agentset

Why we included this project

For teams building retrieval-augmented generation from scratch, this is a practical way to avoid stitching together a dozen separate services. The full pipeline lives in one place: ingestion across many document formats, chunking, embedding, indexing into the vector store you already use, and retrieval exposed through an API and typed SDKs. It stays model-agnostic, so you bring your own LLM, embedding model, and vector database rather than committing to one vendor's stack. The chat playground lets you edit messages and see the citations, which makes it easy to confirm answers actually trace back to the source documents before anything ships. Multi-tenancy is built in and the platform can be self-hosted, which suits teams that want to own their retrieval infrastructure.