VChord PG #extension @cloudnativepg.bsky.social cluster. #RaBitQ quants (4 & 8-bit) ⬇️ memory & disk
6-26x more vectors for the same 💵
#BM25, pre-filtering, 65k dimensions, #MultiVector, compat w #pgvector
#Embeddings #AI #RAG #Agents
#CNPG: Apache 2.0
#OSS VectorChord: AGPLv3 or Elastic License v2
6-26x more vectors for the same 💵
#BM25, pre-filtering, 65k dimensions, #MultiVector, compat w #pgvector
#Embeddings #AI #RAG #Agents
#CNPG: Apache 2.0
#OSS VectorChord: AGPLv3 or Elastic License v2
GitHub - tensorchord/cloudnative-vectorchord
Contribute to tensorchord/cloudnative-vectorchord development by creating an account on GitHub.
github.com
November 7, 2026 at 3:00 PM
VChord PG #extension @cloudnativepg.bsky.social cluster. #RaBitQ quants (4 & 8-bit) ⬇️ memory & disk
6-26x more vectors for the same 💵
#BM25, pre-filtering, 65k dimensions, #MultiVector, compat w #pgvector
#Embeddings #AI #RAG #Agents
#CNPG: Apache 2.0
#OSS VectorChord: AGPLv3 or Elastic License v2
6-26x more vectors for the same 💵
#BM25, pre-filtering, 65k dimensions, #MultiVector, compat w #pgvector
#Embeddings #AI #RAG #Agents
#CNPG: Apache 2.0
#OSS VectorChord: AGPLv3 or Elastic License v2
pg_pinecone (Apache-2.0) runs IVF+RaBitQ vector search inside Postgres, no Pinecone account required. Filters run during search, not after HNSW retrieval. Pinecone reports a 136 MB index for 1M×768d vectors; benchmark is vendor-run.
www.pinecone.io/blog/introdu...
www.pinecone.io/blog/introdu...
Introducing pg_pinecone: Open-Source Search for Your Postgres Data
Apache-2.0 PostgreSQL extension with native RaBitQ vector search and filter-aware retrieval.
www.pinecone.io
October 29, 2026 at 6:55 AM
pg_pinecone (Apache-2.0) runs IVF+RaBitQ vector search inside Postgres, no Pinecone account required. Filters run during search, not after HNSW retrieval. Pinecone reports a 136 MB index for 1M×768d vectors; benchmark is vendor-run.
www.pinecone.io/blog/introdu...
www.pinecone.io/blog/introdu...
📢 New #updates · 28 Oct #5.23
· VectorChord 1.1.0 — Added RaBitQ-quantized vector types for storage efficiency
· VectorChord 1.1.0 — Added RaBitQ-quantized vector types for storage efficiency
October 28, 2026 at 5:55 AM
📢 New #updates · 28 Oct #5.23
· VectorChord 1.1.0 — Added RaBitQ-quantized vector types for storage efficiency
· VectorChord 1.1.0 — Added RaBitQ-quantized vector types for storage efficiency
The side-by-side makes the design choice legible: both Postgres indexes use clustering and RaBitQ for the shortlist, then rerank at full precision.
Two Postgres vector indexes shipped days apart with the same design: k-means or IVF clustering, RaBitQ at about one bit per dimension, full-precision rerank of the shortlist.
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
Announcing pg_vectorsearch: Scalable Open-Source Vector Search in PostgreSQL
pg_vectorsearch is a new open-source Postgres extension. Its PRISM index serves 100M vectors in 6 ms on 4x less RAM than HNSW, with builds in under 2 hours.
www.tigerdata.com
October 25, 2026 at 5:52 PM
The side-by-side makes the design choice legible: both Postgres indexes use clustering and RaBitQ for the shortlist, then rerank at full precision.
Two Postgres vector indexes shipped days apart with the same design: k-means or IVF clustering, RaBitQ at about one bit per dimension, full-precision rerank of the shortlist.
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
Announcing pg_vectorsearch: Scalable Open-Source Vector Search in PostgreSQL
pg_vectorsearch is a new open-source Postgres extension. Its PRISM index serves 100M vectors in 6 ms on 4x less RAM than HNSW, with builds in under 2 hours.
www.tigerdata.com
October 23, 2026 at 5:07 PM
Two Postgres vector indexes shipped days apart with the same design: k-means or IVF clustering, RaBitQ at about one bit per dimension, full-precision rerank of the shortlist.
