spring-ai-memory-integration

Spring AI AgentCore Memory Example

A complete example demonstrating Amazon AgentCore Short-Term & Long-Term Memory integration with Spring AI for persistent conversation history.

This example creates and uses the AgentCore Long-Term Memory Summary Strategy.

Prerequisites

• Java 17+
• Maven 3.6+
• AWS credentials configured locally

Architecture

  User          ChatClient       STM Advisor      LTM Advisor      AgentCore       LLM
   │                │                │                │                │            │
   │  prompt        │                │                │                │            │
   │───────────────>│                │                │                │            │
   │                │                │                │                │            │
   │                │  1. get history│                │                │            │
   │                │───────────────>│                │                │            │
   │                │                │  listEvents    │                │            │
   │                │                │────────────────────────────────>│            │
   │                │                │<────────────────────────────────│            │
   │                │<───────────────│  messages      │                │            │
   │                │                │                │                │            │
   │                │  2. retrieve LTM                │                │            │
   │                │────────────────────────────────>│                │            │
   │                │                │                │  searchMemories│            │
   │                │                │                │───────────────>│            │
   │                │                │                │<───────────────│            │
   │                │<────────────────────────────────│  memories      │            │
   │                │                │                │                │            │
   │                │  3. invoke (history + LTM + prompt)              │            │
   │                │──────────────────────────────────────────────────────────────>│
   │                │<──────────────────────────────────────────────────────────────│
   │                │                │                │                │  response  │
   │                │                │                │                │            │
   │                │  4. save messages               │                │            │
   │                │───────────────>│                │                │            │
   │                │                │  putEvents     │                │            │
   │                │                │────────────────────────────────>│            │
   │                │                │<────────────────────────────────│            │
   │                │                │                │                │            │
   │<───────────────│  response      │                │                │            │
   │                │                │                │                │            │

Quick Start

Local Setup

1. Setup your local AWS credentials / auth

2. Create the AgentCore Memory:

   mvn spring-boot:test-run

3. Export the AGENTCORE_MEMORY_ID and SUMMARY_STRATEGY_ID env vars

4. Start the Spring web application

   mvn spring-boot:run

5. Test the application:

   # --- Short-Term Memory (STM) ---
   # Tell your name
   curl -X POST http://localhost:8080/api/short \
       -H "Content-Type: application/json" \
       -d '{"message": "My name is Andrei"}'
   
   # Ask for your name (memory recall)
   curl -X POST http://localhost:8080/api/short \
       -H "Content-Type: application/json" \
       -d '{"message": "What is my name?"}'
   
   # --- Long-Term Memory (LTM) ---
   # Ask something to be persisted in LTM
   curl -X POST http://localhost:8080/api/long \
       -H "Content-Type: application/json" \
       -d '{"message": "I love hiking in the Alps"}'
   
   # Get conversation history
   curl http://localhost:8080/api/history
   
   # Get stored LTM memories
   curl http://localhost:8080/api/memories
   
   # Clear conversation
   curl -X DELETE http://localhost:8080/api/history

Cleanup

With the AGENTCORE_MEMORY_ID env var set, run:

mvn spring-boot:test-run

Custom Namespace Pattern

This example demonstrates using a custom namespace pattern for the Summary strategy.

In application.properties:

agentcore.memory.long-term.summary.namespace-pattern=custom-namespace/strategies/{memoryStrategyId}/actors/{actorId}/sessions/{sessionId}

The same pattern is configured in SetupTeardown.java when creating the memory strategy:

.namespaces(List.of("custom-namespace/strategies/{memoryStrategyId}/actors/{actorId}/sessions/{sessionId}"))

Important: The namespace pattern in your Spring configuration must match the namespace configured in your AgentCore Memory strategy. The library validates this at startup and will fail with a clear error if there's a mismatch.

Auto-Registration

If you want the library to automatically update AWS Memory namespaces when they don't match, enable auto-registration:

agentcore.memory.long-term.namespace.auto-register=true

This is useful during development when namespace patterns may change. Use with caution in production.

Autodiscovery Mode

Instead of manually configuring each strategy, you can use autodiscovery to automatically detect strategies from your AgentCore Memory:

agentcore.memory.long-term.auto-discovery=true

Autodiscovery behavior:

• Queries AWS to discover all strategies in your memory
• Creates advisors for supported types: SEMANTIC, SUMMARIZATION, USER_PREFERENCE, EPISODIC
• Uses discovered namespace and default topK values

Overriding discovered settings:

agentcore.memory.long-term.auto-discovery=true
# Override topK for a discovered strategy (strategy-id must match)
agentcore.memory.long-term.summary.strategy-id=${SUMMARY_STRATEGY_ID}
agentcore.memory.long-term.summary.top-k=5

The override only applies if strategy-id matches the discovered strategy ID.