feat(automl): time-series backtesting curve support#120
Conversation
https://redhat.atlassian.net/browse/RHOAIENG-65711 Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
|
Warning Review limit reached
More reviews will be available in 58 minutes and 21 seconds. Learn how PR review limits work. Your organization has run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After more reviews become available, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available. Please see our Fair Usage Limits Policy for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Central YAML (base), Organization UI (inherited) Review profile: CHILL Plan: Enterprise Run ID: 📒 Files selected for processing (8)
📝 WalkthroughWalkthroughAdds a back_testing builder that produces a schema-aligned back_testing.json from predictor backtest APIs, integrates it into the training component (writes metrics/back_testing.json and conditionally records its path), expands the timeseries notebook with a Back-testing summary and improved known-covariates handling, adds tests for the builder and integration, and updates a CI import exception and several pipeline description strings. Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Security & Code Quality Issues
🚥 Pre-merge checks | ✅ 3 | ❌ 1❌ Failed checks (1 inconclusive)
✅ Passed checks (3 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 4
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
components/training/automl/autogluon_timeseries_models_training/component.py (1)
392-400:⚠️ Potential issue | 🟡 Minor | ⚡ Quick win
back_testingpath written unconditionally even when generation failed.The backtest block (lines 351-371) swallows failures and skips writing
back_testing.json, butmodel.jsonalways advertisesmetrics/back_testing.json. On the warning path the referenced file does not exist, so any consumer trustingmodel.jsonmetadata will hit a missing file. Gate the metadata entry on actual write success.🛠️ Track write success and conditionally add the path
try: back_testing_payload = build_back_testing_json( ... ) with (metrics_path / "back_testing.json").open("w", encoding="utf-8") as f: json.dump(back_testing_payload, f, indent=2) + back_testing_written = True except Exception as backtest_exc: logger.warning( "Could not generate back_testing.json for model %r: %s. Skipping backtest artifact.", model_name_full, backtest_exc, ) + back_testing_written = False"metrics": str(Path(model_name_full) / "metrics"), - "back_testing": str(Path(model_name_full) / "metrics" / "back_testing.json"), },Then add
back_testingtolocationonly whenback_testing_writtenis true.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/autogluon_timeseries_models_training/component.py` around lines 392 - 400, The model metadata always includes a back_testing path even when the backtest write failed; update the code that builds model_metadata so it only inserts the "back_testing": str(Path(model_name_full) / "metrics" / "back_testing.json") entry into the "location" dict when the backtest write actually succeeded (use the existing back_testing_written flag from the backtest block); i.e., build the base location dict first, then if back_testing_written is True add the back_testing key before assigning model_metadata.
🧹 Nitpick comments (6)
.github/scripts/check_imports/import_exceptions.yaml (1)
14-17: ⚡ Quick winNarrow the
numpyimport exception scope to avoid policy overreach.Line 14 currently grants
numpyto every file matchingcomponents/**/shared/*, which weakens least-privilege controls for import governance and can hide unintended dependency spread. Scope this exception to the specific shared file(s) that requirenumpy.Proposed policy tightening
"components/**/shared/*": - pandas - - numpy - component +"components/training/automl/shared/back_testing.py": + - numpyAs per coding guidelines: “REVIEW PRIORITIES: 1. Security vulnerabilities ... 2. Architectural issues and anti-patterns”.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In @.github/scripts/check_imports/import_exceptions.yaml around lines 14 - 17, The current import_exceptions.yaml grants "numpy" to the broad pattern "components/**/shared/*", which is too permissive; restrict the exception by removing "numpy" from that wildcard entry and add explicit exception entries for only the specific shared module file(s) that actually import numpy (e.g., replace the wildcard pattern with exact filenames under the shared directory that need numpy or add new keys for "components/.../shared/specific_shared_file.py"), updating the "components/**/shared/*" block to keep only allowed libs and adding targeted "numpy" entries for the exact shared file paths so the rule maintains least-privilege.components/training/automl/shared/tests/test_back_testing.py (2)
53-60: ⚡ Quick winCover the actual per-item branch here.
