howso-engine 68.2.0__tar.gz → 69.0.0__tar.gz

{howso_engine-68.2.0 → howso_engine-69.0.0}/LICENSE-3RD-PARTY.txt

@@ -4,7 +4,7 @@ MIT License
 joke2k
 https://github.com/joke2k/faker
 Faker is a Python package that generates fake data for you.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/faker-40.43.0.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/faker-40.43.0.dist-info/licenses/LICENSE.txt
 Copyright (c) 2012 Daniele Faraglia
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
@@ -32,7 +32,7 @@ MIT License
 Kirill Simonov
 https://pyyaml.org/
 YAML parser and emitter for Python
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pyyaml-6.0.3.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pyyaml-6.0.3.dist-info/licenses/LICENSE
 Copyright (c) 2017-2021 Ingy döt Net
 Copyright (c) 2006-2016 Kirill Simonov
 
@@ -61,7 +61,7 @@ BSD-2-Clause
 Georg Brandl <georg@python.org>
 https://pygments.org
 Pygments is a syntax highlighting package written in Python.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pygments-2.21.0.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pygments-2.21.0.dist-info/licenses/LICENSE
 Copyright (c) 2006-2022 by the respective authors (see AUTHORS file).
 All rights reserved.
 
@@ -95,7 +95,7 @@ MIT
 Filipe Laíns <lains@riseup.net>, Bernát Gábor <gaborjbernat@gmail.com>, layday <layday@protonmail.com>, Henry Schreiner <henryschreineriii@gmail.com>
 https://build.pypa.io
 A simple, correct Python build frontend
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/build-1.6.1.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/build-1.6.1.dist-info/licenses/LICENSE
 Copyright © 2019 Filipe Laíns <filipe.lains@gmail.com>
 
 Permission is hereby granted, free of charge, to any person obtaining a
@@ -119,12 +119,12 @@ DEALINGS IN THE SOFTWARE.
 
 
 certifi
-2026.10.18
+2026.10.23
 Mozilla Public License 2.0 (MPL 2.0)
 Kenneth Reitz
 https://github.com/certifi/python-certifi
 Python package for providing Mozilla's CA Bundle.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/certifi-2026.10.18.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/certifi-2026.10.23.dist-info/licenses/LICENSE
 This package contains a modified version of ca-bundle.crt:
 
 ca-bundle.crt -- Bundle of CA Root Certificates
@@ -153,7 +153,7 @@ MIT
 "Ahmed R. TAHRI" <tahri.ahmed@proton.me>
 https://github.com/jawah/charset_normalizer/blob/master/CHANGELOG.md
 The Real First Universal Charset Detector. Open, modern and actively maintained alternative to Chardet.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/charset_normalizer-3.5.2.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/charset_normalizer-3.5.2.dist-info/licenses/LICENSE
 MIT License
 
 Copyright (c) 2025 TAHRI Ahmed R.
@@ -183,7 +183,7 @@ BSD-3-Clause
 UNKNOWN
 https://github.com/pallets/click/
 Composable command line interface toolkit
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/click-8.5.0.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/click-8.5.0.dist-info/licenses/LICENSE.txt
 Copyright 2014 Pallets
 
 Redistribution and use in source and binary forms, with or without
@@ -220,7 +220,7 @@ BSD License
 The cloudpickle developer team
 https://github.com/cloudpipe/cloudpickle
 Pickler class to extend the standard pickle.Pickler functionality
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/cloudpickle-3.1.2.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/cloudpickle-3.1.2.dist-info/licenses/LICENSE
 This module was extracted from the `cloud` package, developed by
 PiCloud, Inc.
 
@@ -261,7 +261,7 @@ MIT
 Jason Moiron <jmoiron@jmoiron.net>
 https://github.com/python-humanize/humanize
 Python humanize utilities
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/humanize-4.16.0.dist-info/licenses/LICENCE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/humanize-4.16.0.dist-info/licenses/LICENCE
 Copyright (c) 2010-2020 Jason Moiron and Contributors
 
 Permission is hereby granted, free of charge, to any person obtaining
@@ -290,7 +290,7 @@ BSD-3-Clause
 Kim Davies <kim+pypi@gumleaf.org>
 https://github.com/kjd/idna
 Internationalized Domain Names in Applications (IDNA)
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/idna-3.20.dist-info/licenses/LICENSE.md
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/idna-3.20.dist-info/licenses/LICENSE.md
 BSD 3-Clause License
 
