C6. `docs/consumption-contract.md` now says what the pre-pass emits:
- SS 2.5 point 2: sub-questions are written in the bundle's terms and given to
the pre-pass in ONE run where it takes several;
- SS 8's example is `okf-consumption/2` as shipped: `withheld` is the
`{total, by_rule, nearest, complete}` mapping, `coverage` carries
`absent_terms` and `weak`, an excerpt carries `text`, `text_sha256`,
`passage` and `own_title`; point 1 closes the identity on `withheld.total`;
- point 7 keeps the falsified verdicts on record and states the one reading
carried since v1.1, `weak`, with its rule: a word held in no form, or
nothing delivered -- whether a word exists at all, which does not move with
question style or corpus size the way the two shares did;
- point 8 the passage (`{start, end, of}`, `text_sha256` over the passage,
`sha256` still the file's), point 9 several sub-questions in one run (one
cut, `questions`, `subquestions`, per-sub-question coverage), point 10 a
title inherited from the heading above, with `own_title`;
- SS 10 names this library's default ranking (BM25) and `--ranking fusion`.
`okf check` holds the two new forms a reader acts on: `passage_malformed`
(not whole numbers with 0 <= start < end <= of) and `subquestions_unindexed`
(not distinct indices into `questions`, or indices in a payload listing none).
17 -> 19 rules; the two tests that pin the published count move with it.
Each rule is held against a real payload (0 findings) and against that
payload broken six ways.
Editing the contract moved the SS 7.4 known-positive, measured once after the
edit: 19 837 -> 23 672 encoded, 19 358 -> 23 092 raw, delta 479 -> 580. The
example payload and `skills/okf-consume` are regenerated by the published
recipe and `okf check` reports 19 rules, 0 findings on them.
Suite on a clean tree after `git add`: 2443 passed, 2 skipped, 4 xfailed.
ruff, ruff format, mypy --strict clean.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
203 lines
8.8 KiB
Python
203 lines
8.8 KiB
Python
"""Which bundle was this skill generated for? The checker had no rule for it.
|
|
|
|
Measured 2026-09-10 on this repository's HEAD, before this file existed: `okf
|
|
check` reported `conformant: 15 rules over 8 excerpts and 438 withheld entries,
|
|
0 findings` for a skill generated from one corpus against a payload assembled
|
|
from a different one -- and the same line for the UNFILLED template against
|
|
that payload, and the same line again when only the `ref` was foreign. Three
|
|
forms, one hole: nothing compared the identity the skill declares with the
|
|
identity the payload declares.
|
|
|
|
**The `ref` half is what makes this a rule rather than an id check.** Three
|
|
distinct builds on this machine carry the same `bundle_id`, so `bundle_id` does
|
|
not identify the bytes. The contract says so in SS 3.3 -- "a version is the
|
|
producer's assertion; a ref is a fact about bytes" -- and the generated skill
|
|
says it about itself: "If the bundle moves, the ref moves with it and this file
|
|
is stale -- regenerate". A rule comparing ids alone would pass a stale skill.
|
|
|
|
**An identity the rule cannot read is a finding, never a silent pass.** That is
|
|
the template arm: `<CORPUS>` and `<REF>` left in place are not an identity, and
|
|
the template's own rule says a copy leaving a placeholder unfilled "is not
|
|
configured, it is unfinished".
|
|
|
|
Two bundles already tracked here supply the arms, so no fixture carries a
|
|
sentence from any corpus: the segmented golden bundle and the provenance
|
|
fixture differ in `bundle_id` and in `ref`.
|
|
|
|
**What this rule does NOT do**, stated here rather than implied: it compares a
|
|
DECLARED identity against a DECLARED identity and never opens the bundle. A
|
|
payload that misreports its own `ref` passes. Proving a ref against bytes is
|
|
`okf consume --ref`'s job and needs a bundle path this checker deliberately
|
|
does not take.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
import subprocess
|
|
import sys
|
|
from pathlib import Path
|
|
from typing import Any
|
|
|
|
import pytest
|
|
|
|
from llm_ingestion_okf import contract_check, skill
|
|
|
|
PROJECT_ROOT = Path(__file__).resolve().parents[1]
|
|
TEMPLATE = PROJECT_ROOT / "skills" / "okf-consume-template" / "SKILL.md"
|
|
|
|
#: Two bundles that differ in both halves of the identity. One bundle would let
|
|
#: an id comparison masquerade as an identity comparison.
