zvec
ActiveDescription
A lightweight, lightning-fast, in-process vector database by Alibaba with C++ core, Node.js and Python bindings, designed for RAG, agent memory, and vector search use cases.
Key Features
- Blazing fast in-process vector database searching billions of vectors in milliseconds
- Dense and sparse vector support with native multi-vector queries in a single call
- Hybrid search combining semantic similarity with structured filters for precise results
- Durable storage via write-ahead logging (WAL) guaranteeing persistence across crashes
- Python and Node.js bindings with simple pip/npm install — zero config, zero servers
- Battle-tested within Alibaba Group for production-grade low-latency similarity search
Use Cases
💡 Embedding vector search for RAG (Retrieval-Augmented Generation) pipelines
💡 Agent memory storage and retrieval for long-term conversational context
💡 Real-time similarity search in edge devices, notebooks, and server applications
💡 Multi-modal embedding search combining text, image, and audio vectors
Strengths & Limitations
✅ Strengths
- • Actively maintained, recent updates
- • High community interest (15.6k stars)
- • Permissive open-source license (Apache-2.0)
- • Responsive to issues, low backlog
Categories
Quick Start
Install via pip: `pip install zvec` (Python 3.10-3.14) or npm: `npm install @zvec/zvec`. Define a schema, create a collection, insert documents with vectors, and run similarity queries. Example: `collection.query(zvec.VectorQuery("embedding", vector=[0.4,0.3,0.3,0.1]), topk=10)`.