llm-ingestion-okf/tests/test_contract_check.py
Kjell Tore Guttormsen bc0b4130f1 feat(tools): okf_contract_check reads the contract's checkable half
Fourteen rules, each with its own code: one "invalid" verdict over fourteen
different defects is a diagnostic no caller can act on. The report quotes its
own denominators -- rules run, excerpts and withheld entries examined -- because
a checker that exempted itself from section 5 would be stating the rule it
breaks.

Three exit codes, not two. "The check did not run" and "the check failed" are
different outcomes, and an unread file reported as a failed check is the fourth
face of the verification law.

The shipped template plus its example payload is the known-positive arm, so a
checker that refuses everything cannot be green on the thirteen negative ones.
A last test asserts every literal the checker enforces appears in the contract
document: two copies of a closed set drift, and the copy nobody reads is the
one that goes wrong.

996 -> 1023 tests.
2026-09-02 16:09:27 +02:00

243 lines
9.3 KiB
Python

"""The consumption contract, checked rather than described.
`docs/consumption-contract.md` states what a consumption skill and its pre-pass
payload MUST carry. Prose has no test, and this repository has already paid for
that once: a published promise without a test goes false without anyone
noticing. So the contract's mechanically checkable half is checked here.
Two disciplines the checker itself is held to:
- **A known-positive, shipped.** The template and its example payload must
PASS. Without that arm a checker that refuses everything is green on every
negative case -- the harness-lies-red failure, one level down.
- **One mutation per arm, each with its own code.** A single "invalid" verdict
over fourteen different defects is a diagnostic no caller can act on, so each
mutation asserts the code it produces, not merely that something failed.
The last test is the anti-drift gate: every literal the checker enforces must
appear in the contract document. Two copies of a closed set drift, and the
copy nobody reads is the one that goes wrong.
"""
from __future__ import annotations
import json
import subprocess
import sys
from pathlib import Path
from typing import Any
import pytest
PROJECT_ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(PROJECT_ROOT / "tools"))
import okf_contract_check # noqa: E402
TEMPLATE = PROJECT_ROOT / "skills" / "okf-consume-template" / "SKILL.md"
EXAMPLE = TEMPLATE.parent / "references" / "example-payload.json"
CONTRACT = PROJECT_ROOT / "docs" / "consumption-contract.md"
def load_example() -> dict[str, Any]:
payload: dict[str, Any] = json.loads(EXAMPLE.read_text(encoding="utf-8"))
return payload
def codes(skill_text: str, payload: Any) -> list[str]:
return [finding.code for finding in okf_contract_check.check(skill_text, payload).findings]
# --- The known-positive, which every negative arm depends on ----------------
def test_shipped_template_and_example_payload_are_conformant() -> None:
"""The CONTROL. A checker that cannot pass anything proves nothing below."""
report = okf_contract_check.check(TEMPLATE.read_text(encoding="utf-8"), load_example())
assert report.findings == ()
def test_the_report_carries_its_denominators() -> None:
""" "Conformant" without a denominator is unmeasured, per the contract's own
section 5. The report says how many rules ran over how many units."""
report = okf_contract_check.check(TEMPLATE.read_text(encoding="utf-8"), load_example())
assert report.rules_evaluated == len(okf_contract_check.RULES)
assert report.rules_evaluated > 0
assert report.excerpts_examined == 3
assert report.withheld_examined == 2
rendered = report.render()
assert str(report.rules_evaluated) in rendered
assert "3" in rendered and "2" in rendered
# --- Payload mutations, one per arm -----------------------------------------
def test_broken_denominator_identity_is_named() -> None:
payload = load_example()
payload["denominators"]["considered"] = 6
assert "denominator_identity" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_delivered_count_disagreeing_with_the_list_is_named() -> None:
payload = load_example()
payload["excerpts"] = payload["excerpts"][:2]
assert "denominator_list_mismatch" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_missing_adjudication_key_is_not_absence() -> None:
"""The contract's sharpest rule: `unknown` is written, never omitted."""
payload = load_example()
del payload["excerpts"][0]["adjudication"]
assert "state_missing" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_an_adjudication_value_outside_the_closed_set_is_named() -> None:
payload = load_example()
payload["excerpts"][0]["adjudication"] = "absent"
assert "state_not_in_set" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_trust_tier_outside_the_closed_set_is_named() -> None:
payload = load_example()
payload["excerpts"][1]["trust_tier"] = "trusted"
