Files
gbrain/test/e2e/takes-scorecard-parity.test.ts
T
1399e519c0 v0.30.0 feat: calibration scorecards (Slice A1 of v0.30 wave) (#731)
* feat(schema): migration v40 — takes_resolved_quality + drift_decisions

Slice A1 of the v0.30 wave. Bundles all wave schema in one migration so
A2/B1/C1 carry no schema of their own (codex F6 schema-first ordering).

- takes.resolved_quality TEXT with CHECK (correct/incorrect/partial).
- takes_resolution_consistency CHECK enforces (quality, outcome) tuple
  consistency at the DB layer. partial → outcome=NULL.
- One-shot backfill maps legacy resolved_outcome → resolved_quality so
  v0.28 brains keep working with no manual reclassification.
- idx_takes_scorecard partial index on (holder, kind, resolved_quality)
  WHERE resolved_quality IS NOT NULL — scorecard hot path.
- drift_decisions audit table (consumed by Slice C1 in v0.30.3).
- PGLite branch via sqlFor.pglite mirrors the same shape; RLS DO-block
  is Postgres-only.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(takes-fence): extend ParsedTake + parser + conditional renderer (codex F3)

A codex consult on the v0.30 plan caught a real bug: the v0.28 parser had
no concept of resolution columns, so every cmdUpdate after a cmdResolve
silently deleted resolution data on the next render. This commit kills
that data-loss path.

ParsedTake gains optional resolvedAt, resolvedQuality, resolvedOutcome,
resolvedEvidence, resolvedValue, resolvedUnit, resolvedBy. parseTakesFence
detects v0.30-shape headers and reads resolution cells when present;
v0.28 7-column fences round-trip byte-identical. renderTakesFence emits
the resolution columns ONLY when at least one row on the page has
resolvedQuality set — pages with no resolved rows keep the narrow shape
exactly as before.

11 new test cases including the round-trip preservation regression gate.
Without those tests, the silent-delete bug returns the moment the parser
shape drifts. Tests cover: parsing v0.30 + v0.28 shapes, conditional
rendering, partial quality round-trip, upsertTakeRow + supersedeRow
preservation when a page already has resolved rows.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(engine): getScorecard + getCalibrationCurve + 3-state TakeResolution

Adds the calibration aggregate methods on BrainEngine. Both engines
implement them with SQL-level allow-list filtering inside the GROUP BY
(D4 fail-closed): hidden-holder rows contribute zero to aggregates.

TakeResolution gains optional `quality` (correct|incorrect|partial). When
both quality and outcome are supplied AND inconsistent, the engine throws
TAKE_RESOLUTION_INVALID rather than silently overwriting. resolveTake
writes both columns: quality directly, outcome derived (correct→true,
incorrect→false, partial→NULL). Schema CHECK is the defense-in-depth
backstop.

Brier scope (D5 + D11): the SQL aggregation excludes partial rows from
the Brier denominator — partial isn't a binary outcome to compare a
probability against. partial_rate is reported alongside as a separate
counter so hedging behavior stays visible. The 20% threshold lives in
src/core/takes-resolution.ts and the CLI surfaces it in v0.30.0's
cmdScorecard.

New module src/core/takes-resolution.ts holds shared pure helpers
(deriveResolutionTuple, finalizeScorecard) consumed by both engines so
the math stays identical across backends. takeRowToTake (utils.ts) reads
resolved_quality through to the Take row shape.

23 new test cases: 16 for the helpers (Brier hand-calc against a 4-bet
reference at 0.205, n=0 no-divide, contradictory-input rejection,
partial-exclusion contract, threshold constant); 7 against PGLite for
the engine path (3-state quality writes, contradictory throws, scorecard
hand-calc, n=0, SQL-level allow-list privacy filter).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(cli): gbrain takes resolve --quality, takes scorecard, takes calibration

cmdResolve widened: --quality correct|incorrect|partial is the new primary
input. --outcome true|false stays as a back-compat alias auto-mapping to
quality, with a stderr deprecation warning on use. Mutually exclusive
with --quality. --evidence is a semantic alias for --source on the
resolve subcommand.

cmdResolve mirrors resolution metadata into the takes-fence on disk via
the page-lock-aware path. Round-trip preservation through parseTakesFence
+ renderTakesFence keeps resolution data intact across unrelated edits to
other rows on the same page. Removes the v0.28 deferred-rendering warning.

cmdScorecard prints `correct | incorrect | partial`, accuracy, Brier
(correct ∨ incorrect only; lower is better; 0.25 = always-50% baseline),
and partial_rate. When partial_rate > 20% the CLI prints
"[!] partial_rate is high — calibration may be optimistic" so hedging
behavior stays visible even though it doesn't enter the math (D11). Small-N
note when resolved < 100. JSON output via --json.

cmdCalibration bins resolved correct/incorrect bets by stated weight
(--bucket-size, default 0.1) and prints observed vs predicted vs delta
per bucket. Diagonal alignment = perfect calibration.

