Nye ref-filer fodes best-practice-konforme sa neste KB-update ikke reverserer TOC-arbeidet (write-path-regresjon). MA fore 1b. - buildToc(body): deterministisk ## Innhold fra ##-seksjoner, GitHub-slugs, fence-aware, idempotent (lister aldri seg selv) - composeKbFile(meta, body): generatoren — header + (store filer >100 linjer) TOC + brodtekst; erstatter bar buildKbHeader+body-konkatenering - validateKbFile: stor fil uten TOC -> missing:['toc'] (write-path-gaten) - Terskel 100 speiler eval.mjs N4_LARGE_FILE_LINES; bundet atferdsmessig via test som importerer ekte checkN4 (anti-regresjons-orakel), ikke cross-import - Wiret composeKbFile inn i transform-prompt.md + kb-update.md (create + in-place) TDD: +11 transform-tester. kb-update 349/349, kb-eval 154/154 (503 totalt). Fikset pre-eksisterende skjor assertion i criterion-testen (stale 'Azure AI Foundry' -> 'AI Foundry'; produktet omdopt) — var rod for 1c.
253 lines
11 KiB
JavaScript
253 lines
11 KiB
JavaScript
// tests/kb-update/test-transform.test.mjs
|
|
// TDD for lag 4 — TRANSFORMASJON (doc → KB-fil). lib/transform.mjs is a pure
|
|
// helper library (mirrors decisions-io / verify-out): it builds the deterministic
|
|
// KB-file header, validates the mandatory-field contract, routes the file to its
|
|
// owning skill via the taxonomy (lag 0), and constructs the lag-5 `change` object.
|
|
// It NEVER writes — applying the file is the operator gate's job (commands/kb-update.md).
|
|
//
|
|
// Why Source lives in the HEADER block: build-registry / kb-headers.parseSourceHeader
|
|
// scan only the top 500 bytes, so a source buried in a bottom section (as the old
|
|
// prompt-template did) is invisible → the 0% authority_source coverage. Putting
|
|
// **Source:** in the header is what lets lag 3 backfill authority_source and lag-5
|
|
// rule 3 (authority mismatch) eventually fire.
|
|
|
|
import { test } from 'node:test';
|
|
import assert from 'node:assert/strict';
|
|
import { readFileSync } from 'node:fs';
|
|
import { dirname, join } from 'node:path';
|
|
import { fileURLToPath } from 'node:url';
|
|
import {
|
|
buildKbHeader,
|
|
validateKbFile,
|
|
buildChange,
|
|
resolveTargetPath,
|
|
buildToc,
|
|
composeKbFile,
|
|
} from '../../scripts/kb-update/lib/transform.mjs';
|
|
import { parseSourceHeader } from '../../scripts/kb-update/lib/kb-headers.mjs';
|
|
import { classifyChange } from '../../scripts/kb-update/lib/verify-out.mjs';
|
|
import { loadTaxonomy } from '../../scripts/kb-update/lib/taxonomy.mjs';
|
|
// The N4 (TOC-in-large-files) check is the cross-subsystem best-practice contract.
|
|
// Fase 1c births files that satisfy it; we assert parity against the REAL check so
|
|
// a future drift in N4's definition fails this test rather than silently diverging.
|
|
import { checkN4 } from '../../scripts/kb-eval/eval.mjs';
|
|
|
|
// A body long enough that header+body exceeds N4's 100-line "large file" threshold,
|
|
// carrying real ## section headings the TOC must pick up.
|
|
const LARGE_BODY =
|
|
'## Introduksjon\n\nNoe innledning.\n\n' +
|
|
'## Kjernekomponenter\n\n' +
|
|
'Mye innhold ...\n'.repeat(110) +
|
|
'\n## Beslutningsveiledning\n\nVelg X når ...\n';
|
|
const SHORT_BODY = '## Introduksjon\n\nKort.\n';
|
|
|
|
const __dirname = dirname(fileURLToPath(import.meta.url));
|
|
|
|
const META = {
|
|
title: 'Azure AI Search - Hybrid Retrieval',
|
|
status: 'GA',
|
|
category: 'rag-architecture',
|
|
source: 'https://learn.microsoft.com/azure/search/hybrid-search-overview',
|
|
lastUpdated: '2026-06',
|
|
};
|
|
|
|
// --- buildKbHeader: mandatory fields, parseable by the real header scanners ----
|
|
|
|
test('buildKbHeader emits title, Last updated, Status, Source, Category in the header block', () => {
|
|
const h = buildKbHeader(META);
|
|
assert.match(h, /^# Azure AI Search - Hybrid Retrieval/m);
|
|
assert.match(h, /\*\*Last updated:\*\*\s*2026-06/);
|
|
assert.match(h, /\*\*Status:\*\*\s*GA/);
|
|
assert.match(h, /\*\*Category:\*\*\s*rag-architecture/);
|
|
assert.match(h, /\*\*Source:\*\*\s*https:\/\/learn\.microsoft\.com/);
|
|
});
|
|
|
|
test('buildKbHeader Source is within the top-500-byte region kb-headers scans', () => {
|
|
const h = buildKbHeader(META);
|
|
// The whole point: parseSourceHeader (top 500 bytes) must find it.
|
|
assert.equal(parseSourceHeader(h), META.source);
|
|
});
|
|
|
|
test('buildKbHeader throws when a mandatory field is missing (status-felt obligatorisk)', () => {
|
|
assert.throws(() => buildKbHeader({ ...META, status: undefined }), /status/i);
|
|
assert.throws(() => buildKbHeader({ ...META, source: '' }), /source/i);
|
|
assert.throws(() => buildKbHeader({ ...META, title: null }), /title/i);
|
|
assert.throws(() => buildKbHeader({ ...META, lastUpdated: undefined }), /last.?updated/i);
|
|
});
|
|
|
|
// --- validateKbFile: catches files missing the mandatory contract -------------
|
|
|
|
test('validateKbFile accepts a header-conformant file', () => {
|
|
const file = buildKbHeader(META) + '\n## Introduksjon\n\nNoe innhold.\n';
|
|
const r = validateKbFile(file);
|
|
assert.equal(r.valid, true);
|
|
assert.deepEqual(r.missing, []);
|
|
});
|
|
|
|
test('validateKbFile reports each missing mandatory field', () => {
|
|
// No Status, no Source, no Last updated, no title.
|
|
const r = validateKbFile('Bare brødtekst uten header.\n');
|
|
assert.equal(r.valid, false);
|
|
for (const f of ['title', 'last_updated', 'status', 'source']) {
|
|
assert.ok(r.missing.includes(f), `expected '${f}' in missing: ${r.missing}`);
|
|
}
|
|
});
|
|
|
|
test('validateKbFile flags a file with body-only source (the 0%-coverage failure mode)', () => {
|
|
// Source in a bottom section, NOT the header → invisible to top-500-byte scan.
