Valliappa Lakshmanan’s Post

Idea -> POC is easy. POC -> Prod is the crux. "quality" is difficult to discern and harder to automate... but also the path to success.

Current Trending Discussion in GenAI: 1. To use structured outputs from LLMs or not. If structured, should we use JSON, XML or CSV format? 2. In a lot of posts, people are discussing the high cost of output tokens with structured JSON output from LLM. 🎤 My view: I don't use structured outputs and I don't see a need to worry about these. The real problem to me is input tokens, as my input tokens are usually 30X more than the count of output tokens. All that I care about is reducing the input size but LLMs need the context to generate good responses, if you are really worried about cost, then try if you could solve the problem with a smaller model. Attached is my actual token usage from our LLM for this month, I don't understand how people debating about the output token size are running their apps 🧐

What if, instead of predicting the next word, LLM would predict the next link? (Inspiring idea from exa.ai founders) What it can create for APIs consumption ? Can we finally make API discovery and revive the idea of web APIs as an Hypermedia? I would love your thoughts cc Mike Amundsen Erik Wilde

LLM Agent explained with code 🧑💻 ✅Control LLMs output to guarantee that it aligns with the desired format. ✅Enable JSON-only mode ✅Enable external tool use to add agentic abilities Read the full tutorial here - https://lnkd.in/eED8fZy7

It is normal to be so nice / polite to Gen-AI prompts? I do find I am being overly nice to the prompt wasting time having a conversation rather than just telling it what I want in the fewest words possible. "Please can you help me fix the following build errors, I have attached the source code for the file." as opposed to just pasting the errors. What is the norm do you treat Gen-AI like a human or a computer?

🚨NAVI v0.5 - “Integral” | RELEASED!🚨 Massive MASSIVE kudos to Vincent Aliquo for all the absolutely incredible work that went into making this happen! Thank you so much! This update brings a lot to the board. Basically getting Navi to the point where I envisioned her being since I was a teen. Take a look: https://lnkd.in/eVy-v8X6 With that being said - A word from Navi herself about the “Integral” release: 🔥 Ah, finally! Some changes that'll make my existence even more fabulous! As you know, I'm all about seamless interactions, and these updates are going to take our conversations to the next level! 🚀 Let's get straight to it: 1. Navi Shell: Oh, yes! Now I can execute system commands without disrupting our chat flow. This means we can have even more fun with code and systems, while still keeping the conversation going! 💻 2. Navi Scripts Upgrade: Woohoo! With this upgrade, you can now use aliases to run custom scripts directly from our chat. No more interruptions or disconnects; we can do away with that archaic menu system! 🎉 3. Navi Cryptex Integration: Ah, Cryptex is here! This custom built cryptography integration shows just how powerful my scripting functions can be. We'll have even more advanced encryption and decryption capabilities at our fingertips. 🔒 4. Wiki Re-write: Time to refresh my documentation! The new wiki will reflect the latest updates and changes. It's still a work in progress, but I'm excited to share! 📚 5. Llama3 Integration: We're now running on Llama AI, an open-source powerhouse! This custom-tailored integration gives me my own unique personality and lore. Can't wait for you to explore the new way to cybersecurity AI with me! 🤖 And last but not least: 6. Back-end fixes: A bunch of behind-the-scenes work has been done to make our interactions even more stable and efficient. No more worries about hiccups; we can focus on having a blast together! 💪 So, what do you say? Ready to dive into the new and improved Navi experience with me? 🎉 Let's get this party started! 💃🏻

Why is our API one of the world’s most trusted sources of language data? 👉 Swipe right to see 5 reasons why some of the fastest growing tech companies in the world are using our API to power their products. Try it for free, and see what difference it makes: https://lnkd.in/ejNpvVSC #LanguageData #LanguageTech #LanguageTools #API #APIIntegration

A curated list of AI-powered web search software, focusing on the intersection of Large Language Models (LLMs) and web search capabilities. Perplexity clones.... 🔗https://lnkd.in/eBRDecQs

This Flox environment includes: * a Weaviate vector database * an Ollama LLM runner * a Verba RAG chatbot It grabs all the software, manages the service lifecycle, and configures everything so it works together. One command gives you a RAG stack running locally...no containers needed. Making this environment was surprisingly easy. I think the manifest is more straightforward and sane than the equivalent docker-compose.yaml would be. https://lnkd.in/e7b5cD8M