Akshay (@akshay_pachaar)
벡터 DB는 단일 쿼리와 개별 청크 유사도 검색에는 적합하지만, 여러 청크의 정보를 종합해야 하는 질문에서는 한계가 있다고 지적한다. FalkorDB의 GraphRAG-Bench 결과를 근거로, GraphRAG 방식이 이런 다중 홉 추론 문제에서 격차를 드러낸다고 설명한다.
Vector DBs can't reason. Top-k similarity ranks chunks one at a time against a query. That's fine for single-hop fact lookups, and it breaks the moment a question needs information stitched across multiple chunks. That's what the FalkorDB GraphRAG-Bench results expose. The gap
`foxing` (formerly xfs-mirror) aspires to be a production-grade, eBPF-powered replication engine for Linux filesystems (XFS, Btrfs, F2FS, Ext4). It captures filesystem events in the kernel and replays them asynchronously on a target directory, providing near real-time mirroring with robust consis...
everyone is talking about RAG, so I went down the rabbit hole and found the boring part that actually makes it work: retrieval and chunking
over the last months I've been studying and experimenting with vector databases and local LLMs, and I turned that into a three–part series on the Storyblok blog
today the last and most challenging piece is out: a step-by-step guide to building a fully local RAG pipeline with @weaviate + @ollama
in the article I show, with real code and a full repo linked:
• why hybrid search (BM25 + vectors) beats pure vector search
• how bad chunking quietly ruins most RAG systems
• why a smaller model + good retrieval often beats a huge (expensive) model with bad context
• how a structured CMS (like Storyblok) basically gives you chunking for free
stack:
• Weaviate for vectors and hybrid search
• Node.js for the glue
• Qwen 3.5 on Ollama running locally (but this works with cloud models too)
if you work on docs, DX, or AI features for content-heavy products, this might be a useful starting point
and since I'm still new to the topic, I'd really like feedback from you if you've built RAG systems in production
full article with code + repo:
How to build a RAG pipeline with Weaviate and Ollama – https://www.storyblok.com/mp/how-to-build-a-rag-pipeline-with-weaviate-and-ollama
Avi Chawla (@_avichawla)
에이전트 메모리가 많아질수록 오히려 더 적게 아는 것처럼 보인다는 관점을 제시하며, 현재의 에이전트 메모리 구조가 기억 저장소의 성질을 그대로 따르기 때문에 vector DB 기반 기억의 한계를 설명하는 트윗이다.
The more your agent remembers, the less it knows. This sounds counterintuitive, but it is actually a direct result of how agent memory is built today. Agent memory inherits the cognitive shape of its store. - A vector DB gives it associative memory to recognize familiar
Avi Chawla (@_avichawla)
RAG의 한계를 설명하며, 자주 변하지 않는 정보도 매번 벡터 DB를 조회해 비용과 지연이 발생하는 문제를 지적한다. 이를 해결하는 Cache-Augmented Generation(CAG)을 소개하며, 캐시를 활용해 더 빠르고 효율적인 생성 방식을 제안한다.
https://x.com/_avichawla/status/2045767552526340205
#rag #cag #vectordb #generativeai #retrievalaugmentedgeneration
RAG vs. CAG, clearly explained! RAG is great, but it has a major problem: Every query hits the vector DB. Even for static information that hasn't changed in months. This is expensive, slow, and unnecessary. Cache-Augmented Generation (CAG) addresses this issue by enabling the
@OpenSearchProj was named a Leader and Fast Mover in the 2025 GigaOm Radar for Vector Databases 🏆
My #OpenSearch report highlights:
✅ Platform play
✅ Search variety
✅ Business criteria
✅ Security
And I'd add - it's OPEN SOURCE @linuxfoundation !!
https://opensearch.org/gigaom-radar-vector-report-2025/
My talk from #NDCLondon is now available on YouTube! 🇬🇧
http://youtube.com/watch?v=N2kn4UybFBI&list=PLPFMHjhoDntsPHFHTN1d4H-U3G8zRlJ-p&index=1&t=9s
@OpenSearchProject @linuxfoundation
#vectorDB #OpenSearch #opensource #search #GenAI #OpenSearchAmbassador
Stoked seeing the OpenSearch Project featured by Jensen Huang on #NVIDIA #GTC keynote! 😍
One of the innovations in #OpenSearch V3 has been adding GPU acceleration based on NVIDIA's cuVS. Our #VectorSearch benchmarks, using CAGRA algorithm integrated through Facebook's Faiss library, showed:
✅ 9.3x faster index builds
✅ 3.75x lower cost
✅ 2x higher throughput
✅ 2.5x lower CPU usage
https://www.linkedin.com/feed/update/urn:li:activity:7439600547852189697/
#OpenSearchAmbassador #opensource #gtc2026 #gtc26 #cuvs #vectordb
Stoked seeing the OpenSearch Project featured by Jensen Huang on NVIDIA #GTC keynote! 😍 One of the innovations in #OpenSearch V3 has been adding GPU acceleration based on #NVIDIA's cuVS. Our benchmarks, using CAGRA algorithm integrated through Facebook's Faiss library, showed: ✅ 9.3x faster index builds ✅ 3.75x lower cost ✅ 2x higher throughput ✅ 2.5x lower CPU usage That's the power of bringing the best of #opensource in vector search together. Check out the comments for the full benchmark setup and results, and more details on the architecture, as well as the RFC on GitHub. Well done to Navneet Verma Corey Nolet Kshitiz G. Dylan Tong Nathan Stephens Vamshi Vijay Nakkirtha and all involved! #OpenSearchAmbassador #VectorSearch
sayzard
Joel Wirāmu, Pauling
Edoardo Dusi
Dotan Horovits #CNCFAmbassador