alibaba/zvec

A lightweight, lightning-fast, in-process vector database

What it solves

Zvec provides a lightweight, high-performance vector database that embeds directly into applications. It eliminates the need for complex server infrastructure, allowing developers to perform fast similarity searches, full-text searches, and hybrid queries locally or on edge devices.

How it works

Zvec operates as an in-process library, meaning it runs within your application's own memory space rather than as a separate network service. It supports both dense and sparse vector embeddings, utilizes write-ahead logging (WAL) for data durability, and offers various indexing strategies (such as HNSW and DiskANN) to balance speed and memory usage. It provides native SDKs for multiple programming languages and supports cross-platform deployment, including mobile environments.

Who it’s for

This project is for developers building AI-powered applications—such as RAG systems, search tools, or agentic workflows—who need a fast, portable, and easy-to-deploy vector storage solution without the overhead of managing a dedicated database server.

Highlights

  • In-process architecture: Runs directly inside your application with no external server dependencies.
  • Hybrid search capabilities: Combines vector similarity search with full-text search and structured filtering.
  • Broad platform support: Compatible with Linux, macOS, Windows, Android, and iOS.
  • Production-ready: Includes durability features like write-ahead logging and is battle-tested in large-scale environments.

Related

  • Project
    RyanCodrai/turbovec

    An in-process vector search library for Python and Rust that uses TurboQuant to compress embeddings by up to 15.5x, enabling fast, memory-efficient similarity search without the need for a dedicated database server.

  • Project
    zilliztech/VectorDBBench

    VectorDBBench is a Python benchmark suite for comparing the performance, recall, and cost of many vector‑search databases and cloud services using standard insertion and search workloads.

  • Project
    milvus-io/milvus-lite

    Milvus Lite is a pure‑Python, local‑only version of the Milvus vector database. It provides a drop‑in `pymilvus`‑compatible API for storing dense and sparse vectors, scalar fields, and full‑text data, and supports similarity search, filtering, and hybrid (vector + text) queries. Designed for notebooks, tests, and small prototypes, it runs either as a local `.db` file, an embedded engine, or a lightweight gRPC server, and can later be swapped to a full Milvus deployment with minimal code changes.

  • Project
    datastax/jvector

    JVector is a Java library for fast, scalable approximate nearest‑neighbor search. It merges HNSW’s hierarchical graph with DiskANN’s Vamana routing, supports two‑pass compressed/on‑disk searches, and uses SIMD (Google Highway or Panama) for high performance on modern JVMs.

  • Project
    milvus-io/milvus

    Milvus is a high-performance, distributed vector database built for scaling AI applications that require efficient search and organization of massive amounts of unstructured data.