|
|
const file =
|
|
'# Tittel\n\n**Last updated:** 2026-06\n**Status:** Preview\n\n---\n\n' +
|
|
'Mye innhold ...\n'.repeat(40) +
|
|
'\n## Kilder\n\n**Source:** https://learn.microsoft.com/x\n';
|
|
const r = validateKbFile(file);
|
|
assert.equal(r.valid, false);
|
|
assert.ok(r.missing.includes('source'));
|
|
});
|
|
|
|
// --- buildChange: the lag-4 → lag-5 contract ----------------------------------
|
|
|
|
test('buildChange produces a classifyChange-compatible object with safe defaults', () => {
|
|
const c = buildChange({
|
|
field: 'status',
|
|
oldValue: 'Preview',
|
|
newValue: 'GA',
|
|
sourceUrl: META.source,
|
|
});
|
|
assert.equal(c.field, 'status');
|
|
assert.equal(c.old_value, 'Preview');
|
|
assert.equal(c.new_value, 'GA');
|
|
assert.equal(c.source_url, META.source);
|
|
assert.equal(c.authority_source, null); // pre-lag-3 default
|
|
assert.deepEqual(c.refutations, []);
|
|
});
|
|
|
|
test('buildChange output flows through lag 5 — a status change is flagged, never auto-applied', () => {
|
|
const c = buildChange({ field: 'status', oldValue: 'Preview', newValue: 'GA', sourceUrl: META.source });
|
|
const verdict = classifyChange(c);
|
|
assert.equal(verdict.verdict, 'flagged');
|
|
assert.equal(verdict.status_claim, true);
|
|
});
|
|
|
|
test('buildChange output flows through lag 5 — a non-status change can auto-apply', () => {
|
|
const c = buildChange({
|
|
field: 'description',
|
|
oldValue: 'En kort beskrivelse.',
|
|
newValue: 'En litt lengre, oppdatert beskrivelse.',
|
|
sourceUrl: META.source,
|
|
});
|
|
assert.equal(classifyChange(c).verdict, 'auto-applied');
|
|
});
|
|
|
|
// --- resolveTargetPath: taxonomy-driven routing (replaces category-skill-map) --
|
|
|
|
test('resolveTargetPath routes via the taxonomy category_skill (disk-truth)', () => {
|
|
const tax = loadTaxonomy();
|
|
assert.equal(
|
|
resolveTargetPath(tax, 'rag-architecture', 'hybrid-retrieval.md'),
|
|
'skills/ms-ai-engineering/references/rag-architecture/hybrid-retrieval.md',
|
|
);
|
|
// a former divergence — taxonomy says advisor, not engineering
|
|
assert.equal(
|
|
resolveTargetPath(tax, 'prompt-engineering', 'few-shot.md'),
|
|
'skills/ms-ai-advisor/references/prompt-engineering/few-shot.md',
|
|
);
|
|
});
|
|
|
|
test('resolveTargetPath returns null for an unknown category (caller must gate)', () => {
|
|
const tax = loadTaxonomy();
|
|
assert.equal(resolveTargetPath(tax, 'does-not-exist', 'x.md'), null);
|
|
});
|
|
|
|
// --- buildToc: deterministic TOC from the body's ## section headings -----------
|
|
|
|
test('buildToc emits a ## Innhold heading with anchor links for each ## section', () => {
|
|
const toc = buildToc('## Introduksjon\n\ntekst\n\n## Kostnad og lisensiering\n\ntekst\n');
|
|
assert.match(toc, /^##\s+Innhold/m);
|
|
assert.match(toc, /- \[Introduksjon\]\(#introduksjon\)/);
|
|
assert.match(toc, /- \[Kostnad og lisensiering\]\(#kostnad-og-lisensiering\)/);
|
|
});
|
|
|
|
test('buildToc slugifies GitHub-style: lowercase, punctuation stripped, spaces→hyphens', () => {
|
|
const toc = buildToc('## 1. Introduksjon & Oversikt\n\ntekst\n');
|
|
assert.match(toc, /- \[1\. Introduksjon & Oversikt\]\(#1-introduksjon--oversikt\)/);
|
|
});
|
|
|
|
test('buildToc returns empty string when the body has no ## sections', () => {
|
|
assert.equal(buildToc('Bare brødtekst, ingen overskrifter.\n'), '');
|
|
assert.equal(buildToc('# Bare en H1\n\ntekst\n'), ''); // H1 is the header title, not a section
|
|
});
|
|
|
|
test('buildToc never lists the Innhold heading itself (idempotent re-runs)', () => {
