Recent Releases of trustpy-tools
trustpy-tools - Release v2.0.14
Version 2.0.14
Release Date: 2025-06-21
Introduced --outputdir flag to the CLI, allowing users to specify a custom directory for saving plots and CSV summaries. Both NTS and CNTS now accept an outputdir argument to programmatically control output location. If output_dir is not set, results default to:
trustpy/nts/ (for NTS) trustpy/cnts/ (for CNTS)
Added unit tests and CLI tests to ensure output is correctly saved and isolated. Updated README examples, CLI section, and usage notes.
- Python
Published by yaniker 8 months ago
trustpy-tools - Release v2.0.13
Version 2.0.13
Release Date: 2025-06-07
Updated CLI help message for --mode. Now it clearly explains selecting nts and cnts with reasoning.
Added .bumpversion.cfg for consistent version management across files.
Integrated flake8 linting into CI for code quality enforcement.
!!! Upcoming in v3.0.0 !!!
The next release will be a major update that restructures the codebase to adopt full scikit-learn API conventions to ensure compatibility with scikit-learn pipelines and common conventions. TrustPy v2.x will remain usable, but will no longer receive new features once v3.0.0 is released.
- Python
Published by yaniker 9 months ago
trustpy-tools - Release V2.0.12
Release Date: 2025-06-02
Added repr() methods to NTS and CNTS classes for object introspection. Objects now display key configuration details when printed.
- Python
Published by yaniker 9 months ago
trustpy-tools - Release V2.0.11
- version = "2.0.11" in trustpy/init.py (was not updated properly the last time)
- Added a command-line interface (CLI) for quick evaluation of trustworthiness metrics using .npy inputs
- README.md: added CLI usage section
- Python
Published by yaniker 9 months ago
trustpy-tools - Release v2.0.10
Release Date: 2025-05-21
- Migrated from os.path to pathlib.Path for all file and plot path handling (in NTS, CNTS, and plot scripts).
- Added post-installation one-liner test commands in the README to verify package functionality and plot output. These validate both in-memory logic and disk write permissions using real synthetic data.
- Introduced MANIFEST.in to ensure README.md, LICENSE, and CITATION.cff are bundled with PyPI/Conda-Forge distributions.
- Backwards compatible.
- Python
Published by yaniker 9 months ago
trustpy-tools - v2.0.9 - Patch Release
Release Date: 2025-05-19
Enhancements & Fixes
1. Included version = "2.0.9" in trustpy/init.py for programmatic version access:
bash
import trustpy
print(trustpy.__version__) # 2.0.9
- Added all = ['NTS', 'CNTS'] to restrict from trustpy import * to only public classes.
Before writing plots or CSV files, the code now checks for write permissions using:
bash if not os.access(output_dir, os.W_OK): raise PermissionError(...)This helps prevent runtime crashes on read-only filesystems or restricted environments.In README documentation clarified where plots are saved, added instructions to run tests post-install.
Updated incorrect error string in CNTS init from 'Predictions must be a 1D array' → 'Predictions must be a 2D array'.
alpha, beta, and other optional parameters in NTS and CNTS constructors must now be passed by keyword.
- Python
Published by yaniker 9 months ago
trustpy-tools - v2.0.8 – Improved Input Validation and Consistency
Description: This release introduces stricter input validation to improve reliability and prevent silent misuse:
- Ensures predictions include at least 2 classes.
- Ensures oracle includes at least 2 unique labels.
- Enforces match between number of classes in oracle and prediction outputs.
- Updated error messages for consistency and clarity.
- Removed single-sample test case (non-generalizable).
- Adjusted test assertions for compatibility with
pytests match.
This is a patch release with no changes to core logic or outputs.
- Python
Published by yaniker 10 months ago
trustpy-tools - Release v2.0.7 – Native Float Output for Easier Usage
What's Changed
NTS and CNTS now return float values instead of strings in the .compute() output dictionary. This makes the results easier to use in downstream code, comparisons, and logging — without needing to cast or parse string values. Example: {'class0': 0.667} instead of {'class0': '0.667'}.
## Internal Notes Summary printing and CSV exports still round and format values cleanly (.3f) as before. Behavior is fully backward-compatible.
Why v2.0.6 Was Skipped
v2.0.6 was reserved for an internal change and did not introduce a public-facing release. v2.0.7 is the next available stable version.
- Python
Published by yaniker 10 months ago
trustpy-tools - Version 2.0.5 – Major Usability and Testing Upgrade
Release Date: 05/06/2025
What's New
1. trust_spectrum() now saves Trust spectrum plots and conditional density plots in the appropriate trustpy/nts or trustpy/cnts folders.
2. Plot filenames now auto-append extensions (e.g., .png) if missing.
3. show_summary() now illustrates the results in table form.
4. export_summary() now saves the results in .csv form in the appropriate trustpy/nts or trustpy/cnts folders.
5. Output directories are auto-created for consistent saving.
6. Comprehensive unit tests covering constructor assertions, shape validations, and output correctness for both NTS and CNTS are added via pytest. Tests run via pytest --cov=trustpy with 100% coverage of critical checks.
7. Polished and clarified docstrings and inline comments.
8. Updated README with simplified setup instructions, usage examples, and confirmation that unit tests have already been executed prior to release.
- Python
Published by yaniker 10 months ago
trustpy-tools - Version 2.0.4
Matching PyPI version. No changes.
- Python
Published by yaniker 10 months ago
trustpy-tools - Release v2.0.3
- Package renamed from nettrustscore to trustpy-tools
- New import path: import trustpy instead of trustquant 3.Ready for Conda-Forge distribution
- Python
Published by yaniker 10 months ago
trustpy-tools - Release v2.0.2 — PyPI Publishing and Minor Metadata Updates
Changes in v2.0.2: - Published NetTrustScore package to PyPI (pip install nettrustscore now available) - Added description to setup.py to fix PyPI/Conda-Forge rendering - Added .gitignore to clean build artifacts - Synced versioning across Conda-Forge and PyPI - No code changes — functionality remains identical to v2.0.0
Thanks for using NetTrustScore!
- Python
Published by yaniker 10 months ago
trustpy-tools - v2.0.0
This release: 1. introduces conditional NetTrustScore (CNTS) that measures trustworthiness for true and false predictions for each class. 2. simplifies the import and call processes, (Trustworthiness.computeNTS() -> NTS.compute() and CNTS.compute()) 3. adds assert conditions, 4. sets trust spectrum plot to False by default, 5. adds plots.py for user interaction under assets, 6. adds conditionaltrust_densities.png under assets, 7. has several minor bug fixes and cosmetic improvements
- Python
Published by yaniker 11 months ago
trustpy-tools - v1.0.1
Moved the example image to assets. Simplified commentary in the scripts. Updated Readme
- Python
Published by yaniker 11 months ago