LangChain LLM Framework

LangChain is an open-source framework for building applications powered by large language models, providing composable tools for prompt management, chains, agents, memory, and integration with external data sources and tools. Created by Harrison Chase in 2022, the project has over 100,000 stars as of 2026 and is the most popular LLM application framework. Key features include: Core abstractions (base interfaces for prompts, LLM interfaces, chat models, output parsers, and runnable components forming a unified interface across providers), Chains (sequential composition of prompts, model calls, and parsers into reusable pipelines with automatic input and output type inference), Agents (autonomous agents that use LLMs to decide which tools to call, supporting OpenAI Functions, ReAct, and custom agent architectures), Memory (conversation buffer, summary, and entity memory for maintaining context across interactions), retrieval-augmented generation (document loaders for 80+ sources, text splitters, embedding models, and vector store integrations including Pinecone, Chroma, Weaviate, Qdrant, Milvus, and pgvector), LangSmith (observability and evaluation platform for tracing, debugging, and testing LLM applications), LangServe (deployment of runnables as REST APIs), LangGraph (stateful, multi-actor LLM applications with cyclic graphs for complex agent workflows), Expression Language (declarative composition syntax for chains and pipelines), and model providers (OpenAI, Anthropic, Google, Cohere, local models via Ollama, vLLM, AWS Bedrock, and Azure).