# VectorDB

> Lightweight Python package for storing and retrieving text using chunking, embeddings, and vector search. Powers AI features in Kagi Search with low latency and small memory footprint. This is an OSS library.

- **URL:** https://github.com/kagisearch/vectordb
- **Category:** Sdks & Libraries
- **Tags:** Open Source, Python, Lightweight
- **Updated:** 2026-03-06 18:05
- **Canonical page:** https://vectordb.ever.works/items/vectordb

## Details

## Overview

VectorDB is a minimal Python package for storing and retrieving text using chunking, embeddings, and vector search. Designed and used by Kagi Search to power AI features with optimal performance.

## Key Features

- **Simple and Lightweight**: Minimal dependencies and straightforward API
- **Fully Local**: All operations handled locally including embeddings and search
- **Performance Optimized**: Automatically selects best algorithm based on data size
- **Low Latency**: Optimized for fast query responses
- **Small Memory Footprint**: Efficient resource usage
- **End-to-End Solution**: Complete package from chunking to retrieval

## Automatic Algorithm Selection

VectorDB intelligently chooses the best approach:
- **Faiss**: For low number of chunks (<4000)
- **MRPT**: For high number of chunks (≥4000)

This ensures maximum performance across different use cases without manual configuration.

## Technical Capabilities

- Text chunking and preprocessing
- Embedding generation (local)
- Vector indexing
- Similarity search
- Result retrieval and ranking

## Production Use

Powers AI features inside Kagi Search, demonstrating:
- Production-ready stability
- Real-world performance at scale
- Reliability for user-facing features

## Use Cases

- Semantic search in applications
- Document retrieval systems
- Question answering
- Context retrieval for LLMs
- Content recommendation
- Knowledge base search

## Installation

Available via PyPI:
```bash
pip install vectordb
```

## API

Simple, intuitive Python API for:
- Creating vector stores
- Adding documents
- Querying for similar content
- Managing embeddings

## Integration

Easy integration with:
- Language models (OpenAI, local models)
- Python applications
- Web services
- CLI tools

## Design Philosophy

- **Simplicity**: Easy to understand and use
- **Local-First**: No external dependencies for core functionality
- **Performance**: Optimized for speed and efficiency
- **Transparency**: Clear operations without hidden complexity

## Repository

- GitHub: kagisearch/vectordb
- PyPI: vectordb
- Documentation: In repository README
- Active maintenance by Kagi Search team

## Comparison to Alternatives

### vs. Heavy Solutions
- Much lighter weight
- Faster startup time
- Lower resource usage
- Simpler deployment

### vs. External Services
- No network latency
- Complete data privacy
- No API costs
- Offline capable

## Pricing

Free and open-source. No licensing costs. Available on GitHub under permissive license.

---

_This Markdown mirror is generated by Ever Works for AI agents. The canonical HTML page is at https://vectordb.ever.works/items/vectordb._