SmartTechLabs
banner
juergen-fey.bsky.social
SmartTechLabs
@juergen-fey.bsky.social
Consulting in the high tech realm. Industrial solutions from embedded edge devices to scaled processing systems. Big Data, machine learning and LLM. SmartCity/Smart vehicles. WebRTC. Kafka. Golang > Java > Python
https://smarttechlabs.de
If you into #Golang, this might be a treat for you. Anton provides interactive #Go release notes on his site. Run examples directly on the page to get the idea. Nice :-)

antonz.org/go-1-23/
Go 1.23: Interactive release notes
Iterators, timers and canonical values.
antonz.org
December 15, 2024 at 5:07 PM
Can PostgreSQL and LLMs work together? Absolutely!

Paul Ramsey's blog provides easy to follow examples of how to use #PostgreSQL and it's #plpsql to combine database content with #LLM use cases - directly "inside" the database. The examples can run locally using #Ollama.

Blog: bit.ly/4frqsMV
Smarter Postgres LLM with Retrieval Augmented Generation | Crunchy Data Blog
As a follow up to Paul's last post on an OpenAI connector to Postgres, Paul shows you how to add new data for your LLM queries to make them more accurate.
www.crunchydata.com
December 10, 2024 at 12:10 PM
A simple but powerfull approach to allow #LLMs to navigate through your web content: #llms.txt . It’s like #robots.txt, but for those mighty AI-enhanced crawlers! The idea: Provide the site structure and the relevant content in two distinct files.

More: towardsdatascience.com/llms-txt-414...
LLMs.txt Explained
Your guide to the web’s new LLM-ready content standard
towardsdatascience.com
December 6, 2024 at 9:09 AM
We are experiencing a slew of interesting #LLM based code editing tools. Since the waves are constantly coming, it makes sense to check the space frequently.

My current leading contenders:
Open Source: #oTToDev (Bolt.new fork): github.com/coleam00/bol...

Commercial: #Windsurf: codeium.com/windsurf
bolt.new
Prompt, run, edit & deploy web apps
Bolt.new
December 5, 2024 at 12:44 PM
Checking potential #LLM model candidates one might be overwhelmed by the amount of options. Besides the commercial API-services some open source models are serious contenders. Check out this #HuggingFace blog. He ran and niclely documented his model benchmark.

Highly recommended

bit.ly/4ie8rnE
🐺🐦‍⬛ LLM Comparison/Test: 25 SOTA LLMs (including QwQ) through 59 MMLU-Pro CS benchmark runs
A Blog post by Wolfram Ravenwolf on Hugging Face
huggingface.co
December 5, 2024 at 12:35 PM
We recently helped to define and implement the "Conferencing" component for a interesting #AR/ #VR/ #XR solution. Of cause we did not start from scratch. We always come back to the mighty #Livekit framework, which stands on the big shoulders of the #Pion project.

A big shoutout!
livekit.io
pion.ly
LiveKit
Instantly transport audio and video between LLMs and your users.
livekit.io
November 26, 2024 at 5:02 PM
Luring a LLM-based bot out of its #guardrails using an "optimized" prompt to let a robot drop a bomb in the end. To simplify the task, a team used a #LLM tool called #RoboPAIR which improves the #prompt till it works.

Overview: robopair.org
Detailled: lnkd.in/eC8-ZUX6
#PAIR: lnkd.in/eNBhKd3
Jailbreaking LLM-controlled robots
Robots can be jailbroken. Read on if you want to learn how.
robopair.org
November 26, 2024 at 4:54 PM