Architecture, security, deployment and cost, written for the reviewers who will ask the hard questions. Everything here is free to read and share. No registration.
Why a read-only, permission-aware index over existing storage beats copying data into yet another platform, what such a layer has to do, and how K-Lake does it.
How per-file trimming, identity federation and an audit trail let you connect assistants to sensitive data without widening anyone's access, and why application-layer filtering is not enough.
Running retrieval and answers entirely inside your own boundary: self-hosted extraction, offline licensing, fully air-gapped operation, and what to ask any vendor who says "private".
What we measured when an assistant answered the same questions from the same filings through K-Lake, careful and naive command-line searching, and a standard vector-chunk store. Method, numbers and limits, including the "up to 47x" claim.
The product documentation is public, versioned to each release, and the canonical place for detail.
Trust model, isolation, identity, encryption, audit and supply chain, written for security, risk and compliance teams.
How source-native permissions are captured, correlated to directory identities and enforced on every query.
Enabling the server, OAuth 2.1 and identity providers, connecting Claude and other clients, the console view and audit.
Stand up the full stack on a single VM with a demo tenant, and what to size it at.
Helm-based deployment into your cluster, bringing your own database, sizing the fleet and autoscaling.
A rolling report of vulnerability scan results across every shipped container image.
Architecture reviews, security questionnaires and sizing for your estate. We would rather answer them directly than have you guess.
Get in touch