assert "state_not_in_set" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
@pytest.mark.parametrize("key", ["bundle_id", "concept_id", "sha256"])
def test_an_excerpt_missing_its_source_marking_is_named(key: str) -> None:
payload = load_example()
del payload["excerpts"][0][key]
assert "source_marking_incomplete" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_bundle_without_a_ref_is_named() -> None:
payload = load_example()
payload["bundle"]["ref"] = ""
assert "ref_missing" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_withheld_concept_without_a_rule_is_an_undeclared_cut() -> None:
payload = load_example()
del payload["withheld"][0]["rule"]
assert "cut_undeclared" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_spending_over_the_limit_is_the_gate_firing() -> None:
payload = load_example()
payload["budget"]["spent"] = payload["budget"]["limit"] + 1
assert "budget_exceeded" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_an_instrument_that_missed_its_known_positive_is_named() -> None:
payload = load_example()
payload["budget"]["known_positive"]["measured"] = 10405
assert "instrument_unvalidated" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_budget_without_an_instrument_is_named() -> None:
payload = load_example()
payload["budget"]["instrument"] = ""
assert "budget_undeclared" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_payload_naming_no_contract_revision_is_named() -> None:
payload = load_example()
del payload["contract"]
assert "contract_unversioned" in codes(TEMPLATE.read_text(encoding="utf-8"), payload)
def test_a_payload_that_is_not_a_mapping_is_refused_not_crashed() -> None:
assert "payload_invalid" in codes(TEMPLATE.read_text(encoding="utf-8"), ["not", "a", "mapping"])
# --- Skill mutations --------------------------------------------------------
def test_a_missing_required_section_is_named() -> None:
text = TEMPLATE.read_text(encoding="utf-8").replace("## Denominators", "## Counts")
assert "skill_section_missing" in codes(text, load_example())
def test_a_missing_required_marking_is_named() -> None:
text = TEMPLATE.read_text(encoding="utf-8").replace("[sourced-not-sufficient]", "[thin]")
assert "skill_marking_missing" in codes(text, load_example())
def test_a_translated_marking_literal_does_not_count() -> None:
"""One literal string, no variants, no translations."""
text = TEMPLATE.read_text(encoding="utf-8").replace(
"[unverifiable-from-bundle]", "[ikke-verifiserbar-fra-bundle]"
)
assert "skill_marking_missing" in codes(text, load_example())
def test_a_skill_omitting_an_adjudication_value_is_named() -> None:
text = TEMPLATE.read_text(encoding="utf-8").replace("`unknown`", "`missing`")
assert "skill_state_missing" in codes(text, load_example())
# --- The command line -------------------------------------------------------
def test_the_command_exits_zero_on_the_shipped_pair() -> None:
exit_code = okf_contract_check.main(["--skill", str(TEMPLATE), "--payload", str(EXAMPLE)])
assert exit_code == 0
def test_the_command_exits_one_on_a_non_conformant_payload(tmp_path: Path) -> None:
payload = load_example()
payload["denominators"]["considered"] = 6
broken = tmp_path / "payload.json"
broken.write_text(json.dumps(payload), encoding="utf-8")
assert okf_contract_check.main(["--skill", str(TEMPLATE), "--payload", str(broken)]) == 1
def test_the_command_separates_could_not_run_from_non_conformant(tmp_path: Path) -> None:
"""Exit 2, never 1: "the check did not run" and "the check failed" are
different outcomes, and collapsing them is the fourth face of the
verification law."""
missing = tmp_path / "absent.json"
assert okf_contract_check.main(["--skill", str(TEMPLATE), "--payload", str(missing)]) == 2
unreadable = tmp_path / "bad.json"
unreadable.write_text("{not json", encoding="utf-8")
assert okf_contract_check.main(["--skill", str(TEMPLATE), "--payload", str(unreadable)]) == 2
def test_the_module_runs_as_a_script() -> None:
result = subprocess.run(
[
sys.executable,
str(PROJECT_ROOT / "tools" / "okf_contract_check.py"),
"--skill",
str(TEMPLATE),
"--payload",
str(EXAMPLE),
],
capture_output=True,
text=True,
check=False,
)
assert result.returncode == 0, result.stderr
assert "conformant" in result.stdout
# --- The anti-drift gate ----------------------------------------------------
def test_every_literal_the_checker_enforces_is_in_the_contract() -> None:
"""The checker and the document state one closed set each. If they can
disagree, the copy nobody reads is the one that goes wrong."""
contract = CONTRACT.read_text(encoding="utf-8")
for literal in (
*okf_contract_check.REQUIRED_MARKINGS,
*okf_contract_check.ADJUDICATION_STATES,
*okf_contract_check.TRUST_TIERS,
):
assert literal in contract, literal
for heading in okf_contract_check.REQUIRED_SECTIONS:
assert heading.lower() in contract.lower(), heading
def test_the_contract_document_is_reachable_from_the_template() -> None:
assert "docs/consumption-contract.md" in TEMPLATE.read_text(encoding="utf-8")