/Engineering :: Information Analysis
-Requires-Python: <3.14,>=3.10
+Requires-Python: <3.15,>=3.10
 Description-Content-Type: text/markdown
 License-File: LICENSE.txt
 Requires-Dist: anyio<5,>=3.5.0
@@ -378,7 +380,7 @@ In Snowflake Warehouse notebook environments, snowflake-ml-python can be install
 Follow the [installation instructions](https://docs.snowflake.com/en/developer-guide/snowpark-ml/index#installing-snowpark-ml)
 in the Snowflake documentation.
 
-Python versions 3.10 to 3.13 are supported. You can use [miniconda](https://docs.conda.io/en/latest/miniconda.html) or
+Python versions 3.10 to 3.14 are supported. You can use [miniconda](https://docs.conda.io/en/latest/miniconda.html) or
 [anaconda](https://www.anaconda.com/) to create a Conda environment (recommended),
 or [virtualenv](https://docs.python.org/3/tutorial/venv.html) to create a virtual environment.
 
@@ -424,7 +426,63 @@ NOTE: Version 1.7.0 is used as example here. Please choose the the latest versio
 
 # Release History
 
-## 2.2.0
+## 2.4.0
+
+### New Features
+
+* Generic: Support Python 3.14.
+
+* Registry: `FeatureSpec(dtype=DataType.OBJECT)` now accepts schema-free Python dictionaries as a
+  model input or output (SQL `OBJECT`). Callers can pass a dict to `ModelVersion.run`, an object
+  literal in SQL, or a JSON object over REST, without stringifying to JSON. Use `shape=(-1,)` for
+  a list of dictionaries (`ARRAY(OBJECT)`). Closed structs with a fixed field list remain
+  `FeatureGroupSpec`.
+
+* Experiment Tracking: Added `snowflake.ml.experiment.callback.catboost.SnowflakeCatboostCallback`, a
+  CatBoost training callback that logs metrics and, optionally, parameters to a Snowflake experiment.
+  Pass `params=model.get_params()` to log parameters. Model logging is handled separately after
+  training via ``ExperimentTracking.log_model``.
+
+* Experiment Tracking: `ExperimentTracking` accepts an optional `warnings_as_errors` parameter. When True,
+  conditions that are normally reported as warnings are raised instead.
+
+### Bug Fixes
+
+### Behavior Changes
+
+* Experiment Tracking: Re-constructing the `ExperimentTracking` singleton now logs an informational
+  message instead of emitting a `UserWarning`.
+
+* Registry: `log_model` no longer raises when `options={'relax_version': True}` is used with pip
+  requirements or Snowpark Container Services-only targets. The option is ignored, a warning is
+  emitted, and exact pip version specifications are kept.
+
+### Deprecations
+
+## 2.3.0 (2026-09-28)
+
+### New Features
+
+* Experiment Tracking: `ExperimentTracking` accepts an optional `model_registry` parameter. When provided,
+  `log_model` stores models in the registry's database and schema instead of the experiment's, so experiments
+  and models can live in separate schemas.
+
+### Bug Fixes
+
+* Feature Store: Offline reads of a feature view with several list aggregations or several secondary-key
+  aggregation windows used to return incomplete results in some scenarios. The bug has been fixed
+  and the existing feature views pick up the fix on their next read, with no re-registration or backfill required.
+
+### Behavior Changes
+
+* Feature Store: `update_feature_view()` now rejects a `target_lag` change on a feature view whose online store
+  is `POSTGRES`, instead of reporting success for a change that is not applied. Set `target_lag` when online
+  storage is first enabled, or delete and re-register the feature view to change it. Enabling or disabling
+  online storage, and `target_lag` changes on hybrid-table online stores, are unaffected.
+
+### Deprecations
+
+## 2.2.0 (2026-09-21)
 
 ### New Features
 
@@ -438,8 +496,17 @@ NOTE: Version 1.7.0 is used as example here. Please choose the the latest versio
 