 Copyright (c) 2013-2026, Kim Davies and contributors.
@@ -330,7 +330,7 @@ BSD-3-Clause
 Gael Varoquaux <gael.varoquaux@normalesup.org>
 https://joblib.readthedocs.io
 Lightweight pipelining with Python functions
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/joblib-1.6.0.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/joblib-1.6.0.dist-info/licenses/LICENSE.txt
 BSD 3-Clause License
 
 Copyright (c) 2008-2021, The joblib developers.
@@ -368,7 +368,7 @@ MIT License
 Chris Sewell <chrisj_sewell@hotmail.com>
 https://github.com/executablebooks/markdown-it-py
 Python port of markdown-it. Markdown parsing, done right!
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/markdown_it_py-4.2.0.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/markdown_it_py-4.2.0.dist-info/licenses/LICENSE
 MIT License
 
 Copyright (c) 2020 ExecutableBookProject
@@ -398,7 +398,7 @@ MIT License
 Taneli Hukkinen <hukkin@users.noreply.github.com>
 https://github.com/executablebooks/mdurl
 Markdown URL utilities
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/mdurl-0.1.2.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/mdurl-0.1.2.dist-info/LICENSE
 Copyright (c) 2015 Vitaly Puzrin, Alex Kocharin.
 Copyright (c) 2021 Taneli Hukkinen
 
@@ -453,7 +453,7 @@ MIT License
 Hajime Senuma <hajime.senuma@gmail.com>
 https://pypi.org/project/mmh3/
 Python extension for MurmurHash (MurmurHash3), a set of fast and robust hash functions.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/mmh3-5.3.1.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/mmh3-5.3.1.dist-info/licenses/LICENSE
 MIT License
 
 Copyright (c) 2011-2026 Hajime Senuma
@@ -482,7 +482,7 @@ BSD-3-Clause AND 0BSD AND MIT AND Zlib AND CC0-1.0
 Travis E. Oliphant et al.
 https://numpy.org
 Fundamental package for array computing in Python
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/numpy-2.4.6.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/numpy-2.4.6.dist-info/licenses/LICENSE.txt
 Copyright (c) 2005-2025, NumPy Developers.
 All rights reserved.
 
@@ -1426,7 +1426,7 @@ Apache-2.0 OR BSD-2-Clause
 Donald Stufft <donald@stufft.io>
 https://github.com/pypa/packaging
 Core utilities for Python packages
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/packaging-26.3.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/packaging-26.3.dist-info/licenses/LICENSE
 This software is made available under the terms of *either* of the licenses
 found in LICENSE.APACHE or LICENSE.BSD. Contributions to this software is made
 under the terms of *both* these licenses.
@@ -1438,7 +1438,7 @@ BSD License
 The Pandas Development Team <pandas-dev@python.org>
 https://pandas.pydata.org
 Powerful data structures for data analysis, time series, and statistics
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pandas-2.3.3.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pandas-2.3.3.dist-info/LICENSE
 BSD 3-Clause License
 
 Copyright (c) 2008-2011, AQR Capital Management, LLC, Lambda Foundry, Inc. and PyData Development Team
@@ -2696,7 +2696,7 @@ BSD License
 Vincent Driessen <me@nvie.com>
 https://github.com/jazzband/pip-tools/
 pip-tools keeps your pinned dependencies fresh.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pip_tools-7.4.1.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pip_tools-7.4.1.dist-info/LICENSE
 BSD 3-Clause License
 
 Copyright (c). All rights reserved.
@@ -2733,7 +2733,7 @@ BSD-3-Clause
 Giampaolo Rodola
 https://github.com/giampaolo/psutil
 Cross-platform lib for process and system monitoring.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/psutil-7.2.2.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/psutil-7.2.2.dist-info/LICENSE
 BSD 3-Clause License
 
 Copyright (c) 2009, Jay Loden, Dave Daeschler, Giampaolo Rodola
@@ -2771,7 +2771,7 @@ MIT
 Thomas Kluyver <thomas@kluyver.me.uk>
 https://github.com/pypa/pyproject-hooks
 Wrappers to call pyproject.toml-based build backend hooks.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pyproject_hooks-1.3.3.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pyproject_hooks-1.3.3.dist-info/licenses/LICENSE
 The MIT License (MIT)
 
