USearch
USearch is a C++11-based similarity search and clustering engine for vectors and random objects, consisting of a single header file. It is based on the HNSW algorithm and offers low-precision data type support independent of hardware with SIMD optimizations. By enabling multi-programming language bindings, it allows large directories over disk to be displayed without installing RAM.
unum-cloud · Search Engines
Common commercial products in this category: Algolia, Elastic Cloud, Azure AI Search
- Using the HNSW algorithm, you can search for similarities on vectors and objects.
- You can view large folders on the disk directly through the file without uploading them to memory.
- With JIT compilation support, you can run user-defined metrics and SIMD optimized calculations.
- ↓ NegativeHacker News
USearch: Smaller and faster single-file vector search engine
“I will reply with a meme from my recent talk on Vector Databases, if that is in order: § Supporting advanced filtering in USearch In the low-level C++ interface we already support arbitrary predicates (callbacks) evaluated during HNSW graph traversal. JIT-ing them from the…”
- ↓ NegativeHacker News
USearch: Smaller and faster single-file vector search engine
“Sure! You can pass exact=True to the search interface. > Adding to index appears to be very slow. Interesting. Can you please elaborate? We benchmark it on daily basis, but there is always a chance we forgot some corner case :) PS: Thanks for considering us! USearch is already…”
My note
Alternative pages
More in: Search Engines
- Meilisearch★ 59,725 · yesterday
- ZincSearch★ 17,927 · 1 mo ago
- xerj★ 17,772 · today
- Tantivy★ 16,368 · yesterday
- zvec★ 16,164 · yesterday
- OpenSearch★ 13,937 · today