Database Setup

Memory Service supports multiple database backends. This guide covers setup and configuration for each option.

PostgreSQL (Recommended)

PostgreSQL with pgvector is the recommended setup for most deployments.

Installation

Using Docker:

docker run -d \
  --name postgres \
  -e POSTGRES_DB=memoryservice \
  -e POSTGRES_USER=postgres \
  -e POSTGRES_PASSWORD=postgres \
  -p 5432:5432 \
  pgvector/pgvector:pg16

Configuration

quarkus.datasource.db-kind=postgresql
quarkus.datasource.jdbc.url=jdbc:postgresql://localhost:5432/memoryservice
quarkus.datasource.username=postgres
quarkus.datasource.password=postgres

# Schema management
quarkus.hibernate-orm.database.generation=update

# Connection pool
quarkus.datasource.jdbc.max-size=20
quarkus.datasource.jdbc.min-size=5

pgvector Extension

Enable pgvector for semantic search:

CREATE EXTENSION IF NOT EXISTS vector;

Configure in Memory Service:

memory-service.vector-store.type=pgvector
memory-service.vector-store.dimension=1536
memory-service.vector-store.index-type=ivfflat
memory-service.vector-store.lists=100

Index Types

-- Create HNSW index for better performance
CREATE INDEX ON entries USING hnsw (embedding vector_cosine_ops);

MongoDB

MongoDB is supported for teams already using MongoDB infrastructure.

Installation

docker run -d \
  --name mongodb \
  -e MONGO_INITDB_ROOT_USERNAME=admin \
  -e MONGO_INITDB_ROOT_PASSWORD=password \
  -p 27017:27017 \
  mongo:7

Configuration

quarkus.mongodb.connection-string=mongodb://admin:password@localhost:27017
quarkus.mongodb.database=memoryservice

Atlas Vector Search

For semantic search with MongoDB Atlas:

1. Create a vector search index in Atlas:

{
  "fields": [
    {
      "type": "vector",
      "path": "embedding",
      "numDimensions": 1536,
      "similarity": "cosine"
    }
  ]
}

2. Configure Memory Service:

memory-service.vector-store.type=mongodb
memory-service.vector-store.mongodb.index=vector_index
memory-service.vector-store.dimension=1536

Embedding Configuration

Configure the embedding model for vector generation:

OpenAI

memory-service.embedding.provider=openai
memory-service.embedding.model=text-embedding-ada-002
memory-service.embedding.api-key=${OPENAI_API_KEY}
memory-service.embedding.dimension=1536

Azure OpenAI

memory-service.embedding.provider=azure
memory-service.embedding.azure.endpoint=https://your-resource.openai.azure.com
memory-service.embedding.azure.deployment=text-embedding-ada-002
memory-service.embedding.azure.api-key=${AZURE_OPENAI_API_KEY}

Local Model

memory-service.embedding.provider=local
memory-service.embedding.local.model-path=/models/all-MiniLM-L6-v2
memory-service.embedding.dimension=384

Performance Tuning

Connection Pooling

# PostgreSQL
quarkus.datasource.jdbc.max-size=50
quarkus.datasource.jdbc.min-size=10
quarkus.datasource.jdbc.initial-size=10
quarkus.datasource.jdbc.acquisition-timeout=30

# MongoDB
quarkus.mongodb.max-pool-size=100
quarkus.mongodb.min-pool-size=10

Batch Operations

# Batch embedding requests
memory-service.embedding.batch-size=100

# Batch database inserts
memory-service.storage.batch-size=50

Backup and Recovery

PostgreSQL

# Backup
pg_dump -h localhost -U postgres memoryservice > backup.sql

# Restore
psql -h localhost -U postgres memoryservice < backup.sql

MongoDB

# Backup
mongodump --uri="mongodb://localhost:27017" --db=memoryservice --out=backup/

# Restore
mongorestore --uri="mongodb://localhost:27017" backup/

Next Steps

• Return to Configuration
• Learn about Docker Deployment