How to Build an Enterprise AI Knowledge Base Directly From Office Documents
From traditional RAG to a full agent — a real enterprise knowledge experiment we ran with SourceLens, an open-source project that skips the vector database entirely.
Blog
Writing from the HyperFileLens and SourceLens team, plus articles about the projects published on other platforms.
From traditional RAG to a full agent — a real enterprise knowledge experiment we ran with SourceLens, an open-source project that skips the vector database entirely.
Kopia's dedup and encryption are solid, but its CLI runs on memorized flags and KopiaUI isn't much better — once something's backed up, it disappears into a black box. HyperFileLens builds its Agent directly on Kopia, collapses setup into a few clicks, and wires in an AI engine (SourceLens) that reads snapshots directly. A full hands-on walkthrough, from deployment to asking AI questions about your own backups.
From recovering data to rediscovering its value — the open-source HyperFileLens skips Embedding, Vector DB, and pre-indexing entirely, letting an agent search, read, and reason over raw files directly. It's already running in production across HyperBDR and AGIOne's global sales support, product support, and engineering analysis.