MotteMB provides sophisticated memory management using vector embeddings for context-aware agent memory and retrieval.
MotteMB implements a sophisticated memory management system that allows agents to store, retrieve, and reason about past interactions. Using vector embeddings and semantic search, agents can access relevant historical context to improve decision-making and maintain coherent long-term behavior.
Find relevant memories using natural language queries with vector similarity search.
Store memories as high-dimensional vectors for efficient similarity matching and retrieval.
Memories are automatically categorized by type and context for better organization.
Import/export memories in various formats and perform batch operations efficiently.
Start by adding some context or knowledge that your agents can reference:
1. Navigate to MotteMB 2. Click "Create Memory" 3. Enter content: "Our company policy states that refunds are available within 30 days of purchase for digital products, and 60 days for physical products. Customers need to provide proof of purchase and reason for return."
Use natural language to find relevant memories:
Bulk import your existing knowledge base or documentation:
MotteMB uses OpenAI's text-embedding-3-large model to create vector representations of memories. This enables semantic search that understands context and meaning, not just keyword matching.
Query: "customer return policy"
↓
Vector Embedding (1536 dimensions)
↓
Similarity Search (cosine similarity)
↓
Ranked Results (similarity > 0.7)Keep your memory bank efficient with built-in optimization tools:
Automatically identify and remove duplicate or near-duplicate memories.
Compress and optimize vector storage for better performance.
Rebuild search indexes for optimal query performance.
POST /api/memory/store
{
"content": "Memory content here",
"metadata": {
"category": "policy",
"source": "handbook",
"priority": "high"
}
}Stores a new memory with optional metadata for categorization.
POST /api/memory/search
{
"query": "refund policy for digital products",
"limit": 5,
"threshold": 0.7
}Searches for relevant memories using semantic similarity.
POST /api/memory/import Content-Type: multipart/form-data file: [uploaded file]
Bulk import memories from JSON, CSV, or text files.
Write clear, descriptive memories that include context and relevant keywords.
Add metadata like categories, sources, and priorities to improve organization.
Run optimization regularly to maintain search performance and remove duplicates.
Export your memories regularly as backups and for version control.