ui/tests/test_scoring_live.py

"""Integration test for the REAL scorer seam (ObservableScorer -> backend pipeline -> live model).

Skipped unless a scoring endpoint is configured (OPENBMB_BASE_URL / OPENBMB_TOKEN). Runs base-only
(DISABLE_LORA) so it doesn't require Modal. Proves the adopted scoring path end-to-end on a fixture export.

  OPENBMB_BASE_URL=... OPENBMB_TOKEN=... python -m pytest ui/tests/test_scoring_live.py -q
"""
from __future__ import annotations

import os

import pytest

pytestmark = pytest.mark.skipif(
    not (os.environ.get("OPENBMB_BASE_URL") and os.environ.get("OPENBMB_TOKEN")),
    reason="no scoring endpoint configured (set OPENBMB_BASE_URL / OPENBMB_TOKEN)",
)

FIXTURE = os.path.join(os.path.dirname(__file__), "fixtures", "chatgpt_sample.json")


def test_observable_scorer_end_to_end():
    os.environ.setdefault("DISABLE_LORA", "1")  # base-only: no Modal dependency
    from ui.parsing.parser import parse_export
    from ui.scoring.observable import ObservableScorer
    from ui.scoring import score_to_card
    from ui.data import AXES

    parsed = parse_export(FIXTURE)
    card = score_to_card(ObservableScorer().score(parsed), "Test")

    assert [a.name for a in card.axes] == AXES               # canonical UI axis order
    assert all(0.0 <= a.score <= 10.0 for a in card.axes)    # scores in range
    assert all(a.confidence in ("high", "medium", "low") for a in card.axes)
    assert 0.0 <= card.overall <= 10.0
    assert card.tier[0] in ("D", "C", "B", "A", "S")
    # critical counts are real non-negative tallies
    assert all(n >= 0 for _, n in card.critical.as_pairs())
    assert card.improvement