mirror of
https://github.com/garrytan/gbrain.git
synced 2026-07-28 14:59:47 +00:00
* feat: search quality boost — compiled truth ranking, detail parameter, cosine re-scoring Compiled truth chunks now rank 2x higher in hybrid search via RRF normalization + source boost. New --detail flag (low/medium/high) controls timeline inclusion. Cosine re-scoring blends query-chunk similarity before dedup for query-specific ranking. Also: remove DISTINCT ON from keyword search (dedup handles per-page capping), add chunk_id + chunk_index to SearchResult, add getEmbeddingsByChunkIds to BrainEngine interface. Inspired by Ramp Labs' "Latent Briefing" paper (April 2026). * feat: RRF normalization, source-aware dedup, detail param in operations RRF scores normalized to 0-1 before 2.0x compiled truth boost. Source-aware dedup guarantees compiled truth chunk per page. Detail parameter added to query operation, dedupResults added to bare search operation. Debug logging via GBRAIN_SEARCH_DEBUG=1. * chore: bump version and changelog (v0.8.1) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * fix: CJK word count in query expansion CJK text is not space-delimited. A query like "向量搜索优化" was counted as 1 word and silently skipped expansion. Now counts characters for CJK queries instead of space-separated tokens. Co-Authored-By: YIING99 <yiing99@users.noreply.github.com> * feat: retrieval evaluation harness — P@k, R@k, MRR, nDCG@k + gbrain eval Full IR evaluation framework: precisionAtK, recallAtK, mrr, ndcgAtK metrics with runEval() orchestrator. gbrain eval CLI with single-run table and A/B comparison mode (--config-a / --config-b) for parameter tuning. HybridSearchOpts now accepts rrfK and dedupOpts overrides. Co-Authored-By: 4shut0sh <4shut0sh@users.noreply.github.com> * test: search quality tests — RRF boost, dedup guarantee, cosine similarity, E2E benchmark 42 new tests across 3 files: - test/search.test.ts: RRF normalization, compiled truth 2x boost, dedup key collision prevention, cosine similarity edge cases, CJK word count detection - test/dedup.test.ts: source-aware compiled truth guarantee, layer interactions, custom maxPerPage, empty/single result edge cases - test/e2e/search-quality.test.ts: full pipeline against PGLite with basis vector embeddings — chunk_id/chunk_index fields, detail parameter filtering, getEmbeddingsByChunkIds, keyword multi-chunk, vector ordering Also: export rrfFusion + cosineSimilarity for unit testing, fix PGLite getEmbeddingsByChunkIds to parse string vectors from pgvector. * test: search quality benchmark with A/B comparison (baseline vs PR#64) Benchmark measures P@1, MRR, nDCG@5, and source accuracy across 8 queries against 5 seeded pages. Key finding: boost helps entity lookups but over-corrects temporal queries. Validates the --detail parameter as the right control mechanism. Output at docs/benchmarks/2026-04-13.md. * feat: query intent classifier — auto-selects detail level, 100% source accuracy Zero-latency heuristic classifier detects query intent from text patterns: - "Who is Pedro?" → entity → detail=low (compiled truth only) - "When did we last meet?" → temporal → detail=high (no boost, natural ranking) - "Variant fund announcement" → event → detail=high - General queries → detail=medium (default with boost) The key insight: skip the 2.0x compiled truth boost for detail=high queries. Temporal/event queries want natural ranking where timeline entries can win. Benchmark results (source accuracy = does the top chunk match expected type): - Baseline: 100% (already good, no boost needed) - Boost only: 71.4% (boost over-corrects temporal queries) - Boost + intent classifier: 100% (best of both worlds) 35 unit tests for the classifier. 590 total tests pass. * feat: query intent classifier — auto-selects detail level, 100% source accuracy Heuristic classifier detects query intent from text patterns (zero latency, no LLM call). Maps temporal queries ("when did we last meet") to detail=high, entity queries ("who is X") to detail=low, events to detail=high. Benchmark results (29 pages, 20 queries, graded relevance): - Baseline: P@1=0.947, MRR=0.974, source accuracy=89.5% - Boost only: P@1=0.895, MRR=0.939, source accuracy=63.2% (over-correction) - Boost + intent: P@1=0.947, MRR=0.974, source accuracy=89.5% (fully recovered) The intent classifier eliminates the boost's over-correction on temporal queries while preserving its benefits for entity lookups. 35 unit tests for the classifier. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * test: search quality benchmark with A/B comparison (baseline vs PR#64) Rich benchmark: 29 pages, 58 chunks, 20 queries with graded relevance. Now measures CHUNK-LEVEL quality, not just page-level retrieval. Key findings (C. Boost+Intent vs A. Baseline): - Unique pages in top-10: 7.2 → 8.7 (+21% broader coverage) - Compiled truth ratio: 51.6% → 66.8% (+15pp more signal) - CT-first rate: 100% (compiled truth leads for entity queries) - Timeline accessible: 100% (temporal queries still find dates) - Source accuracy: 89.5% maintained (intent classifier prevents regression) The boost alone (B) causes -26pp source accuracy regression. Intent classifier (C) recovers it fully. * docs: clean benchmark report — ELI10 search quality analysis for PR#64 Replaces two drafts with one clean report. Explains what changed, why it matters, and what the numbers mean. All fictional data, no private info. Key findings: 21% more page coverage per query, 29% more compiled truth in results. Intent classifier prevents boost from burying timeline for temporal queries. Full per-query breakdown with before/after comparison. * chore: remove auto-generated benchmark file (clean version is 2026-04-14-search-quality.md) * docs: update project documentation for search quality boost CLAUDE.md: added search/intent.ts, search/eval.ts, commands/eval.ts to key files. Added 5 new test files (search, dedup, intent, eval, e2e/search-quality). Updated test count from 23+4 to 28+5. Added docs/benchmarks/ to key files. README.md: updated search pipeline diagram with intent classifier, RRF normalization, compiled truth boost, cosine re-scoring, and 5-layer dedup. Added --detail flag explanation and benchmark instructions. CHANGELOG.md: added search quality entries to v0.9.3 (intent classifier, --detail flag, gbrain eval, CJK fix). Credited @4shut0sh and @YIING99. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> * docs: headline benchmark gains in changelog * docs: add community attribution rule to CHANGELOG voice section --------- Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com> Co-authored-by: YIING99 <yiing99@users.noreply.github.com> Co-authored-by: 4shut0sh <4shut0sh@users.noreply.github.com>
157 lines
6.8 KiB
TypeScript
157 lines
6.8 KiB
TypeScript
/**
|
|
* Dedup pipeline unit tests — source-aware guarantee, layer interactions,
|
|
* and compiled truth preservation.
|
|
*/
|
|
|
|
import { describe, test, expect } from 'bun:test';
|
|
import { dedupResults } from '../src/core/search/dedup.ts';
|
|
import type { SearchResult } from '../src/core/types.ts';
|
|
|
|
function makeResult(overrides: Partial<SearchResult> = {}): SearchResult {
|
|
return {
|
|
slug: 'test-page',
|
|
page_id: 1,
|
|
title: 'Test',
|
|
type: 'concept',
|
|
chunk_text: 'unique chunk text ' + Math.random(),
|
|
chunk_source: 'compiled_truth',
|
|
chunk_id: Math.floor(Math.random() * 10000),
|
|
chunk_index: 0,
|
|
score: 0.5,
|
|
stale: false,
|
|
...overrides,
|
|
};
|
|
}
|
|
|
|
describe('dedupResults', () => {
|
|
test('basic dedup caps per page to 2', () => {
|
|
const results = [
|
|
makeResult({ slug: 'a', score: 0.9, chunk_text: 'first' }),
|
|
makeResult({ slug: 'a', score: 0.8, chunk_text: 'second' }),
|
|
makeResult({ slug: 'a', score: 0.7, chunk_text: 'third' }),
|
|
makeResult({ slug: 'a', score: 0.6, chunk_text: 'fourth' }),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
const aChunks = deduped.filter(r => r.slug === 'a');
|
|
expect(aChunks.length).toBeLessThanOrEqual(2);
|
|
});
|
|
|
|
test('removes text-similar chunks', () => {
|
|
const results = [
|
|
makeResult({ slug: 'a', score: 0.9, chunk_text: 'the quick brown fox jumps over the lazy dog' }),
|
|
makeResult({ slug: 'b', score: 0.8, chunk_text: 'the quick brown fox jumps over the lazy cat' }),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
// These share high Jaccard similarity, one should be removed
|
|
expect(deduped.length).toBeLessThanOrEqual(2);
|
|
});
|
|
|
|
test('enforces type diversity when mixed types present', () => {
|
|
// Mix of person and concept types — diversity should cap person
|
|
const results = [
|
|
...Array.from({ length: 8 }, (_, i) =>
|
|
makeResult({ slug: `p${i}`, page_id: i, score: 1 - i * 0.05, type: 'person', chunk_text: `person ${i} unique text content here` })
|
|
),
|
|
...Array.from({ length: 4 }, (_, i) =>
|
|
makeResult({ slug: `c${i}`, page_id: 100 + i, score: 0.4 - i * 0.05, type: 'concept', chunk_text: `concept ${i} unique text content here` })
|
|
),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
const personCount = deduped.filter(r => r.type === 'person').length;
|
|
const conceptCount = deduped.filter(r => r.type === 'concept').length;