 ### Bug Fixes
 
+* Registry: `ModelVersion.load()` now compares the model owner to `CURRENT_ROLE()` (the role that
+  executes SQL) instead of `Session.get_current_role()`. Owner's-rights Streamlit apps and
+  `EXECUTE AS OWNER` procedures can load when statements run as the owner even if the caller's
+  primary role is not the owner.
+
 ### Behavior Changes
 
+* Dependencies: `scipy` is no longer a required dependency of `snowflake-ml-python`. It is still
+  installed transitively with `pip install snowflake-ml-python[scikit-learn]`. Direct `scipy` usage
+  in `snowflake.ml.modeling.preprocessing` already requires that extra.
+
 ### Deprecations
 
 ## 2.1.0 (2026-09-14)
@@ -461,21 +528,12 @@ NOTE: Version 1.7.0 is used as example here. Please choose the the latest versio
 
 ### Bug Fixes
 
-* Registry: `ModelVersion.load()` now compares the model owner to `CURRENT_ROLE()` (the role that
-  executes SQL) instead of `Session.get_current_role()`. Owner's-rights Streamlit apps and
-  `EXECUTE AS OWNER` procedures can load when statements run as the owner even if the caller's
-  primary role is not the owner.
-
 * Registry: Fixed logging of MLflow models created with `mlflow.sklearn.save_model()`. The model is
   now re-logged from the scikit-learn estimator using the serialization format it was saved with,
   instead of from MLflow's PyFunc wrapper with the version-dependent default format.
 
 ### Behavior Changes
 
-* Dependencies: `scipy` is no longer a required dependency of `snowflake-ml-python`. It is still
-  installed transitively with `pip install snowflake-ml-python[scikit-learn]`. Direct `scipy` usage
-  in `snowflake.ml.modeling.preprocessing` already requires that extra.
-
 ### Deprecations
 
 ## 2.0.0 (2026-09-09)

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/README.md

@@ -57,7 +57,7 @@ In Snowflake Warehouse notebook environments, snowflake-ml-python can be install
 Follow the [installation instructions](https://docs.snowflake.com/en/developer-guide/snowpark-ml/index#installing-snowpark-ml)
 in the Snowflake documentation.
 
-Python versions 3.10 to 3.13 are supported. You can use [miniconda](https://docs.conda.io/en/latest/miniconda.html) or
+Python versions 3.10 to 3.14 are supported. You can use [miniconda](https://docs.conda.io/en/latest/miniconda.html) or
 [anaconda](https://www.anaconda.com/) to create a Conda environment (recommended),
 or [virtualenv](https://docs.python.org/3/tutorial/venv.html) to create a virtual environment.
 

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/pyproject.toml

@@ -5,8 +5,8 @@ build-backend = "setuptools.build_meta"
 [project]
 name = "snowflake-ml-python"
 description = "The machine learning client library that is used for interacting with Snowflake to build machine learning solutions."
-classifiers = [ "Development Status :: 5 - Production/Stable", "Environment :: Console", "Environment :: Other Environment", "Intended Audience :: Developers", "Intended Audience :: Education", "Intended Audience :: Information Technology", "Intended Audience :: System Administrators", "License :: OSI Approved :: Apache Software License", "Operating System :: OS Independent", "Programming Language :: Python :: 3.10", "Programming Language :: Python :: 3.11", "Programming Language :: Python :: 3.12", "Topic :: Database", "Topic :: Software Development", "Topic :: Software Development :: Libraries", "Topic :: Software Development :: Libraries :: Application Frameworks", "Topic :: Software Development :: Libraries :: Python Modules", "Topic :: Scientific/Engineering :: Information Analysis",]
-requires-python = ">=3.10, <3.14"
+classifiers = [ "Development Status :: 5 - Production/Stable", "Environment :: Console", "Environment :: Other Environment", "Intended Audience :: Developers", "Intended Audience :: Education", "Intended Audience :: Information Technology", "Intended Audience :: System Administrators", "License :: OSI Approved :: Apache Software License", "Operating System :: OS Independent", "Programming Language :: Python :: 3.10", "Programming Language :: Python :: 3.11", "Programming Language :: Python :: 3.12", "Programming Language :: Python :: 3.13", "Programming Language :: Python :: 3.14", "Topic :: Database", "Topic :: Software Development", "Topic :: Software Development :: Libraries", "Topic :: Software Development :: Libraries :: Application Frameworks", "Topic :: Software Development :: Libraries :: Python Modules", "Topic :: Scientific/Engineering :: Information Analysis",]
+requires-python = ">=3.10, <3.15"
 dynamic = [ "version", "readme",]
 dependencies = [ "anyio>=3.5.0,<5", "cachetools>=3.1.1", "cloudpickle>=2.0.0", "fsspec[http]>=2024.6.1,<2026", "importlib_resources>=6.1.1, <7", "jinja2>=3.1.0,<4", "numpy>=1.23,<3", "packaging>=20.9,<26", "pandas>=2.1.4,<3", "platformdirs<5", "pyarrow", "pydantic>=2.8.2, <3", "pyjwt>=2.0.0, <3", "pytimeparse>=1.1.8,<2", "pyyaml>=6.0,<7", "retrying>=1.3.3,<2", "snowflake-connector-python[pandas]>=3.17.3,<5", "snowflake-snowpark-python>=1.37.0,<2", "sqlparse>=0.4,<1", "tqdm<5", "typing-extensions>=4.1.0,<5",]
 [[project.authors]]

