v1.2.0
What's Changed
- native-core: stop pinning the version test to a stale literal by @Davidobot in #49
- feat(embeddings): make the embedding runtime an opt-in extra by @Davidobot in #48
- fix(native-core): drop the engine on its worker at interpreter exit by @Davidobot in #51
- MaxScore BM25 serving index for lexical + hybrid latency by @Davidobot in #56
- refactor(native-core): drop the unused tvim_present commit-manifest field by @Davidobot in #58
- native-core: concurrent multi-writer WAL append + FFI/Python/Swift bindings by @Davidobot in #52
- Opt-in ANN cluster-prune index with exact re-score by @Davidobot in #57
- native-core: ANN query-path perf and cross-layer consolidation (#60) by @Davidobot in #61
- native-core: finish the multi-producer WAL path (appender hot-path cleanup + reader freshness) by @Davidobot in #64
- native-core: multi-producer text ingest + running checkpointer lease by @Davidobot in #65
- Deployment/performance docs, GPU CPU-fallback warning, and dependency caps by @Davidobot in #68
Full Changelog: v1.1.0...v1.2.0