Urgent.News

What's breaking now, across thousands of outlets.

More in AI

Why my 0.8 cosine threshold failed in an offline AI memory

I'm a self-taught robotics engineering student in Bengaluru. I built Neo, a fully offline AI assistant that remembers facts about you across sessions. It runs llama3.1 through Ollama, stores vectors in LanceDB and facts in SQLite, and does web search through a self-hosted SearXNG. No cloud, no API keys.

  • Student created offline AI assistant Neo using llama3.1 and LanceDB
  • 0.8 cosine similarity threshold ineffective for paraphrases and updates
  • Corrected metric improved Recall@k to 0.95 and MRR to 0.925

More from Friday 6 November →