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
pg_vectorsearch → www.tigerdata.com/blog/announc...
Lakebase Search → www.databricks.com/blog/lakebas...
pgexperts.com
Experiments:
RaBitQ v remex: github.com/oaustegard/e...
Matryoshka v Quantization retrieval performance: github.com/oaustegard/e...
RaBitQ v remex: github.com/oaustegard/e...
Matryoshka v Quantization retrieval performance: github.com/oaustegard/e...
October 14, 2026 at 11:44 AM
Experiments:
RaBitQ v remex: github.com/oaustegard/e...
Matryoshka v Quantization retrieval performance: github.com/oaustegard/e...
RaBitQ v remex: github.com/oaustegard/e...
Matryoshka v Quantization retrieval performance: github.com/oaustegard/e...
This argument had me scratching my head a bit, and led to a deeper dive on the RaBitQ embedding, which led to a new remex version
pypi.org/project/remex/
pypi.org/project/remex/
October 14, 2026 at 11:44 AM
This argument had me scratching my head a bit, and led to a deeper dive on the RaBitQ embedding, which led to a new remex version
pypi.org/project/remex/
pypi.org/project/remex/
TIL about www.lancedb.com/blog/feature...
Good to see that I'm not insane and that the remex-like vector quantization approach has real implementations and demonstrated performance benefits
(remex: pypi.org/project/remex/)
Good to see that I'm not insane and that the remex-like vector quantization approach has real implementations and demonstrated performance benefits
(remex: pypi.org/project/remex/)
LanceDB's RaBitQ Quantization for Blazing Fast Vector Search
Introducing RaBitQ quantization in LanceDB for higher compression, faster indexing, and better recall on high‑dimensional embeddings.
www.lancedb.com
October 14, 2026 at 12:08 AM
TIL about www.lancedb.com/blog/feature...
Good to see that I'm not insane and that the remex-like vector quantization approach has real implementations and demonstrated performance benefits
(remex: pypi.org/project/remex/)
Good to see that I'm not insane and that the remex-like vector quantization approach has real implementations and demonstrated performance benefits
(remex: pypi.org/project/remex/)
But yeah, as fun as it was to develop remex, realistically RaBitQ is the way to go, here’s yet another commercial implementation: www.databricks.com/blog/announc...
Announcing Lakebase Search: agent-native retrieval built into Lakebase Postgres
Toda
www.databricks.com
October 14, 2026 at 12:07 PM
But yeah, as fun as it was to develop remex, realistically RaBitQ is the way to go, here’s yet another commercial implementation: www.databricks.com/blog/announc...
Keeping the query in float while the index is binary is the same asymmetry RaBitQ leans on, so it's nice to see both land in the same place. Did a rerank pass over the full vectors still pay off at your corpus size, or did the asymmetric distance get close enough on its own?
October 14, 2026 at 2:45 AM
Keeping the query in float while the index is binary is the same asymmetry RaBitQ leans on, so it's nice to see both land in the same place. Did a rerank pass over the full vectors still pay off at your corpus size, or did the asymmetric distance get close enough on its own?
The new pg_vectorsearch extension introduces the PRISM index for PostgreSQL, offering high-performance approximate nearest neighbor search. Currently in alpha, it uses RaBitQ quantization and hierarchical routing to outperform existing methods.
Pg_vectorsearch: An index that is faster than pgvector's IVFFlat and HNSW (152)
High-performance vector search for PostgreSQL. Contribute to timescale/pg_vectorsearch development by creating an account on GitHub.
news.ycombinator.com
October 14, 2026 at 6:35 PM
The new pg_vectorsearch extension introduces the PRISM index for PostgreSQL, offering high-performance approximate nearest neighbor search. Currently in alpha, it uses RaBitQ quantization and hierarchical routing to outperform existing methods.
