Skip to content

perf: pre-compute query phrases once per rankAnswerEvidence call#99

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
BigSimmo merged 1 commit into from
Jun 29, 2026
Merged

perf: pre-compute query phrases once per rankAnswerEvidence call #99

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
16 changes: 11 additions & 5 deletions src/lib/answer-ranking.ts
View file
Open in desktop
Original file line numberDiff line numberDiff line change
Expand Up@@ -105,14 +105,19 @@ function tokenCoverage(tokens: string[], haystack: string) {
return hits.length / tokens.length;
}

function phraseScore(tokens: string[], haystack: string) {
if (tokens.length < 2 || !haystack) return 0;
function buildQueryPhrases(tokens: string[]): string[] {
if (tokens.length < 2) return [];
const phrases: string[] = [];
for (let size = Math.min(4, tokens.length); size >= 2; size -= 1) {
for (let index = 0; index <= tokens.length - size; index += 1) {
phrases.push(tokens.slice(index, index + size).join(" "));
}
}
return phrases;
}

function phraseScore(phrases: string[], haystack: string) {
if (!phrases.length || !haystack) return 0;
const hits = phrases.filter((phrase) => haystack.includes(phrase)).length;
return Math.min(1, hits / Math.max(1, Math.min(phrases.length, 4)));
}
Expand DownExpand Up@@ -167,7 +172,7 @@ function sourceQualityScore(result: SearchResult) {
return score;
}

function answerEvidenceScore(tokens: string[], result: SearchResult, queryClass: RagQueryClass) {
function answerEvidenceScore(tokens: string[], phrases: string[], result: SearchResult, queryClass: RagQueryClass) {
const texts = resultTexts(result);
const base = Math.min(1, result.hybrid_score ?? result.similarity ?? 0) * 0.28;
const contentCoverage = tokenCoverage(tokens, texts.content);
Expand All@@ -176,7 +181,7 @@ function answerEvidenceScore(tokens: string[], result: SearchResult, queryClass:
const metadataCoverage = tokenCoverage(tokens, texts.metadata);
const combinedCoverage = tokenCoverage(tokens, texts.combined);
const adjacentCoverage = tokenCoverage(tokens, texts.adjacent);
const phraseCoverage = phraseScore(tokens, texts.combined);
const phraseCoverage = phraseScore(phrases, texts.combined);
const directnessScore =
contentCoverage * 0.34 + titleCoverage * 0.12 + sectionCoverage * 0.1 + metadataCoverage * 0.08;
const weakOverlapPenalty = combinedCoverage < 0.2 ? -0.18 : combinedCoverage < 0.34 ? -0.07 : 0;
Expand DownExpand Up@@ -242,11 +247,12 @@ export function rankAnswerEvidence(
queryClass: RagQueryClass = classifyRagQuery(query).queryClass,
) {
const tokens = uniqueQueryTokens(query);
const phrases = buildQueryPhrases(tokens);
const ranked = results
.map((result, index) => ({
result,
index,
score: answerEvidenceScore(tokens, result, queryClass),
score: answerEvidenceScore(tokens, phrases, result, queryClass),
}))
.sort(
(a, b) =>
Expand Down