Skip to content

fix(search): dedupe repeated query tokens - #686

Merged
nicobailon merged 3 commits into
mainfrom
backlog/pr-678-promotion
Sep 27, 2026
Merged

nicobailon merged 3 commits into
mainfrom
backlog/pr-678-promotion

Conversation

@nicobailon

Copy link
Copy Markdown
Owner

Summary

  • Deduplicate normalized query tokens before scoring and coverage checks.
  • Keep repeated terms such as search search from hiding an otherwise matching tool.
  • Exercise both the scorer and the production ranking path.

This is a current-main promotion of #678. It preserves @kojizada490-wq’s implementation commits and adds visible changelog credit. The original fork branch was not modified.

Validation

  • npx vitest run __tests__/search-ranking.test.ts
  • npm run typecheck
  • git diff --check

Review evidence

Fresh semantic review found no ranking-code defect. The promoted executable tree matches the clean three-way composition of current main and #678; the only promotion delta is changelog credit.

@greptile-apps

greptile-apps Bot commented Sep 27, 2026 •

Copy link
Copy Markdown

RetriggerConfidence Score: 5/5

[Medium risk] Changes how search queries rank matching tools.

The PR appears safe to merge; no actionable issue was identified.

Summary

The PR deduplicates normalized query tokens in both search-scoring paths so repeated words no longer reduce coverage and hide matching tools.

  • Adds a regression test for the scorer and production ranking path.
  • Credits the contributor in the changelog.

Reviews (1) · Last reviewed commit: "docs(changelog): credit search token fix"

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants