vchord

Vector database plugin for Postgres, written in Rust

Overview

IDExtensionPackageVersionCategoryLicenseLanguage
1810
vchord
vchord
0.5.3
RAG
AGPL-3.0
Rust
AttributeHas BinaryHas LibraryNeed LoadHas DDLRelocatableTrusted
--sLd-r
No
Yes
Yes
Yes
yes
no
Relationships
Requires
vector
See Also
vectorscale
vectorize
vchord_bm25
pg_tiktoken
pgml
pg_bestmatch
pg_similarity
smlar

Packages

TypeRepoVersionPG Major AvailabilityPackage PatternDependencies
EL
PIGSTY
0.5.3
18
17
16
15
14
13
vchord_$vpgvector_$v
Debian
PIGSTY
0.5.3
18
17
16
15
14
13
postgresql-$v-vchordpostgresql-$v-pgvector
Linux / PGPG18PG17PG16PG15PG14PG13
el8.x86_64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
el8.aarch64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
el9.x86_64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
el9.aarch64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
el10.x86_64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
el10.aarch64
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
PIGSTY 0.5.3
MISS
d12.x86_64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
d12.aarch64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
d13.x86_64
MISS
MISS
MISS
MISS
MISS
MISS
d13.aarch64
MISS
MISS
MISS
MISS
MISS
MISS
u22.x86_64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
u22.aarch64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
u24.x86_64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
u24.aarch64
MISS
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
PIGSTY 0.5.1
MISS
PackageVersionOSORGSIZEFile URL
vchord_180.5.3el8.x86_64pigsty2.1 MiBvchord_18-0.5.3-1PIGSTY.el8.x86_64.rpm
vchord_180.5.3el8.aarch64pigsty1.8 MiBvchord_18-0.5.3-1PIGSTY.el8.aarch64.rpm
vchord_180.5.3el9.x86_64pigsty2.2 MiBvchord_18-0.5.3-1PIGSTY.el9.x86_64.rpm
vchord_180.5.3el9.aarch64pigsty1.9 MiBvchord_18-0.5.3-1PIGSTY.el9.aarch64.rpm
vchord_180.5.3el10.x86_64pigsty2.1 MiBvchord_18-0.5.3-1PIGSTY.el10.x86_64.rpm
vchord_180.5.3el10.aarch64pigsty1.8 MiBvchord_18-0.5.3-1PIGSTY.el10.aarch64.rpm
PackageVersionOSORGSIZEFile URL
vchord_170.5.3el8.x86_64pigsty2.1 MiBvchord_17-0.5.3-1PIGSTY.el8.x86_64.rpm
vchord_170.5.3el8.aarch64pigsty1.8 MiBvchord_17-0.5.3-1PIGSTY.el8.aarch64.rpm
vchord_170.5.3el9.x86_64pigsty2.2 MiBvchord_17-0.5.3-1PIGSTY.el9.x86_64.rpm
vchord_170.5.3el9.aarch64pigsty1.9 MiBvchord_17-0.5.3-1PIGSTY.el9.aarch64.rpm
vchord_170.5.3el10.x86_64pigsty2.1 MiBvchord_17-0.5.3-1PIGSTY.el10.x86_64.rpm
vchord_170.5.3el10.aarch64pigsty1.8 MiBvchord_17-0.5.3-1PIGSTY.el10.aarch64.rpm
postgresql-17-vchord0.5.1d12.x86_64pigsty1.0 MiBpostgresql-17-vchord_0.5.1-1PIGSTY~bookworm_amd64.deb
postgresql-17-vchord0.5.1d12.aarch64pigsty851.0 KiBpostgresql-17-vchord_0.5.1-1PIGSTY~bookworm_arm64.deb
postgresql-17-vchord0.5.1u22.x86_64pigsty1.1 MiBpostgresql-17-vchord_0.5.1-1PIGSTY~jammy_amd64.deb
postgresql-17-vchord0.5.1u22.aarch64pigsty1002.8 KiBpostgresql-17-vchord_0.5.1-1PIGSTY~jammy_arm64.deb
postgresql-17-vchord0.5.1u24.x86_64pigsty1.1 MiBpostgresql-17-vchord_0.5.1-1PIGSTY~noble_amd64.deb
postgresql-17-vchord0.5.1u24.aarch64pigsty997.5 KiBpostgresql-17-vchord_0.5.1-1PIGSTY~noble_arm64.deb
PackageVersionOSORGSIZEFile URL
vchord_160.5.3el8.x86_64pigsty2.1 MiBvchord_16-0.5.3-1PIGSTY.el8.x86_64.rpm