snowflake_ml_python-2.4.0/snowflake/cortex/_classify_text.py

@@ -0,0 +1,93 @@
+import json
+from typing import Any, cast
+
+from typing_extensions import deprecated
+
+from snowflake import snowpark
+from snowflake.cortex._util import (
+    CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+    call_sql_function,
+    warn_cortex_deprecated,
+)
+from snowflake.ml._internal import telemetry
+from snowflake.snowpark import functions
+
+_AI_CLASSIFY_FUNCTION_NAME = "AI_CLASSIFY"
+# Legacy CLASSIFY_TEXT always returned exactly one label. AI_CLASSIFY defaults to single-label mode, but
+# we set output_mode explicitly so the single-label contract can't drift and multi-label results can
+# never leak through.
+_AI_CLASSIFY_SINGLE_LABEL_CONFIG = {"output_mode": "single"}
+
+
+def _adapt_ai_classify_result(result: Any) -> str:
+    """Map an AI_CLASSIFY result to the legacy CLASSIFY_TEXT {"label": ...} JSON string."""
+    # Legacy classify_text returned the raw CLASSIFY_TEXT OBJECT, which the immediate (non-Column) path
+    # collects as a JSON string such as '{"label": "positive"}'. We reproduce that exact shape so
+    # existing callers that json.loads the result and read ["label"] keep working.
+    parsed: Any = result
+    if isinstance(result, str):
+        try:
+            parsed = json.loads(result)
+        except json.JSONDecodeError:
+            return result
+    label: Any = None
+    if isinstance(parsed, dict):
+        labels = parsed.get("labels")
+        if isinstance(labels, list) and labels:
+            label = labels[0]
+        elif parsed.get("label") is not None:
+            label = parsed.get("label")
+    return json.dumps({"label": label})
+
+
+@telemetry.send_api_usage_telemetry(
+    project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+)
+def classify_text(
+    str_input: str | snowpark.Column,
+    categories: list[str] | snowpark.Column,
+    session: snowpark.Session | None = None,
+) -> str | snowpark.Column:
+    """Use the LLM inference service to classify the INPUT text into one of the target CATEGORIES.
+
+    Args:
+        str_input: A Column of strings to classify.
+        categories: A list of candidate categories to classify the INPUT text into.
+        session: The snowpark session to use. Will be inferred by context if not specified.
+
+    Returns:
+        A column of classification responses.
+    """
+    warn_cortex_deprecated("classify_text", "ai_classify")
+    result = _classify_text_impl(
+        _AI_CLASSIFY_FUNCTION_NAME, str_input, categories, _AI_CLASSIFY_SINGLE_LABEL_CONFIG, session=session
+    )
+    if isinstance(result, snowpark.Column):
+        # AI_CLASSIFY returns {"labels": [...]}; re-wrap as the legacy CLASSIFY_TEXT {"label": ...} OBJECT
+        # so downstream SQL that reads col["label"] is unchanged. Subscript access (not functions.get with
+        # a bare string, which Snowpark treats as a column identifier) extracts the field by literal key.
+        label = functions.coalesce(result["labels"][0], result["label"])
+        return functions.object_construct(functions.lit("label"), label)
+    return _adapt_ai_classify_result(result)
+
+
+def _classify_text_impl(
+    function: str,
+    str_input: str | snowpark.Column,
+    categories: list[str] | snowpark.Column,
+    config: dict[str, Any] | None = None,
+    session: snowpark.Session | None = None,
+) -> str | snowpark.Column:
+    args: list[Any] = [str_input, categories]
+    if config is not None:
+        args.append(config)
+    return cast(str | snowpark.Column, call_sql_function(function, session, *args))
+
+
+ClassifyText = deprecated(
+    "ClassifyText() is deprecated and will be removed in a future release. Please use classify_text() instead."
+)(
+    telemetry.send_api_usage_telemetry(
+        project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+    )(classify_text)
+)

