Bog — A toolkit for building custom
high-performance queryless databases.

So, what is Bog?

Bog is a pre-release database runtime and programming language for building custom compiled databases. It's the first higher-order system for creating DBMS applications. It leverages compile-time correctness guarantees (à la Rust) to represent any set of data access patterns as a purpose-built binary.

Bog produces databases where ‘queries’ are more accurately represented as function calls. Bog is built in Rust, so thanks to Rust's trait system and expressive generics, complex dataflow pipelines (similar to Differential Dataflow or DBSP) are monomorphized at compile time into a single concrete type, avoiding expensive runtime graph interpretation.

Bog’s core subsystems are distributed together as BogKit:

• Fold, an incremental programming kernel
• ANNy, a high-performance HNSW implementation
• ESE, the fastest static embedding model in the world

Why should I use Bog?

Bog databases have the following properties due to being compiled:

• Bog databases have already proven to save 60% of token costs on real-world agent tasks.
• Bog databases are very fast, very simple, and easy to integrate/build for humans and agents alike. They are literally defined in code, no ORMs here.
• Bog databases resolve into smaller binaries than SQLite databases.
• Bog databases conform to any data-shape: graph, relational, semantic, document, etc.
• Bog databases have performant semantic and full-text search built in by virtue of us building the fastest static embedding model in the world. No external services or runtimes are needed.
• And much more!