VectorChord 0.4.3 Release Notes

• use mimalloc on aarch64-linux
• fix compilation with gcc on x86_64
• fix compilation with clang on Windows
• prompt the user to rebuild the index after the upgrade

Full Changelog: 0.4.2...0.4.3

VectorChord 0.4.2 Release Notes

• fix compilation on aarch64 macos
• add support for pgxnclient: you can install VectorChord with pgxnclient install vchord==0.4.2 now

Full Changelog: 0.4.1...0.4.2

VectorChord 0.4.1 Release Notes

• Fix potential precision issue if the dimension of vectors is 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, 4096, 8192, 16384, 32768.

Full Changelog: 0.4.0...0.4.1

VectorChord 0.4 Release Notes

Major Improvements

• Streaming I/O & Page Prefetch

  • Complete rewrite of page layout to enable pipelined computation with streaming I/O.
  • On PostgreSQL 17, uses fadvise to prefetch buffers into the OS page cache, eliminating per-buffer read waits and fully leveraging disk throughput.
  • In upcoming PostgreSQL 18, direct support for io_uring will further streamline asynchronous I/O.
  • Benchmarks: 2–3× lower latency on cold queries (no buffer or page cache), translating to significantly improved tail latency in production.

• Prefilter Acceleration

  • Introduces true prefilter support for vector + filter queries.
  • Previous postfilter approach ranked full result sets and then applied filters—inefficient when selectivity is low (e.g., 1% filter rate).
  • Applies SQL filters before full precision vector distance computations, avoiding unnecessary work.
  • Benchmarks: Up to 3× faster end-to-end search on highly selective filters without any additional tuning.

Other Improvements

• Optimized Residual Quantization

  • Collaboration with RaBitQ author Jianyang: refactored distance term $|⟨o, q–c⟩|$ into $⟨o, q⟩ – ⟨o, c⟩$, so the query vector is quantized only once.
  • Result: ~20% QPS improvement over 0.3.
  • Recommendation: Enable residual quantization for L2 workloads.

• Fast Walsh-Hadamard Transform for Rotation

  • Collaboration with RaBitQ author Jianyang: replaced manual vchordrq.prewarm_dim GUC with an on-the-fly Fast Walsh-Hadamard Transform.
  • Removes the need to configure a prewarmed dimension list and yields marginal speed gains during setup.

Thank you for using VectorChord! As always, we welcome feedback and contributions on GitHub.

Full Changelog: 0.3.0...0.4.0

Features

Native support for the maxsim operator and efficient indexing inspired by XTR-WARP project. This makes it possible to build ColBERT- or ColPaLI-style multi-vector retrieval applications seamlessly within PostgreSQL.

Improvements

• More KMeans parameters can be configured
• Better progress report for internal KMeans build

What's Changed

• feat: check quals to skip rerank if rerank_in_table is enabled by @usamoi in #206
• feat: maxsim operator and indexing on maxsim operator by @usamoi in #197
• chore: update dependencies by @usamoi in #216
• feat: use mimalloc by @usamoi in #217
• feat: emit INFO when performing kmeans by @usamoi in #219
• feat: encode kmeans progress to phase name by @usamoi in #220
• readme: move docs to docs repo by @usamoi in #222
• chore: change default lists and probes to empty by @usamoi in #223
• feat: allows skipping rerank by @usamoi in #225
• delete cnpg image by @xieydd in #224
• refactor: remove allows_skipping_rerank by @usamoi in #226
• fix: apply epsilon for default search by @usamoi in #228
• chore: remove patch of crate half by @usamoi in #231
• fix: release env vars by @cutecutecat in #232
• chore: update 0.3.0 schema script by @usamoi in #230

