Memory Types in LangChain

LangChain provides several memory patterns to make stateless LLMs maintain conversational context, including Transcript (full buffer), Window (sliding window), Summary, Entity, and Vector Retrieval patterns. Each pattern balances context accuracy against token cost, ranging from verbatim storage for short interactions to selective retrieval for knowledge-intensive systems.