• Subscribe
  • Have you heard about Langchain 🦜🔗 ? A framework to connect GPT-4 with data sources and APIs

    Maxime Dolores
    1 reply
    👋 Hey makers! I wanted to get your opinion on frameworks like Langchain or Dust.tt. As you know GPT-4 is extraordinary but it cannot perform actions and lacks context for things that have happened after 2021. Langchain / Dust.tt allow you to fix this by connecting GPT-4 and other LLMs to other data sources and APIs They handle everything (prompt templates, parsing results, interacting with your data sources, splitting the context in multiple queries to not go over the limit, ...) Even Zapier announced a new Natural Language API to interface with those kind of systems last week (https://community.zapier.com/show-tell-5/nla-natural-language-actions-with-ai-api-by-zapier-22084) So with Langchain you could write a prompt like "What is the weather in Paris today" and Langchain would go on Google, find the weather, feed it to GPT and get you an accurate answer. This tutorial is probably the greatest 1 minute showcase of the capabilities it gives to GPT: https://youtu.be/kYRB-vJFy38?t=483 Langchain: https://github.com/hwchase17/langchain Dust.tt: https://dust.tt I haven't seen a lot of people talk about those frameworks so I thought I'd share them here! So what great products are you gonna build with this ? 😄

    Replies

    Nischal
    Frameworks like Langchain have definitely made shipping on top of LLMs faster. But the system is still evolving pretty fast. We at https://collectivai.com built an internal tool to query langchain & its docs in natural language & it has made our development process 10x faster. We even use it to generate code to integrate langchain with other tools. We thought to open up the tool to the world so that the entire community can benefit. We call it LangChainX - https://langchainx.web.app/