What's Changed
- Feature: Add onnx support by @donhardman in #145
- Boost the performance of the inference by @donhardman in #146
- Hnsw performance improvements by @glookka in #147
- ci: pin protoc version to 29.3 in build workflow by @donhardman in #148
- ci: add repo-token to setup-protoc action in workflow by @donhardman in #151
- fix(embeddings): make ort downloads optional via feature flag by @donhardman in #152
- ci(linux): add library compatibility checks by @donhardman in #153
- fixed CI to work with embedding build in the previous stages by @tomatolog in #160
- added manual workflow dispatch trigger by @tomatolog in #161
- fixed CI to work with embedding build in the previous stages by @tomatolog in #162
- Ci failures by @tomatolog in #163
- si new library version by @tomatolog in #158
- Fix embeddings perf by @donhardman in #157
- fix: fixed a crash that could occur when running knn queries with 1-bit quantized vectors from multiple threads by @glookka in #164
- perf(model): implement fast path for single-text embeddings by @donhardman in #171
- Fix/mkl build candle single batch by @donhardman in #170
- feat(embeddings): support passthrough remote model ids by @sanikolaev in #156
- CI changes to download ORT in cmake by @sanikolaev in #175
- Fix/clean stack fix by @donhardman in #178
- VX2 type mismatch in quantizer.cpp by @midasi in #180
- Fix crash on Rosetta emulation in docker by @donhardman in #182
New Contributors
Full Changelog: release-13.0.0...release-13.6.1