v11 (2025-10-16) - Backend supported version: 15.0
Python engine version supporting 15.0 backend, with added libraries and updated major versions of dependencies.
Added
-
Python libraries:
-
gaminet -
interpret -
polars
-
-
Added
pricingscience.winrate -
push_pandasmethod has new optional parametersmanual_fields_specsandkeep_index. Former one allows to specify FieldTypes for columns. seeFieldSpecdocumentation inpyfx.internal.pandasutilfor more details.
Changed
-
Image is built from
ubuntu 24.04(previously 22.04) -
Python version is changed to version
python 3.12(previously 3.10) -
Bugfixed retry on retryable errors in
push_pandasandupdate_pandas, removed retry frompushandupdatemethods -
endpoint utilized by
updateandupdate_pandaschanged fromdatamart.importfiletodatamart.loadfc -
Multiple packages with update of major version:
-
lightgbm -
neuralprophet -
optuna -
pandas -
sentence-transformers -
SQLAlchemy -
xgboost
-
Removed
-
eli5library was removed, as project is no longer developed
v10 (2024-4-2) - Backend supported version: 12.0
Python engine version supporting Clover Club 12.0.X backend, with updated minor versions of dependencies
Changed
-
Updated minor versions of dependencies
v9 (2024-03-01) - Backend supported version: 11.1
This is the official version supporting Bee’s Knees 11.1 backend.
It replaces the faulty append_xxx methods with more accurate update_xxx.
These methods were relying on a REST API endpoint with a silent faulty behavior regarding duplicate keys columns. They could replace data instead of appending them if usage of unique keys/index was not handled.
Support for the methods is removed immediately as the faulty behavior is hard to detect, easy to trigger and impossible to fix properly.
Users should switch to the corresponding update_xxx methods introduced in this release.
Documentation now available at https://developer.pricefx.eu/pricefx-api/.
pricingscience.similarity has been improved.
Added
-
Added data updating methods
-
connection.update_table -
TableMutable.update -
TableMutable.update_pandas -
Added
pricingscience.similarity.process_new_products -
Official documentation deployment
Changed
-
Updated dependencies
Removed
-
The following methods from the API (replaced with the corresponding
update_xxxmethods) have been removed:-
connection.append_to_table -
TableMutable.append -
TableMutable.append_pandas
-
v8 (2023-07-20) - Backend supported version: 11.0
Official version supporting Bee Knees 11.0.X backend.
This release adds the new pricingscience.similarity module.
Added
-
Add
pricingscience.similarity -
Add
inflectlibrary to neural image
Changed
-
Updated dependencies
-
Improved error handling from backend
v7 (2023-05-25) - Backend supported version: 10.3
This release changes 'importance' function from 'feature_selection' by adding a reliability measure.
This release adds the pricingscience.clustering module, providing hierarchical clustering features.
Added
-
Add
pricingscience.clustering
Changed
-
importance now returns a tuple[importance, reliability_measure].
v6 (2023-05-02) - Backend supported version: 10.3
Official version supporting Bee’s Knees 10.3.x backend.
v5 (2023-04-18) - Backend supported version: 10.2
Official version supporting Bee’s Knees 10.2.x backend.
Added
-
Added
leidenalglibrary to datascience and neural images -
Added
umap-learnlibrary to datascience and neural images -
Added
BERTopiclibrary to neural images
v4 (2023-02-27) - Backend supported version: 10.1
Official version supporting Bee’s Knees 10.1.x backend.
Added
-
push_pandasmethod now supports the optionalcolumn_labelsparameter. -
Added
dythonlibrary to datascience and neural images. -
Added
optunalibrary to datascience and neural images. -
Added
igraphlibrary to datascience and neural images.
Changed
-
Bump version of
werkzeugpackage.
v3 (2023-01-31) - Backend supported version: 10.0
Official version supporting Bee’s Knees 10.0.x backend. No addition or change since v2.
v2 (2023-01-24) - Backend supported version: 9.4
This release upgrades Python version to 3.10 and changes the API vocabulary to be more consistent with the backend one. This is a major breaking change.
Added
-
Data Source, Datamarts, Model tables and Model attachments collection can now be indexed by name or by index (MR45).
pfx.get_context().datasources().get_by_name("my_data_source")can now be writtenpfx.get_context().datasources()["my_data_source"] -
pynndescentlibrary to datascience and neural images (MR51).
Changed
-
Uses Python 3.10 (MR48).
-
Improved backend vocabulary consistency by renaming API parameters.
Warning: this involves many breaking changes if you were referring to these parameters by name when calling the corresponding methods (MR38).-
Mainly
idoruniqueidrenamed totypedidandjstrenamed tojst_idwhen relevant. -
Refer to the API documentation for more information.
-
-
When a backend exception occurs, the entire exception stack trace is printed in the job messages instead of the exception message only (MR47).
v1 (2022-12-20) - Backend supported version: 9.4
First version of the pricefx-python-engine.