core: add the API surface MSCommunity needs from ModelSEEDpy

[freiburgermsu]

Why

MSCommunity cannot run against released ModelSEEDpy. It imports six names that exist only in the freiburgermsu fork, and calls sixteen MSModelUtil/FBAHelper members that main does not define, so import mscommunity raises on the first line of commhelper.py against every published version:

ImportError: cannot import name 'minimizeFlux_withGrowth' from 'modelseedpy.core'

That blocks distributing MSCommunity on PyPI, because PyPI does not accept direct-URL/VCS dependencies — there is no way to express "install the fork" in package metadata. This PR closes the gap so MSCommunity can depend on plain modelseedpy.

What

Exactly what MSCommunity needs, nothing else. Every change is additive — no existing signature, behaviour or export is altered.

core/exceptions.py — five classes, in the style of the four already there:
ParameterError, ObjectAlreadyDefinedError, NoFluxError, ObjectiveError, ModelError.

core/fbahelper.py — FBAHelper.isnumber, .mediaName, .rxn_mets_list, .solution_to_variables_dict, .convert_kbase_media; plus module-level bioFlux_check and minimizeFlux_withGrowth. The fork's minimizeFlux_withGrowth logged through icecream; that call is dropped so this adds no new dependency.

core/msmodelutl.py — MSModelUtil gains add_medium, add_minimal_objective_cons, add_objective, carbon_exchange_list, carbon_exchange_mets_list, create_constraint, exchange_mets_list, metabolites_set, remove_constraint, remove_cons_vars, run_fba, standard_exchanges.

__init__ gains four optional keyword arguments — copy, environment, climit, o2limit — all defaulting to the current no-op behaviour, so existing MSModelUtil(model) callers are unaffected. The uptake-limit logic is factored into apply_uptake_limits() so it can be called independently. get_pkg_mgr is now passed self.model rather than model: identical when copy=False, and correct when copy=True.

core/__init__.py — re-export bioFlux_check and minimizeFlux_withGrowth. The new exceptions need no change here; from modelseedpy.core.exceptions import * already covers them.

Verification

Run on this branch against main @ a5ea486:

Suite	Before	After
ModelSEEDpy tests/	39 passed	39 passed (identical baseline)
MSCommunity tests/ (ModelSEED/MSCommunity main @ b5f37c4)	7 collection errors	72 passed

Formatted with black, matching .pre-commit-config.yaml.

Notes for review

• The branch is cut from ModelSEED/ModelSEEDpy main, not from the fork's main, so the diff is only these four files rather than the fork's accumulated drift.
• minimizeFlux_withGrowth is imported by MSCommunity but never called. It is included so MSCommunity's import line resolves and because it pairs naturally with bioFlux_check; if you would rather not carry it, MSCommunity can drop the import instead and it can come out of this PR.
• The two module-level helpers were placed in core/fbahelper.py rather than reviving the fork's core/msminimalmedia.py, to keep the change small. Say the word if you would prefer that module.

🤖 Generated with Claude Code

https://claude.ai/code/session_01HFX6x9QKfUuHSuAdhmgL12

[freiburgermsu]

On the red checks — they are pre-existing on main, not from this PR

Both workflows fail identically on main itself, and have on every run for months:

Date	Workflow	Result	Commit
2026-07-21	Run Tox	failure	Testing for conflicts
2026-07-21	Run Pre-Commit	failure	Testing for conflicts
2026-06-17	Run Tox	failure	Testing for conflicts
2026-06-17	Run Pre-Commit	failure	Testing for conflicts
2026-05-08	Run Tox	failure	Testing for conflicts
2026-05-08	Run Pre-Commit	failure	Testing for conflicts

The tox failure is dependency rot rather than anything in the source:

ERROR: Failed to build 'scikit-learn' when getting requirements to build wheel
ModuleNotFoundError: No module named 'pkg_resources'

setup.py:38 pins scikit-learn == 0.24.2, whose build requires pkg_resources; setuptools removed that module in 81.0. So the environment can no longer be created, and the suite never runs. Worth a separate PR — either unpin, or pin setuptools<81 for the build.

Two things I checked so the formatting question does not muddy the review:

• black==22.8.0, the version pinned in .pre-commit-config.yaml, reports "4 files would be left unchanged" on the files this PR touches.
• Running both suites locally against this branch: ModelSEEDpy tests/ gives 39 passed, byte-identical to the unmodified main baseline, and MSCommunity tests/ gives 72 passed where main produces 7 collection errors.

Happy to rebase once CI is healthy again if you would rather see it green before merging.