Savflux AI

Question

Can a visitor to Savflux ask natural-language questions about the site’s content and receive accurate, cited answers?

Examples:

• “What has Freedom built with Cloudflare Workers?”
• “Show me his AI projects.”
• “Explain the PrimeShieldHub architecture.”
• “What is he currently working on?”

Hypothesis

A retrieval-augmented generation (RAG) system over Savflux’s own content could answer these — with citations back to the source pages, and without hallucinating.

Approach

• Index all projects, experiments, and articles at build time
• Embed the content into a vector store (Cloudflare Vectorize, once available)
• On query, retrieve the top-k relevant chunks
• Feed them into a language model with strict instructions to cite sources
• Return an answer with links to the source pages

Current State

Research. Evaluating Cloudflare’s AI stack — Workers AI, Vectorize, and the available embedding models.

Why This Is Interesting

An AI assistant that cites its sources is trustworthy. An AI assistant that invents facts is a liability. The technical problem is not generation — it’s retrieval and grounding.

If this works, it becomes a template for any content-driven site that wants to offer conversational access to its knowledge.