turbovec
Verifiedby Ryan Codrai
Indexes and searches embeddings with 2- to 4-bit TurboQuant compression in a Rust library with Python bindings, for local RAG pipelines that need a small memory footprint.
- Access
- API
- Auth
- None
- Skills
- 3
Skills
Filtered Search
Restricts search() to an id allowlist or slot bitmask inside the kernel, returning up to k results from the allowed set without over-fetching.
Incremental Saves
Persists only what changed since the last sync(path) call with one fsync, so small appends or removals stay fast on large indexes.
Framework Integrations
Replaces the in-memory vector stores of LangChain, LlamaIndex, Haystack and Agno through optional pip extras such as turbovec[langchain].
Related Agents
Qdrant
by Qdrant
High-performance open-source vector database. Supports dense, sparse, and multi-vector search. On-premise or cloud.…
Pinecone
by Pinecone
Fully managed vector database for AI applications. Store and search billions of high-dimensional embeddings with low…
LightRAG
by HKU Data Science Lab
Builds a graph-based retrieval index over documents so agents can answer questions spanning many sources, served…
PageIndex
by Vectify AI
Indexes long documents as a reasoning-based hierarchical tree so agents retrieve by navigating structure instead of…