LangChain: the framework for LLM applications is born
In one sentence Harrison Chase releases LangChain, an open-source Python library to chain LLMs with prompt templates, memory, tools and external data sources. It will become the default stack of the first LLM apps.
When you work with a language model like GPT-3, you soon realize "just ask it a question" isn't enough. You need it to talk to a database, to documents, to other tools. You need it to remember what was said. You need to orchestrate steps.
Harrison Chase, a developer, publishes a Python library that puts order into all this. It's called LangChain. You define a "chain": take a document, ask the model to summarize it, pass the result into another call, and so on.
In a few weeks LangChain becomes the standard way thousands of developers build their first AI applications. Later the criticism will come (too "magical", too many abstractions), but the framework defines a vocabulary: chain, agent, tool, memory.
Companies
LangChain
Tools
LangChain
Tags
Sources