with RaBitQ, each quantized score includes an error bound. to reduce rescoring without recall loss, we gossip minimum scores among query threads: if a candidate's error bounds don't overlap the global min, we prune it to skip rescoring
blog post 🔜
blog post 🔜
October 8, 2026 at 3:24 PM
with RaBitQ, each quantized score includes an error bound. to reduce rescoring without recall loss, we gossip minimum scores among query threads: if a candidate's error bounds don't overlap the global min, we prune it to skip rescoring
blog post 🔜
blog post 🔜
Yuexuan Xu, Jianyang Gao, Michael Norris, Alibek Zhakubayev, Pankaj Singh, Junjie Qi, Matthijs Douze, Cheng Long: RaBitQ-SSD: Split Codes and Pipelined I/O for SSD-Resident Vector Search https://arxiv.org/abs/2610.02652 https://arxiv.org/pdf/2610.02652 https://arxiv.org/html/2610.02652
October 5, 2026 at 6:40 AM
Yuexuan Xu, Jianyang Gao, Michael Norris, Alibek Zhakubayev, Pankaj Singh, Junjie Qi, Matthijs Douze, Cheng Long: RaBitQ-SSD: Split Codes and Pipelined I/O for SSD-Resident Vector Search https://arxiv.org/abs/2610.02652 https://arxiv.org/pdf/2610.02652 https://arxiv.org/html/2610.02652
New blog post as my HNSW #duckdb plugin hits v0.2 with support for quantisation (rabitq) and join+where pushdown and groupby+aggregations. cigrainger.com/blog/duckdb-...
Making HNSW actually work with WHERE clauses — Christopher Grainger
Transparent filtered vector search in DuckDB, or: why post-filtering is broken and what to do about it.
cigrainger.com
March 29, 2026 at 10:26 PM
New blog post as my HNSW #duckdb plugin hits v0.2 with support for quantisation (rabitq) and join+where pushdown and groupby+aggregations. cigrainger.com/blog/duckdb-...
DuckDB with HNSW, ACORN & RaBitQ
nice blogpost about implementing ACORN in a DuckDB plugin and why it makes so much sense.
cigrainger.com/blog/duckdb-...
nice blogpost about implementing ACORN in a DuckDB plugin and why it makes so much sense.
cigrainger.com/blog/duckdb-...
Making HNSW actually work with WHERE clauses — Christopher Grainger
Transparent filtered vector search in DuckDB, or: why post-filtering is broken and what to do about it.
cigrainger.com
April 16, 2026 at 12:56 PM
DuckDB with HNSW, ACORN & RaBitQ
nice blogpost about implementing ACORN in a DuckDB plugin and why it makes so much sense.
cigrainger.com/blog/duckdb-...
nice blogpost about implementing ACORN in a DuckDB plugin and why it makes so much sense.
cigrainger.com/blog/duckdb-...
Bring Vector Compression to the Extreme: How Milvus Serves 3× More Queries with RaBitQ
Nide to see more RabitQ adoption, after Lucene, Elasticsearch, CockroachDB and VectorChord.
Nide to see more RabitQ adoption, after Lucene, Elasticsearch, CockroachDB and VectorChord.
Bring Vector Compression to the Extreme: How Milvus Serves 3× More Queries with RaBitQ - Milvus Blog
Discover how Milvus leverages RaBitQ to enhance vector search efficiency, reducing memory costs while maintaining accuracy. Learn to optimize your AI solutions today!
milvus.io
June 10, 2025 at 12:56 PM
Bring Vector Compression to the Extreme: How Milvus Serves 3× More Queries with RaBitQ
Nide to see more RabitQ adoption, after Lucene, Elasticsearch, CockroachDB and VectorChord.
Nide to see more RabitQ adoption, after Lucene, Elasticsearch, CockroachDB and VectorChord.
I came across your computational SSD post and, down the rabbit hole last night, I found these.
SmartANNS (USENIX ATC 2024): www.usenix.org/conference/a...
RaBitQ (SIGMOD 2024): arxiv.org/abs/2405.12497
SmartANNS (USENIX ATC 2024): www.usenix.org/conference/a...
RaBitQ (SIGMOD 2024): arxiv.org/abs/2405.12497
Scalable Billion-point Approximate Nearest Neighbor Search Using SmartSSDs | USENIX
Bing Tian, Haikun Liu, Zhuohui Duan, Xiaofei Liao, Hai Jin, and Yu Zhang, Huazhong University of Science and Technology
www.usenix.org
April 10, 2026 at 2:49 PM
I came across your computational SSD post and, down the rabbit hole last night, I found these.
SmartANNS (USENIX ATC 2024): www.usenix.org/conference/a...
RaBitQ (SIGMOD 2024): arxiv.org/abs/2405.12497
SmartANNS (USENIX ATC 2024): www.usenix.org/conference/a...
