Create semantic vectors from new content records for search

Your content records lack semantic tags, so finding precedent snippets and templates is slow. Adding embeddings makes each entry searchable so you find relevant documents within minutes.