|
|
const body = '## Innhold\n\n- [Introduksjon](#introduksjon)\n\n## Introduksjon\n\ntekst\n';
|
|
const toc = buildToc(body);
|
|
// exactly one anchor — to Introduksjon — never an anchor back to Innhold
|
|
assert.doesNotMatch(toc, /\(#innhold\)/);
|
|
assert.match(toc, /\(#introduksjon\)/);
|
|
});
|
|
|
|
// --- composeKbFile: the deterministic generator (header + TOC-if-large + body) --
|
|
|
|
test('composeKbFile births a large file with canonical header AND a TOC', () => {
|
|
const file = composeKbFile(META, LARGE_BODY);
|
|
assert.match(file, /^# Azure AI Search - Hybrid Retrieval/m); // canonical header first
|
|
assert.equal(parseSourceHeader(file), META.source); // Source in top-500-byte region
|
|
assert.match(file, /^##\s+Innhold/m); // TOC present
|
|
assert.equal(validateKbFile(file).valid, true);
|
|
});
|
|
|
|
test('composeKbFile output for a large file passes the REAL N4 check (anti-regression oracle)', () => {
|
|
const file = composeKbFile(META, LARGE_BODY);
|
|
const n4 = checkN4([{ path: 'skills/x/references/y/z.md', content: file }]);
|
|
assert.equal(n4.largeFiles, 1);
|
|
assert.equal(n4.withToc, 1);
|
|
assert.equal(n4.pass, true);
|
|
});
|
|
|
|
test('composeKbFile does NOT add a TOC to a small file (TOC is for large files only)', () => {
|
|
const file = composeKbFile(META, SHORT_BODY);
|
|
assert.doesNotMatch(file, /^##\s+Innhold/m);
|
|
assert.equal(validateKbFile(file).valid, true);
|
|
});
|
|
|
|
test('composeKbFile is idempotent: a body that already carries a TOC is not double-TOCd', () => {
|
|
const once = composeKbFile(META, LARGE_BODY);
|
|
const body = once.slice(once.indexOf('---\n') + 4); // strip header, keep TOC+body
|
|
const twice = composeKbFile(META, body);
|
|
assert.equal((twice.match(/^##\s+Innhold/gm) || []).length, 1);
|
|
assert.equal(validateKbFile(twice).valid, true);
|
|
});
|
|
|
|
// --- validateKbFile: large files must carry a TOC (the write-path regression gate)
|
|
|
|
test('validateKbFile flags a large file missing a TOC', () => {
|
|
const file = buildKbHeader(META) + '\n' + 'En linje med innhold.\n'.repeat(120);
|
|
const r = validateKbFile(file);
|
|
assert.equal(r.valid, false);
|
|
assert.ok(r.missing.includes('toc'), `expected 'toc' in missing: ${r.missing}`);
|
|
});
|
|
|
|
test('validateKbFile accepts a large file that carries a TOC', () => {
|
|
const file = composeKbFile(META, LARGE_BODY);
|
|
const r = validateKbFile(file);
|
|
assert.equal(r.valid, true);
|
|
assert.ok(!r.missing.includes('toc'));
|
|
});
|
|
|
|
test('validateKbFile does NOT require a TOC for a small file', () => {
|
|
const file = buildKbHeader(META) + '\n## Introduksjon\n\nKort innhold.\n';
|
|
const r = validateKbFile(file);
|
|
assert.equal(r.valid, true);
|
|
assert.ok(!r.missing.includes('toc'));
|
|
});
|
|
|
|
// --- ARCHITECTURE INVARIANT: transform.mjs is a pure lib, imports NO write-utils
|
|
|
|
test('transform.mjs imports NO write-utils (pure transformer, never writes)', () => {
|
|
const src = readFileSync(join(__dirname, '../../scripts/kb-update/lib/transform.mjs'), 'utf8');
|
|
const importLines = src.split('\n').filter((l) => /^\s*import\b/.test(l)).join('\n');
|
|
assert.doesNotMatch(importLines, /\bsaveDecisions\b/);
|
|
assert.doesNotMatch(importLines, /\bsaveRegistry\b/);
|
|
assert.doesNotMatch(importLines, /\bsaveReport\b/);
|
|
assert.doesNotMatch(importLines, /from\s+'[^']*atomic-write[^']*'/);
|
|
assert.doesNotMatch(importLines, /from\s+'[^']*\/backup\.mjs'/);
|
|
});
|