Full Changelog: 0.2.2...0.3.0

What's Changed

• Support linux/arm64 for vchord-cnpg docker image by @xieydd in #184
• refactor: split quantized data vectors to two tapes by @usamoi in #196
• chore: add ghcr image by @cutecutecat in #200
• chore: update readme and scripts with url by @cutecutecat in #198
• fix: lint by ruff by @cutecutecat in #201
• fix: wrong operators for halfvec by @cutecutecat in #202
• doc: add new options by @usamoi in #207
• fix: do not leak memory in heap fetch and fix reading tuples in prewarm by @usamoi in #205
• fix: recycle pages in maintainance by @usamoi in #204
• chore: update v0.2.2 schema script by @usamoi in #210
• fix: correct logic of marking free pages by @usamoi in #211
• fix: remove heapify by @usamoi in #213
• fix error in cnpg amd64 Dockerfile by @xieydd in #214
• ci: fix release docker build by @usamoi in #215

Full Changelog: 0.2.1...0.2.2

Major Improvement

We optimize the external centroid index building speed, about 30%. Now it takes about 30h to build index for 100M vectors with only 4 vcpu on i4i.xlarge.

What's Changed

• refactor: move algorithm to a crate by @usamoi in #172
• feat: pinning index in memory when building, second try by @usamoi in #181
• fix: use linked list of vectors to skip realloc by @usamoi in #182
• feat: use select algorithm to replace heap, if k in top-k is expected to be small by @usamoi in #183
• ci: install pg13 in docker image by @usamoi in #186
• ci: use less docker by @usamoi in #187
• feat: rerank by fetching vectors in heap table by @usamoi in #189
• ci: enable CI for pg13 by @usamoi in #185
• chore: update dependencies by @usamoi in #190
• refactor: remove meaningless target feature requirements by @usamoi in #192
• fix: test simd operations in emulator by @usamoi in #193
• feat: neon impl of u8::reduce_sum_of_x by @usamoi in #194
• chore: update 0.2.1 schema (upgrade) script by @usamoi in #195

Full Changelog: 0.2.0...0.2.1

VectorChord 0.2 Release Notes

We are thrilled to announce the release of VectorChord 0.2, advancing vector search capabilities within PostgreSQL.

🚀 New Features

Optimized Storage Layout

• Long Cross-Page Vector Support: Redesigned internal storage allows vectors to span multiple 8KB PostgreSQL pages, enabling support for vectors with over 2000 dimensions, up to 16000 dim.
• Enhanced Storage Efficiency: Achieves higher storage density by minimizing wasted space, reducing index size by up to 50% compared to version 0.1.

Additional Data Types

• Float16 Support: Introduces Float16 data type, allowing users to halve the storage space required with a slight decrease in recall. Note that Float16 does not reduce the size of quantized vectors, maintaining 1 bit per dimension for original vector representation.

Architecture Enhancements

• ARM Architecture Support: Rewritten distance calculations and Fast Scan implementations using the Scalable Vector Extension (SVE) instruction set for optimal performance on ARM-based systems.
• AWS Graviton4 Compatibility: Leverage the latest i8g platform based on Graviton4 processors for improved performance at the same cost as i4i models.

⚡ Performance Improvements

• Reduced Index Size: Up to 50% reduction in index size compared to version 0.1.

🔧 Getting Started

• Comprehensive getting started guides will be available soon.

📝 Summary

VectorChord 0.2 introduces support for high-dimensional vectors, Float16 data type, ARM architecture optimizations, and a more compact storage layout. These enhancements collectively improve storage efficiency and query performance, providing a superior vector retrieval experience within PostgreSQL.

