ADR-001: Use PostgreSQL + pgvector as sole data store for v1

Status: Accepted
Date: 2026-04-24

Context

We need a durable store for episodes, memories, and embeddings. Options considered: Postgres, Postgres+pgvector, dedicated vector DB, Redis, multi-store.

Decision

Use PostgreSQL with the pgvector extension as the single data store for v1.

Rationale

• One database to operate, backup, and reason about
• pgvector handles embedding similarity search adequately for MVP scale
• Postgres JSONB handles flexible payloads
• Alembic provides reliable migrations
• Avoids operational complexity of multiple stores
• Easy to swap or add a dedicated vector DB later if needed

Consequences

• Embedding search performance is bounded by pgvector capabilities
• Acceptable for MVP; revisit at scale