Both new subcommands wire allow-list as undefined for local CLI callers
(trusted); MCP path will thread it from access_tokens.permissions.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* feat(mcp): register takes_scorecard + takes_calibration ops

Both ops are read-scope, MCP-callable, allow-list-honoring. Handlers
thread ctx.takesHoldersAllowList into the engine method's required
allowList parameter, which applies WHERE holder = ANY at SQL aggregation
level (D4 fail-closed). Local CLI callers leave the allow-list
undefined and see all holders.

Updates the OperationContext.takesHoldersAllowList contract comment to
list the new aggregate ops alongside takes_list, takes_search, query.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* v0.30.0 release: calibration core (Slice A1 of v0.30 wave)

VERSION + package.json bump. CHANGELOG entry covers the release-summary
(headline + math table + privacy note + data-loss-bug-killed note),
"## To take advantage of v0.30.0" upgrade path, and itemized changes.

llms-full.txt regenerated to capture the v0.28.x annotations that had
been merged but not yet rolled into the docs bundle.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* test(e2e): scorecard + calibration parity on real Postgres + NUMERIC fix

Adds end-to-end coverage for v0.30.0 (Slice A1) against real Postgres:

- test/e2e/takes-scorecard-parity.test.ts (new): seeds the same 6-bet
  fixture (4 binary garry + 1 partial garry + 1 binary harj) into both
  Postgres and PGLite, asserts getScorecard + getCalibrationCurve return
  byte-identical results across engines, runs the 4-bet hand-calc Brier
  reference (0.205) on real PG, and verifies the SQL-level allow-list
  filter strictly subtracts hidden-holder rows on both engines.

- test/e2e/takes-postgres.test.ts: extended with 8 v0.30 cases — quality
  semantics (correct/partial/back-compat) writes the expected (quality,
  outcome) tuple on real PG; the takes_resolution_consistency CHECK
  constraint actually fires on a contradictory raw UPDATE; getScorecard
  + getCalibrationCurve coherent shape + ordered-bucket invariants;
  PRIVACY allow-list filter on real PG; MCP dispatch path for
  takes_scorecard + takes_calibration with allow-list threading.

While writing the parity test, the e2e harness caught a real bug PGLite
tolerated: postgres.js sends scalar `${bucketSize}` params as text by
default, so `FLOOR(weight / $N)` tried to coerce '0.1' to integer and
threw `invalid input syntax for type integer: "0.1"`. The NUMERIC fix
also kills a separate FP-precision divergence — `FLOOR(0.7 / 0.1)`
returns 6 on real PG (IEEE 754 rounds 0.7/0.1 to 6.9999...) and 7 on
PGLite. Both engines now bucket via `weight::numeric / $N::numeric`
which is exact decimal arithmetic and engine-agnostic.

This is the v0.30.0 wave's first cross-engine parity test. Same shape
will guard A2's getTrajectory + getAnnualReview when those land.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* docs(changelog): correct v40→v43 reference and expand v0.30.0 test note

Two fixes to the v0.30.0 entry after the master merge renumbered the
migration:

- The "#### Added" bullet still said "Schema migration v40"; bumped to v43.
- The "#### Tests" section only enumerated unit tests. The PR also ships
  19 E2E cases (11 in takes-scorecard-parity, 8 extending takes-postgres)
  that exercise the calibration math against real Postgres and the
  PG↔PGLite engine parity. Added the count + a note about the two real
  bugs the parity test caught (postgres.js string-typed scalar params
  and IEEE 754 bucketing divergence) that PGLite tolerated.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-07 23:07:00 -07:00