What's Changed

• fix: Unlock the conversation in CLA bot by @gaocegege in #86
• fix: CI env SEMVER by @kemingy in #87
• fix: format by @cutecutecat in #90
• feat: max_scan_tuples by @usamoi in #94
• test: fix push down tests by @cutecutecat in #95
• fix: set max dimension to 1600 in readme by @usamoi in #97
• fix: set max dimension to 1600 by @usamoi in #98
• docs: Update README by @VoVAllen in #103
• feat: support up to 60000 dimensions by @usamoi in #100
• chore: directory structure by @usamoi in #104
• feat: vchordrqfscan by @usamoi in #105
• chore: add CI to build the pgrx image by @kemingy in #107
• chore: fix the pgrx ci by @kemingy in #108
• fix: optimize insertions in building when lists = 1 by @usamoi in #106
• chore: build multiarch docker images by @kemingy in #112
• docs(readme): fix markdown style for docker run by @kemingy in #113
• feat: improve internal build by @usamoi in #115
• add enterprise image build step to ci by @xieydd in #114
• chore(README): Add some benchmark data by @gaocegege in #126
• fix: use stable toolchain by default by @usamoi in #128
• feat: scalar8 & indexing on halfvec by @usamoi in #131
• feat: Use dual license (AGPLv3 and ELv2) by @gaocegege in #130
• chore: update base by @usamoi in #137
• fix: remove sudo in dockerfile by @usamoi in #138
• fix: ci by @usamoi in #139
• fix: preprocess for halfvec by @usamoi in #140
• docs: update README for clarity and new features by @VoVAllen in #142
• fix: set an implicit root in external build if parents are not set by @usamoi in #147
• chore: type check and external test by @cutecutecat in #149
• chore: update dependencies by @usamoi in #151
• fix: update docker image in ci by @usamoi in #152
• Update enterprise dockerfile by @xieydd in #148
• feat: build multiarch pgrx image by @kemingy in #153
• chore: impl dereference traits for page guards by @usamoi in #156
• chore: fix the psql & release CI target, update readme by @kemingy in #158
• chore: add postgres sqllogicaltest for arm by @kemingy in #159
• chore: update link in readme by @VoVAllen in #160
• refactor: move pgvecto.rs base to this repo by @usamoi in #161
• fix: pick feat back to vchordrqfscan by @usamoi in #162
• chore: fix discord and x badge by @kemingy in #170
• feat: unify vchordrq and vchordrqfscan by @usamoi in #167
• fix: respect aliasing rule by not reading past of reference by @usamoi in #169
• fix: correct output of prewarm by @usamoi in #173
• fix: add magic number and version number to meta tuple by @usamoi in #174
• Release 0.2.0 by @cutecutecat in #177
• chore: install zip in tensorchord-pgrx by @usamoi in #178
• fix: release package name, version and licenses by @usamoi in #179

New Contributors

• @xieydd made their first contribution in #114

Full Changelog: 0.1.0...0.2.0

Highlights

• Support fp16 vec
• Support vector longer than 2000 dim

What's Changed

• fix: Unlock the conversation in CLA bot by @gaocegege in #86
• fix: CI env SEMVER by @kemingy in #87
• fix: format by @cutecutecat in #90
• feat: max_scan_tuples by @usamoi in #94
• test: fix push down tests by @cutecutecat in #95
• fix: set max dimension to 1600 in readme by @usamoi in #97
• fix: set max dimension to 1600 by @usamoi in #98
• docs: Update README by @VoVAllen in #103
• feat: support up to 60000 dimensions by @usamoi in #100
• chore: directory structure by @usamoi in #104
• feat: vchordrqfscan by @usamoi in #105
• chore: add CI to build the pgrx image by @kemingy in #107
• chore: fix the pgrx ci by @kemingy in #108
• fix: optimize insertions in building when lists = 1 by @usamoi in #106
• chore: build multiarch docker images by @kemingy in #112
• docs(readme): fix markdown style for docker run by @kemingy in #113
• feat: improve internal build by @usamoi in #115
• add enterprise image build step to ci by @xieydd in #114
• chore(README): Add some benchmark data by @gaocegege in #126
• fix: use stable toolchain by default by @usamoi in #128
• feat: scalar8 & indexing on halfvec by @usamoi in #131
• feat: Use dual license (AGPLv3 and ELv2) by @gaocegege in #130
• chore: update base by @usamoi in #137
• fix: remove sudo in dockerfile by @usamoi in #138
• fix: ci by @usamoi in #139
• fix: preprocess for halfvec by @usamoi in #140

New Contributors

• @xieydd made their first contribution in #114

Full Changelog: 0.1.0...0.1.1-alpha.1