Rad: Open-Source Database Redesigned for Modern OLTP with Schema-First Architecture
Key Takeaways
- ▸Schema-first approach where all database components (migrations, type-safe clients, relationships) auto-generate from a single declarative schema definition
- ▸Eliminates ORMs and query builders with generated clients that provide typed queries, rows, and relationships without string concatenation
- ▸OLTP backed by durable object storage with stateless compute, aligning with modern cloud-native infrastructure patterns
Summary
Rad is an open-source relational database that fundamentally rethinks database design around a declarative schema-first approach. Rather than starting with code and building schema around it, Rad inverts the relationship: you define your data model once, and the database automatically generates type-safe clients, handles migrations, and enforces contracts between data and application logic.
The platform eliminates traditional ORMs and query builders by generating structured clients directly from your schema definition. Rad includes the database server, CLI developer tools, and code generation all in a single executable that runs on Windows, macOS, and Linux. It supports nested relationships, automatic migrations from schema diffs, and uses a simple JSON-over-HTTP protocol designed for generated clients rather than hand-written requests.
Underlying the entire system is durable object storage, enabling stateless compute and modern cloud-native architecture patterns. The toolkit handles the complete loop from schema definition through client generation and migrations, reducing boilerplate and enabling developers to focus on business logic rather than database plumbing.
- Single unified executable containing database server, CLI tools, and code generation—no separate orchestration required
- Available as open-source project with support for automatic migrations derived from schema diffs and nested relationship queries
Editorial Opinion
Rad represents a thoughtful reimagining of database architecture that challenges decades of assumptions about how data layers should be structured. By inverting the typical approach—starting with the schema and generating everything from it—Rad elegantly solves a long-standing friction point in modern development. The focus on durable object storage and stateless compute positions it well for cloud-native deployments, though adoption will depend on how the project navigates the ecosystem of existing database tools and developer familiarity.
