Skip to content

A fast efficient in-memory vector database designed for edge devices.

License

Notifications You must be signed in to change notification settings

adamped/edge_vector_index

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Edge Vector Index

A fast efficient in-memory vector index designed for use on edge devices.

WARNING

  • This library is experimental and uses simple cosine similarity for comparison

BENCHMARKS

C#

Local is a C# implementation using very similar code. Interop is using the Rust library. It shows that even with a slight FFI overhead the interop to the Rust library performs approximately 7x faster.

Method Mean Error StdDev
FindClosestMatch_Benchmark_Interop 594.7 us 7.16 us 6.34 us
FindClosestMatch_Benchmark_Local 4,331.7 us 12.97 us 12.13 us

Dart

Dart produces similar benchmarks

Non AOT

Method Mean
Benchmark_Interop(RunTime) 598.3645 us
Benchmark_Local(RunTime) 5733.4972 us

AOT

Method Mean
Benchmark_Interop(RunTime) 589.7700 us
Benchmark_Local(RunTime) 4741.7123 us

Python

Python had the worst runtime for local but similar for interop

Method Mean StdDev
Find Closest Match Interop 612 us 12 us
Find Closest Match Local 26.9 ms 0.7 us

JS vs WASM

This is for 10 iterations.

Method Mean
Find Closest Match Interop (WASM) 700 us
Find Closest Match Local (JS) 12400 us

This is for 1000 iterations. The speed of JS improves a lot due to caching of the code. This was run in a chromium engine.

Method Mean
Find Closest Match Interop (WASM) 619 us
Find Closest Match Local (JS) 2342 us

About

A fast efficient in-memory vector database designed for edge devices.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published