Deployment & Infrastructure

Infrastructure

AWS Setup

• Compute: EC2 t2.micro for admin access
• Database: PostgreSQL 17 RDS (t3.micro) with pgvector extension
• Stores both metadata (pointers, relationships) and vector embeddings

• pgvector enables fast similarity search: embedding <-> query_vector

• Storage: S3 bucket with encrypted, deduplicated blobs

• Security: Custom security groups for EC2↔RDS access

Key Terraform Resources

• VPC with default networking
• S3 bucket with public access blocked
• RDS PostgreSQL instance (20GB allocated) with pgvector extension
• Security groups for SSH access and database connectivity

Privacy & Data Retention

• All user-generated data stored in S3 with encryption
• Metadata in PostgreSQL with retention policies per event
• Redaction flags for sensitive moments
• Privacy classes: public (share), personal (archive), sensitive (delete after N days)

Future Enhancements

• Real-time vector database integration (Pinecone, Weaviate)
• Neo4j for graph-based entity relationships
• Normalized schema for canonical entities and facts
• Federated learning for on-device model improvement
• Multi-modal embeddings (text, audio, video)
• Search UI for memory exploration