One of the main key choice points when doing GenAI app development is choosing your foundation model. Should you go “closed-source” frontier model, like GPT-4 or Claude 3.5, or try to stick with an open source model like Llama3? Or does your use case fit a more purpose-trained model like Gorrilla or Stability? This mostly comes down to testing, and understanding the economics of your app (assuming it’s a commercial effort) to decide which model to build upon. But waiting behind this decision is another key choice - choosing what level of *abstraction* on which to build. The simple approach is to use the native API provided by the model itself. This can be a good choice, but strongly limits your ability to support other models in your app. Prompt instruction syntax, tool calling syntax, and JSON or markdown generation can vary pretty wildly across different models. So introducing some abstraction to represent the LLM offers the possibility of supporting or swapping different models depending on your needs, without requiring heavy surgery to your app. Unsurprisingly there are lots of options vying to abstract your LLM interface, including LangChain, LLamaIndex, vllm and many more. Testing and picking amongst these is a project in and of itself. Generally you evaluate each solution across a few different axes: Does it support commercial or open source models, or both? Does it look to support “advanced” features of different models, or rather attempt a “lowest denominator” approach to preserve compatibility? What other features (agent behavior, RAG, indexing…) is it trying to offer? How much will you be bound to those implementations if you choose this abstraction? I haven’t seen it yet, but this would be a great opportunity for a “bake-off” type evaluation to produce a matrix to help you understand the different options and trade-offs. Maybe I’ll go ask ChatGPT to write it up… #ai #genai #llm
Scott Persinger’s Post
More Relevant Posts
-
𝐀𝐬 𝐝𝐞𝐯𝐞𝐥𝐨𝐩𝐞𝐫𝐬, 𝐰𝐞’𝐫𝐞 𝐚𝐥𝐰𝐚𝐲𝐬 𝐚𝐬𝐤𝐢𝐧𝐠—𝐜𝐚𝐧 𝐭𝐡𝐢𝐬 𝐛𝐞 𝐟𝐚𝐬𝐭𝐞𝐫, 𝐛𝐞𝐭𝐭𝐞𝐫, 𝐨𝐫 𝐦𝐨𝐫𝐞 𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭? Optimizing code is one of the most satisfying parts of development. Improving things like load times, database queries, or cleaning up messy code can be thrilling. Finding a better solution never gets old. I remember when a small change to our caching strategy cut API response times in half. At first, it seemed like a small thing, but it had a big impact on the user experience. These moments show why attention to detail and curiosity are so important in this field. Optimization isn't just about speed. It's about creating systems that are strong, can handle growth, and are enjoyable to use. It's an ongoing process, but it's always worth doing. What's the most satisfying optimization you've ever made? #CodeOptimization #FullStackDeveloper #TechLeadership #StartupJourney #NeverStopImproving #ViralContent #MobileApp #AI #ChatGPT #CodeWebster
To view or add a comment, sign in
-
Have you ever wondered how to use Large Language Models (#LLMs) like #ChatGPT in your app? Yes, there is the #OpenAI #API. You can use it to generate responses based on prompts. But what if you want to build a complex application with LLMs integrated tightly into the user experience? What if you want to use multiple services like OpenAI, Mistral, and others? The answer is #Langchain – a powerful tool designed to streamline and enhance AI workflows. #LangChain is an open source framework for building applications based on large language models (LLMs). LLMs are large deep-learning models pre-trained on large amounts of data that can generate responses to user queries—for example, answering questions or creating images from text-based prompts. #Chatbots are one of the most popular use-cases for LLMs. The core features of chatbots are that they can have long-running, stateful conversations and can answer user questions using relevant information. Optimizations like this can make your chatbot more powerful, but add latency and complexity. The aim of this guide is to give you an overview of how to implement various features and help you tailor your chatbot to your particular use-case. Could be use chatbot approach to increase #dpp into the #fashion ecosystem ? https://lnkd.in/ghJi8_iq #langchain #scanner #ui #bot #microlearning #landing #web #machinelearning #question #utah #deeplearning #machine #chatgpt #questionui #chatbot #langchain #openai
To view or add a comment, sign in
-
🚀 Replit’s New AI Agent: The Future of App Development? 🤖💻 🎉 Picture this: You're sipping your morning coffee ☕, brainstorming your next big app idea. Before you know it, your AI agent buddy is there, not just suggesting code snippets but building the entire app for you – from setup to deployment – all while you sit back and watch the magic happen. Well! Replit has just launched an AI agent that automates end-to-end application development from natural language prompts. 📝This feels close to a tool that made big buzz some time ago called Devin AI created by Cognition labs. 🛠️, which also streamlined development, but Replit takes it further with real-time collaboration and self-debugging. What do you think ? What's the future of app development with tools like this? Let’s chat! 😉 #AI #ChatGPT #Jobsearch #innovation #datascience
To view or add a comment, sign in
-
I've been working on a versatile app that combines the capabilities of LLaMA 2 and LangChain to deliver a range of practical AI features: 🤖 𝗔𝗜 𝗔𝘀𝘀𝗶𝘀𝘁𝗮𝗻𝘁: A conversational bot ready to help with any queries. 📄 𝗣𝗗𝗙 𝗦𝘂𝗺𝗺𝗮𝗿𝗶𝘇𝗮𝘁𝗶𝗼𝗻: Quickly extract key insights from lengthy documents. ❓ 𝗤𝗔 𝗳𝗿𝗼𝗺 𝗣𝗗𝗙 & 𝗪𝗲𝗯 𝗟𝗶𝗻𝗸: Get precise answers from both PDFs and web content. 💻 𝗖𝗼𝗱𝗲 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗼𝗿: Simplify coding tasks with automated code generation. 🌐 𝗟𝗮𝗻𝗴𝘂𝗮𝗴𝗲 𝗧𝗿𝗮𝗻𝘀𝗹𝗮𝘁𝗼𝗿: Seamlessly translate content across multiple languages. This app brings together multiple AI features into a single, user-friendly platform, and I'd really appreciate your feedback on it. https://lnkd.in/dkjxaUuT #AI #LangChain #LLaMA2 #MachineLearning #Innovation #Tech
