Reference
Authoritative documentation for every public surface. If a name shows up here, that's the contract.
CLI
init, validate, run, resume, inspect, state, reset-state, evolve, diff, ultra, ultra-revise, graph, schema, doctor, demo, plus the registry commands.
Common flows:
oe init my-flow # scaffold a new experience (+ editor autocomplete schema)
oe validate # check the YAML schema
oe graph . # render the experience DAG as Mermaid
oe run . --tui # run with the TUI dashboard
oe inspect <id> --html # event timeline + per-node metrics → HTML run report
oe evolve <id> # advisor proposes graph upgradesNew since v0.1.0: oe graph (Mermaid DAG), oe schema (JSON Schema for editor autocomplete), and oe ultra-revise (apply feedback to an existing draft).
Programmatic API
9 modules exposing the same primitives the CLI uses. Useful when embedding OE in a service.
import { runExperience, EventBus } from '@openexpertise/core'
const events = new EventBus()
events.subscribe((e) => /* … */)
await runExperience({ specPath: './experience.yaml', events })Recent additions: UltraExpertise gains the critique → revise quality loop (the loop field + maxRounds/criticModel/exemplars opts) and a reviseDraft() method; EvolutionAdvisor gains analyzeAcrossRuns() + renderMarkdownCrossRun() for cross-run analysis.
YAML schema
The canonical shape of experience.yaml. AJV-validated by oe validate and the runtime, and available as a JSON Schema for editor autocomplete via oe schema.
Top-level sections: name, version, description, state.schema, phases, graph (nodes + edges + optional pipelines + loops), runtime.