 Copyright (c) 2017 Thomas Kluyver
@@ -2796,12 +2796,12 @@ THE SOFTWARE.
 
 
 python-dateutil
-2.9.0.post0
+3.0.0
 Apache Software License; BSD License
 Gustavo Niemeyer
 https://github.com/dateutil/dateutil
 Extensions to the standard Python datetime module
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/python_dateutil-2.9.0.post0.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/python_dateutil-3.0.0.dist-info/licenses/LICENSE
 Copyright 2017- Paul Ganssle <paul@ganssle.io>
 Copyright 2017- dateutil contributors (see AUTHORS file)
 
@@ -2863,7 +2863,7 @@ MIT License
 Stuart Bishop
 http://pythonhosted.org/pytz
 World timezone definitions, modern and historical
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/pytz-2026.5.dist-info/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/pytz-2026.5.dist-info/LICENSE.txt
 Copyright (c) 2003-2019 Stuart Bishop <stuart@stuartbishop.net>
 
 Permission is hereby granted, free of charge, to any person obtaining a
@@ -2891,7 +2891,7 @@ Apache-2.0 AND CNRI-Python
 Matthew Barnett <regex@mrabarnett.plus.com>
 https://github.com/mrabarnett/mrab-regex
 Alternative regular expression module, to replace re.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/regex-2026.9.29.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/regex-2026.9.29.dist-info/licenses/LICENSE.txt
 This work was derived from the 're' module of CPython 2.6 and CPython 3.1,
 copyright (c) 1998-2001 by Secret Labs AB and licensed under CNRI's Python 1.6
 license.
@@ -3108,7 +3108,7 @@ Apache Software License
 Kenneth Reitz <me@kennethreitz.org>
 https://github.com/psf/requests
 Python HTTP for Humans.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/requests-2.34.2.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/requests-2.34.2.dist-info/licenses/LICENSE
 
                                  Apache License
                            Version 2.0, January 2004
@@ -3292,7 +3292,7 @@ MIT License
 Will McGugan
 https://github.com/Textualize/rich
 Render rich text, tables, progress bars, syntax highlighting, markdown and more to the terminal
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/rich-15.0.0.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/rich-15.0.0.dist-info/licenses/LICENSE
 Copyright (c) 2020 Will McGugan
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
@@ -3320,7 +3320,7 @@ MIT License
 UNKNOWN
 https://github.com/hplt-project/sacremoses
 SacreMoses
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/sacremoses-0.2.0.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/sacremoses-0.2.0.dist-info/licenses/LICENSE
 MIT License
 
 Copyright (c) 2020 alvations
@@ -3350,7 +3350,7 @@ BSD License
 Raphaël Barrois
 https://github.com/rbarrois/python-semanticversion
 A library implementing the 'SemVer' scheme.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/semantic_version-2.10.0.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/semantic_version-2.10.0.dist-info/LICENSE
 Copyright (c) The python-semanticversion project
 All rights reserved.
 
@@ -3381,7 +3381,7 @@ MIT License
 Benjamin Peterson
 https://github.com/benjaminp/six
 Python 2 and 3 compatibility utilities
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/six-1.17.0.dist-info/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/six-1.17.0.dist-info/LICENSE
 Copyright (c) 2010-2024 Benjamin Peterson
 
 Permission is hereby granted, free of charge, to any person obtaining a copy of
@@ -3408,7 +3408,7 @@ MPL-2.0 AND MIT
 UNKNOWN
 https://tqdm.github.io
 Fast, Extensible Progress Meter
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/tqdm-4.70.1.dist-info/licenses/LICENCE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/tqdm-4.70.1.dist-info/licenses/LICENCE
 `tqdm` is a product of collaborative work.
 Unless otherwise stated, all authors (see commit logs) retain copyright
 for their respective work, and release the work under the MIT licence
@@ -3466,7 +3466,7 @@ PSF-2.0
 "Guido van Rossum, Jukka Lehtosalo, Łukasz Langa, Michael Lee" <levkivskyi@gmail.com>
 https://github.com/python/typing_extensions
 Backported and Experimental Type Hints for Python 3.9+
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/typing_extensions-4.16.0.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/typing_extensions-4.16.0.dist-info/licenses/LICENSE
 A. HISTORY OF THE SOFTWARE
 ==========================
 