vchord_160.5.3el8.aarch64pigsty1.7 MiBvchord_16-0.5.3-1PIGSTY.el8.aarch64.rpm
vchord_160.5.3el9.x86_64pigsty2.2 MiBvchord_16-0.5.3-1PIGSTY.el9.x86_64.rpm
vchord_160.5.3el9.aarch64pigsty1.9 MiBvchord_16-0.5.3-1PIGSTY.el9.aarch64.rpm
vchord_160.5.3el10.x86_64pigsty2.0 MiBvchord_16-0.5.3-1PIGSTY.el10.x86_64.rpm
vchord_160.5.3el10.aarch64pigsty1.8 MiBvchord_16-0.5.3-1PIGSTY.el10.aarch64.rpm
postgresql-16-vchord0.5.1d12.x86_64pigsty1008.7 KiBpostgresql-16-vchord_0.5.1-1PIGSTY~bookworm_amd64.deb
postgresql-16-vchord0.5.1d12.aarch64pigsty830.1 KiBpostgresql-16-vchord_0.5.1-1PIGSTY~bookworm_arm64.deb
postgresql-16-vchord0.5.1u22.x86_64pigsty1.1 MiBpostgresql-16-vchord_0.5.1-1PIGSTY~jammy_amd64.deb
postgresql-16-vchord0.5.1u22.aarch64pigsty978.6 KiBpostgresql-16-vchord_0.5.1-1PIGSTY~jammy_arm64.deb
postgresql-16-vchord0.5.1u24.x86_64pigsty1.1 MiBpostgresql-16-vchord_0.5.1-1PIGSTY~noble_amd64.deb
postgresql-16-vchord0.5.1u24.aarch64pigsty970.2 KiBpostgresql-16-vchord_0.5.1-1PIGSTY~noble_arm64.deb
PackageVersionOSORGSIZEFile URL
vchord_150.5.3el8.x86_64pigsty2.1 MiBvchord_15-0.5.3-1PIGSTY.el8.x86_64.rpm
vchord_150.5.3el8.aarch64pigsty1.7 MiBvchord_15-0.5.3-1PIGSTY.el8.aarch64.rpm
vchord_150.5.3el9.x86_64pigsty2.2 MiBvchord_15-0.5.3-1PIGSTY.el9.x86_64.rpm
vchord_150.5.3el9.aarch64pigsty1.9 MiBvchord_15-0.5.3-1PIGSTY.el9.aarch64.rpm
vchord_150.5.3el10.x86_64pigsty2.0 MiBvchord_15-0.5.3-1PIGSTY.el10.x86_64.rpm
vchord_150.5.3el10.aarch64pigsty1.8 MiBvchord_15-0.5.3-1PIGSTY.el10.aarch64.rpm
postgresql-15-vchord0.5.1d12.x86_64pigsty1008.0 KiBpostgresql-15-vchord_0.5.1-1PIGSTY~bookworm_amd64.deb
postgresql-15-vchord0.5.1d12.aarch64pigsty830.8 KiBpostgresql-15-vchord_0.5.1-1PIGSTY~bookworm_arm64.deb
postgresql-15-vchord0.5.1u22.x86_64pigsty1.1 MiBpostgresql-15-vchord_0.5.1-1PIGSTY~jammy_amd64.deb
postgresql-15-vchord0.5.1u22.aarch64pigsty978.6 KiBpostgresql-15-vchord_0.5.1-1PIGSTY~jammy_arm64.deb
postgresql-15-vchord0.5.1u24.x86_64pigsty1.1 MiBpostgresql-15-vchord_0.5.1-1PIGSTY~noble_amd64.deb
postgresql-15-vchord0.5.1u24.aarch64pigsty971.3 KiBpostgresql-15-vchord_0.5.1-1PIGSTY~noble_arm64.deb
PackageVersionOSORGSIZEFile URL
vchord_140.5.3el8.x86_64pigsty2.1 MiBvchord_14-0.5.3-1PIGSTY.el8.x86_64.rpm
vchord_140.5.3el8.aarch64pigsty1.7 MiBvchord_14-0.5.3-1PIGSTY.el8.aarch64.rpm
vchord_140.5.3el9.x86_64pigsty2.2 MiBvchord_14-0.5.3-1PIGSTY.el9.x86_64.rpm
vchord_140.5.3el9.aarch64pigsty1.9 MiBvchord_14-0.5.3-1PIGSTY.el9.aarch64.rpm
vchord_140.5.3el10.x86_64pigsty2.0 MiBvchord_14-0.5.3-1PIGSTY.el10.x86_64.rpm
vchord_140.5.3el10.aarch64pigsty1.8 MiBvchord_14-0.5.3-1PIGSTY.el10.aarch64.rpm
postgresql-14-vchord0.5.1d12.x86_64pigsty1008.9 KiBpostgresql-14-vchord_0.5.1-1PIGSTY~bookworm_amd64.deb
postgresql-14-vchord0.5.1d12.aarch64pigsty830.8 KiBpostgresql-14-vchord_0.5.1-1PIGSTY~bookworm_arm64.deb
postgresql-14-vchord0.5.1u22.x86_64pigsty1.1 MiBpostgresql-14-vchord_0.5.1-1PIGSTY~jammy_amd64.deb
postgresql-14-vchord0.5.1u22.aarch64pigsty978.2 KiBpostgresql-14-vchord_0.5.1-1PIGSTY~jammy_arm64.deb
postgresql-14-vchord0.5.1u24.x86_64pigsty1.1 MiBpostgresql-14-vchord_0.5.1-1PIGSTY~noble_amd64.deb
postgresql-14-vchord0.5.1u24.aarch64pigsty970.9 KiBpostgresql-14-vchord_0.5.1-1PIGSTY~noble_arm64.deb

