

TurboVec
β Text source-checked Nov 19, 2026Fast vector quantization with 2-4 bit compression and SIMD search
Source-checked against the vendor's site Β· rebuilt from the vendor's own pagesProfile updated Link checked How we rate
In short
TurboVec is a free, in-process vector index for Python and Rust using 2-4 bit compression. It offers fast SIMD search with no training step and supports immediate vector updates.
TurboVec is an in-process vector index for Python and Rust. TurboVec stores each vector at 2 or 4 bits per dimension using Google Research's TurboQuant and searches that compressed form directly with agent-optimised SIMD kernels.
The vendor says that TurboVec requires no training step or rebuilds, as the codebook comes from the maths rather than user data. TurboVec allows users to add a vector and make it immediately searchable, and remove one in microseconds. TurboVec's sync function writes only what changed since the last call, using one fsync to remain crash-safe at any byte.
TurboVec is a library rather than a service, ensuring nothing leaves the user's process. TurboVec supports filtered search by passing an allowlist of ids or a bitmask, returning up to k results from the allowed set.
TurboVec provides drop-in framework integrations for LangChain, LlamaIndex, Haystack, and Agno.
TurboVec requires Python 3.9 or higher. TurboVec is available for MacOS, Windows, and Linux operating systems.
According to TurboVec: TurboVec is an in-process vector index for Python and Rust. Its pages list this strength: TurboVec is 7.8 times smaller than float32 at 4-bit and 15.5 times smaller at 2-bit. They also state this limitation: TurboVec requires vector dimensions to be multiples of 8 and no wider than 16,384.
β The AI Tool Bible editorial team
Pros
- β TurboVec is 7.8 times smaller than float32 at 4-bit and 15.5 times smaller at 2-bit.
- β TurboVec operates without a training step or rebuilds.
- β TurboVec provides incremental, crash-safe persistence via its sync function.
- β TurboVec runs locally as a library, keeping data within the user's process.
- β TurboVec offers drop-in integrations for LangChain, LlamaIndex, Haystack, and Agno.
Cons
- β οΈ TurboVec requires vector dimensions to be multiples of 8 and no wider than 16,384.
Facts
- Platforms
- macos, linux (not source-checked)
- Open source
- Yes (license: MIT)
- GitHub stars
- 17,491 (checked Nov 13, 2026) (not source-checked)
- Last GitHub push
- Oct 7, 2026 (not source-checked)
- First commit
- Mar 2026 (not source-checked)
- Package downloads (30 d)
- 39,739 Β· pypi turbovec Β· checked Nov 22, 2026 (not source-checked)
Where this tool sits
Use cases
Frequently asked
- How much does TurboVec cost?
- TurboVec is free. The provided documentation does not list any specific pricing details or tiers, indicating it is available at no cost to users.
- Which frameworks does TurboVec integrate with?
- TurboVec provides drop-in integrations for LangChain, LlamaIndex, Haystack, and Agno. It is designed as a library to fit directly into existing Python and Rust workflows.
- Does TurboVec require a training step?
- No, TurboVec requires no training step or rebuilds. The codebook is derived from mathematics rather than user data, allowing vectors to be added and made immediately searchable.
- What are the system requirements for TurboVec?
- TurboVec requires Python 3.9 or higher. It is available for MacOS, Windows, and Linux operating systems and functions as an in-process library.
- How does TurboVec handle data privacy?
- TurboVec is a library rather than a service, ensuring that data does not leave the user's process. It uses a sync function that writes only changes to remain crash-safe.