Awesome ClickHouse › ClickHouse and derivatives

myscale/MyScaleDB

⭐ 1043 C++ repository created 2024-03-14

MyScaleDB is a high-performance SQL vector database built as a fork of ClickHouse, designed to support scalable AI applications with powerful vector search and full-text search capabilities. It enables developers to build production-ready AI solutions using familiar SQL syntax, eliminating the need to learn new tools or frameworks. MyScaleDB integrates vector search, filtered search, and SQL-vector join queries, providing a unified platform to manage and process diverse data types including structured data, text, vectors, JSON, geospatial, and time-series data. The project leverages ClickHouse's columnar storage and advanced compression techniques, optimizing for analytical workloads and big data processing. MyScaleDB enhances ClickHouse by adding advanced vector algorithms and modifications to the execution and storage engine to support fast, cost-effective SQL vector queries. This architecture allows for millisecond-level search performance on billion-scale vector datasets, making it suitable for large-scale AI applications. MyScaleDB also supports hybrid search combining text and vector data, improving retrieval-augmented generation (RAG) accuracy by filtering vectors with rich metadata and full-text search. The system is highly scalable, reliable, and designed for production environments. It offers a cloud service called MyScale Cloud for fully managed deployments with premium features and high data density indexing using the MSTG algorithm. The project emphasizes cost-effectiveness and simplicity compared to specialized vector databases and integrated solutions like PostgreSQL with pgvector or ElasticSearch with vector extensions. It unifies SQL database, vector database, and full-text search engine functionalities into a single system, reducing infrastructure and maintenance overhead while enabling complex joint queries and analytics. MyScaleDB is open source, with contributions back to the ClickHouse community, and provides extensive documentation, benchmarks, and community support through Discord and social media. It supports complex SQL vector queries, LLM observability, and offers quick start options via Docker and cloud services.

https://github.com/myscale/MyScaleDB

ai-applicationsanalytical-databaseannbig-dataclickhousecloud-servicecolumnar-storagedockerembeddingfiltered-searchfull-text-searchhigh-performancehybrid-searchimage-searchllmllm-observabilitymstg-algorithmmyscaledbopen-sourceproduction-readyragscalablesearch-enginesimilarity-searchsqlsql-vectorsql-vector-databaseunstructured-analyticsvector-searchvectordb