This fixture only has one series, so it never proves
_holdout_frame()keeps the lastprediction_lengthrows per item. A regression totsdf.tail(prediction_length)would still pass this test. Add at least a two-item panel and assert each item retains its own final horizon.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/shared/tests/test_back_testing.py` around lines 53 - 60, Update the test_holdout_frame_takes_last_prediction_length_per_item to include multiple series so it verifies per-item behavior: create a panel with at least two distinct items using _make_panel (e.g., item IDs "A" and "B") with different time-series values, call _holdout_frame(panel, prediction_length=2), then assert that the resulting holdout contains exactly the last prediction_length rows for each item (check both length and that holdout["target"] contains the expected tail values for each item individually, and optionally verify holdout["item_id"] or equivalent grouping to ensure rows belong to the correct series).
129-158: ⚡ Quick winMake the ranking fixture discriminate between metrics.
This does not actually verify
eval_metricdispatch because the same series wins under every common lower-is-better point metric with the current numbers. Use scale-skewed targets/predictions soMAPEandRMSErank different items; otherwise an implementation that ignoreseval_metricstill passes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/shared/tests/test_back_testing.py` around lines 129 - 158, The test currently uses symmetric errors so every point metric ranks the same series; change the synthetic targets/predictions in test_ranks_by_point_metric_matching_eval_metric so RMSE and MAPE would choose different winners (e.g., make one series have small absolute errors but large relative errors by using very small target values for one series and larger targets for the other, or scale predictions so one series has low RMSE but high MAPE), then keep eval_metric="RMSE" and assert the expected best_performer (item_id "good") still wins; this forces build_back_testing_json (and its use of predictor.evaluate/predictor.backtest_predictions) to actually dispatch on the eval_metric rather than relying on identical ranking under all metrics. Ensure the modified predictions/targets are applied to window_targets, good_preds, bad_preds (the DataFrames used to build predictions and window_targets) so the test discriminates between metrics.components/training/automl/shared/back_testing.py (2)
291-324: 💤 Low valueRedundant recomputation of per-window metrics and forecast rows.
_item_window_metrics(which itself calls_forecast_data_for_item) runs once per item during ranking, then_series_payloadrecomputes both again for the best/worst series. For the selected performers, forecast extraction is effectively done three times. Caching the per(item_id, window_id)result would remove the duplication. Low priority given the 500-point cap per window.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/shared/back_testing.py` around lines 291 - 324, The code recomputes per-window metrics and forecast rows multiple times: first when building per_item_window_metrics using _item_window_metrics (which calls _forecast_data_for_item), and again inside _series_payload for best/worst series; this is wasteful. Fix by computing and caching results keyed by (item_id, window_id) when you first iterate over zip(predictions_windows, targets_windows) (the same loop that fills per_item_window_metrics), storing both the metrics and forecast_data; then have _series_payload read from that cache instead of calling _item_window_metrics or _forecast_data_for_item again. Ensure the cache is populated before calling _series_ranking_metric/_select_best_worst and used by _series_payload to build windows_payload.
98-104: 💤 Low valueNumeric-column detection diverges from
_quantile_bounds.
_mean_prediction_columnusesto_finite_float(c), which rejects string column labels (AutoGluon emits quantile columns as strings like"0.1"), sonumeric_colsstays empty and it falls back tocolumns[0]._quantile_boundsinstead usesfloat(col), which parses those strings. The two helpers will disagree about which columns are numeric. Harmless while"mean"is present, but brittle if it isn't.♻️ Align parsing with _quantile_bounds
- numeric_cols = [c for c in predictions.columns if to_finite_float(c) is not None] - if numeric_cols: - return str(numeric_cols[0]) + def _as_float(col: Any) -> float | None: + try: + return float(col) + except (TypeError, ValueError): + return None + numeric_cols = [c for c in predictions.columns if _as_float(c) is not None] + if numeric_cols: + # Prefer the median (0.5) if present, else the first numeric column. + median = next((c for c in numeric_cols if _as_float(c) == 0.5), numeric_cols[0]) + return str(median)🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/shared/back_testing.py` around lines 98 - 104, The _mean_prediction_column helper currently uses to_finite_float(c) which rejects numeric-looking string labels and therefore can disagree with _quantile_bounds; change the numeric-column detection in _mean_prediction_column to mirror _quantile_bounds by attempting to parse each column label with float(col) (catching ValueError/TypeError) to consider it numeric, so string labels like "0.1" are recognized; preserve the existing fallbacks (prefer "mean", then first numeric-parsed column, then columns[0]) and keep references to the function name _mean_prediction_column and the differing helper _quantile_bounds when making this change.components/training/automl/autogluon_timeseries_models_training/component.py (1)
349-349: 💤 Low valueMove the import out of the per-model loop.
