Repolode
Submit
alibaba

alibaba/zvec

An open-source, in-process vector database that embeds directly into applications for low-latency similarity search without running a server.

Open-source alternative to:Pinecone

View on GitHubWebsite
Stars
16,127
Forks
1,014
Open issues
87
Last commit
20 hours ago

What is it?

Zvec is an in-process vector database: it runs as a library inside your application rather than as a separate server, so it works in notebooks, backend services, CLI tools, and edge devices. It stores dense and sparse vectors and supports similarity search with a range of index types, from in-memory structures to disk-based ones.

Beyond vector search, Zvec provides full-text search over string fields and hybrid queries that combine vector similarity, keyword matching, and structured filters in one request. Writes are protected by a write-ahead log, and multiple processes can read the same collection at once, while writes are restricted to a single process.

Official SDKs are available for Python, Node.js, Go, Rust, and Dart/Flutter, with a C++ core and a C API. Prebuilt binaries cover Linux (x86_64 and ARM64, glibc and musl), macOS, and Windows.

Key features

  • 01Dense and sparse vector storage with multiple index types
  • 02Full-text search over string fields
  • 03Hybrid queries combining vector, keyword, and filter conditions
  • 04Write-ahead logging for durable storage
  • 05Multiple processes can read the same collection concurrently
  • 06Official SDKs for Python, Node.js, Go, Rust, and Dart/Flutter

Use cases

  • Adding semantic search or retrieval to a RAG pipeline without deploying a database server
  • Storing and querying embeddings for an AI agent's long-term memory
  • Running similarity search inside a desktop, mobile, or edge application
  • Combining vector search with keyword and metadata filters in one query

Quick start

  1. 1

    Install the Python SDK

    pip install zvec
  2. 2

    Run the one-minute example (create, insert, and query a collection)

    python

Things to know

  • Building from source is a separate, more involved process described in its own guide.
  • Writes are single-process exclusive, so concurrent writing from multiple processes is not supported.
  • Python SDK requires 64-bit Python 3.10 to 3.14.

Topics

#vector-database#embedded#similarity-search#rag#hnsw#full-text-search#semantic-search#agent-skills#db#faiss

FAQ

What is zvec?
An open-source, in-process vector database that embeds directly into applications for low-latency similarity search without running a server. Zvec is an in-process vector database: it runs as a library inside your application rather than as a separate server, so it works in notebooks, backend services, CLI tools, and edge devices. It stores dense and sparse vectors and supports similarity search with a range of index types, from in-memory structures to disk-based ones.
Is zvec free to use? What license does it have?
zvec is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is zvec still maintained?
Actively maintained. The last commit to zvec was on Oct 19, 2026.
What can I use zvec for?
Adding semantic search or retrieval to a RAG pipeline without deploying a database server Storing and querying embeddings for an AI agent's long-term memory Running similarity search inside a desktop, mobile, or edge application Combining vector search with keyword and metadata filters in one query
How do I install zvec?
Install the Python SDK: pip install zvec Run the one-minute example (create, insert, and query a collection): python
Where can I get zvec?
The source code and installation instructions are on GitHub: https://github.com/alibaba/zvec

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects