Drag and drop LLM pipeline development

🧰 AI tools to help you code

We start with three drag and drop LLM pipeline design tools

Harness the Power of AI Pipelines with AgentHub
AgentHub is an innovative platform for building and hosting AI-powered pipelines using open-source operators. AgentHub simplifies the process: just drag, drop, link operators, and run! The platform takes care of the infrastructure and hosting, letting us focus on creating powerful AI applications.

Flowise - Drag & Drop UI for Customized LLM Flow
Flowise is an open-source UI visual tool built on top of LangChainJS that allows you to build customized LLM flows using a drag & drop interface. It aims to make it easy for people to visualize and build LLM apps

Stack AI - No-Code Tools for Large Language Models
Stack AI is a no-code tool that allows you to design, test, and deploy AI workflows using models such as ChatGPT. It enables teams to create various applications, including chatbots, assistants, document processing, and content creation.

Some additional, equally impressive tools

Chat with any Github repo
This a a great workflow. If you want to work with a new library (which is in Github). Simply point Collectiv. to it (if someone hasn’t beat you to it), and you can the chat with the content of that repo. This gives the chatbot precise contact to git clear and relevant answers.

Locofy: A Design to Code Automation Tool
Locofy is a design to code automation tool that helps developers build web and mobile applications faster. It works by converting Figma or Adobe XD designs into production-ready code in React, React Native, HTML-CSS, Next.js, Gatsby, and more. Locofy also provides pre-built components and a community forum for developers to collaborate.

AI workshop; text, image & audio models, all in one place
LastMile AI, a platform that provides interactive workbooks for developers to experiment with and combine various LLMs, image, and audio models to build powerful applications and workflows in a single environment. Build AI-powered apps with text, image, and audio. Leverage models such as OpenAI’s ChatGPT, Google’s PaLM, Stable Diffusion and many more.

🏗️ Utilizing AI in what you build

Emerging Architectures for LLM Applications
LLMs are a powerful new tool for developers, but they can be difficult to use. This article from Andreessen Horowitz provides a reference architecture for building LLM applications. The architecture includes a number of key components, such as data preprocessing, embedding, prompt construction, and inference. The article also discusses some of the challenges of building LLM applications and how to overcome them.

Framework to easily create LLM powered bots over any dataset
Embedchain abstracts away the complexity of loading, chunking, embedding, and storing data, so you can focus on building your bot.

BentoML: Unified AI Application Framework
BentoML is an open-source platform for building AI applications with any pre-trained models, shipping to production in minutes, and scaling with confidence.

Self-Hosted, Community-Driven, Local OpenAI-Compatible APl
LocalAI is a self-hosted, community-driven, local OpenAI-compatible API. It is a drop-in replacement for OpenAI running LLMs on consumer-grade hardware. It is a free open-source OpenAI alternative that does not require a GPU. I found it interesting that it supports multiple model families that are compatible with the ggml format.

📰 AI news for devs

Google DeepMind's New AI Algorithm Could Be More Powerful Than ChatGPT
In a Wired interview, Google DeepMind CEO Demis Hassabis revealed they're developing a new AI algorithm, codenamed Gemini, that may outperform OpenAI's ChatGPT. Using techniques from AlphaGo, the AI that beat a Go champion, Hassabis believes Gemini will generate more innovative and insightful text.

Databricks Acquires MosaicML
Databricks, a leading provider of data analytics and machine learning platforms, has acquired MosaicML, an open source startup that specializes in generative AI. The deal is worth $1.3 billion, and it will give Databricks access to MosaicML's platform for training and deploying large language models.

Finetuning large language models for GDScript generation
Godot-dodo is a project that finetunes large language models for GDScript generation. GDScript is the scripting language used by the Godot game engine. The project's goal is to create a tool that can be used to generate GDScript code automatically.

LLMs, meet my data
LlamaIndex is a framework that allows developers to connect data from files like PDFs, PowerPoints, apps such as Notion and Slack, and databases like Postgres and MongoDB to LLMs. It includes connectors to ingest data sources and data formats, as well as ways to structure data so that it can be easily used with LLMs ¹. I found it interesting that investors pledged $8.5 million toward LlamaIndex in a recently closed seed funding round.

🏫 Read/Listen/Watch: Learn

This week’s learning theme is all about embeddings

Getting Started with Embeddings [read 10min]
This article/workshop from Hugging Face provides an introduction to embeddings, numerical representations of information that capture semantic meaning. It shows how to use the Hugging Face Inference API to easily embed a dataset and demonstrates how to create a small Frequently Asked Questions (FAQs) engine using embeddings

A Complete Overview of Word Embeddings [watch 17min]
In this video by AssemblyAI, Learn about different techniques like one-hot encoding, count-based approaches, and Word Embeddings. Explore how words can be converted into numbers or vectors to feed into your models, the creation process, and their use in your projects.

The Pinecone learning center [read, multiple, multi-chapter workbooks]
The fine folks at Pinecone have an extensive learning center which offers guides to help you conquer topics such as vector embeddings and building better applications. It features series on natural language processing for semantic search, Faiss, vector search in the wild, embedding methods for image search, and the LangChain AI Handbook

😂 Fun

Make Music! - Audio Processing and Generation with Deep Learning
Kick the tires on MusicGen, Meta’s state-of-the-art controllable text-to-music model. No need to get this running locally, head over to a hosted instance on Huggingface

Quirky projects fueled by AI magic
lumine AI - AI-powered playground, where artsy vibes, games, and other interactive experiences cross the singularity frontier! Still very much in the beta phase but some very interesting experiments going on here.