Source

pig build get vchord; # get vchord source code
pig build dep vchord; # install build dependencies
pig build pkg vchord; # build extension rpm or deb
pig build ext vchord; # build extension rpms

Install

To add the required PGDG / PIGSTY upstream repository, use:

pig repo add pgsql -u   # add PGDG + Pigsty repo and update cache (leave existing repos)

Install this extension with:

pig ext install vchord; # install by extension name, for the current active PG version
pig ext install vchord; # install via package alias, for the active PG version
pig ext install vchord -v 18;   # install for PG 18
pig ext install vchord -v 17;   # install for PG 17
pig ext install vchord -v 16;   # install for PG 16
pig ext install vchord -v 15;   # install for PG 15
pig ext install vchord -v 14;   # install for PG 14

Create this extension with:

CREATE EXTENSION vchord;

Usage

Add this extension to shared_preload_libraries in postgresql.conf

CREATE EXTENSION vchord CASCADE;

Create Index on embedding:

CREATE INDEX ON gist_train USING vchordrq (embedding vector_l2_ops) WITH (options = $$
residual_quantization = true
[build.internal]
lists = [4096]
spherical_centroids = false
$$);

Docs

Query

The query statement is exactly the same as pgvector. VectorChord supports any filter operation and WHERE/JOIN clauses like pgvecto.rs with VBASE.

SELECT * FROM items ORDER BY embedding <-> '[3,1,2]' LIMIT 5;

Supported distance functions are:

  • <-> - L2 distance
  • <#> - (negative) inner product
  • <=> - cosine distance

Due to the limitation of postgresql query planner, we cannot support the range query like SELECT embedding <-> '[3,1,2]' as distance WHERE distance < 0.1 ORDER BY distance directly.

To query vectors within a certain distance range, you can use the following syntax.

-- Query vectors within a certain distance range
-- sphere(center, radius) means the vectors within the sphere with the center and radius, aka range query
-- <<->> is L2 distance, <<#>> is inner product, <<=>> is cosine distance
SELECT vec FROM t WHERE vec <<->> sphere('[0.24, 0.24, 0.24]'::vector, 0.012) 

Query Performance Tuning

You can fine-tune the search performance by adjusting the probes and epsilon parameters:

-- Set probes to control the number of lists scanned. 
-- Recommended range: 3%–10% of the total `lists` value.
SET vchordrq.probes = 100;

-- Set epsilon to control the reranking precision.
-- Larger value means more rerank for higher recall rate.
-- Don't change it unless you only have limited memory.
-- Recommended range: 1.0–1.9. Default value is 1.9.
SET vchordrq.epsilon = 1.9;