@@ -3754,7 +3754,7 @@ Apache-2.0
 Python Software Foundation
 https://github.com/python/tzdata
 Provider of IANA time zone data
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/tzdata-2026.5.dist-info/licenses/LICENSE
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/tzdata-2026.5.dist-info/licenses/LICENSE
 Apache Software License 2.0
 
 Copyright (c) 2020, Paul Ganssle (Google)
@@ -3779,7 +3779,7 @@ MIT
 Andrey Petrov <andrey.petrov@shazow.net>
 https://github.com/urllib3/urllib3/blob/main/CHANGES.rst
 HTTP library with thread-safe connection pooling, file post, and more.
-/home/runner/.pyenv/versions/3.14.8/lib/python3.14/site-packages/urllib3-2.8.0.dist-info/licenses/LICENSE.txt
+/home/runner/.pyenv/versions/3.14.9/lib/python3.14/site-packages/urllib3-2.8.0.dist-info/licenses/LICENSE.txt
 MIT License
 
 Copyright (c) 2008-2020 Andrey Petrov and contributors.

{howso_engine-68.2.0/howso_engine.egg-info → howso_engine-69.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: howso-engine
-Version: 68.2.0
+Version: 69.0.0
 Summary: The Howso Engine™ is a natively and fully explainable ML engine, serving as an alternative to black box AI neural networks.
 Author: Howso Incorporated
 Author-email: support@howso.com
@@ -705,7 +705,7 @@ Requires-Dist: sacremoses
 Requires-Dist: semantic-version~=2.0
 Requires-Dist: typing-extensions~=4.9
 Requires-Dist: urllib3>=2.6.3
-Requires-Dist: amalgam-lang==36.0.1
+Requires-Dist: amalgam-lang==37.0.0
 Provides-Extra: dev
 Requires-Dist: dill~=0.0; extra == "dev"
 Requires-Dist: flake8-docstrings; extra == "dev"

{howso_engine-68.2.0 → howso_engine-69.0.0}/howso/client/base.py

@@ -73,6 +73,13 @@ if TYPE_CHECKING:
     from .configuration import HowsoConfiguration
 
 
+def _is_multithreaded_library_type(library_type: str | None) -> bool:
+    """Report whether a Trainee runtime ``library_type`` names a multi-threaded build."""
+    if not isinstance(library_type, str):
+        return False
+    return "mt" in library_type.strip().split("-")
+
+
 class AbstractHowsoClient(ABC):
     """The base definition of the Howso client interface."""
 
@@ -464,6 +471,37 @@ class AbstractHowsoClient(ABC):
         """
         return None
 
