Weaviate launches vector embeddings generation service - Blocks and Files

By Chris Mellor

Weaviate launches vector embeddings generation service - Blocks and Files

Vector database supplier Weaviate says its Weaviate Embeddings enable data items to be converted into vectors and stored for use by GenAI apps. It launches with Snowflake's Arctic-Embed open source text embedding model.

Weaviate provides an open source vector database for AI training and inference applications. It was founded in June 2019 in the Netherlands by CEO Bob van Luijt, CTO Etienne Dilocker, and CFO/COO Micha Verhagen, and is funded to the tune of $67.7 million by Battery Ventures, Cortical Ventures, Index Ventures, ING Ventures, NEA and Zetta Venture Partners with seed, A, and B rounds. Verhagen left in May 2022.

A vector database stores vector embeddings. These mathematical representations (dimensions with coordinates) of multiple aspects of data items have to be created from the base unstructured data, and stored in a so-called vector space, with similar vectors nearer to each other than dissimilar vectors. The vectors are created by embedding services.

CEO van Luijt stated: "Our goal is to equip developers with the tools and operational support to bring their models closer to their data. Weaviate Embeddings makes it simpler for developers to build and manage AI-native applications. For those who prefer a custom approach, our open source database supports any way they want to work. It's all about giving developers the freedom to choose what's best for them."

In Weaviate's view, existing embedding services can be a bottleneck for developers, with restrictive rate limits that slow down operations, and remote API calls that hinder performance. They can also use proprietary models to lock developers into their ecosystem.

Weaviate says its Embeddings product combines the flexibility of open source with the convenience and scalability of a managed service and pay-as-you-go pricing. The company says it "provides access to open source or proprietary models fully hosted in Weaviate Cloud, eliminating the need to connect to an external embedding provider or bear the burden of self-hosting. Users maintain full control of their embeddings and can easily switch between models."

Weaviate's Embeddings software runs on GPUs and brings machine learning models closer to the data to minimize latency. There are no rate limits or caps on embeddings per second in production environments. It says simple pricing reduces the cost of model inference.

The Weaviate Embeddings service is currently available in preview on the Weaviate Cloud. Weaviate plans to add new models and modalities to the service on an ongoing basis starting in early 2025.

Weaviate's database runs in its own cloud, AWS, Google Cloud, and Snowflake's Snowpark Container Services. There have been more than 13 million Weaviate software downloads and it has gained more than 10,500 GitHub stars. The company competes with Milvus, Pinecone, Qdrant, Redis, SingleStore, and other vector database suppliers.

Previous articleNext article

POPULAR CATEGORY

corporate

10658

tech

11464

entertainment

13093

research

5972

misc

13901

wellness

10592

athletics

13937