The Runtime Layer is the backbone of every FRAKTIΛ agent. It handles all internal logic, action sequencing and coordination with Add-Ons, memory and communication layers. It interprets the character.json file, evaluates triggers and routes events to appropriate modules. (inference, webhook, memory, etc.)

The default runtime in FRAKTIΛ is ElizaOS, with support for other execution environments on the roadmap.

Responsibilities of the Runtime

Runtime Configuration Block (character.json)

"runtime": {
  "engine": "elizaos@2.2",
  "onBoot": ["greetUser"],
  "onMessage": "parseCommand",
  "onTrigger": "routeToAddOn",
  "fallback": {
    "onFailure": "respondWithDefault",
    "onTimeout": "retryOnce"
  }
}

Trigger Routing Example

Input: "What's the current ETH price?"
Runtime sequence:

Triggers onMessage → parseCommand.
Determines intent: getPrice.
Routes to Add-On: PriceOracle → getPrice().
Returns result.
Formats output and responds via voice + text

Multi-Runtime (Future Support)

You’ll be able to choose between runtime engines for specialized agents:

✦ elizaos: general-purpose deterministic kernel.
✦ chaosos: probabilistic runtime with event branching.
✦ roboticaos: real-time robotics execution.
✦ complianceos: audit-focused, deterministic logging.

Runtime Isolation & Trust

✦ Each agent runs in its own containerized runtime.
✦ Memory is scoped per runtime instance.
✦ Runtime decisions can be versioned and audited.
✦ Swarm-wide coordination is sandboxed by design.

Strategic Role

The Runtime Layer is where structured logic meets autonomy.
It ensures every agent remains reactive, traceable and upgradeable, regardless of cognitive engine, Add-Ons or deployment environment.