Fenja AI
++ The sovereign AI platform. Open-source models — Llama 3, Mistral — deployed entirely + inside your infrastructure. No data leaves your network. No external API calls. + No foreign cloud dependencies. +
++ You interact with it through a clean interface: ask a question, get an answer, + see exactly which documents informed that answer and with what confidence. + The audit trail is not an afterthought — it is the design. +
++ What Bifrost is testing is the core of this: document ingestion, contextual retrieval, + traceable responses. What comes after — agentic workflows, self-service automation, + developer APIs — builds on the same foundation. +
+Fenja Knowledge
++ An evolving knowledge layer that learns the context you work in — your organisation, + your regulatory obligations, your domain-specific language — and makes it available + to the model without requiring you to re-explain it in every query. +
++ Think of it as a structured, permission-controlled wiki that the AI reads before + answering. Not a prompt. Not a system message. A maintained, versioned context + that grows with your work. +
++ The concept was previewed at the pilot kickoff. The response from participants + was clear: not a question of whether, but when. It is on the roadmap. +
+When the MVP ships
++ This page will become a gateway: screenshots, a walkthrough, and a path + for pilot participants to request early access to the interface. + Until then, this is the honest picture of where things stand. +
+