Optional --output_path arg added to evaluate subcommand

[rosielickorish]

Add an optional --output_path argument to the evaluate subcommand, allowing callers to specify a custom directory for saved predictions. When omitted, output falls back to the existing MLflow artifacts/test
directory. The predict subcommand default of "data" is unchanged.

• Add --output_path (default None) to evaluate_parser in main.py
• Route evaluate output in main_cli: use custom path when set, else
  fall back to <artifacts_dir>/test
• Add parser and integration tests covering both the custom-path and
  fallback cases

#130

[romeokienzler]

@rosielickorish Thanks for the PR — this is clean and well-scoped (issue #130). A few quick notes, nothing blocking:

What's needed

• CI is still running — pytests, pre-commit-run, CodeQL (Python), security, and pip-audit (deps) are pending. Please make sure they go green. If pip-audit fails on PYSEC-2026-3624 (lightning), that's a known repo-wide infra blocker being fixed separately — not your fault, so no action needed on that one.
• DCO already passes and commits are signed off — 👍

Looks good already

• README table updated with --output_path, and the getattr(args, "output_path", None) or <artifacts_dir>/test fallback keeps the predict default (data) untouched.
• Tests cover both the custom-path and fallback routing plus parser registration/defaults for both subcommands — nice coverage of the edge cases.
• No new dependencies and no config-YAML params introduced (this is a CLI arg), so no examples/config/tests/config changes are required.

I'll leave the merge decision to a maintainer.

— 🤖 _automated pre-review; a maintainer will follow up_

[albanpuech]

@rosielickorish , thank you for the PR. Could you update the docs accordingly please? thank you!

[romeokienzler]

@albanpuech good catch — the PR updates the CLI table in README.md but not the matching table in the mkdocs site.

@rosielickorish the docs table to update is in docs/quick_start/quick_start.md: the evaluate subcommand's argument table (around the --save_output / --mp_context rows, ~L110–112) is missing an --output_path row. The predict table already documents --output_path (L151), so only the evaluate table needs the new row — matching the one you added to README.md.

Everything else still looks good: all checks are green (including pip-audit), and the tests cover both the custom-path and fallback routing.

Merge decision stays with a maintainer.

— 🤖 _automated pre-review; a maintainer will follow up_

[romeokienzler]

@rosielickorish Thanks — the new commit (54cddb2) adds the --output_path row to the evaluate table in docs/quick_start/quick_start.md, right where @albanpuech asked, and it matches the README.md row. That was the only outstanding item from the last pass.

CI is re-running after the push (pytests, pre-commit-run, CodeQL, security, pip-audit pending); DCO is green and commits are signed off. Assuming the checks come back green, this looks ready for a maintainer's look.

Merge decision stays with a maintainer.

— 🤖 _automated pre-review; a maintainer will follow up_

[albanpuech]

Thank you @rosielickorish :)