Operator architecture for reliable automation systems.

Define the task. Select the capability. Provide context. Run with a real runtime. Verify the result deterministically.

`operator-runtime` is still in active development.

Read the docs Start with the first operator guide

Task

The operator goal states the outcome the run should produce.

Capability

The capability names the bounded task domain for the run.

Context

Context carries structured execution input like URLs and IDs.

Runtime

Runtime holds the resources and state the engine can use.

Engine

The engine performs execution with the runtime resources.

Verification

Verification defines deterministic acceptance logic.

Trace

Trace records attempts, failures, and accepted results.

Execution model

The operator coordinates a run. The capability defines what kind of task the run is. The engine executes. The runtime supplies resources. Verification decides whether the result is accepted.

OPERATOR
  │
  │ task + context + verify
  ▼
Capability-selected execution
  │
  ▼
Engine
  │
  │ uses runtime
  ▼
Runtime (Browser / APIs / Test Runtime)
  │
  ▼
Environment
  │
  ▼
Verification
  │
  ▼
Result + Trace

Current state

The current implementation includes a typed operator runtime, a browser-backed example using Playwright, and a verified browser integration test. The project is still in active development.

The docs call out what exists today and what is planned so the abstraction stays honest.