Blog

First research into RAG and embeddings, written while teaching a Rust CLI to work out which folder a file belongs in.

1. 01
   Adding RAG and ML to AI files reorganization CLI ↗

   Version 0.2: a Qdrant vector store and clustering replace plain LLM prompting for matching files to existing folders.

2. 02
   How cosine similarity helped my CLI decide where files belong ↗

   Why cosine distance is the metric that fits file and folder embeddings, and what the alternatives got wrong.

3. 03
   Making embeddings understand files and folders with simple sentences ↗

   Wrapping raw names in a short natural sentence measurably improved classification quality.

4. 04
   Embeddings clustering with agglomerative hierarchical clustering ↗

   Normalization, cosine distance and step-by-step merging — grouping leftover files without knowing the cluster count up front.