t

turbovec

A vector index built on TurboQuant, written in Rust with Python bindings

open-sourcememory-knowledge
17.3k
Stars
+1439
Stars/month
210
Commits (90d)
0
Releases (6m)

Star Growth

+4.3k (33.3%)estimated from velocity
12.7k15.2k17.6kJul 2Sep 30

Overview

turbovec is a vector index built on Google Research's TurboQuant algorithm, offering online ingest without a training phase and fast SIMD search. It provides incremental saves, filtered search capabilities, and is designed for local, privacy-focused RAG stacks.

Deep Analysis

Key Differentiator

Fits a 10M document corpus in 4GB RAM with faster search than FAISS using TurboQuant's data-oblivious quantization.

⚡ Capabilities

  • • vector indexing
  • • online ingest without training
  • • fast SIMD search
  • • incremental saves
  • • filtered search
  • • hybrid retrieval

🔗 Integrations

Python bindingsany open-source embedding model

✓ Best For

  • ✓ RAG applications where privacy matters
  • ✓ memory-constrained environments
  • ✓ local/air-gapped AI agent stacks

✗ Not Ideal For

  • ✗ managed vector database services
  • ✗ end-user AI applications

⚠ Known Limitations

  • ⚠ Pure local solution (no managed service)
  • ⚠ Requires casting to float32 arrays

Alternatives

See all 8 turbovec alternatives →

Compare turbovec

Maintain turbovec?

Show your live rank in your README, or put turbovec in front of every visitor to AgentoolRank.