Persist agent knowledge beyond the context window — vector stores, key-value caches, and structured stores for cross-session recall.
Five passes over the same idea, each from a different angle. Do them in order, or jump to whichever you need.
Long-term memory gives agents continuity. Facts learned in session 1 should be retrievable in session 100. Implementation options range from simple key-value stores (user preferences) through vector databases (semantic recall) to structured stores (task history). The engineering challenge is deciding what to remember, how to retrieve it relevantly, and when to forget.
