Mem0LongTermMemory

Long-term memory implementation using Mem0 as the store.

This implementation integrates with Mem0, a memory layer for AI applications that provides persistent, searchable memory storage using vector embeddings and LLM-powered memory extraction.

Key Features:

• Semantic memory search using vector embeddings

• LLM-powered memory extraction and inference

• Multi-tenant memory isolation (agent, user, run)

• Custom metadata support for tagging and filtering memories

• Automatic fallback mechanisms to ensure reliable memory storage

• Reactive, non-blocking operations

Memory Isolation: Memories are organized using three levels of metadata:

• agentId: Identifies the agent (optional)

• userId: Identifies the user/workspace (optional)

• runId: Identifies the session/run (optional)

• metadata: Custom key-value pairs for additional filtering (optional)

At least one identifier is required. During retrieval, only memories with matching metadata are returned.

Usage Example:

`// Create memory instance with authentication
Mem0LongTermMemory memory = Mem0LongTermMemory.builder()
.agentName("Assistant")
.userId("user_123")
.apiBaseUrl("http://localhost:8000")
.apiKey(System.getenv("MEM0_API_KEY"))
.build();

// For local deployments without authentication
Mem0LongTermMemory localMemory = Mem0LongTermMemory.builder()
.agentName("Assistant")
.userId("user_123")
.apiBaseUrl("http://localhost:8000")
.build();

// For self-hosted Mem0
Mem0LongTermMemory selfHostedMemory = Mem0LongTermMemory.builder()
.agentName("Assistant")
.userId("user_123")
.apiBaseUrl("http://localhost:8000")
.apiType(Mem0ApiType.SELF_HOSTED) // Specify self-hosted API type
.build();

// With custom metadata for filtering
Map<String, Object> metadata = new HashMap<>();
metadata.put("category", "travel");
metadata.put("project_id", "proj_001");

Mem0LongTermMemory memoryWithMetadata = Mem0LongTermMemory.builder()
.agentName("Assistant")
.userId("user_123")
.apiBaseUrl("http://localhost:8000")
.metadata(metadata) // Custom metadata for storage and filtering
.build();

// Record a message (metadata will be stored with the memory)
Msg msg = Msg.builder()
.role(MsgRole.USER)
.content("I prefer homestays when traveling")
.build();

memory.record(List.of(msg)).block();

// Retrieve relevant memories (metadata will be used as filter)
Msg query = Msg.builder()
.role(MsgRole.USER)
.content("What are my travel preferences?")
.build();

String memories = memory.retrieve(query).block();
// Result: "User prefers homestays when traveling"
`