Building my 2x3090 home AI setup few months ago was the best decision ever. All used parts from Facebook market, very good bang for the buck. It's my vehicle to ramble the Hugging Face streets. Without it I would never develop RestAI github.com/apocas/restai :)

RestAI now features InstantID feel free to play with it at github.com/apocas/restai :) #restai #ai #avatar #GenAI

RestAI v4.1.0 now uses ollama for LLM execution. Awesome to finally delegate execution and just scale horizontally using N Ollama instances. Was waiting for keep_alive :) #ollama #genai #restai

Down the rabbit hole... #restai

If you are struggling on how to update your codebase to LlamaIndex 🦙 v0.10.x check this VERY good update example (it features rag chat and question, system prompts on both, multiple vector stores, etc): github.com/apocas/restai/… #llamaindex #restai

Finishing RestAI's stream based interface today. It's being a real pleasure to write Python after so many years. Less "bureaucratics" than Node. github.com/apocas/restai #restai #llamaindex #ollama

RestAI now features stream based interface on public and local models (rag and pure inference). Amazing how streaming tokens instead of waiting for total inference completely changes the notion of speed. #restai #llamaindex #ollama

I really suck at frontend, everything looks "auto-magic" to me, frameworks built on other frameworks like a matryoshka doll... I miss the good ol times of jquery, those projects that didn't scale and always ended in spaghetti 🤣

RestAI now features blazing fast Mixtral and Llama70b from the AI rocket ship Groq Inc for fast RAG. Groq removes the usual speed degradation of using LLM Reranking in RAG. Check it out at github.com/apocas/restai #llamaindex #groq

Main take away from this thread: People are insanely corrupted by money and will stab you in the back when possible (YES even those you trusted now/before) So, either get lucky your co-founder is a sane person with integrity OR make sure you do everything possible legally

BREAKING: Wright brother’s forced to make emergency landing just 12 seconds after takeoff.

US Innovates, China copies, and EU regulates.

Cool example on how to use RestAI and ollama to generate a description from a video (youtube or mp4). github.com/apocas/restai/… #restai #ollama

Anthropic's Claude LLMs are now available in RestAI github.com/apocas/restai #restai #llamaindex #anthropic

We’re excited to feature RestAI, a nifty platform that offers RAG, advanced text-to-SQL, multimodal inference as a service with a nifty UI 🤖📊 Fun highlights: ✅ Supports advanced features for data ingestion (chunk, splitter) and retrieval (cutoff, K, reranking) and evaluation

Cool video about RestAI from WorldofAI 📺 youtube.com/watch?v=aVYe4u… #restai

I’m loving the inference speed of llama3 70b way faster than llama2 in my 2x3090 home setup. #llama3

Good to see Google back.

If cars were invented today the European Union would have regulated them out of existence in favor of horses

Finally going to give dockerode some love this weekend. #dockerode #docker #nodejs