Graph-Aware Retrieval for AI Agent Memory: Linking Memories Across Time

[srinidhi-sat]

Summary:
This is a companion notebook for the Graph-Aware Retrieval for AI Agent Memory article.
This notebook demonstrates how Oracle AI Agent Memory preserves evolving facts by linking durable memories and retrieving related context through graph-aware search.

Covers:

• Oracle Database connection setup, with local Oracle Free validation/ FreeSQL setup
• oracleagentmemory 26.8 package and API checks
• Managed schema initialization and graph-memory privilege preflight
• Durable memory creation for an evolving support case
• Typed memory links: supersedes, refines, and supports
• Conceptual coverage for duplicates and contradicts
• Persisted lifecycle/status inspection for valid and invalid memories
• Direct search versus graph-aware search
• num_hops=0 versus num_hops=1 with matching search options
• linked_results inspection and prompt-ready graph context generation
• Scoped retrieval with user_id, agent_id, tenant, and case metadata
• Automatic memory linking during extraction as part of the main runnable flow
• Deterministic local validation providers for reproducible Podman execution