v0.3.0

What's Changed

• Add benchmark regression CI workflows by @hunhee98 in #1
• Generalize retrieval quality datasets by @hunhee98 in #2
• Inline tiny bodies in read outlines by @hunhee98 in #3
• Publish nightly benchmark release assets by @hunhee98 in #4
• Update GitHub Actions to Node 24 runtimes by @hunhee98 in #5
• Report recall quality by language by @hunhee98 in #6
• Filter BM25 stopwords for natural-language queries by @hunhee98 in #7
• Add CJK retrieval quality coverage by @hunhee98 in #8
• Add two-stage hybrid search cascade by @hunhee98 in #9
• Expand natural-language queries with indexed BM25 vocab by @hunhee98 in #10
• Infer hybrid search alpha from query embeddings by @hunhee98 in #11
• Expand recall suite to 100 repo-backed queries by @hunhee98 in #12
• Boost ranking with symbol and path components by @hunhee98 in #13
• Prepare v0.3.0 release by @hunhee98 in #14
• ci: run benchmark gate only on PRs by @hunhee98 in #15
• ci: ship release artifacts for supported targets by @hunhee98 in #16
• ci: update Intel macOS release runner by @hunhee98 in #17

New Contributors

• @hunhee98 made their first contribution in #1

Full Changelog: v0.2.0...v0.3.0