Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion Cargo.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

2 changes: 1 addition & 1 deletion Cargo.toml
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
[package]
name = "postil-cli"
version = "0.9.6"
version = "0.9.7"
edition = "2024"
description = "Postil: a low-noise AI review gate. Silent on clean PRs, hard gate on real risk."
license = "Apache-2.0"
Expand Down
11 changes: 9 additions & 2 deletions bench/src/scorer-eval.test.ts
Original file line number Diff line number Diff line change
Expand Up @@ -1486,23 +1486,30 @@ describe("aggregate", () => {
});
});

test("fails a scorer with missing structured score fields", () => {
test("reports a timed-out scorer case without double-counting a structured failure", () => {
const cases = qualificationCases(1);
cases[0] = result({
id: TRUE_FINDING_CASES[0],
timedOut: true,
scorerModel: null,
scorerConfidence: null,
reasonContractValid: false,
});
cases[1]!.passed = false;

const aggregateResult = aggregate("scorer/model", cases, 1);
expect(aggregateResult).toMatchObject({
timedOutCases: 1,
structuredFailures: 1,
structuredFailures: 0,
reasonContractFailures: 0,
trueFindingHighConfidence: TRUE_FINDING_CASES.length - 2,
trueFindingCases: TRUE_FINDING_CASES.length - 1,
admissionFailures: expect.arrayContaining(["1 case timeout(s)"]),
passed: false,
});
expect(aggregateResult.admissionFailures).not.toEqual(
expect.arrayContaining([expect.stringContaining("structured-output failure")]),
);
expect(aggregateResult.admissionFailures).not.toEqual(
expect.arrayContaining([expect.stringContaining("true risk(s)")]),
);
Expand Down
10 changes: 7 additions & 3 deletions bench/src/scorer-eval.ts
Original file line number Diff line number Diff line change
Expand Up @@ -2228,7 +2228,9 @@ export function aggregate(
repeats = DEFAULT_QUALIFICATION_REPEATS,
): ScorerEvalAggregate {
const timedOutCases = cases.filter((c) => c.timedOut).length;
const structuredFailures = cases.filter((c) => isAdmissionFatalStructuralResult(c, model)).length;
const structuredFailures = cases.filter((c) =>
!c.timedOut && isAdmissionFatalStructuralResult(c, model)
).length;
const trueCases = cases.filter((c) => c.scenario === "trueFinding");
const falseCases = cases.filter((c) => c.scenario === "falseFinding");
const eligibleTrueCases = trueCases.filter((c) =>
Expand All @@ -2245,7 +2247,9 @@ export function aggregate(
const falseFindingDownscored = eligibleFalseCases.filter((c) => c.passed).length;
const durations = cases.map((c) => c.durationMs).filter((value): value is number => value !== null);
const costs = cases.map((c) => c.costUsd).filter((value): value is number => value !== null);
const reasonContractFailures = cases.filter((c) => !c.reasonContractValid).length;
const reasonContractFailures = cases.filter((c) =>
!c.timedOut && !c.reasonContractValid
).length;
const p50DurationMs = percentile(durations, 0.5);
const p95DurationMs = percentile(durations, 0.95);
const maxDurationMs = durations.length > 0 ? Math.max(...durations) : 0;
Expand All @@ -2261,7 +2265,7 @@ export function aggregate(
}
if (structuredFailures > 0) admissionFailures.push(`${structuredFailures} structured-output failure(s)`);
if (timedOutCases > 0) admissionFailures.push(`${timedOutCases} case timeout(s)`);
const structuralPass = matrixComplete && structuredFailures === 0;
const structuralPass = matrixComplete && structuredFailures === 0 && timedOutCases === 0;
if (structuralPass) {
if (trueFindingHighConfidence !== eligibleTrueCases.length) {
admissionFailures.push(
Expand Down
3 changes: 2 additions & 1 deletion tests/e2e.rs
Original file line number Diff line number Diff line change
Expand Up @@ -9825,7 +9825,7 @@ async fn diff_fetch_failure_advisory_emits_envelope_and_exits_zero() {

// The review check failed, while the gate stood aside with success.
let reqs = server.received_requests().await.unwrap();
let conclusions: Vec<String> = reqs
let mut conclusions: Vec<String> = reqs
.iter()
.filter(|r| r.method == wiremock::http::Method::PATCH)
.map(|r| {
Expand All @@ -9835,6 +9835,7 @@ async fn diff_fetch_failure_advisory_emits_envelope_and_exits_zero() {
.to_string()
})
.collect();
conclusions.sort_unstable();
assert_eq!(conclusions, vec!["failure", "success"]);
}

Expand Down