203 lines
8.5 KiB
TypeScript

/**
* v0.30.0 (Slice A1) E2E: scorecard + calibration parity between Postgres
* and PGLite. Same fixture, same query, same numbers.
*
* Seeds 4 binary bets + 1 partial bet into both engines (mirrors the
* `takes-resolution.test.ts` 4-bet hand-calc reference: Brier=0.205).
* Asserts getScorecard returns byte-identical numeric output and
* getCalibrationCurve emits the same buckets.
*
* Privacy gate: also asserts the SQL-level allow-list filter (D4
* fail-closed) returns identical results across engines — hidden-holder
* rows must contribute zero on both sides.
*
* Skips gracefully when DATABASE_URL is unset.
*/
import { describe, test, expect, beforeAll, afterAll } from 'bun:test';
import { PGLiteEngine } from '../../src/core/pglite-engine.ts';
import { hasDatabase, setupDB, teardownDB, getEngine } from './helpers.ts';
import type { BrainEngine } from '../../src/core/engine.ts';
import type { TakesScorecardOpts, CalibrationCurveOpts } from '../../src/core/engine.ts';
const SKIP_PG = !hasDatabase();
const d = SKIP_PG ? describe.skip : describe;
let pglite: PGLiteEngine;
let pgEngine: BrainEngine;
interface FixtureBet {
rowNum: number;
claim: string;
holder: string;
weight: number;
resolveAs?: 'correct' | 'incorrect' | 'partial';
}
// Same fixture used by takes-resolution.test.ts hand-calc.
// 4 binary garry bets at varied weights (Brier=0.205) + 1 partial garry +
// 1 binary harj bet so the allow-list assertion has signal.
const FIXTURE_BETS: FixtureBet[] = [
{ rowNum: 1, claim: 'b1', holder: 'garry', weight: 0.9, resolveAs: 'correct' },
{ rowNum: 2, claim: 'b2', holder: 'garry', weight: 0.6, resolveAs: 'correct' },
{ rowNum: 3, claim: 'b3', holder: 'garry', weight: 0.7, resolveAs: 'incorrect' },
{ rowNum: 4, claim: 'b4', holder: 'garry', weight: 0.4, resolveAs: 'incorrect' },
{ rowNum: 5, claim: 'b5', holder: 'garry', weight: 0.5, resolveAs: 'partial' },
{ rowNum: 6, claim: 'h1', holder: 'harj-taggar', weight: 0.8, resolveAs: 'correct' },
];
const SCORECARD_PAGE = 'companies/scorecard-parity-fixture';
async function seedFixture(engine: BrainEngine): Promise<number> {
const page = await engine.putPage(SCORECARD_PAGE, {
title: 'Scorecard parity fixture',
type: 'company' as const,
compiled_truth: '## Takes\n',
});
await engine.addTakesBatch(
FIXTURE_BETS.map(b => ({
page_id: page.id,
row_num: b.rowNum,
claim: b.claim,
kind: 'bet' as const,
holder: b.holder,
weight: b.weight,
})),
);
for (const b of FIXTURE_BETS) {
if (b.resolveAs === undefined) continue;
await engine.resolveTake(page.id, b.rowNum, {
quality: b.resolveAs,
resolvedBy: b.holder,
});
}
return page.id;
}
beforeAll(async () => {
if (SKIP_PG) return;
// PGLite (in-memory, no DATABASE_URL needed; runs even on the skipped
// suite path so we'd still know if its seed broke).
pglite = new PGLiteEngine();
await pglite.connect({});
await pglite.initSchema();
await seedFixture(pglite);
// Real Postgres
await setupDB();
pgEngine = getEngine();
await seedFixture(pgEngine);
});
afterAll(async () => {
if (pglite) await pglite.disconnect();
if (!SKIP_PG) await teardownDB();
});
d('v0.30.0 e2e: scorecard parity (PG vs PGLite)', () => {
const queries: Array<{ name: string; opts: TakesScorecardOpts; allowList?: string[] }> = [
{ name: 'all garry, no allow-list', opts: { holder: 'garry', domainPrefix: SCORECARD_PAGE } },
{ name: 'all rows, no holder filter', opts: { domainPrefix: SCORECARD_PAGE } },
{ name: 'allow-list garry-only', opts: { domainPrefix: SCORECARD_PAGE }, allowList: ['garry'] },
{ name: 'allow-list world-only (zero match)', opts: { domainPrefix: SCORECARD_PAGE }, allowList: ['world'] },
];
for (const q of queries) {
test(`scorecard: ${q.name}`, async () => {
const pgCard = await pgEngine.getScorecard(q.opts, q.allowList);
const pgliteCard = await pglite.getScorecard(q.opts, q.allowList);
expect(pgCard.total_bets).toBe(pgliteCard.total_bets);
expect(pgCard.resolved).toBe(pgliteCard.resolved);
expect(pgCard.correct).toBe(pgliteCard.correct);