from ... import build_back_testing_jsonre-runs the import machinery on everytop_modelsiteration. Hoist it next to the other deferred imports (e.g., near line 255) so it executes once.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@components/training/automl/autogluon_timeseries_models_training/component.py` at line 349, The import of build_back_testing_json is inside the per-model loop (iterating over top_models); hoist it so it runs once by moving "from kfp_components.components.training.automl.shared.back_testing import build_back_testing_json" out of the loop and placing it with the other deferred imports (near where other late imports are declared) and then use the imported build_back_testing_json within the loop; update any local references to ensure they still point to the same symbol.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@components/training/automl/shared/back_testing.py`:
- Around line 386-400: per_window_metrics currently stores window_scores
returned by predictor.evaluate() which uses AutoGluon’s "higher-is-better"
convention (so error metrics like MAPE/RMSE/MAE are negated) causing sign
mismatch with series_analysis/_point_errors() which emits positive error values;
fix by normalizing those metrics before serialization: after window_scores =
filter_finite_metrics(predictor.evaluate(...)) (and before appending to
per_window_metrics) detect the known error metric keys (e.g.,
"MAPE","RMSE","MAE") and multiply their values by -1 (or otherwise flip their
sign) so stored per_window_metrics["metrics"] contains positive error
magnitudes, or alternatively add a boolean flag (e.g., "higher_is_better":
true/false) alongside metrics to explicitly document the convention. Ensure
changes reference window_scores, predictor.evaluate, filter_finite_metrics,
per_window_metrics and are consistent with series_analysis/_point_errors().
In
`@components/training/automl/shared/notebook_templates/timeseries_notebook.ipynb`:
- Around line 159-163: The _download_model function uses model_name directly to
form base_dir and later targets, allowing path traversal; validate and sanitize
model_name before using it as a local extraction root by rejecting or
normalizing any input containing path separators or parent-traversal (e.g.,
'..') and only allowing a safe basename (use os.path.basename or a strict
whitelist/regex). Then construct base_dir from a fixed workspace directory
joined with that sanitized basename (not the raw model_name), and perform
canonicalization (os.path.abspath/os.path.realpath) on both base_dir and any
computed abs_target and enforce containment (abs_target.startswith(base_dir +
os.sep)) to ensure downloads cannot escape the intended workspace. Ensure checks
reference the existing symbols model_name, base_dir, target, relative, and
abs_target in _download_model.
- Around line 58-60: The global suppression using
warnings.filterwarnings("ignore") mutes important security and TLS warnings;
replace it with scoped or targeted suppression instead: remove the global call
and either wrap only the specific noisy code blocks in a
warnings.catch_warnings() context or call warnings.filterwarnings("ignore",
category=<SpecificWarningClass>) for the exact non-security category you want
silenced; search for the warnings.filterwarnings("ignore") call and update it so
that warnings.catch_warnings() is used around benign code sections or the filter
is restricted to a precise warning category/message rather than globally
silencing all warnings.
- Around line 184-188: The current exception handler for SSLError around
_download_model/_get_s3_bucket silently falls back to an insecure verify=False
retry; remove that insecure retry and instead fail closed: when catching
SSLError (from _download_model or _get_s3_bucket), log the error with details
and re-raise or surface the exception to the caller so the download fails; if an
insecure bypass is absolutely required, require an explicit, local-only opt-in
flag (e.g., a clearly named parameter) and document it, but do not perform an
automatic verify=False retry.
---
Outside diff comments:
In
`@components/training/automl/autogluon_timeseries_models_training/component.py`:
- Around line 392-400: The model metadata always includes a back_testing path
even when the backtest write failed; update the code that builds model_metadata
so it only inserts the "back_testing": str(Path(model_name_full) / "metrics" /
"back_testing.json") entry into the "location" dict when the backtest write
actually succeeded (use the existing back_testing_written flag from the backtest
block); i.e., build the base location dict first, then if back_testing_written
is True add the back_testing key before assigning model_metadata.
