Two-tier hybrid search for Rust: sub-millisecond initial results via potion-128M, quality-refined rankings in 150ms via MiniLM-L6-v2. Combines lexical (Tantivy BM25) and semantic (vector cosine) search with Reciprocal Rank Fusion. Progressive iterator API, f16 SIMD vector index, feature-gated compilation.
Stars
51
Forks
11
Watchers
51
Open Issues
0
Overall repository health assessment
No package.json found
This might not be a Node.js project
401
commits
feat: add CJK bigram tokenization for Chinese/Japanese/Korean search
b7a6cf6View on GitHubfeat(rerank,fsfs,runtime): propagate raw reranker logits, implement config set/reset, and add WAL compaction on shutdown
fa94560View on GitHubfeat(lexical): add conversation_id field to cass schema; fix: use git dep for asupersync
9961c0eView on GitHubchore(deps): switch asupersync to local path dependency and refine error formatting
7086d96View on GitHubrefactor(embed,lexical,rerank): extract map_lock_error helpers and handle PolledAfterCompletion explicitly
ac4e650View on GitHubperf(fusion): parallelize embedding and lexical search with rayon::join; add poll_immediate for sync-in-async futures
cba3632View on GitHubperf(fusion): replace filter+collect with in-place retain for negation filtering
1824bbbView on GitHubperf(fusion,index): compute in_both_sources inline during RRF insertion and pre-screen WAL hashes in search
a840c57View on GitHubfix(ci): increase Quality job timeout from 60 to 90 minutes
1e15295View on GitHubfix: add tempfile dev-dependency for interaction_unit tests
05988bcView on GitHubfix(tests): use tempdir for interaction_unit test index
64d95b0View on GitHubfix(ops): widen write latency budget for CI runner variability
7ea06cdView on GitHubfix(storage): correct snapshot isolation test for WAL mode semantics
5d2dce4View on GitHub