Lion Cipher Lab
Build a semantic search engine for your own documents
- AI Security
- Beginner
- ~50 min
- Free · no account required
Why it matters
Keyword search fails the moment someone asks for a thing using words that are not in the document — and that is most of the time. Semantic search fixes that by matching meaning instead of characters, and it is the same engine sitting under site search, support-ticket routing, note apps that 'find related', and the retrieval half of every RAG system. Building one over your own files, fully offline, turns 'AI that understands my documents' from a product you rent into a mechanism you understand: you can see why a result came back, tune how you split your text, and keep private material on your…
What you'll learn
- semantic search vs keyword search
- running a local embedding model with sentence-transformers
- chunking a folder of documents for retrieval
- building and querying a FAISS vector index
- recognising the limits of embedding search (bias, chunking, retrieval is not storage)
Steps in this lab
- The problem: keyword search does not understand you
- Set up: two libraries, both local
- Point it at your own folder and split into chunks
- Turn every chunk into a vector
- Build the vector index
- Search your documents by meaning
- The limits: what this will and will not do
- What you built, and what to do with it