Whenever you're ready, walk me through how you'd approach this.
How do you handle long-term memory and context retrieval for agents operating across fragmented data sources?
