Optimize recompute latency: Add query embedding cache and reusable ZMQ connections - #226

Open

VedantMadane wants to merge 15 commits into

StarTrail-org:mainfrom

VedantMadane:optimize-recompute-latency

Commits on Aug 11, 2026

Merge upstream main: Add query embedding cache and reusable ZMQ connections (PR #226)
VedantMadane
committed
1502b8f
fix(ci): stop Linux pytest exit 127 in prompt-template integration test
VedantMadane
committed
4853b95
style: fix ruff import blank-line in query cache unit tests
VedantMadane
committed
2492b90
fix(ci): stub native HNSW in prompt-template persistence tests (Linux exit 127)
VedantMadane
committed
38adbef
fix(ci): stub HNSW native path on Linux CI to avoid pytest exit 127
VedantMadane
committed
bcd86cd
fix(ci): install zeromq for Arch Linux smoke (libzmq.so.5)
VedantMadane
committed
b7fe060
fix(ci): provide MKL runtime for Arch smoke via PyPI mkl
VedantMadane
committed
ae85ba8
fix(ci): discover MKL lib dirs with find for Arch smoke
VedantMadane
committed
75d3ea6
ci: re-trigger full matrix after macos-14 setup-uv infra flake
VedantMadane
committed
9be5d53
fix(ci): run Arch smoke even when some matrix cells fail
VedantMadane
committed
9a52832
fix(ci): use absolute paths in Arch smoke LD_LIBRARY_PATH
VedantMadane
committed
23be2e8
fix(ci): symlink MKL .so.3 to .so.2 for Arch FAISS load
VedantMadane
committed
6c37b17
fix(ci): pin Arch smoke MKL to 2025.3.1 for .so.2 soname
VedantMadane
committed
75c3666

Commits on Aug 12, 2026