Swiftide

Swiftide is a Rust library for building LLM applications that index data, answer questions over it and run agents. It is designed around Rust's async and streaming features, with minimal abstractions and a modular structure that developers extend by implementing simple traits. The library has three main parts. Indexing pipelines load data, chunk it, generate metadata such as synthetic questions and answers, embed it with a provider like FastEmbed and store it in a vector database such as Qdrant. Query pipelines generate subquestions, embed the transformed query, retrieve matching documents and generate an answer with an LLM client. Agents are built with a builder that takes an LLM and a list of tools, and tools are defined with a macro, for example a code search function that runs a shell command against the working directory. Swiftide is meant for Rust developers who want retrieval augmented generation and agent workflows in a typed, performant language and a stable base to move experiments into production. The documentation site includes a getting started guide, reference material and links to GitHub and Discord. It is a developer library and the site publishes no pricing.