moai-domain-backend

Define Pydantic response models, then implement endpoints protected by an HTTPBearer security dependency. The skill's pattern shows list and create endpoints returning typed UserResponse models with JWT-based access control.

Configure the SQLAlchemy engine with QueuePool, pool_size of 20, max_overflow of 30, pool_pre_ping enabled, and pool_recycle of 3600 seconds. Event listeners on cursor execution can log queries exceeding a 100ms threshold.

Use a CacheManager class with a cache_result decorator that accepts a ttl parameter. It builds cache keys from the function name and arguments, returns cached values on hit, and stores serialized results with expiration on miss.

Yes, it includes a GraphQL implementation pattern using Strawberry, defining typed objects and async resolvers, alongside REST with FastAPI and gRPC for high-performance services.

Frontend validation is a user-experience concern, while backend validation is a security requirement. Every endpoint must validate input because requests can bypass the frontend entirely, and unvalidated input risks injection and data corruption.

Unbounded endpoint growth creates a monolith-in-disguise. Evaluate service boundaries first; if the new functionality belongs to a distinct domain, a separate microservice with its own discovery and event flow may be the better pattern.