|
|
BUNDLE_A = PROJECT_ROOT / "examples" / "ingest-golden-segmented-okf-v0-2" / "expected-bundle"
|
|
BUNDLE_B = PROJECT_ROOT / "tests" / "fixtures" / "consume-provenance"
|
|
|
|
|
|
def instantiate(bundle: Path) -> tuple[str, dict[str, Any]]:
|
|
"""One generated skill and the payload generated beside it, in memory."""
|
|
text, payload = skill.render(bundle, out=bundle.parent / "unwritten")
|
|
return text, dict(payload)
|
|
|
|
|
|
def codes(skill_text: str, payload: Any) -> list[str]:
|
|
return [finding.code for finding in contract_check.check(skill_text, payload).findings]
|
|
|
|
|
|
def run(tmp_path: Path, skill_text: str, payload: Any) -> int:
|
|
"""The exit code, taken from `main` directly. A pipe would report the
|
|
status of the last command in it, which is not the checker's."""
|
|
skill_file = tmp_path / "SKILL.md"
|
|
payload_file = tmp_path / "payload.json"
|
|
skill_file.write_text(skill_text, encoding="utf-8")
|
|
payload_file.write_text(json.dumps(payload), encoding="utf-8")
|
|
return contract_check.main(["--skill", str(skill_file), "--payload", str(payload_file)])
|
|
|
|
|
|
# --- The control, which every arm below depends on ---------------------------
|
|
|
|
|
|
def test_the_two_bundles_declare_different_identities() -> None:
|
|
"""The known-positive for the arms: if these agreed, every mismatch below
|
|
would be measuring nothing."""
|
|
_, payload_a = instantiate(BUNDLE_A)
|
|
_, payload_b = instantiate(BUNDLE_B)
|
|
assert payload_a["bundle"]["bundle_id"] != payload_b["bundle"]["bundle_id"]
|
|
assert payload_a["bundle"]["ref"] != payload_b["bundle"]["ref"]
|
|
|
|
|
|
def test_a_skill_and_its_own_payload_stay_conformant(tmp_path: Path) -> None:
|
|
"""KNOWN-POSITIVE. A rule that refuses everything is green on every
|
|
negative arm above it."""
|
|
text, payload = instantiate(BUNDLE_A)
|
|
assert contract_check.check(text, payload).findings == ()
|
|
assert run(tmp_path, text, payload) == 0
|
|
|
|
|
|
# --- The three measured forms ------------------------------------------------
|
|
|
|
|
|
def test_a_skill_for_another_bundle_is_a_finding(tmp_path: Path) -> None:
|
|
"""Row A: different `bundle_id` AND different `ref`."""
|
|
text, _ = instantiate(BUNDLE_A)
|
|
_, payload = instantiate(BUNDLE_B)
|
|
assert codes(text, payload) == ["bundle_mismatch"]
|
|
assert run(tmp_path, text, payload) == 1
|
|
|
|
|
|
def test_the_same_id_at_another_ref_is_a_finding(tmp_path: Path) -> None:
|
|
"""Row C, the arm that separates a whole rule from half of one: the
|
|
`bundle_id` agrees and only the `ref` is foreign."""
|
|
text, payload = instantiate(BUNDLE_A)
|
|
_, other = instantiate(BUNDLE_B)
|
|
payload["bundle"] = dict(payload["bundle"], ref=other["bundle"]["ref"])
|
|
assert codes(text, payload) == ["bundle_mismatch"]
|
|
assert run(tmp_path, text, payload) == 1
|
|
|
|
|
|
def test_an_unfilled_template_is_a_finding(tmp_path: Path) -> None:
|
|
"""Row B: `<CORPUS>` and `<REF>` are not an identity. An identity the rule
|
|
cannot read must fail, or the template passes again and the rule is a
|
|
comparison nobody reaches."""