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/snowflake/cortex/_complete.py

@@ -14,6 +14,7 @@ from snowflake.cortex._util import (
     CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
     SnowflakeAuthenticationException,
     SnowflakeConfigurationException,
+    warn_cortex_deprecated,
 )
 from snowflake.ml._internal import telemetry
 from snowflake.snowpark import context, functions
@@ -331,12 +332,46 @@ def _return_stream_response(
                 pass
 
 
-def _complete_call_sql_function_snowpark(
-    function: str, *args: str | snowpark.Column | CompleteOptions
-) -> snowpark.Column:
+def _complete_call_sql_function_snowpark(function: str, *args: Any) -> snowpark.Column:
     return cast(snowpark.Column, functions.builtin(function)(*args))
 
 
+_SQL_MODEL_PARAMETER_KEYS = ("max_tokens", "temperature", "top_p", "guardrails")
+
+
+def _sql_ai_complete_args(
+    model: str | snowpark.Column,
+    prompt: str | list[ConversationMessage] | snowpark.Column,
+    options: CompleteOptions | snowpark.Column | None,
+) -> tuple[Any, ...]:
+    """Build positional SQL arguments for AI_COMPLETE.
+
+    Args:
+        model: Model name or Column of model names.
+        prompt: Prompt string, conversation history, or Column of prompts.
+        options: Complete options, a Column of options, or None.
+
+    Returns:
+        Positional AI_COMPLETE arguments with response format separate from
+        model parameters.
+    """
+    if options is None:
+        return (model, prompt)
+    if isinstance(options, snowpark.Column):
+        # Can't introspect a Column's keys at build time, so split response_format out of the options
+        # object in SQL: model_parameters is the object minus response_format, response_format is read
+        # out separately (NULL when absent, which AI_COMPLETE treats as "no response format").
+        model_parameters_col = functions.builtin("object_delete")(options, functions.lit("response_format"))
+        response_format_col = functions.get(options, functions.lit("response_format"))
+        return (model, prompt, model_parameters_col, response_format_col)
+    option_values = cast(dict[str, Any], options)
+    model_parameters = {key: option_values[key] for key in _SQL_MODEL_PARAMETER_KEYS if key in option_values}
+    response_format = options.get("response_format")
+    if response_format is None:
+        return (model, prompt, model_parameters) if model_parameters else (model, prompt)
+    return (model, prompt, model_parameters, response_format)
+
+
 def _complete_non_streaming_immediate(
     snow_api_xp_request_handler: Callable[..., dict[str, Any]] | None,
     model: str,
@@ -366,10 +401,7 @@ def _complete_non_streaming_impl(
     deadline: float | None = None,
 ) -> str | snowpark.Column:
     if isinstance(prompt, snowpark.Column):
-        if options is not None:
-            return _complete_call_sql_function_snowpark(function, model, prompt, options)
-        else:
-            return _complete_call_sql_function_snowpark(function, model, prompt)
+        return _complete_call_sql_function_snowpark(function, *_sql_ai_complete_args(model, prompt, options))
     if isinstance(model, snowpark.Column):
         raise ValueError("'model' cannot be a snowpark.Column when 'prompt' is a string.")
     if isinstance(options, snowpark.Column):
@@ -414,7 +446,7 @@ def _complete_impl(
     model: str | snowpark.Column,
     prompt: str | list[ConversationMessage] | snowpark.Column,
     snow_api_xp_request_handler: Callable[..., dict[str, Any]] | None = None,
-    function: str = "snowflake.cortex.complete",
+    function: str = "AI_COMPLETE",
     options: CompleteOptions | None = None,
     session: snowpark.Session | None = None,
     stream: bool = False,
@@ -480,7 +512,7 @@ def complete(
     Returns:
         A column of string responses.
     """
-
+    warn_cortex_deprecated("complete", "ai_complete")
     # Set the XP snow api function, if available.
     snow_api_xp_request_handler = None
     if is_in_stored_procedure():  # type: ignore[no-untyped-call]

