Woke up a 2015 Maxwell GPU (GTX 970M) to run embeddings inside k3s. CPU indexing took coffee-break long. GPU does it in seconds. TEI said no. Ollama said yes. The dragon breathes fire again 🐉 https://t.co/t4N9vRMWDj #homelab#k3s#gpu
An AI that can't read docs is just a confident liar 🧠
Self-hosted RAG on k3s: docs-mcp-server + local Ollama embeddings. My agents read real docs instead of hallucinating APIs 😎
https://t.co/Vt96QUT6FJ
#rag#selfhosted#homelab#mcp
Three AI agents, one shared API key = chaos 😅 Gave each its own LiteLLM key. Fought Cloudflare, https://t.co/Niav0EZbX0, and a ghost variable
https://t.co/vf2MWisVKM
#homelab#litellm#ai
One AI front door 🚪 Every agent, the wife's chat UI and my CLI route through a single LiteLLM gateway. Cost: 9 bugs + a weekend. Worth it
https://t.co/NqzexOwhfV
AI-assisted #homelab#k3s#LiteLLM#OpenWebUI
One docker compose led to k3s in my living room. homelab-2nd: GitOps from day one, SOPS+age, CNPG, MinIO, Cloudflare Tunnel — to host a chat app for 5 people. Overkill? Absolutely 😎
https://t.co/bVkSqZtkmZ
#homelab#k3s#gitops
Got the embedding model running on the homelab last Saturday — CLEVO box, k3s, text-embeddings-inference. Slow? Yes. Working? Also yes. #homelab#gitops#embeddings#selfhosted
Realia F1.
Gasly: „Przed i po GP Węgier zespół mówił, że zostaję do końca sezonu”.
Hulkenberg: „Dobrze wiadomo, że dobór kierowców nie opiera się tylko na osiągach. Chodzi o szereg czynników”.
Taka jest rzeczywistość. Nigdy nie dajcie sobie wmówić, że jest inaczej.
#ElevenF1