Claude can now process PDF, XLSX, DOCX, PPT with the new Sandbox feature.

But what's interesting is that it uses skills under the hood. And there is a way you can extend/add new skills: First of all, what are skills? As answered by claude:

🚀Launching BrainyRead – Convert YouTube videos into Notion-like pages using AI!

Built using:
@nextjs- Web Framework
@LangChainAI - LLM Framework
@supabase - Auth & DB
@GroqInc - Transcription
@cursor_ai + Claude 3.5- IDE
@v0 - UI
Novel (@steventey) + @tiptap_editor - Editor How it works:

The video is transcribed using @GroqInc Whisper and then passed to a vision model that analyzes the content and generates the article in the backend.

The generated articles are displayed in a Notion-like interface using Novel and @tiptap_editor.

🚀 Introducing RAGTune: An Open-Source tool for tuning and optimizing RAG pipelines!
Curious about the best combo of LLMS, embeddings, retrievers, etc. for your RAG app? Now you can easily experiment and see what works best.

1/ Start with uploading documents and providing some Q&As These Q&As act as ground truth for evaluations.

2/ Compare LLMs: Select from a wide range of LLMs available in LangChain and see their performance for your data.

For no-code builders looking to make AI apps using @LangChainAI together with hundreds of integrations, @n8n_io's latest update includes LangChain components built-in. Let's look at some cool AI automation you can do:

1/ Automated newsletter generation from a podcast episode 2/ Our famous Chat-with-document app: Read files from Google Drive, embed using @cohere's embedding model, save them to @pinecone, and then use a Retrieval QA chain to retrieve and answer questions using LLM from @cohere .

A few ways you can benefit from the @LangChainAI Hub. 🧵↓

1 - Benefit from the top prompts shared by the community. Filter based on likes, views, downloads, and tags. These prompts can be used anywhere, from a simple translation app to prompts powering RAG and agent apps. 2 - Test the prompt response right in the browser before deploying it as part of your app. You can edit the prompt response, check for susceptibility to prompt injection, or add context to check if the response is as you expect.

🔥AI Agents are currently a hot topic in the AI community. With @LangChainAI, you can work with different types of agents out of the box.

1) ReAct Agent: This agent utilizes the Reason-Action framework to determine the appropriate tool based on its description. 2) Conversational Agent - As the name suggests, these agents are specialized for conversations, equipped with added memory, and utilize the ReAct framework.

Prompt Templates can be quite powerful if used the right way. Let's explore three types of prompt templates you can use in @LangChainAI.

1) Basic prompt template - Uses a single input from the user and combines it with the template string to send the combined prompt to the LLM. Basic prompt template can instruct LLM to:
- Respond in a certain format eg. poem, essay, blog, social media post
- Respond as a certain person eg. Shakespeare, Aristotle
- Act as an expert in an area eg. Physicist, Business Analyst, Coach

"Hey LangChain!"

These 4 blocks in the Shortcuts app let you connect Siri with a @LangChainAI app API.

Learn how to use Siri for chatting with documents in 100+ languages using @CohereAI 's multilingual model & @FlowiseAI /LangChain in this walkthrough!

Video link below.🧵 ↴ We start with building a Chat with PDF app in @FlowiseAI, by using @CohereAI embeddings, to embed in 100+ languages. The upsert flow is as below:

With these few lines of code, you can build a chat app with Falcon LLM using @LangChainAI and @chainlit_io

Video implementation and repo link below ↓ Chat UI is built using @chainlit_io, which makes it easy to run LangChain apps. Thanks, @derekcheungsa for the inspiration.

📢 Calling all aspiring AI builders!🚀

I'm thrilled to announce the pre-launch of my course, "Build AI Apps with No-Code" using @LangChainAI. Join me on this exciting journey to learn how to create incredible LangChain apps without any coding experience. Join using link below↓ The goal is to help you create ready-to-launch consumer apps, with step-by-step instructions, using @bubble and LangChain. With the power of @FlowiseAI and @logspace_ai's LangFlow, you'll build ChatData apps, Agent apps, and so much more in a no-code manner.

How would Retrieval change with increased LLM token limits? Do we need vectorDBs? That's what triggered this experiment. 🧵↴
For a PDF-search app, what if we send the entire text content to the LLM and ask questions based on that. Our usual retrieval chain would modify as such: I wanted to test with the @MosaicML MPT-7B-Storywriter model, with a token limit of ~65k. This is already deployed on @replicatehq so it was easy to call using @LangChainAI with the following code:

Want to create @LangChainAI apps without coding? No problem!
Join us in our latest video tutorial where we'll guide you through building a chat-with-PDF app. Upload docs, filter searches, and embed chatbots effortlessly. Built using @FlowiseAI @bubble and @render We will build around 3 scenarios - Scenario 1: building a chat widget to embed on any website.

Did you know @CohereAI open-sourced 94 million vectors of Wikipedia embeddings from 10 languages? And @weaviate_io hosted 10 million of them to search for free?
With a few lines of code using @LangChainAI you can retrieve them. But we face one issue, let's solve 🧵↓ Building a typical document search system begins with first extracting the text (Wikipedia in our case), splitting it into chunks, embedding them, and then saving it to a vector database.

Let's build a @LangChainAI agent in under 30 minutes that can access the internet, chat with a PDF, and help us understand code. All without a single line of code using @FlowiseAI and @bubble.

This agent will have access to 3 tools and a memory as shown. Tools explained in 🧵👇 We start with a conversational agent and connect it to the serpAPI tool available in LangChain to perform a Google search. This returns a list of search results, that is then passed to the LLM to generate an answer based on the results and the user query.

With these few lines of code, you can deploy a backend for multilingual semantic search over a PDF using @CohereAI and @qdrant_engine, with the ability to embed and perform a search in 100+ languages. Certainly, you can see the power of @LangChainAI here. It basically divides a document retrieval system into two API functions corresponding to document ingestion and retrieval/search query.