How to deploy AI Copilots with CopilotKit — High Signal AI #3

Nvidia’s developer contest, HuggingFace’s paper on FineWeb, fabric the open source framework that crowdsources prompts and more!

This is an archived (1 month old) post from my weekly newsletter called High Signal AI where I aggregate the best of AI Engineering resources to help you build faster.

To get these fresh out of oven to your inbox, every Friday, consider subscribing to the newsletter.

Here’s the link to the original post:

Build Custom In-App Copilots with CopilotKit - High Signal AI #3

The video version of this:

On my YT channel: https://youtu.be/dOyVnq5xY0g

This week’s action-prompting insights:

1. Top find of the week 🏆: CopilotKit
2. Developer tool of the week
3. Nvidia’s Developer Contest
4. GitHub repositories you must check out
5. What people are building with AI
6. AI startups that raised funds
7. AI tutorials and videos on YouTube worth your time
8. Reading list for the week
9. Weekly updates from tools, libraries and hubs

Top find of the week 🏆: CopilotKit

CopilotKit is a versatile framework designed for building custom AI copilots, including in-app AI chatbots, AI agents, and AI-enhanced text areas.

• It provides a variety of components that allow developers to integrate AI capabilities directly into applications, managing both frontend and backend interactions.
• Application you should try to develop: An example application that can be built using CopilotKit is an AI-powered customer support chatbot that integrates directly into a CRM platform, offering dynamic responses and actions based on the CRM data it accesses.
• This chatbot could enhance customer interaction by providing personalized support and automating routine tasks.

I gave it a spin, here’s how I created AI-powered Google Spreadsheet clone:

Dev tools you should bookmark

pip install marker-pdf

marker (10k stars pon GitHub) — is designed to convert PDF documents into markdown format quickly and with high accuracy.

• It supports a variety of document types, handles multiple languages, and can effectively format complex elements like tables, code blocks, and equations.
• It uses a combination of deep learning models and heuristics to optimize performance and accuracy, and offers a setup for both local use and a hosted API.

Generative AI Agents Developer Contest by NVIDIA and LangChain

NVIDIA announced a Generative AI Agents developer contest in partnership with LangChain. Key details:

• Objective: Encourage the development of generative AI agents that can enhance user interactions in various applications.
• Participation: Open to developers worldwide, providing an opportunity to showcase their skills in generative AI.
• Technology: Utilizes NVIDIA’s cutting-edge AI and computing resources, as well as LangChain’s language model technologies.
• Prizes: Significant rewards for top entries, aiming to incentivize groundbreaking solutions.
• Application Areas: The contest emphasizes applications across diverse domains such as healthcare, customer service, and more.

GitHub repositories you must check out

• fabric (14k stars) — an open-source framework that offers a modular approach to solving specific problems using a crowdsourced set of AI prompts called “Patterns,” which can be applied across various applications.
• The repository provides tools and a structure for integrating these AI solutions into everyday tasks effectively. It emphasizes a community-driven development where users contribute and refine these prompts to enhance functionality and applicability.
• ToonCrafter(1.6k stars) project leverages pre-trained image-to-video diffusion models to interpolate between two cartoon images, transforming them into a fluid, animated sequence.
• This technology is primarily research-oriented and serves as a tool for those interested in AI-driven video generation, particularly in the animation domain.

What people are building with AI

• Perplexica(7k stars) — A developer on GitHub created a clone of Perplexity, an AI-powered searching tool. It goes deep into the internet to find answers.
• Inspired by Perplexity AI, it’s an open-source option that not just searches the web but understands your questions. It uses advanced machine learning algorithms like similarity searching and embeddings to refine results and provides clear answers with sources cited.

Learning resources:

• Anthropic’s tool use tutorial notebooks — Across six lessons, you will learn everything you need to know to implement tool use successfully in your workflows with Claude. We recommend that you start from the beginning with the tool use overview, as each lesson builds on key concepts taught in previous ones.

AI startups that raised funds

• Nashville-based Ducky, provider of an AI-powered support platform, raised $2.7M in Pre-Seed funding.
• Nevada’s Play AI has raised $4.3 million in seed funding round to build the middleware for AI models and agents in gaming.
• Prague’s Product Fruits Raises $1.6M AI-driven Digital Adoption

AI tutorials and videos on YouTube worth your time

• Bonnie got a shoutout from LangChain for writing a neat guide showing how to use LangGraph.js and @CopilotKit to build a Next.js app that can make a presentation about an arbitrary topic.
• A cool application of LangGraph and a fun Sunday read!
• TechwithTim published a tutorial on Building a Coding Assistant w/ RAG & LangChain

Reading list for the week

• FineWeb: decanting the web for the finest text data at scale — a detailed paper written by Guilherme Penedo et al.
• FineWeb is a project focused on extracting high-quality text data from the web at scale.
• It’s a part of the Hugging Face Space initiative by HuggingFaceFW.
• The project aims to provide fine text data for training machine learning models.
• Andreessen Horowitz’s blog discusses their thesis on AI voice agents, focusing on both B2C and B2B applications. Key points include:
• Evolution of AI Voice Agents: Transition from traditional phone trees to more dynamic, AI-driven voice agents.
• Opportunities in B2B: Emphasis on vertical-specific agents with high customization to handle specialized tasks, driven by sector-specific demands.
• Consumer Potential: Exploration of AI agents replacing traditional roles like therapists or coaches, with potential for unique consumer engagement.