[bug] Labels and annotations parsed from notebook tags but never rendered in KFP DSL output

[StefanoFioravanzo]

Kale supports tagging notebook cells with label: and annotation: prefixes, parses them in nbprocessor.py, validates them via K8sLabelsValidator/K8sAnnotationsValidator, and stores them in StepConfig — but they are never applied to the generated KFP v2 DSL.

There are two issues:

1. Labels don't even reach the Step object. In nbprocessor.py, cell_labels is parsed (line ~479) but never added to parsed_tags, unlike cell_annotations which is correctly added (line ~509-515). This means tags.get("labels", {}) at Step creation always returns {}.

2. Neither labels nor annotations are rendered in the templates. pipeline_template.jinja2 handles step.config.limits, step.config.enable_caching, and step.config.base_image, but has no code to emit add_pod_label() or add_pod_annotation() calls on the KFP task.

Expected behavior:

When a user tags a cell with e.g. label:team:ml-infra or annotation:prometheus.io/scrape:true, the generated DSL should include:

step_task.add_pod_label("team", "ml-infra")
step_task.add_pod_annotation("prometheus.io/scrape", "true")

Relevant files:

• kale/processors/nbprocessor.py — tag parsing (labels bug at ~line 509)
• kale/step.py:41-54 — StepConfig with labels/annotations fields
• kale/templates/pipeline_template.jinja2 — missing rendering logic
• kale/config/validators.py:142-181 — validators exist but output is unused

[ada333]

/assign