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/snowflake/cortex/_embed_text_1024.py

@@ -1,11 +1,17 @@
-from typing import Union, cast
+from typing import cast
 
 from typing_extensions import deprecated
 
 from snowflake import snowpark
-from snowflake.cortex._util import CORTEX_FUNCTIONS_TELEMETRY_PROJECT, call_sql_function
+from snowflake.cortex._util import (
+    CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+    call_sql_function,
+    warn_cortex_deprecated,
+)
 from snowflake.ml._internal import telemetry
 
+_AI_EMBED_FUNCTION_NAME = "AI_EMBED"
+
 
 @telemetry.send_api_usage_telemetry(
     project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
@@ -26,8 +32,8 @@ def embed_text_1024(
     Returns:
         A column of vectors containing embeddings.
     """
-
-    return _embed_text_1024_impl("snowflake.cortex.embed_text_1024", model, text, session=session)
+    warn_cortex_deprecated("embed_text_1024", "ai_embed")
+    return _embed_text_1024_impl(_AI_EMBED_FUNCTION_NAME, model, text, session=session)
 
 
 def _embed_text_1024_impl(
@@ -36,7 +42,7 @@ def _embed_text_1024_impl(
     text: str | snowpark.Column,
     session: snowpark.Session | None = None,
 ) -> list[float] | snowpark.Column:
-    return cast(Union[list[float], snowpark.Column], call_sql_function(function, session, model, text))
+    return cast(list[float] | snowpark.Column, call_sql_function(function, session, model, text))
 
 
 EmbedText1024 = deprecated(

{snowflake_ml_python-2.2.0 → snowflake_ml_python-2.4.0}/snowflake/cortex/_embed_text_768.py

@@ -1,11 +1,17 @@
-from typing import Union, cast
+from typing import cast
 
 from typing_extensions import deprecated
 
 from snowflake import snowpark
-from snowflake.cortex._util import CORTEX_FUNCTIONS_TELEMETRY_PROJECT, call_sql_function
+from snowflake.cortex._util import (
+    CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+    call_sql_function,
+    warn_cortex_deprecated,
+)
 from snowflake.ml._internal import telemetry
 
+_AI_EMBED_FUNCTION_NAME = "AI_EMBED"
+
 
 @telemetry.send_api_usage_telemetry(
     project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
@@ -26,8 +32,8 @@ def embed_text_768(
     Returns:
         A column of vectors containing embeddings.
     """
-
-    return _embed_text_768_impl("snowflake.cortex.embed_text_768", model, text, session=session)
+    warn_cortex_deprecated("embed_text_768", "ai_embed")
+    return _embed_text_768_impl(_AI_EMBED_FUNCTION_NAME, model, text, session=session)
 
 
 def _embed_text_768_impl(
@@ -36,7 +42,7 @@ def _embed_text_768_impl(
     text: str | snowpark.Column,
     session: snowpark.Session | None = None,
 ) -> list[float] | snowpark.Column:
-    return cast(Union[list[float], snowpark.Column], call_sql_function(function, session, model, text))
+    return cast(list[float] | snowpark.Column, call_sql_function(function, session, model, text))
 
