lancedb/lancedb is an embedded (running directly inside an application) multimodal vector search library written in Rust and released under the Apache-2.0 license. It stores and searches multimodal data such as text, images, and video together with vectors. The latest release is v0.40.0, and the repository has 11718 GitHub stars.
According to the README, the project is a multimodal data platform built on Lance, a columnar format that stores values column by column. It supports vector similarity search (comparing distances between vectors to find nearby items), full-text search (finding entries inside a document that match a given word), and SQL. Text, image, video, and point cloud data can be stored together with vectors, indexed, and then searched; it runs both locally and in the cloud; and it provides Python, Typescript, and Rust SDKs plus a REST API. Bundling vector search and SQL in one place in embedded form is the core composition of this repository.
That said, the verifiable scope is narrow. The given material contains nothing about what changed in v0.40.0, and no figures comparing search speed or performance by scale. The freshness, tech, and quality scores are empty, and the only filled entry is the momentum score of 4.3 out of 10. Taking into account that the repository was created on February 28, 2023, the direction of embedded multimodal search itself is clear, but how it behaves at real scale and what changed in v0.40.0 can only be known by reading the GitHub release notes directly.