How to manage context window and full history persistence in a LangGraph Agent with SqliteSaver.

[mo-ham-mad]

How can I configure memory for my LangGraph agent (built with create_react_agent and SqliteSaver) so that:

• The LLM prompt only includes the most recent N messages plus a summary of older turns, and
• The SQLite database still persists the full raw history for my frontend to display?

Does LangGraph provide any built-in memory or summarization utilities—similar to LangChain’s ConversationSummaryBufferMemory—to automate this, or what’s the recommended pattern to achieve it?