0.6.0

What's Changed

• Support using NEON for SIMD distance calculations on aarch64 by @knudtty in #127
• Support for inner product distance metric by @tjgreen42 in #167
• Raise diskann maximum dimension from 2K to 16K by @tjgreen42 in #181
• Automatically set AVX2/FMA flags on x86 by @syvb in #188
• Fix Missing/Incorrect Index Statistics by @tjgreen42 in #180
• Upload built .deb packages to packagecloud by @syvb in #198

Internal changes

• Mark distance_type_* as CREATE OR REPLACE by @syvb in #169
• Find repo root properly when upgrade testing in a Cargo workspace by @syvb in #170
• Properly check if PG17 is supported in upgrade tester by @syvb in #171
• Use workspace target directory location in Debian install file by @syvb in #172
• Release 0.5.1 by @syvb in #173
• [chore] fix clippy warning by @tjgreen42 in #174
• [chore] Fix clippy errors in test code, and adjust precommit to check for these by @tjgreen42 in #175
• Run clippy as part of pgrx tests by @tjgreen42 in #178
• Fix CI - rely on ubuntu-22.04 as we used to by @adolsalamanca in #183
• Have cargo-pgrx install command specify version by @syvb in #187
• Move helper function to proper spot; fix new clippy warnings by @tjgreen42 in #189
• Hygiene: bump default pg version and update several dependencies by @tjgreen42 in #192
• Remove io_optimized / sbq_speedup storage mode by @tjgreen42 in #195
• Cleanup and traits for nodes by @tjgreen42 in #201
• Use LLVM 18 in GH Actions instead of 14 by @syvb in #202
• Release 0.6.0 by @syvb in #206

New Contributors

• @knudtty made their first contribution in #127
• @adolsalamanca made their first contribution in #183

Full Changelog: 0.5.1...0.6.0

0.5.1

What's Changed

• Bump version number for 0.5.0 postrelease by @tjgreen42 in #163
• Only normalize vectors if index uses cosine distance by @tjgreen42 in #165
• Fix check on number of dimensions by @cevian in #168

Full Changelog: 0.5.0...0.5.1

0.5.0

What's Changed

• Run Debian packager for PG13/PG14 too by @syvb in #141
• chore: add github issue template by @alejandrodnm in #145
• Introduce pre-commit hooks template and fix cargo clippy issues by @tjgreen42 in #149
• Fixes to nodes at distance 0 by @cevian in #150
• Avoid creating a new page for every insert by @syvb in #148
• fix: change dimension limit from 1999->2000 by @cevian in #159
• Add upgrade tests from all released versions by @tjgreen42 in #160
• Support for L2 distance by @tjgreen42 in #156
• Fit and finish before 0.5.0 release PR by @tjgreen42 in #161
• 0.5.0 release by @tjgreen42 in #162

New Contributors

• @tjgreen42 made their first contribution in #149

Full Changelog: 0.4.0...0.5.0

0.4.0

What's Changed

• feat: support pg13, pg14 as pgvector does by @yihong0618 in #94
• Bump pgrx version to 0.12.5 by @syvb in #137
• Add PG17 support by @syvb in #139
• Pre 0.4.0 by @syvb in #140

New Contributors

• @yihong0618 made their first contribution in #94
• @syvb made their first contribution in #137

Full Changelog: 0.3.0...0.4.0

0.3.0

This release fixes a few bugs, add some safety checks and fixes an issue with storing many nodes in the same location. The latter should improve recall for low-dimension data.

What's Changed

• Build on ARM by @cevian in #102
• Fix handling of many graph nodes at the same location by @cevian in #111
• Fix hot update handling by @cevian in #117
• Panic if AVX or FMA is missing by @cevian in #119

New Contributors

• @mcadariu made their first contribution in #89

Full Changelog: 0.2.0...0.3.0

0.2.0

This is a major release

This release contains major algorithmic improvements to make the index much more performant. The (now default) SBQ mode allows for very fast retrieval over a compressed index. Please see the accompanying benchmark posts for more information.

• improved DiskANN algorithm including better stopping criteria and other algorithmic changes
• A new quantization algorithm called Statistical Binary Quantization (SBQ) for improved accuracy on quantized retrieval. This is SOTA for binary quantization.
• Rescoring as part of index retrieval -- which greatly improves accuracy for partitioned tables.
• Deprecating PQ
• New extension name vectorscale and project name pgvectorscale
• New index name diskann

0.0.2

Prepare for release 0.0.2

Release 0.0.1

Prepare 0.0.1

0.0.1-dev.1

First release for dev