To view or add a comment, sign in
-
Dear AI apps: Search & Folders We need an in-app way to find and organize everything we make or are working on. App-Specific Examples: Suno has the ability to like songs and create playlists, which is nice, but I’d also like to: - Group somgs in generating as I work on refining an idea - Save specific genres, descriptions, voices (I have a lot of UX design ideas around this. Message me if you’re interested.) OpenAI ChatGPT & Anthropic Claude keep previous chats until you delete them, and that’s great! But, Odd also like to: - Group chats about like items into folders - Have a way for all the items in the folder to share all of the knowledge that each separate conversation has in it (message me and I’m happy to share UX ideas around this) Pika I know I’m the master of the obvious with these, but they need to be stated: - Generations need a “Continue from”. More may be in the comments as I think of them…
To view or add a comment, sign in
-
The rise of Generative AI has revolutionized how we think about building smart applications. Since the introduction of tools like ChatGPT, developers and innovators are exploring ways to harness the true potential of LLMs for various tasks — from coding assistants to creating personalized user experiences. But what’s next? How can we build even more powerful and scalable apps with LLMs? In my latest blog, I dive into LangChain — a game-changing framework that simplifies working with LLMs, making it easier to develop robust, flexible, and intelligent applications. Whether you're a developer, student, or entrepreneur looking to explore the possibilities of AI-driven applications, this guide offers practical insights to get started. https://lnkd.in/g_BZ9rJM #AI #LLM #LangChain #GenerativeAI #MachineLearning #TechInnovation #AIApplications #Developers #FutureOfWork
To view or add a comment, sign in
-
Bright Eye: free mobile AI app to generate art, text, analyzes photos, and more! (GPT-4 powered) Hi, all! I’m the cofounder of a multipurpose, all-in-one AI app to generate text, images, code, stories, poems, and to analyze image and text, and much more. Sort of like the Swiss Army knife of AI. It can generate poems, short stories, code, essays, math, and more via GPT4! In addition, it can generate art via stable diffusion v2. On a smaller scale, we have analytical tools that provides text extraction, and a small social environment. We’re looking for feedback on the functionality, design, and user experience of the app. Check it out below and give me your thoughts: https://lnkd.in/dP7EDMWn #chatgpt #ai #aiart #midjourney #openai #digitalart #chatbot #nft #aigenerated #dalle
To view or add a comment, sign in
-
LLMs are not a "killer app" says Joon Sung Park in this @Forbes article. He has a point, but is it the right point? He continues with the example of Excel as a killer app because it transformed how we work with tabular data. He argues that LLMs fail to reach that point because it doesn't fully embrace the power of LLMs. "Its blank text box places the burden on users to articulate their intent, provide all relevant context, and evaluate the model’s responses..." I beg to differ. It puts the focus on the wrong thing. LLMs represent a fundamental shift in how we interact with technology, not just from the perspective of the UI but how we even think of apps. Traditionally, if something doesn't work, it's the app's fault, but with LLMs...it's probably our fault. If you are talking with another person and they don't quite understand or give you a wrong response, you often think "okay, how do I explain better." Joon sees this as a deficiency, but I see it as a movement towards humanity. You have to learn to use it and understand how to get the best out of it as you do any other person. How you interact with it matters. The more context you supply the better the response, same with interacting with each other. Joon discusses how LLMs are better suited to soft problems, those that don't have a clear answer, and that are more subjective. I agree with that and that makes it more human. Ask it to do better and it does. Maybe it should just do better the first time? I was recently proudly called a "futurist" for saying "Don't focus on what it can do today, but what it will be able to a year from now, two years from now". The pace of technological innovation is accelerating. AGI (look it up 😉 ) is not that far away. oh...and here's that article - https://lnkd.in/eXHgSHfQ #genai #future #chatgpt #bard #llm #llms
How to Build Generative AI Agents with Joon Sung Park
forbes.com
To view or add a comment, sign in
-
Builder.ai's most popular blog is out and it's a BEAST 🦁! But there's a reason for that, we're providing more FACT than FICTION in the fascinating world of AI. As AI transforms traditional app development, from entrepreneurs needing a prototype to enterprises tired of doing things the old way, Builder.ai is here to help. Join the AI hype cycle and learn how AI is changing the game. Check out our blog now! @builder.ai @disruption @composablesoftware #AI #appdevelopment #innovation Read more here: https://lnkd.in/daiJnwcS
How to Create an App With AI in 8 (Surprisingly Simple) Steps
builder.ai
To view or add a comment, sign in
-
🔮 Users in 2024 expect their favorite apps to include AI-powered features. Have you joined this trend yet? 📲 If not, here's a detailed guide on where to start: https://lnkd.in/dii6NNhE #ai #techtrends2024 #aidevelopment #buildaiapp #digitaltransformation
Ultimate guide to building an AI app step-by-step
syndicode.com
To view or add a comment, sign in
CEO at Supercog AI, prev at Tatari, Stripe, Heroku, founder at CloudConnect
5moHeh, ok GPT-4o did a passable job, with a very simple prompt: https://chatgpt.com/share/a8582be8-4ace-4d9f-b3ab-f4dbff8f7c09 Maybe you can come up with someone better?