Memory tools: manage and search

create_manage_memory_tool and create_search_memory_tool give an agent memory as tools: it decides during the conversation what to save and when to look.

save = create_manage_memory_tool(namespace=("memories", "{user_id}"), store=store)
print(save.name, "|", save.description.splitlines()[0])
print(list(save.args))

asha = {"configurable": {"user_id": "asha"}}
print(save.invoke({"content": "Order A-1001 arrived broken", "action": "create"}, config=asha)[:15])

A tool is what a model sees: a name, a description and arguments. manage_memory takes content, an action of create, update or delete, and an id for the last two. Invoking it directly, as here, does what an agent's call would do, and returns the text the agent reads back, with the new memory's id.

find = create_search_memory_tool(namespace=("memories", "{user_id}"), store=store)
results = json.loads(find.invoke({"query": "broken order"}, config=asha))
for row in results:
    print(row["value"], round(row["score"], 3))

search_memory returns JSON the model reads: each item with its namespace, key, value, timestamps and score, best match first.

Hot path or background