|
|
_, payload = instantiate(BUNDLE_A)
|
|
template = TEMPLATE.read_text(encoding="utf-8")
|
|
assert "<CORPUS>" in template and "<REF>" in template
|
|
assert "bundle_mismatch" in codes(template, payload)
|
|
assert run(tmp_path, template, payload) == 1
|
|
|
|
|
|
# --- The same question one level down ----------------------------------------
|
|
|
|
|
|
def test_an_excerpt_from_another_bundle_is_a_finding() -> None:
|
|
"""SS 3.1's identity is the (bundle_id, concept_id) tuple, so an excerpt
|
|
naming a bundle the payload does not is the same defect one level down.
|
|
Measured on the two real payloads reproduced this round: 0 of 16 excerpts
|
|
disagreed, so this arm is built rather than found."""
|
|
text, payload = instantiate(BUNDLE_A)
|
|
excerpts = [dict(excerpt) for excerpt in payload["excerpts"]]
|
|
assert excerpts, "the known-positive: no excerpts would make the zero meaningless"
|
|
excerpts[0]["bundle_id"] = "some-other-bundle"
|
|
payload["excerpts"] = excerpts
|
|
assert codes(text, payload) == ["bundle_mismatch"]
|
|
|
|
|
|
# --- The rule does not reach past what it is given ----------------------------
|
|
|
|
|
|
def test_a_payload_declaring_no_identity_is_left_to_its_own_rule(tmp_path: Path) -> None:
|
|
"""No rule short-circuits another, and none restates another either: a
|
|
payload carrying no `bundle` is `ref_missing`'s defect, not this rule's.
|
|
The count is the known-negative's, written down rather than assumed."""
|
|
text, _ = instantiate(BUNDLE_A)
|
|
found = codes(text, {})
|
|
assert "bundle_mismatch" not in found
|
|
assert "ref_missing" in found
|
|
assert len(found) == 9
|
|
assert run(tmp_path, text, {}) == 1
|
|
|
|
|
|
def test_the_rule_count_is_the_denominator_the_report_quotes() -> None:
|
|
"""The nevner moved 15 -> 16 with this rule, 16 -> 17 with
|
|
`parent_unfollowable` (K3-21) and 17 -> 19 with `subquestions_unindexed`
|
|
and `passage_malformed` (v1.1 C6), and every consumer quoting the old number is
|
|
quoting a number that has changed."""
|
|
text, payload = instantiate(BUNDLE_A)
|
|
report = contract_check.check(text, payload)
|
|
assert report.rules_evaluated == len(contract_check.RULES) == 19
|
|
assert "19 rules" in report.render()
|
|
|
|
|
|
def test_the_generator_writes_an_identity_the_checker_can_read() -> None:
|
|
"""The two halves are in different files, so the coupling gets a test: the
|
|
sentence `okf skill` writes is the sentence the rule parses."""
|
|
for bundle in (BUNDLE_A, BUNDLE_B):
|
|
text, payload = instantiate(bundle)
|
|
read = contract_check.skill_identity(text)
|
|
assert read == (payload["bundle"]["bundle_id"], payload["bundle"]["ref"])
|
|
|
|
|
|
@pytest.mark.parametrize("bundle", (BUNDLE_A, BUNDLE_B), ids=lambda path: path.name)
|
|
def test_the_console_script_reports_the_mismatch(bundle: Path, tmp_path: Path) -> None:
|
|
"""Through the installed entry point, because that is what a consumer
|
|
runs, and its exit code is read directly rather than through a pipe."""
|
|
text, _ = instantiate(bundle)
|
|
other = BUNDLE_B if bundle == BUNDLE_A else BUNDLE_A
|
|
_, payload = instantiate(other)
|
|
skill_file = tmp_path / "SKILL.md"
|
|
payload_file = tmp_path / "payload.json"
|
|
skill_file.write_text(text, encoding="utf-8")
|
|
payload_file.write_text(json.dumps(payload), encoding="utf-8")
|
|
result = subprocess.run(
|
|
[
|
|
sys.executable,
|
|
str(PROJECT_ROOT / "tools" / "okf_contract_check.py"),
|
|
"--skill",
|
|
str(skill_file),
|
|
"--payload",
|
|
str(payload_file),
|
|
],
|
|
capture_output=True,
|
|
text=True,
|
|
check=False,
|
|
)
|
|
assert result.returncode == 1, result.stdout
|
|
assert "bundle_mismatch" in result.stdout
|