+    def is_multithreaded(self, trainee_id: str) -> bool:
+        """
+        Report whether a Trainee's engine runs on a multi-threaded Amalgam build.
+
+        A multi-threaded build tolerates concurrent calls into one engine: it
+        can be polled for progress while it works and can serve requests from
+        several threads at once. A single-threaded build, including the OpenMP
+        build, does neither safely.
+
+        The answer is a snapshot. A client whose Trainees can change library
+        type may answer differently on a later call.
+
+        This implementation reads ``library_type`` from
+        :meth:`get_trainee_runtime` and reports True only for the ``mt``
+        library type. Clients that can identify their build more directly
+        override it.
+
+        Parameters
+        ----------
+        trainee_id : str
+            The identifier of the Trainee.
+
+        Returns
+        -------
+        bool
+            True when the build is confirmed multi-threaded; False for a
+            single-threaded build or an unrecognized library type.
+        """
+        runtime = self.get_trainee_runtime(trainee_id)
+        return _is_multithreaded_library_type(runtime.get("library_type"))
+
     @abstractmethod
     def query_trainees(self, search_terms: str | None = None) -> list[dict]:
         """Query accessible Trainees."""
@@ -2227,6 +2265,12 @@ class AbstractHowsoClient(ABC):
                 'num_most_similar_cases' is not specified) relevant number of
                 similar cases, which will first include the influential cases.
                 Uses only the context features of the reacted case.
+            - most_similar_cases_features : list of str, optional
+                A list of feature names to output values for when outputting
+                most similar, influential, or boundary cases. If an empty list,
+                only the distance and session information of each case is
+                output. If unspecified, outputs all context and action
+                features.
             - num_boundary_cases : int, optional
                 Outputs this manually specified number of boundary cases.
             - num_most_similar_cases : int, optional
@@ -4433,6 +4477,7 @@ class AbstractHowsoClient(ABC):
         features: Collection[str] | None = None,
         filter_fanout_values: bool = False,
         group_id_features: Collection[str] | None = None,
+        isolation_distance_contributions: bool = False,
         kl_divergence_addition: bool = False,
         kl_divergence_removal: bool = False,
         new_cases: TabularData3D | None = None,
@@ -4455,7 +4500,7 @@ class AbstractHowsoClient(ABC):
         action_features : list of str, optional
             A list of features whose values should be predicted for
             each group. Each group of cases gets a single action value for each
-            feature.
+            feature. Predictions are made holding out the rest of the group.
         case_indices : list of lists of lists of tuples of {str, int}, optional
             A list of lists of case indices tuples containing the session ID and
             the session training indices that uniquely identify trained cases.
@@ -4504,7 +4549,8 @@ class AbstractHowsoClient(ABC):
                 predicted for each group.
         distance_contributions : bool, default False
             Calculate and output the average distance contribution for each group
-            holding out rest of the group.
+            without holding out the rest of the group. Each case is only held out
+            of its own query.
         familiarity_conviction_addition : bool, default False
             Calculate and output familiarity conviction of adding the
             specified cases.
@@ -4518,11 +4564,18 @@ class AbstractHowsoClient(ABC):
             holding out other cases that have the same fanned out values
             duplicated. Only used when predicting the given ``action_features``.
         group_id_features : Collection of str, optional
-            List of feature names whose values in the specified cases identify
-            trained cases that should be held out of queries. This parameter is ignored if
-            ``new_cases`` is not specified. It is assumed that all groups of
-            cases have a singular value for each feature in
-            ``group_id_features``.
+            List of feature names whose values identify trained cases that
+            belong to the same group as the given ``new_cases``. These trained
+            cases are held out along with the group, but only in computations
+            that hold out the group: ``isolation_distance_contributions``,
+            ``residual_contributions``, ``similarity_conviction``, and
+            predictions of ``action_features``. They are not held out for
+            ``distance_contributions``. Each group is assumed to have a single
+            value for each of these features. Ignored if ``new_cases`` is not
+            specified.
+        isolation_distance_contributions : bool, default False
+            Calculate and output the average distance contribution for each group
+            holding out the rest of the group.
         kl_divergence_addition : bool, default False
             Calculate and output KL divergence of adding the
             specified cases.
@@ -4548,10 +4601,11 @@ class AbstractHowsoClient(ABC):
         p_value_of_removal : bool, default False
             If true will output p value of removal.
         residual_contributions : bool, default False
-            Calculate and output mean residual contribution for each group's cases.
+            Calculate and output mean residual contribution for each group's cases
+            holding out the rest of the group.
         similarity_conviction : bool, default False
             If true will output the mean similarity conviction of the group's
-            cases.
+            cases holding out the rest of the group.
         task_id : str, optional
             If provided, should be a short, unique identifier that can be used
             with ``get_progress`` if concurrent progress updates are desired.
@@ -4611,6 +4665,7 @@ class AbstractHowsoClient(ABC):
             "features": features,
             "filter_fanout_values": filter_fanout_values,
             "group_id_features": group_id_features,
+            "isolation_distance_contributions": isolation_distance_contributions,
             "kl_divergence_addition": kl_divergence_addition,
             "kl_divergence_removal": kl_divergence_removal,
             "new_cases": serialized_cases,

{howso_engine-68.2.0 → howso_engine-69.0.0}/howso/client/schemas/group_reaction.py

@@ -26,6 +26,7 @@ GroupMetric: TypeAlias = Literal[
     "familiarity_conviction_addition",
     "familiarity_conviction_removal",
     "distance_contribution",
+    "isolation_distance_contribution",
     "residual_contribution",
     "kl_divergence_addition",
     "kl_divergence_removal",