---
Nitpick comments:
In @.github/scripts/check_imports/import_exceptions.yaml:
- Around line 14-17: The current import_exceptions.yaml grants "numpy" to the
broad pattern "components/**/shared/*", which is too permissive; restrict the
exception by removing "numpy" from that wildcard entry and add explicit
exception entries for only the specific shared module file(s) that actually
import numpy (e.g., replace the wildcard pattern with exact filenames under the
shared directory that need numpy or add new keys for
"components/.../shared/specific_shared_file.py"), updating the
"components/**/shared/*" block to keep only allowed libs and adding targeted
"numpy" entries for the exact shared file paths so the rule maintains
least-privilege.
In
`@components/training/automl/autogluon_timeseries_models_training/component.py`:
- Line 349: The import of build_back_testing_json is inside the per-model loop
(iterating over top_models); hoist it so it runs once by moving "from
kfp_components.components.training.automl.shared.back_testing import
build_back_testing_json" out of the loop and placing it with the other deferred
imports (near where other late imports are declared) and then use the imported
build_back_testing_json within the loop; update any local references to ensure
they still point to the same symbol.
In `@components/training/automl/shared/back_testing.py`:
- Around line 291-324: The code recomputes per-window metrics and forecast rows
multiple times: first when building per_item_window_metrics using
_item_window_metrics (which calls _forecast_data_for_item), and again inside
_series_payload for best/worst series; this is wasteful. Fix by computing and
caching results keyed by (item_id, window_id) when you first iterate over
zip(predictions_windows, targets_windows) (the same loop that fills
per_item_window_metrics), storing both the metrics and forecast_data; then have
_series_payload read from that cache instead of calling _item_window_metrics or
_forecast_data_for_item again. Ensure the cache is populated before calling
_series_ranking_metric/_select_best_worst and used by _series_payload to build
windows_payload.
- Around line 98-104: The _mean_prediction_column helper currently uses
to_finite_float(c) which rejects numeric-looking string labels and therefore can
disagree with _quantile_bounds; change the numeric-column detection in
_mean_prediction_column to mirror _quantile_bounds by attempting to parse each
column label with float(col) (catching ValueError/TypeError) to consider it
numeric, so string labels like "0.1" are recognized; preserve the existing
fallbacks (prefer "mean", then first numeric-parsed column, then columns[0]) and
keep references to the function name _mean_prediction_column and the differing
helper _quantile_bounds when making this change.
In `@components/training/automl/shared/tests/test_back_testing.py`:
- Around line 53-60: Update the
test_holdout_frame_takes_last_prediction_length_per_item to include multiple
series so it verifies per-item behavior: create a panel with at least two
distinct items using _make_panel (e.g., item IDs "A" and "B") with different
time-series values, call _holdout_frame(panel, prediction_length=2), then assert
that the resulting holdout contains exactly the last prediction_length rows for
each item (check both length and that holdout["target"] contains the expected
tail values for each item individually, and optionally verify holdout["item_id"]
or equivalent grouping to ensure rows belong to the correct series).
- Around line 129-158: The test currently uses symmetric errors so every point
metric ranks the same series; change the synthetic targets/predictions in
test_ranks_by_point_metric_matching_eval_metric so RMSE and MAPE would choose
different winners (e.g., make one series have small absolute errors but large
relative errors by using very small target values for one series and larger
targets for the other, or scale predictions so one series has low RMSE but high
MAPE), then keep eval_metric="RMSE" and assert the expected best_performer
(item_id "good") still wins; this forces build_back_testing_json (and its use of
predictor.evaluate/predictor.backtest_predictions) to actually dispatch on the
eval_metric rather than relying on identical ranking under all metrics. Ensure
the modified predictions/targets are applied to window_targets, good_preds,
bad_preds (the DataFrames used to build predictions and window_targets) so the
test discriminates between metrics.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Central YAML (base), Organization UI (inherited)
Review profile: CHILL
Plan: Enterprise
Run ID: f2b39a45-fdab-42d5-9006-f5f8c64ae2e1
📒 Files selected for processing (7)
.github/scripts/check_imports/import_exceptions.yamlcomponents/training/automl/autogluon_timeseries_models_training/README.mdcomponents/training/automl/autogluon_timeseries_models_training/component.pycomponents/training/automl/shared/back_testing.pycomponents/training/automl/shared/notebook_templates/timeseries_notebook.ipynbcomponents/training/automl/shared/run_status_templates/pipelines/autogluon-timeseries-training-pipeline.jsoncomponents/training/automl/shared/tests/test_back_testing.py
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
|
/ok-to-test |
|
@dryszka please take a look. 
