| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
Fix Unicode case handling in literal query ranking Co-authored-by: KaranocaVe<karanocave@qq.com> # message auto-generated for no-merge-commit merge: !88 merge fix/20261005-unicode-literal-ranking into master Fix Unicode case handling in literal query ranking Created-by: KaranocaVe Commit-by: KaranocaVe Merged-by: cann-robot Description: # Fix Unicode case handling in literal query ranking Literal search ranks exact title matches ahead of title fragments, descriptions, and body mentions. Its scorer used SQLite lower(), which only handles ASCII case, while indexing and query normalization use Python's Unicode-aware str.lower(). A query such as Θ Exact title could therefore rank a description mention of θ Exact title above the exact title. Use the same Unicode normalization in field scoring and add a CLI regression for the Greek case pair with an ASCII ranking control. Validation: python3 -m unittest evals.test_retrieval_index.RetrievalIndexTests (10 tests passed) and bash check.sh (295 repository tests passed with 1 skipped, 36 trace tests passed, Ruff and shell checks passed). 关联 Issue: https://gitcode.com/cann/cannbot-knowledge/issues/33 See merge request: cann/cannbot-knowledge!88 | 1 小时前 |