{howso_engine-68.2.0 → howso_engine-69.0.0}/howso/direct/client.py

@@ -22,6 +22,7 @@ import warnings
 from amalgam.api import Amalgam
 import certifi
 from packaging.version import parse as parse_version
+from typing_extensions import override
 import urllib3
 from urllib3.util import Retry, Timeout
 
@@ -85,6 +86,13 @@ def squelch_logs(log_level: int) -> Generator[None, None, None]:
         logging.disable(_old_level)
 
 
+def _is_multithreaded_concurrency(concurrency: bytes | str) -> bool:
+    """Report whether an Amalgam concurrency type string names a multi-threaded build."""
+    if isinstance(concurrency, bytes):
+        concurrency = concurrency.decode("utf-8", errors="replace")
+    return concurrency.strip().startswith("MultiThreaded")
+
+
 class HowsoDirectClient(AbstractHowsoClient):
     """
     The direct Howso client.
@@ -254,6 +262,7 @@ class HowsoDirectClient(AbstractHowsoClient):
                 UnsupportedArgumentWarning, stacklevel=2)
         amlg_params = {k: v for k, v in amlg_params.items() if k in allowed_amlg_params}
         self.amlg = Amalgam(**amlg_params)  # pyright: ignore[reportArgumentType]
+        self._multithreaded = _is_multithreaded_concurrency(self.amlg.get_concurrency_type_string())
         if gc_params:
             self.set_garbage_collection_params(gc_params)
 
@@ -1437,6 +1446,30 @@ class HowsoDirectClient(AbstractHowsoClient):
             versions=TraineeVersion(trainee=trainee_version, amalgam=amlg_version)
         )
 
+    @override
+    def is_multithreaded(self, trainee_id: str) -> bool:
+        """
+        Report whether this client's Amalgam library is a multi-threaded build.
+
+        The in-process library reports its own concurrency type, which this
+        client reads once when it loads the library. That is a property of the
+        process, so every Trainee on this client gets the same answer and
+        ``trainee_id`` is not consulted. ``library_type`` from
+        :meth:`get_trainee_runtime` is derived from the library filename and
+        reports ``mt`` for a path carrying no postfix, so it is not used here.
+
+        Parameters
+        ----------
+        trainee_id : str
+            The identifier of the Trainee. Not consulted.
+
+        Returns
+        -------
+        bool
+            True when the loaded library is a multi-threaded build.
+        """
+        return self._multithreaded
+
     def query_trainees(self, search_terms: str | None = None) -> list[dict[str, str]]:
         """
         Query accessible Trainees.

{howso_engine-68.2.0 → howso_engine-69.0.0}/howso/engine/trainee.py

@@ -149,8 +149,6 @@ class Trainee(BaseTrainee):
         self._created: bool = False
         self._updating: bool = False
         self._was_saved: bool = False
-        # Must precede the client assignment that follows.
-        self._supports_engine_progress: bool | None = None  # lazy loaded
         # NOTE: self.client is a accessor property see setter and getter below.
         self.client = client or get_client()
 
@@ -220,9 +218,6 @@ class Trainee(BaseTrainee):
         if not isinstance(client, HowsoPandasClientMixin):
             raise HowsoError("``client`` must be a HowsoPandasClient")
         self._client = client
-        # Engine progress support is a property of the client and the Trainee
-        # it serves, so a new client invalidates the cached answer.
-        self._supports_engine_progress = None
 
     @property
     def supports_engine_progress(self) -> bool:
@@ -231,21 +226,21 @@ class Trainee(BaseTrainee):
 
         Polling a single-threaded engine while it works terminates the
         process, so :func:`~howso.utilities.engine_polling_supported` reports
-        support only for Trainees it can confirm are multi-threaded. The
-        library type is fixed when the Trainee is created, so the answer is
-        resolved once and cached for the life of this instance.
+        support only for Trainees whose client confirms a multi-threaded build.
+
+        The answer is a snapshot taken on each access. A direct client's
+        library type is fixed for the life of the process and is read from the
+        in-process library; other clients may answer differently over the
+        life of a Trainee.
 
         Returns
         -------
         bool
             True only when engine progress polling is confirmed safe.
         """
-        if self._supports_engine_progress is None:
-            # Lazy load: requires a created Trainee to ask the runtime about.
-            if not self._created:
-                return False
-            self._supports_engine_progress = engine_polling_supported(self.client, self.id)
-        return self._supports_engine_progress
+        if not self._created:
+            return False
+        return engine_polling_supported(self.client, self.id)
 
     @property
     def project(self) -> Project | None:
@@ -1759,6 +1754,12 @@ class Trainee(BaseTrainee):
                 'num_most_similar_cases' is not specified) relevant number of
                 similar cases, which will first include the influential cases.
                 Uses only the context features of the reacted case.
+            - most_similar_cases_features : list of str, optional
+                A list of feature names to output values for when outputting
+                most similar, influential, or boundary cases. If an empty list,
+                only the distance and session information of each case is
+                output. If unspecified, outputs all context and action
+                features.
             - num_boundary_cases : int, optional
                 Outputs this manually specified number of boundary cases.
             - num_most_similar_cases : int, optional
@@ -3415,6 +3416,7 @@ class Trainee(BaseTrainee):
         features: Collection[str] | None = None,
         filter_fanout_values: bool = False,
         group_id_features: Collection[str] | None = None,
+        isolation_distance_contributions: bool = False,
         kl_divergence_addition: bool = False,
         kl_divergence_removal: bool = False,
         new_cases: TabularData3D | None = None,
@@ -3437,7 +3439,7 @@ class Trainee(BaseTrainee):
         action_features : list of str, optional
             A list of features whose values should be predicted for
             each group. Each group of cases gets a single action value for each
-            feature.
+            feature. Predictions are made holding out the rest of the group.
         case_indices : list of lists of lists of tuples of {str, int}, optional
             A list of lists of case indices tuples containing the session ID and
             the session training indices that uniquely identify trained cases.
@@ -3486,7 +3488,8 @@ class Trainee(BaseTrainee):
                 predicted for each group.
         distance_contributions : bool, default False
             Calculate and output the average distance contribution for each group
-            holding out rest of the group.
+            without holding out the rest of the group. Each case is only held out
+            of its own query.
         familiarity_conviction_addition : bool, default False
             Calculate and output familiarity conviction of adding the
             specified cases.
@@ -3500,11 +3503,18 @@ class Trainee(BaseTrainee):
             holding out other cases that have the same fanned out values
             duplicated. Only used when predicting the given ``action_features``.
         group_id_features : Collection of str, optional
-            List of feature names whose values in the specified cases identify
-            trained cases that should be held out of queries. This parameter is ignored if
-            ``new_cases`` is not specified. It is assumed that all groups of
-            cases have a singular value for each feature in
-            ``group_id_features``.
+            List of feature names whose values identify trained cases that
+            belong to the same group as the given ``new_cases``. These trained
+            cases are held out along with the group, but only in computations
+            that hold out the group: ``isolation_distance_contributions``,
+            ``residual_contributions``, ``similarity_conviction``, and
+            predictions of ``action_features``. They are not held out for
+            ``distance_contributions``. Each group is assumed to have a single
+            value for each of these features. Ignored if ``new_cases`` is not
+            specified.
+        isolation_distance_contributions : bool, default False
+            Calculate and output the average distance contribution for each group
+            holding out the rest of the group.
         kl_divergence_addition : bool, default False
             Calculate and output KL divergence of adding the
             specified cases.
@@ -3535,10 +3545,11 @@ class Trainee(BaseTrainee):
         p_value_of_removal : bool, default False
             If true will output :math:`p` value of removal.
         residual_contributions : bool, default False
-            Calculate and output mean residual contribution for each group's cases.
+            Calculate and output mean residual contribution for each group's cases
+            holding out the rest of the group.
         similarity_conviction : bool, default False
             If true will output the mean similarity conviction of the group's
-            cases.
+            cases holding out the rest of the group.
         task_id : str, optional
             If provided, should be a short, unique identifier that can be used
             with ``get_progress`` if concurrent progress updates are desired.
@@ -3566,6 +3577,7 @@ class Trainee(BaseTrainee):
                 features=features,
                 filter_fanout_values=filter_fanout_values,
                 group_id_features=group_id_features,
+                isolation_distance_contributions=isolation_distance_contributions,
                 distance_contributions=distance_contributions,
                 familiarity_conviction_addition=familiarity_conviction_addition,
                 familiarity_conviction_removal=familiarity_conviction_removal,