Why should you care?

Hospital assistants need episodic chat plus persistent protocol memory plus authoritative EHR facts — one store is never enough.

See it live — copy this example

Paste into an ASP.NET Core 8+ / AgentNest project, then run with dotnet run (set your API keys in user-secrets).

// AgentNest.Memory/CompositeMemoryFacade.cs
public sealed class CompositeMemoryFacade(
    ThreadMemoryStore threads, IVectorStore vectors, ICrmTimelineRepository timeline)
{
    public async Task<MemoryBundle> LoadAsync(string tenantId, string threadId, string subjectId, CancellationToken ct)
    {
        var chat = await threads.LoadAsync(threadId, ct);
        var docs = await vectors.SearchAsync(tenantId, subjectId, topK: 3, ct);
        var events = await timeline.GetRecentAsync(subjectId, 10, ct);
        return new MemoryBundle(chat, docs, events);
    }
}

What happened?

• CompositeMemoryFacade merges chat history, vector hits, and CRM timeline events into one MemoryBundle for agents.
• Follow the steps below — typing the code yourself is the fastest way to learn.

Practice next

1. Define MemoryBundle DTO with three memory tiers.
2. Rank and dedupe overlapping facts before prompt injection.
3. Encrypt thread cache; classify vector content per retention policy.
4. Add memory summarization worker compressing old thread turns.
5. Tag vector chunks with retention class metadata.

Remember

Enterprise memory tiers: chat, vectors, structured SQL. CompositeMemoryFacade merges them for AgentNest agents. SQL remains authoritative for financial and clinical facts.

CRM copilot context

Rep resumes deal chat expecting prior email thread and playbook hits.

Outcome: CompositeMemoryFacade assembles chat, RAG, and CRM timeline in one agent call.