fix: Break cycle between skore and skore-cli

[thomass-dev]

Fix after #3255 that introduce a bug when installing skore from source:

error: No solution found when resolving dependencies
cause: Because skore-cli>=0.5.0 depends on skore[hub]>=0.27 and skore[hub]==0.0.0+unknown depends on skore-cli==0.5, we can conclude that skore[hub]==0.0.0+unknown depends on skore[hub]>=0.27.
       And because only skore[hub]==0.0.0+unknown is available and you require skore[hub], we can conclude that your requirements are unsatisfiable.

skore-cli being not directly used, remove it from the direct dependencies.

---

Fix compatibility with mlflow>3 and python<3.15.

[github-actions]

Documentation preview @ 697aa45

[glemaitre]

We consider this a temporary fix and we are just migrating the CLI into skore :)

[thomass-dev]

We consider this a temporary fix and we are just migrating the CLI into skore :)

Exactly!

[glemaitre]

Regarding the drop of MLflow 2, I'm all for it because it would be needed for this fix: #3305. Basically, their I saw that the support for polars only start in a certain version of MLflow 3. So we can bump to this version directly.

[github-actions]

Coverage Report for skore/

Warning: Comment truncated due to GitHub's 65536 character limit

File	Stmts	Miss	Branch	BrPart	Cover	Missing
skore/src/skore
__init__.py	26	2	0	0	92%	104–105
_config.py	58	3	12	1	94%	70, 116->117, 117–118
_exceptions.py	4	4	0	0	0%	4, 15, 19, 23
skore/src/skore/_checks
__init__.py	4	0	0	0	100%
accessor.py	35	0	14	0	100%
base.py	103	0	28	2	98%	271->273, 273->277
builtin.py	17	0	0	0	100%
skd001_overfitting.py	21	0	4	0	100%
skd002_underfitting.py	22	0	2	0	100%
skd003_metrics_consistency.py	31	0	8	0	100%
skd004_high_class_imbalance.py	25	0	4	0	100%
skd005_underrepresented_classes.py	25	0	4	0	100%
skd006_coefficients_interpretation.py	22	0	4	0	100%
skd007_mdi_high_cardinality_bias.py	25	0	4	0	100%
skd008_correlated_features.py	28	0	6	0	100%
skd009_worse_than_baseline.py	35	1	10	1	96%	66->67, 67
skd010_slower_than_baseline.py	46	0	12	0	100%
skd011_golden_feature.py	83	2	26	0	98%	207–208
skd012_useless_features.py	24	0	2	0	100%
skd013_train_test_time_overlap.py	33	0	14	0	100%
skd014_hyperparams_at_search_edge.py	53	1	26	1	97%	88->90, 90
skd015_search_params_to_tune.py	43	0	18	1	98%	61->60
skd016_estimator_not_tuned.py	50	1	24	1	97%	46->47, 47
tunable_hyperparameters.py	5	0	0	0	100%
utils.py	199	13	88	11	92%	285->286, 286–288, 306->307, 307, 311->318, 312->313, 313, 318–320, 322–324, 326, 329->330, 330
skore/src/skore/_dispatch
__init__.py	4	0	0	0	100%
compare.py	5	0	0	0	100%
evaluate.py	44	0	22	0	100%
skore/src/skore/_displays
__init__.py	4	0	0	0	100%
base.py	73	2	14	1	97%	72->74, 74–75
utils.py	158	3	76	4	97%	256–257, 276->272, 381->375, 455->456, 456, 489->492
skore/src/skore/_displays/data
__init__.py	4	0	0	0	100%
table_report.py	193	3	62	3	98%	123->124, 124–125, 415->418, 726->729, 729
skore/src/skore/_displays/inspection
__init__.py	0	0	0	0	100%
calibration_curve.py	80	9	14	7	83%	117->119, 119–122, 124, 232->235, 235, 238–239, 268->269, 269
coefficients.py	202	0	100	1	99%	277->280
impurity_decrease.py	101	2	34	3	96%	220->223, 446->447, 447, 488->491, 491
permutation_importance.py	197	1	90	1	99%	662->663, 663
utils.py	32	0	10	0	100%
skore/src/skore/_displays/metrics
__init__.py	4	0	0	0	100%
confusion_matrix.py	197	0	66	2	99%	328->330, 430->433
metrics_summary_display.py	175	0	62	1	99%	143->158
precision_recall_curve.py	116	0	32	1	99%	337->341
prediction_error.py	175	0	58	2	99%	303->305, 326->307
roc_curve.py	120	0	34	2	99%	312->314, 359->353
skore/src/skore/_metrics
__init__.py	3	0	0	0	100%
metrics.py	375	1	86	1	99%	477->478, 478
registry.py	29	0	12	0	100%