-- vchordrq relies on a projection matrix to optimize performance.
-- Add your vector dimensions to the `prewarm_dim` list to reduce latency.
-- If this is not configured, the first query will have higher latency as the matrix is generated on demand.
-- Default value: '64,128,256,384,512,768,1024,1536'
-- Note: This setting requires a database restart to take effect.
ALTER SYSTEM SET vchordrq.prewarm_dim = '64,128,256,384,512,768,1024,1536';

And for postgres’s setting

-- If using SSDs, set `effective_io_concurrency` to 200 for faster disk I/O.
SET effective_io_concurrency = 200;

-- Disable JIT (Just-In-Time Compilation) as it offers minimal benefit (1–2%) 
-- and adds overhead for single-query workloads.
SET jit = off;

-- Allocate at least 25% of total memory to `shared_buffers`. 
-- For disk-heavy workloads, you can increase this to up to 90% of total memory. You may also want to disable swap with network storage to avoid io hang.
-- Note: A restart is required for this setting to take effect.
ALTER SYSTEM SET shared_buffers = '8GB';

Indexing prewarm

To prewarm the index, you can use the following SQL. It will significantly improve performance when using limited memory.

-- vchordrq_prewarm(index_name::regclass) to prewarm the index into the shared buffer
SELECT vchordrq_prewarm('gist_train_embedding_idx'::regclass)"

Index Build Time

Index building can parallelized, and with external centroid precomputation, the total time is primarily limited by disk speed. Optimize parallelism using the following settings:

-- Set this to the number of CPU cores available for parallel operations.
SET max_parallel_maintenance_workers = 8;
SET max_parallel_workers = 8;

-- Adjust the total number of worker processes. 
-- Note: A restart is required for this setting to take effect.
ALTER SYSTEM SET max_worker_processes = 8;

Indexing Progress

You can check the indexing progress by querying the pg_stat_progress_create_index view.

SELECT phase, round(100.0 * blocks_done / nullif(blocks_total, 0), 1) AS "%" FROM pg_stat_progress_create_index;

External Index Precomputation

Unlike pure SQL, an external index precomputation will first do clustering outside and insert centroids to a PostgreSQL table. Although it might be more complicated, external build is definitely much faster on larger dataset (>5M).

To get started, you need to do a clustering of vectors using faiss, scikit-learn or any other clustering library.

The centroids should be preset in a table of any name with 3 columns:

  • id(integer): id of each centroid, should be unique
  • parent(integer, nullable): parent id of each centroid, should be NULL for normal clustering
  • vector(vector): representation of each centroid, pgvector vector type

And example could be like this:

-- Create table of centroids
CREATE TABLE public.centroids (id integer NOT NULL UNIQUE, parent integer, vector vector(768));
-- Insert centroids into it
INSERT INTO public.centroids (id, parent, vector) VALUES (1, NULL, '{0.1, 0.2, 0.3, ..., 0.768}');
INSERT INTO public.centroids (id, parent, vector) VALUES (2, NULL, '{0.4, 0.5, 0.6, ..., 0.768}');
INSERT INTO public.centroids (id, parent, vector) VALUES (3, NULL, '{0.7, 0.8, 0.9, ..., 0.768}');
-- ...

-- Create index using the centroid table
CREATE INDEX ON gist_train USING vchordrq (embedding vector_l2_ops) WITH (options = $$
[build.external]
table = 'public.centroids'
$$);

To simplify the workflow, we provide end-to-end scripts for external index pre-computation, see scripts.


Limitations

  • Data Type Support: Currently, only the f32 data type is supported for vectors.
  • Architecture Compatibility: The fast-scan kernel is optimized for x86_64 architectures. While it runs on aarch64, performance may be lower.
  • KMeans Clustering: The built-in KMeans clustering is not yet fully optimized and may require substantial memory. We strongly recommend using external centroid precomputation for efficient index construction.

Build

Building this extension requires clang-17+

Which is available on EL 8/9, Ubuntu 24.04 directly, but require manual installation on Ubuntu 22.04 / Debian 12.

For example, install clang-18 on Ubuntu 22 / Debian 12 and set it as the default clang:

curl --proto '=https' --tlsv1.2 -sSf https://apt.llvm.org/llvm.sh | bash -s -- 18
sudo update-alternatives --install /usr/bin/clang clang $(which clang-18) 255
Last updated on