AdvancedPython~20 min

Production with ElastiCache

Step 1: ElastiCache for Valkey 8.2+

ElastiCache for Valkey 8.2 includes built-in vector search at no additional cost. Create a cluster via the AWS Console or CLI, then use the cluster endpoint as your Valkey URL.

from mem0 import Memory

config = {
    "vector_store": {
        "provider": "valkey",
        "config": {
            # Use your ElastiCache cluster endpoint
            "valkey_url": "valkey://your-cluster.xxxxx.use1.cache.amazonaws.com:6379",
            "collection_name": "prod_memories",
            "embedding_model_dims": 1536,
            "index_type": "hnsw",
            # HNSW tuning parameters
            "hnsw_m": 16,              # connections per node
            "hnsw_ef_construction": 200, # build-time search width
            "hnsw_ef_runtime": 10,      # query-time search width
        }
    }
}
memory = Memory.from_config(config)

Step 2: HNSW Parameter Tuning

Parameter	Default	Effect	Recommendation
`hnsw_m`	16	More connections = higher recall, more memory	16 for most cases, 32 for >1M memories
`hnsw_ef_construction`	200	Higher = better index quality, slower build	200 default, increase to 400 for critical apps
`hnsw_ef_runtime`	10	Higher = better recall, higher latency	10 for speed, 50+ for maximum recall
`index_type`	hnsw	HNSW=fast approximate, FLAT=exact	HNSW for >1000 memories, FLAT for small sets

Step 3: TLS for ElastiCache

# ElastiCache with TLS enabled
config = {
    "vector_store": {
        "provider": "valkey",
        "config": {
            "valkey_url": "valkeys://your-cluster.xxxxx.use1.cache.amazonaws.com:6379",
            # Note: valkeys:// (with 's') enables TLS
            "collection_name": "prod_memories",
            "embedding_model_dims": 1536,
        }
    }
}

Step 4: Monitoring

import valkey

# Connect directly to check index health
client = valkey.from_url("valkey://your-cluster:6379")

# Check index info
info = client.execute_command("FT.INFO", "prod_memories")
print(info)

# Check memory usage
mem_info = client.info("memory")
print(f"Used: {mem_info['used_memory_human']}")

# Count memories per user
results = client.execute_command(
    "FT.SEARCH", "prod_memories",
    "@user_id:{alice}",
    "LIMIT", "0", "0",  # count only
)
print(f"Alice has {results[0]} memories")

Production Checklist

Area	Recommendation
ElastiCache	Use Valkey 8.2+ for built-in vector search
TLS	Use `valkeys://` URL scheme for encrypted connections
Multi-AZ	Enable for high availability
HNSW M	16 default, 32 for large memory stores
Embedding model	text-embedding-3-small (1536 dims, fast, cheap)
Memory limits	Use maxmemory-policy to handle full memory
Monitoring	Track FT.INFO for index size + memory usage

Source: Mem0's Valkey connector: valkey.py and config: ValkeyConfig

← Previous02 - Multi-User Memory