Underlying our expressive compute precompiles—which enable any type of heterogeneous compute, including AI inference, ZK proof verification, TEE code execution, and beyond—are our optimized execution sidecars.
These sidecars run and scale parallel to the Ritual execution client, invoked
through abstraction via the EVM precompile interface. By default, these sidecars
are purpose-built for execution and proof generation avoiding
blockchain-specific overhead, run as an isolated process or remote to an
execution client, and can be selectively enabled by node operators based on
their available infrastructure, underlying hardware, or firewalled preferences.Because of their generic architecture, sidecars enable Ritual to easily service
heterogeneous compute across any supported execution environment by connecting
to their precompile-like interfaces (EVM precompiles, SVM precompiled programs,
MoveVM system modules, etc.).
Sidecars today are stateless, but will soon be extended to being stateful,
allowing for a minimally invasive approach to enabling asynchronous execution on
our chain for long-running compute precompiles.
