Skip to content

README.md

Latest commit

 

History

History
104 lines (74 loc) · 3.3 KB

README.md

File metadata and controls

104 lines (74 loc) · 3.3 KB

Workshop 4: Agent Memory Server

Give your AI agents persistent memory with Redis-powered Agent Memory Server (AMS).

What You'll Learn

  • Working memory for session context (auto-summarization, TTL)
  • Long-term memory for persistent facts (semantic search)
  • How Redis stores and retrieves AI memories

Prerequisites

  • OpenAI API Key - Required for embeddings and chat

Run with Docker

# From repository root
cd java-springboot/workshop-hub

# Start Redis + Workshop
docker compose -f docker-compose.local.yml --profile infrastructure up -d
docker compose -f docker-compose.local.yml --profile workshop-4_agent_memory up -d

Open http://localhost:8083

Enter your OpenAI API key when prompted.

Workshop Flow

Stage What You Do
1. Problem Understand why AI agents forget between sessions
2. Solution Implement working + long-term memory
3. Demo 8 guided tests to verify memory behavior

Guided Tests

  1. Working Memory Basics - AI remembers within a session
  2. Context Window Growth - Watch context % increase
  3. Store Long-Term Memory - Save facts that persist
  4. Semantic Search - Find memories by meaning
  5. Clear Working Memory - Session resets, facts remain
  6. Cross-Session Recall - New session retrieves old facts
  7. TTL and Lifecycle - Working memory expires
  8. Redis Insight - Explore memory data structures

Your Tasks

Work through the same three files shown in the in-browser editor:

  1. AgentMemoryService.java Implement the core AMS SDK calls for working memory, long-term memory, and the health check.
  2. AmsChatMemoryRepository.java Bridge Spring AI conversation state to AMS, including context percentage lookup, conversation loading, deduplicated writes, and TTL setup.
  3. ChatService.java Pass the conversation ID into the advisor chain and enable the working-memory and long-term-memory advisors used by the chat demo.

The editor guidance follows this order: SDK client setup first, repository integration second, advisor wiring last.

Memory Types

Type Purpose Example
SEMANTIC Facts and preferences "User prefers dark mode"
EPISODIC Events with time context "User went hiking on Saturday"
MESSAGE Conversation excerpts "User mentioned they work in tech"

View in Redis Insight

Open http://localhost:5540

Working Memory:

KEYS working_memory:workshop:*
HGETALL working_memory:workshop:session-default
TTL working_memory:workshop:session-default

Long-Term Memory:

KEYS long_term_memory:workshop:*
FT.INFO idx:long_term_memory:workshop

Key Concepts

  • Working Memory - Session-scoped, auto-summarizes at 70% context, has TTL (1800s default)
  • Long-Term Memory - Persistent, vector-indexed for semantic search
  • Namespace - Isolates memories between applications
  • Session ID - Groups working memory for a conversation

Stopping

docker compose -f docker-compose.local.yml --profile workshop-4_agent_memory down

Resources