memory/episodic

Episodic memory implementation via few-shot prompting

Episodic memories are often implemented through few-shot example prompting, where agents learn from past sequences to perform tasks correctly. Few-shot learning lets developers program their LLM by updating the prompt with input-output examples to illustrate intended behavior. The challenge lies in selecting the most relevant examples based on user input. The memory store is one way to store data as few-shot examples, but you can also use a LangSmith Dataset and implement custom retrieval logic.