skore/src/skore/_plugins
__init__.py	12	0	0	0	100%
requirements.py	66	2	26	0	98%	78–79
serde.py	32	0	8	0	100%
skore/src/skore/_plugins/hub
__init__.py	9	2	0	0	78%	15, 20
exception.py	2	0	0	0	100%
json.py	10	1	2	1	83%	14->16, 16
metric.py	37	2	8	2	91%	84->85, 85, 92->93, 93
skore/src/skore/_plugins/hub/artifact
__init__.py	0	0	0	0	100%
artifact.py	23	0	4	0	100%
environment.py	15	3	0	0	80%	50–52
pickle.py	29	0	6	0	100%
serializer.py	27	0	2	0	100%
upload.py	26	0	4	0	100%
skore/src/skore/_plugins/hub/artifact/media
__init__.py	6	0	0	0	100%
checks.py	13	0	0	0	100%
data.py	23	0	0	0	100%
inspection.py	58	0	10	0	100%
media.py	12	0	0	0	100%
model.py	10	0	0	0	100%
performance.py	57	0	0	0	100%
skore/src/skore/_plugins/hub/authentication
__init__.py	4	0	0	0	100%
key.py	25	0	6	0	100%
login.py	74	2	6	2	95%	81->82, 82, 84->87, 87
uri.py	10	0	0	0	100%
skore/src/skore/_plugins/hub/authentication/registry
__init__.py	2	0	0	0	100%
distant.py	62	0	6	0	100%
local.py	93	0	20	0	100%
skore/src/skore/_plugins/hub/client
__init__.py	0	0	0	0	100%
client.py	100	9	20	3	90%	142, 190–192, 194–195, 197, 199, 201, 255->289
skore/src/skore/_plugins/hub/project
__init__.py	0	0	0	0	100%
project.py	137	6	28	5	93%	87->88, 88, 108->113, 113, 127->128, 128, 324->325, 325, 431, 456->461, 461
skore/src/skore/_plugins/hub/report
__init__.py	3	0	0	0	100%
cross_validation_report.py	191	6	44	3	96%	179->188, 188–192, 206->207, 207, 237->240
estimator_report.py	17	0	0	0	100%
report.py	74	1	8	1	98%	189->191, 191
skore/src/skore/_plugins/local
__init__.py	2	0	0	0	100%
project.py	250	3	58	2	98%	455, 494->495, 495, 515->516, 516
skore/src/skore/_plugins/mlflow
__init__.py	5	0	0	0	100%
project.py	307	26	76	10	91%	51, 53–55, 120–121, 178->179, 179, 279->286, 286, 301->304, 304, 319–320, 389->392, 407->408, 408, 458->459, 459, 556, 579->580, 580, 581->582, 582, 598, 600–605, 607–609
reports.py	160	6	36	5	94%	126->129, 129, 173->176, 176, 209->exit, 214–215, 274->277, 277, 287->288, 288
skore/src/skore/_project
__init__.py	0	0	0	0	100%
dependencies.py	19	0	6	0	100%
git.py	30	0	4	0	100%
plugin.py	9	0	0	0	100%
project.py	76	1	26	1	98%	161->162, 162
protocol.py	5	5	0	0	0%	3, 5, 7, 13, 16
summary.py	118	1	48	1	99%	107->108, 108
sync.py	41	0	10	0	100%
types.py	4	0	0	0	100%
skore/src/skore/_reports
__init__.py	4	0	0	0	100%
base.py	157	6	42	6	94%	71, 157->131, 192->exit, 222, 296->299, 299, 307->308, 308, 313->314, 314, 348->349, 349
skore/src/skore/_reports/comparison
__init__.py	9	0	0	0	100%
inspection.py	25	0	2	0	100%
metrics.py	116	2	26	2	97%	418->419, 419, 673->674, 674
report.py	188	7	90	6	95%	278–279, 477->483, 483, 549->534, 556->534, 590->589, 602, 685–687
skore/src/skore/_reports/cross_validation
__init__.py	11	0	0	0	100%
data.py	40	2	14	2	93%	48->49, 49, 74->75, 75
inspection.py	25	0	2	0	100%
metrics.py	101	5	18	2	94%	106->107, 107–108, 110, 113, 620->621, 621
report.py	203	9	46	5	94%	231->232, 232, 236->237, 237, 247->248, 248, 252->253, 253, 354->355, 355, 597, 736–738
skore/src/skore/_reports/estimator
__init__.py	11	0	0	0	100%
data.py	61	1	28	1	98%	64->65, 65
inspection.py	42	1	8	2	94%	297->298, 298, 347->351
metrics.py	116	0	22	0	100%
report.py	323	10	88	7	96%	64->65, 65, 80->81, 81, 254->258, 299->300, 300, 381->382, 382, 435->436, 436, 685->686, 686, 771, 865–867
skore/src/skore/_sklearn
__init__.py	4	0	0	0	100%

[glemaitre]

So also, I would suggest to pin python<3.15. pyarrow is still not available for this version on conda-forge: https://github.com/conda-forge/pyarrow-feedstock/pulls

So it means that we end-up in a complex situation with weird dependencies. We can remove this constraint once pyarrow is supporting this python version.

[glemaitre]

LGTM