[maximusunc]

Three breaking changes, released together as 0.6.0 so the Tests repo, the Test Harness and the Information Radiator switch over once.

TestAsset.output_id → output_ids

An expected output can now be named by several CURIEs, any of which counts as the expected answer. For example, sirolimus is CHEBI:9168 and PUBCHEM.COMPOUND:5284616; node normalization no longer merges these, so Translator returns them as separate answers with their evidence split.

• output_ids is a multivalued uriorcurie that defaults to [] (performance assets have no expected output).
• Data still using output_id is rejected, and a single CURIE has to be a one-item list.
• The examples are converted, and there's a new two-CURIE sirolimus example (Good-TestAsset-002.json).

test_env removed from test cases

A suite is the same in every environment, so test cases no longer say which one they run in. The Test Harness is told instead (--env).

• test_env is removed from TestCase and PathfinderTestCase, so from every test case class.
• TestRunSession keeps it, since a record of a run should say where it ran.

Python 3.10+

CI was failing because snok/install-poetry installs the latest Poetry, and since 2.3.0 that needs Python 3.10, so the 3.9 job couldn't start.

• The package now requires Python >=3.10,<3.13.
• CI tests 3.10, 3.11 and 3.12 with the latest Poetry, and the docs and release workflows run on 3.10.
• poetry.lock is relocked with Poetry 2.5.1. Every package keeps its version; importlib-metadata and zipp, needed only on 3.9, drop out. The lock format is now 2.1, which needs Poetry 2.0 or later.

Also

• The datamodels and project/ artifacts are regenerated with linkml 1.7.8, the version in poetry.lock. I checked that it reproduces the committed Pydantic model exactly before the change. The large OWL/SHACL diffs are the generators reordering lines; the only real changes in them are the two slots above.
• New tests: tests/test_output_ids.py and tests/test_test_env.py.
• Version bumped to 0.6.0.

Testing

• CI steps: poetry install --no-root, poetry install and make test, run locally with Poetry 2.5.1 on Python 3.10, 3.11 and 3.12. All pass, with 39 tests and 4 skipped as before.
• Release build: the release workflow's poetry build, run with poetry-dynamic-versioning and a v0.6.0 tag, gives translator_testing_model-0.6.0 with Requires-Python: >=3.10,<3.13.
• Docs install: poetry install -E docs resolves.

Rollout

1. Merge this, then create the v0.6.0 release, which publishes to PyPI.
2. Multi-CURIE expected outputs, a run-level --env, and shared suites in the radiator TestHarness#62 pins translator-testing-model==0.6.0; its CI can't install until the release is out. The harness image is on Python 3.11.
3. NCATSTranslator/Tests is converted with a migration script, merged in the same window as the harness deploy.

🤖 Generated with Claude Code

https://claude.ai/code/session_01S2yuREtzxTuRqVGp8Q8a6d