RaBitQ (SIGMOD 2024): arxiv.org/abs/2405.12497
He, Ye, Cai, Lv, Cui, Yan, Zhan, Zhang, Yi, Xiang, Li, Gai, Zhang, Zheng, Du: Ascend-RaBitQ: Heterogeneous NPU-CPU Acceleration of Billion-Scale Similarity Search with 1-bit Quantization https://arxiv.org/abs/2605.16007 https://arxiv.org/pdf/2605.16007 https://arxiv.org/html/2605.16007
May 18, 2026 at 6:41 AM
He, Ye, Cai, Lv, Cui, Yan, Zhan, Zhang, Yi, Xiang, Li, Gai, Zhang, Zheng, Du: Ascend-RaBitQ: Heterogeneous NPU-CPU Acceleration of Billion-Scale Similarity Search with 1-bit Quantization https://arxiv.org/abs/2605.16007 https://arxiv.org/pdf/2605.16007 https://arxiv.org/html/2605.16007
Jifan Shi, Jianyang Gao, James Xia, Tam\'as B\'ela Feh\'er, Cheng Long: GPU-Native Approximate Nearest Neighbor Search with IVF-RaBitQ: Fast Index Build and Search https://arxiv.org/abs/2602.23999 https://arxiv.org/pdf/2602.23999 https://arxiv.org/html/2602.23999
March 2, 2026 at 6:30 AM
Jifan Shi, Jianyang Gao, James Xia, Tam\'as B\'ela Feh\'er, Cheng Long: GPU-Native Approximate Nearest Neighbor Search with IVF-RaBitQ: Fast Index Build and Search https://arxiv.org/abs/2602.23999 https://arxiv.org/pdf/2602.23999 https://arxiv.org/html/2602.23999
In-process embeddable #vectorDB
#Sparse, #Dense, #Multi-Vector
#HybridSearch #BM25, #RabitQ
Python, Node.JS bindings
#OpenSource Apache 2.0 lic
#Sparse, #Dense, #Multi-Vector
#HybridSearch #BM25, #RabitQ
Python, Node.JS bindings
#OpenSource Apache 2.0 lic
GitHub - alibaba/zvec: A lightweight, lightning-fast, in-process vector database
A lightweight, lightning-fast, in-process vector database - alibaba/zvec
github.com
May 3, 2026 at 9:59 PM
In-process embeddable #vectorDB
#Sparse, #Dense, #Multi-Vector
#HybridSearch #BM25, #RabitQ
Python, Node.JS bindings
#OpenSource Apache 2.0 lic
#Sparse, #Dense, #Multi-Vector
#HybridSearch #BM25, #RabitQ
Python, Node.JS bindings
#OpenSource Apache 2.0 lic
📊 RaBitQ by #LanceDB cuts storage up to 32× with only ~5% recall loss — boosting speed & efficiency for #AI workloads.
🔗 Link in first 💬⤵️
Repost 🔁 #RAG #VectorDB #VectorStore #SemanticSearch
🔗 Link in first 💬⤵️
Repost 🔁 #RAG #VectorDB #VectorStore #SemanticSearch
October 16, 2025 at 8:57 PM
📊 RaBitQ by #LanceDB cuts storage up to 32× with only ~5% recall loss — boosting speed & efficiency for #AI workloads.
🔗 Link in first 💬⤵️
Repost 🔁 #RAG #VectorDB #VectorStore #SemanticSearch
🔗 Link in first 💬⤵️
Repost 🔁 #RAG #VectorDB #VectorStore #SemanticSearch
"アルゴリズムのパフォーマンスを段階的に改善する | POSTD" https://postd.cc/rabitq-bench/
December 20, 2024 at 4:08 AM
"アルゴリズムのパフォーマンスを段階的に改善する | POSTD" https://postd.cc/rabitq-bench/
✅ Huawei's Ascend-RaBitQ speeds up billion-scale vector search 100x
Combines NPU coarse ranking with CPU fine ranking for 4.6x throughput gain.
https://theneuralfeed.com/share/post/pj0Flt89
Combines NPU coarse ranking with CPU fine ranking for 4.6x throughput gain.
https://theneuralfeed.com/share/post/pj0Flt89
Huawei's Ascend-RaBitQ speeds up billion-scale vector search 100x
theneuralfeed.com
May 18, 2026 at 7:24 AM
✅ Huawei's Ascend-RaBitQ speeds up billion-scale vector search 100x
Combines NPU coarse ranking with CPU fine ranking for 4.6x throughput gain.
https://theneuralfeed.com/share/post/pj0Flt89
Combines NPU coarse ranking with CPU fine ranking for 4.6x throughput gain.
https://theneuralfeed.com/share/post/pj0Flt89