|
|
// With diversity enforcement, person shouldn't completely dominate
|
|
expect(personCount).toBeGreaterThan(0);
|
|
expect(conceptCount).toBeGreaterThan(0);
|
|
});
|
|
});
|
|
|
|
describe('compiled truth guarantee', () => {
|
|
test('swaps in compiled_truth when page has only timeline in results', () => {
|
|
const results = [
|
|
makeResult({ slug: 'a', chunk_id: 1, score: 0.9, chunk_source: 'timeline', chunk_text: 'timeline entry about meeting' }),
|
|
makeResult({ slug: 'a', chunk_id: 2, score: 0.8, chunk_source: 'timeline', chunk_text: 'another timeline entry here' }),
|
|
makeResult({ slug: 'a', chunk_id: 3, score: 0.3, chunk_source: 'compiled_truth', chunk_text: 'compiled truth assessment of entity' }),
|
|
makeResult({ slug: 'b', chunk_id: 4, score: 0.7, chunk_source: 'compiled_truth', chunk_text: 'page b compiled truth' }),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
const aChunks = deduped.filter(r => r.slug === 'a');
|
|
const hasCompiledTruth = aChunks.some(c => c.chunk_source === 'compiled_truth');
|
|
expect(hasCompiledTruth).toBe(true);
|
|
});
|
|
|
|
test('does not swap when page already has compiled_truth', () => {
|
|
const results = [
|
|
makeResult({ slug: 'a', chunk_id: 1, score: 0.9, chunk_source: 'compiled_truth', chunk_text: 'compiled assessment' }),
|
|
makeResult({ slug: 'a', chunk_id: 2, score: 0.8, chunk_source: 'timeline', chunk_text: 'timeline entry details' }),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
const aChunks = deduped.filter(r => r.slug === 'a');
|
|
// Should still have compiled_truth
|
|
expect(aChunks.some(c => c.chunk_source === 'compiled_truth')).toBe(true);
|
|
});
|
|
|
|
test('does nothing when no compiled_truth exists for page', () => {
|
|
const results = [
|
|
makeResult({ slug: 'a', chunk_id: 1, score: 0.9, chunk_source: 'timeline', chunk_text: 'only timeline chunk one' }),
|
|
makeResult({ slug: 'a', chunk_id: 2, score: 0.8, chunk_source: 'timeline', chunk_text: 'only timeline chunk two' }),
|
|
];
|
|
const deduped = dedupResults(results);
|
|
// All timeline, no compiled_truth to swap in
|
|
const aChunks = deduped.filter(r => r.slug === 'a');
|
|
expect(aChunks.every(c => c.chunk_source === 'timeline')).toBe(true);
|
|
});
|
|
|
|
test('guarantee works across multiple pages', () => {
|
|
const results = [
|
|
// Page A: only timeline in top results, compiled_truth exists lower
|
|
makeResult({ slug: 'a', chunk_id: 1, score: 0.95, chunk_source: 'timeline', chunk_text: 'a timeline high score' }),
|
|
makeResult({ slug: 'a', chunk_id: 2, score: 0.9, chunk_source: 'timeline', chunk_text: 'a timeline medium score' }),
|
|
makeResult({ slug: 'a', chunk_id: 3, score: 0.2, chunk_source: 'compiled_truth', chunk_text: 'a compiled truth low score' }),
|
|
// Page B: has compiled_truth already
|
|
makeResult({ slug: 'b', chunk_id: 4, score: 0.85, chunk_source: 'compiled_truth', chunk_text: 'b compiled truth content' }),
|
|
// Page C: only timeline, no compiled_truth at all
|
|
makeResult({ slug: 'c', chunk_id: 5, score: 0.8, chunk_source: 'timeline', chunk_text: 'c timeline only entry' }),
|
|
];
|
|
|
|
const deduped = dedupResults(results);
|
|
|
|
// Page A should have compiled_truth guaranteed
|
|
const aChunks = deduped.filter(r => r.slug === 'a');
|
|
if (aChunks.length > 0) {
|
|
expect(aChunks.some(c => c.chunk_source === 'compiled_truth')).toBe(true);
|
|
}
|
|
|
|
// Page B already had compiled_truth
|
|
const bChunks = deduped.filter(r => r.slug === 'b');
|
|
if (bChunks.length > 0) {
|
|
expect(bChunks.some(c => c.chunk_source === 'compiled_truth')).toBe(true);
|
|
}
|
|
|
|
// Page C has no compiled_truth to swap in, so all timeline is fine
|
|
const cChunks = deduped.filter(r => r.slug === 'c');
|
|
if (cChunks.length > 0) {
|
|
expect(cChunks.every(c => c.chunk_source === 'timeline')).toBe(true);
|
|
}
|
|
});
|
|
});
|
|
|
|
describe('edge cases', () => {
|
|
test('empty input returns empty', () => {
|
|
expect(dedupResults([])).toEqual([]);
|
|
});
|
|
|
|
test('single result passes through', () => {
|
|
const result = makeResult({ chunk_text: 'single result here' });
|
|
const deduped = dedupResults([result]);
|
|
expect(deduped).toHaveLength(1);
|
|
});
|
|
|
|
test('respects custom maxPerPage option', () => {
|
|
const results = Array.from({ length: 5 }, (_, i) =>
|
|
makeResult({ slug: 'a', chunk_id: i + 100, score: 1 - i * 0.1, chunk_text: `chunk number ${i} with unique content` })
|
|
);
|
|
const deduped = dedupResults(results, { maxPerPage: 3 });
|
|
expect(deduped.filter(r => r.slug === 'a').length).toBeLessThanOrEqual(3);
|
|
});
|
|
});
|