Re in Act Extensions

Optional extensions to Re in Act that add capabilities without changing the core contract.

Re in Act Extensions

Re in Act keeps a small core and allows optional extensions on top.

The core remains:

  • top-level reasoning defines and orchestrates the RAS
  • reason() is required
  • act() is optional
  • deterministic control flow runs inside the RAS runtime

Extensions add extra capability for specific environments while preserving that core.

Why Extensions

Extensions let implementations evolve faster than the core spec:

  • new capabilities can be introduced without forcing all runtimes to adopt them
  • production systems can standardize optional features gradually
  • experimentation can happen without destabilizing the base interfaces

This follows the same design philosophy used by protocols that keep a strict core and a flexible extension surface.

Extension Design Rules

An extension SHOULD:

  • be explicitly opt-in
  • define clear input/output schema and failure semantics
  • keep RAS boundaries clear (what stays local, what leaves the space)
  • avoid redefining the core contract of reason()
  • preserve runtime portability where practical

An extension MUST NOT:

  • change the required status of reason()
  • assume hidden global state as part of its contract
  • bypass sandbox, policy, or budget boundaries defined by the runtime

Current Extensions

IdentifierReference TitleLink
ria.agentAgent Interface ExtensionAgent Interface Extension

See Extension Support Matrix for status and compatibility.