expect(pgCard.incorrect).toBe(pgliteCard.incorrect);
expect(pgCard.partial).toBe(pgliteCard.partial);
// Float fields: byte-equality is too strict due to engine-driver
// numeric coercion (Postgres returns string-coerced floats; PGLite
// returns JS numbers). Use 6-decimal closeness — well below any
// user-visible precision.
const closeOrBothNull = (a: number | null, b: number | null) => {
if (a === null && b === null) return;
expect(a).not.toBeNull();
expect(b).not.toBeNull();
expect(a).toBeCloseTo(b!, 6);
};
closeOrBothNull(pgCard.accuracy, pgliteCard.accuracy);
closeOrBothNull(pgCard.brier, pgliteCard.brier);
closeOrBothNull(pgCard.partial_rate, pgliteCard.partial_rate);
});
}
test('Brier on the 4-bet hand-calc reference matches 0.205 on both engines', async () => {
// Filter to garry bets only — the harj bet would skew the Brier.
const opts: TakesScorecardOpts = { holder: 'garry', domainPrefix: SCORECARD_PAGE };
const pgCard = await pgEngine.getScorecard(opts, undefined);
const pgliteCard = await pglite.getScorecard(opts, undefined);
expect(pgCard.brier).toBeCloseTo(0.205, 4);
expect(pgliteCard.brier).toBeCloseTo(0.205, 4);
// Also: partial_rate = 1/5 = 0.2 (4 binary + 1 partial = 5 resolved garry rows).
expect(pgCard.partial_rate).toBeCloseTo(0.2, 6);
expect(pgliteCard.partial_rate).toBeCloseTo(0.2, 6);
});
test('PRIVACY: allow-list ["garry"] gives same totals on both engines AND excludes harj', async () => {
const opts: TakesScorecardOpts = { domainPrefix: SCORECARD_PAGE };
const pgGarry = await pgEngine.getScorecard(opts, ['garry']);
const pgliteGarry = await pglite.getScorecard(opts, ['garry']);
// Both engines see exactly the 5 garry rows.
expect(pgGarry.resolved).toBe(5);
expect(pgliteGarry.resolved).toBe(5);
// Without allow-list, both see all 6 resolved rows (5 garry + 1 harj).
const pgFull = await pgEngine.getScorecard(opts, undefined);
const pgliteFull = await pglite.getScorecard(opts, undefined);
expect(pgFull.resolved).toBe(6);
expect(pgliteFull.resolved).toBe(6);
// Allow-list strictly subtracts (defense in depth: if SQL filter were
// post-filtered or applied wrong, this assertion catches it).
expect(pgFull.resolved - pgGarry.resolved).toBe(1);
expect(pgliteFull.resolved - pgliteGarry.resolved).toBe(1);
});
});
d('v0.30.0 e2e: calibration curve parity (PG vs PGLite)', () => {
const queries: Array<{ name: string; opts: CalibrationCurveOpts; allowList?: string[] }> = [
{ name: 'garry-only, default bucket', opts: { holder: 'garry' } },
{ name: 'all-rows, default bucket', opts: {} },
{ name: 'allow-list garry-only', opts: {}, allowList: ['garry'] },
];
for (const q of queries) {
test(`calibration: ${q.name}`, async () => {
const pgBuckets = await pgEngine.getCalibrationCurve(q.opts, q.allowList);
const pgliteBuckets = await pglite.getCalibrationCurve(q.opts, q.allowList);
expect(pgBuckets.length).toBe(pgliteBuckets.length);
for (let i = 0; i < pgBuckets.length; i++) {
expect(pgBuckets[i].bucket_lo).toBeCloseTo(pgliteBuckets[i].bucket_lo, 6);
expect(pgBuckets[i].bucket_hi).toBeCloseTo(pgliteBuckets[i].bucket_hi, 6);
expect(pgBuckets[i].n).toBe(pgliteBuckets[i].n);
if (pgBuckets[i].observed !== null) {
expect(pgBuckets[i].observed).toBeCloseTo(pgliteBuckets[i].observed!, 6);
} else {
expect(pgliteBuckets[i].observed).toBeNull();
}
if (pgBuckets[i].predicted !== null) {
expect(pgBuckets[i].predicted).toBeCloseTo(pgliteBuckets[i].predicted!, 6);
} else {
expect(pgliteBuckets[i].predicted).toBeNull();
}
}
});
}
test('partial bet (weight 0.5) is excluded from calibration curve on both engines', async () => {
const pgBuckets = await pgEngine.getCalibrationCurve({ holder: 'garry' }, undefined);
const pgliteBuckets = await pglite.getCalibrationCurve({ holder: 'garry' }, undefined);
const pgTotal = pgBuckets.reduce((s, b) => s + b.n, 0);
const pgliteTotal = pgliteBuckets.reduce((s, b) => s + b.n, 0);
// 4 binary garry rows; partial excluded.
expect(pgTotal).toBe(4);
expect(pgliteTotal).toBe(4);
});
});