Michael Brazda

New AI Project I'm working on

Building a multi-agent RAG system spread across two servers.
Three collaborative agents (two on one machine, one on the other) powered by Llama-3.1-8B + Qwen2.5-7B. Using Qdrant vector DB, BGE-M3 embeddings, semantic chunking, rich metadata extraction, and a critic step for higher accuracy.

Goal: a reliable private knowledge base from all my internal docs.
Ingestion is currently running and will probably take 3 days. The waiting part is painful 😅

Suggestions?

Getting a little rain in Houston. The worse is still coming..

RE: @OpenComputeDesign@linuxrocks.online

I wonder if there is enough left to rebuild.and say running. In the days before repos I deleted /etc and had to extraction everything from rpms and rebuild configs by hand but the server stayed running and rebooted cleanly.

LR

Sorry for the bandwidth issues we are currently working on it. Please be patient.

Wifes pulled me out from behind my computer for another pool day.

The girls play while I am stuck in my office working. WTH! LOL

Musk vs Altman trial just got even spicier. Sam Altman testified today: Elon once mused about handing OpenAI to his kids if he died, demanded a 90% stake, and pushed a Tesla merger. Altman: “It feels difficult to even wrap my head around that framing… We created one of the largest charities in the world.” AI’s future on the line. 🔥 #OpenAI #MuskAltman

I am building what amounts to a AI cluster. 4 servers with 48+ cores 512GB+ of ram and 2x Nvidia A10s per server.

Any suggestions on setup? Anything from models to the software itself? Going to setup a collaborative AI so each will work with each other to determine the best answer. I just got the hardware ready just ready for install to test and play with. Any suggestions Welcome.