Development · GitHub repo
Zvec
Lightweight, embedded in-process vector database with hybrid search and zero configuration
Zvec is an open-source, embedded vector database developed by Alibaba and built on its Proxima search engine. It runs directly inside applications without requiring external server processes, Docker containers, or cloud setup. It supports dense and sparse vectors, native full-text keyword search, and hybrid querying with persistent storage.
What it's for
- Build local RAG applications and agent memory stores without running a separate database server
- Perform hybrid semantic and full-text keyword searches across document collections
- Embed fast vector similarity search directly inside CLI tools, desktop apps, or edge devices
- Index multi-vector representations with structured metadata filtering
Features
- In-process embedded architecture requiring no external server or configuration
- Support for both dense and sparse vector embeddings with multi-vector querying
- Native full-text search (FTS) with customizable tokenizers
- Hybrid search combining vector similarity, full-text keywords, and structured filters
- Write-ahead logging (WAL) ensuring durable persistence across crashes
- Concurrent multi-process read access with single-process exclusive writes
- Multiple index types including HNSW and DiskANN with hardware-accelerated quantizers (AVX2 / AVX512)
- SDK bindings available for Python, Node.js, Go, Rust, C++, and Dart/Flutter
Pricing
Open source. Free and open source under Apache-2.0 license
Details
- Type
- GitHub repo
- Category
- Development
- Topic
- Vector Databases
- Open source
- Yes
- Platforms
- Linux, macOS, Windows, Code library, Command line
- For
- Software engineers and AI developers looking for a fast, serverless vector search library.
- Alternative to
- Chroma, LanceDB, FAISS, Qdrant, Milvus
- Tags
- vectordb, embeddings, rag, similarity search, semantic search, python, cpp
Why it was saved
- @hasantoxr: Production-grade, low-latency similarity search with minimal setup: searches billions of vectors in milliseconds, pip install zvec to start in under a minute, no Docker or cloud costs.
Related
Open-source collection of runnable AI agent, skill, and RAG templates
Lightweight Python toolkit for RAG using DuckDB or PostgreSQL
Turn websites and documents into clean, LLM-ready markdown or structured data.
Semantic search and web crawling API designed for AI agents
Open-source Python library and API for prompt-based web scraping using LLMs
Lightweight open-source Python library for building code-writing and vision AI agents
Updated