spent sunday building a local rag pipeline with llama3

finally got inference running under 2 seconds on my m2 mac. the vector store is just sqlite vec and it handles my entire notes folder without lag. shipping the github repo tonight so others can fork it.