LangChain — фреймворк для разработки LLM-приложений

★ 92k

LangChain is an open-source framework in Python and JavaScript for building applications and agents on top of large language models. Reach for it when you are writing your own AI product in code and want ready building blocks: a unified interface to different LLMs without provider lock-in, tool calling, memory and prompt handling, connections to vector databases, and building RAG pipelines and agents. It is essentially a standard glue library so you don't reinvent integrations with models and stores by hand, with a large ecosystem around it (LangGraph for complex agent graphs, LangSmith for observability). Its focus is a builder library for LLM apps aimed at developers, not a visual no-code tool (for a canvas, Langflow/Flowise) or a finished app: it gives components and abstractions from which you assemble your logic in code. For agents with complex state and branching, people usually use LangGraph on top of it.