|
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@components/training/automl/shared/tests/test_back_testing.py`:
- Around line 357-359: Update the stale comment above the assertions in the test
to reflect that for a single series, best_performer and worst_performer are both
set (non-None) — i.e., the single series is considered both best and worst per
_build_series_analysis — so change the comment to state that best and worst
should be non-None (the same series) to match the assertions
payload["series_analysis"]["best_performer"] and
payload["series_analysis"]["worst_performer"].
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Central YAML (base), Organization UI (inherited)
Review profile: CHILL
Plan: Enterprise
Run ID: 83e4819e-ad5f-447a-9a17-e1866ec19440
📒 Files selected for processing (8)
components/training/automl/autogluon_timeseries_models_training/component.pycomponents/training/automl/autogluon_timeseries_models_training/tests/test_component_unit.pycomponents/training/automl/shared/back_testing.pycomponents/training/automl/shared/tests/test_back_testing.pypipelines/data_processing/autorag/documents_indexing_pipeline/pipeline.pypipelines/training/automl/autogluon_tabular_training_pipeline/pipeline.pypipelines/training/automl/autogluon_timeseries_training_pipeline/pipeline.pypipelines/training/autorag/documents_rag_optimization_pipeline/pipeline.py
✅ Files skipped from review due to trivial changes (2)
- pipelines/training/autorag/documents_rag_optimization_pipeline/pipeline.py
- pipelines/training/automl/autogluon_timeseries_training_pipeline/pipeline.py
🚧 Files skipped from review as they are similar to previous changes (2)
- components/training/automl/autogluon_timeseries_models_training/component.py
- components/training/automl/shared/back_testing.py
| # For single series, best and worst should be None (only one series) | ||
| assert payload["series_analysis"]["best_performer"] is not None | ||
| assert payload["series_analysis"]["worst_performer"] is not None |
There was a problem hiding this comment.
Stale comment contradicts the assertion.
The comment claims best/worst "should be None (only one series)", but the asserts require both to be non-None — and per _build_series_analysis, a single series is both best and worst performer, so the assertions are right and the comment is wrong. Fix the comment to avoid misleading future maintainers into "correcting" the test.
📝 Proposed fix
- # For single series, best and worst should be None (only one series)
+ # Single series is both best and worst performer, so both are populated.
assert payload["series_analysis"]["best_performer"] is not None
assert payload["series_analysis"]["worst_performer"] is not None📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
| # For single series, best and worst should be None (only one series) | |
| assert payload["series_analysis"]["best_performer"] is not None | |
| assert payload["series_analysis"]["worst_performer"] is not None | |
| # Single series is both best and worst performer, so both are populated. | |
| assert payload["series_analysis"]["best_performer"] is not None | |
| assert payload["series_analysis"]["worst_performer"] is not None |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@components/training/automl/shared/tests/test_back_testing.py` around lines
357 - 359, Update the stale comment above the assertions in the test to reflect
that for a single series, best_performer and worst_performer are both set
(non-None) — i.e., the single series is considered both best and worst per
_build_series_analysis — so change the comment to state that best and worst
should be non-None (the same series) to match the assertions
payload["series_analysis"]["best_performer"] and
payload["series_analysis"]["worst_performer"].
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
|
/retest |
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
Signed-off-by: Lukasz Cmielowski <lcmielow@redhat.com> Assisted-by: Cursor
|
/retest |
|
/lgtm |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: dryszka, Mateusz-Switala The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
|
/retest |
openshift-merge-bot
Bot
merged commit 85fee01
into
opendatahub-io:main
3 participants
Description of your changes:
Checklist:
Pre-Submission Checklist
Learn more about the pull request title convention used in this repository.
Additional Checklist Items for New or Updated Components/Pipelines
metadata.yamlincludes freshlastVerifiedtimestampare present and complete
snake_casenaming conventionSummary by CodeRabbit
New Features
Documentation
Tests