[srinidhi-sat]

What changed

• Added a end-to-end guide for evolving a RAG application into an agent memory system with Oracle AI Database.
• Builds separate knowledge, conversation-turn, and durable-memory stores with vector embeddings and identity-based scope metadata.
• Demonstrates RAG-only retrieval, memory-aware context assembly, and scoped retrieval across tenants, users, agents, and threads.
• Implements memory promotion, unsafe-candidate rejection, correction, supersession, and forgetting.
• Verifies database-backed persistence by closing and recreating the connection pool.
• Includes deterministic evaluation checks, local embeddings by default, and an optional OpenAI embedding path.