api/data_io.py:242: from sqlalchemy import create_engine, text as sql_text ModuleNotFoundError: No module named 'sqlalchemy' → ValueError: SQL import requires sqlalchemy (pip install sqlalchemy)

[ranasl62]

Cause: api/data_io.py import_sql() depends on SQLAlchemy, but sqlalchemy is not listed in requirements.txt (CI installs pip install -r requirements.txt then -e .).

Also leaves uncovered: api/data_io.py lines 245–248 (engine connect / read_sql after successful import).

Failure 2 — coverage < 100%

File	Miss	Lines
api/data_io.py	4	245–248
api/jobs.py	4	485, 504–506
api/researcher_routes.py	7	112–113, 137, 154, 166–167, 742
openhealth/trust_pack.py	2	164, 166
TOTAL	17	99.38%

Likely branches: CSV header parse failure, reconcile early-returns, HPO temporal_split → split_by_patient=False, SHAP purge/remirror, trust_pack mirror skip for missing/invalid PNG.

Expected

• CI green with cov-fail-under=100
• Either SQLAlchemy is a declared CI/runtime dep, or the test mocks the ImportError path and separate tests cover the happy path when SQLAlchemy is present

Proposed fix

• Add sqlalchemy to requirements.txt (or requirements-dev.txt + CI install) so import_sql works in CI
• Ensure test_data_io_delete_symlink_and_imports exercises the success path (or split ImportError vs happy-path tests)
• Add small unit tests / fixtures for remaining miss lines (jobs SHAP cleanup, trust_pack.mirror_to_shared early returns, researcher_routes header/reconcile/HPO temporal_split)
• Re-run: PYTHONPATH=. pytest tests/ -m "not e2e" -q --tb=short → 0 fail, coverage ≥ 100%

Notes

• Warnings (Starlette TestClient/httpx, shap colormap, pandas copy=) are noise for this issue; track separately if desired.
• Local runs with --no-cov may hide the coverage gate; CI uses pytest.ini defaults.

Acceptance criteria

• No ModuleNotFoundError: sqlalchemy in CI
• test_data_io_delete_symlink_and_imports passes
• Coverage total ≥ 100% for configured packages (api, openhealth)