Add Memory

Agent calls m.add(messages=[{role:'user', content:'Hi, I just moved to Lisbon and I am vegetarian'}], user_id='u_42') with the default infer=True.

With infer=True (default), Mem0 sends the messages to the extraction LLM and stores distilled facts (e.g., 'Lives in Lisbon', 'Is vegetarian') — not the raw chat turn. Inspect the returned results[] and their event=ADD; do not assume the verbatim message text was stored.

Operator needs to persist exact assistant tool outputs (a JSON config blob) without LLM distillation and calls m.add(messages, user_id='u_1', infer=False).

Set infer=False so Mem0 stores the message content verbatim with no fact extraction or dedup pass. Use this only when raw fidelity matters; expect no ADD/UPDATE/DELETE consolidation events and no LLM cost for extraction.

An add() response returns results=[{memory:'Prefers window seat', event:'ADD', id:'m1'}, {memory:'Allergic to peanuts', event:'UPDATE', id:'m2'}].

Branch on each result's event: ADD = new memory created; UPDATE = an existing memory was revised in place (same id, new history entry); DELETE = a memory was removed as contradicted; NONE = nothing changed. Reconcile the operator's local mirror against these events rather than assuming every add cr…