 
 EmbedText768 = deprecated(

snowflake_ml_python-2.4.0/snowflake/cortex/_extract_answer.py

@@ -0,0 +1,85 @@
+from typing import Any, cast
+
+from typing_extensions import deprecated
+
+from snowflake import snowpark
+from snowflake.cortex._util import (
+    CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+    SnowflakeAuthenticationException,
+    warn_cortex_deprecated,
+)
+from snowflake.ml._internal import telemetry
+from snowflake.snowpark import context, functions
+
+_AI_EXTRACT_FUNCTION_NAME = "AI_EXTRACT"
+_ANSWER_KEY = "answer"
+
+
+def _legacy_extract_answer_column(
+    from_text: str | snowpark.Column,
+    response_format: dict[str, Any] | snowpark.Column,
+) -> snowpark.Column:
+    """Call AI_EXTRACT and reshape it to the legacy EXTRACT_ANSWER [{"answer": ..., "score": ...}] ARRAY."""
+    # AI_EXTRACT returns the score only when the `scores` argument is TRUE, and `scores` is a named-only
+    # argument that the Snowpark builtin wrapper can't pass positionally -- inject it as a raw SQL fragment.
+    # AI_EXTRACT(<text>, <response_format>, scores => TRUE) returns:
+    #   {"response": {"answer": ...}, "scoring": {"scores": {"answer": {"score": ...}}}}
+    extracted = functions.builtin(_AI_EXTRACT_FUNCTION_NAME)(
+        from_text, response_format, functions.sql_expr("scores => TRUE")
+    )
+    # Subscript access (not functions.get with a bare string, which Snowpark treats as a column
+    # identifier) extracts fields by literal key.
+    answer = extracted["response"][_ANSWER_KEY]
+    score = extracted["scoring"]["scores"][_ANSWER_KEY]["score"]
+    return functions.array_construct(
+        functions.object_construct(functions.lit(_ANSWER_KEY), answer, functions.lit("score"), score)
+    )
+
+
+@telemetry.send_api_usage_telemetry(
+    project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT,
+)
+def extract_answer(
+    from_text: str | snowpark.Column,
+    question: str | snowpark.Column,
+    session: snowpark.Session | None = None,
+) -> str | snowpark.Column:
+    """Calls into the LLM inference service to extract an answer from within specified text.
+
+    Args:
+        from_text: A Column of strings representing input text.
+        question: A Column of strings representing a question to ask against from_text.
+        session: The snowpark session to use. Will be inferred by context if not specified.
+
+    Returns:
+        A column of arrays, each [{"answer": ..., "score": ...}], matching legacy EXTRACT_ANSWER.
+
+    Raises:
+        SnowflakeAuthenticationException: if no session is provided and none is active.
+    """
+    warn_cortex_deprecated("extract_answer", "ai_extract")
+    if isinstance(question, snowpark.Column):
+        response_format: dict[str, Any] | snowpark.Column = functions.object_construct(
+            functions.lit(_ANSWER_KEY), question
+        )
+    else:
+        response_format = {_ANSWER_KEY: question}
+
+    column = _legacy_extract_answer_column(from_text, response_format)
+    if isinstance(from_text, snowpark.Column) or isinstance(question, snowpark.Column):
+        return column
+
+    # Immediate (literal) path: evaluate the expression and return the collected ARRAY as a JSON string,
+    # exactly as the legacy EXTRACT_ANSWER immediate path did.
+    session = session or context.get_active_session()
+    if session is None:
+        raise SnowflakeAuthenticationException(
+            """Session required. Provide the session through a session=... argument or ensure an active session is
+            available in your environment."""
+        )
+    return cast(str, session.create_dataframe([snowpark.Row()]).select(column).collect()[0][0])
+
+
+ExtractAnswer = deprecated(
+    "ExtractAnswer() is deprecated and will be removed in a future release. Use extract_answer() instead"
+)(telemetry.send_api_usage_telemetry(project=CORTEX_FUNCTIONS_TELEMETRY_PROJECT)(extract_answer))