LightMem

Why we included this project

Agents that keep forgetting earlier turns in a conversation are a common pain point, and LightMem addresses it directly. It gives LLM applications a small memory layer that stores, retrieves, and updates context, so a chatbot or agent can recall prior exchanges without rereading every message. The API is simple enough to integrate in a few lines, and the modular design lets you plug in different storage backends and retrieval strategies. You can run it against hosted models like OpenAI or DeepSeek, or local runtimes such as Ollama and vLLM, which makes prototyping in different setups straightforward. Tutorial notebooks for a travel-planning agent and a code assistant show how to wire everything together, and the benchmark scripts let you compare it against other memory approaches on LoCoMo and LongMemEval before you commit.