# Open Workflow Graph (OWG)

> Specification v0.92 (draft) — an open specification for describing media
> production workflows as portable, validatable, executable JSON documents,
> built on MovieLabs' Ontology for Media Creation (OMC), with provenance,
> rights, and security carried in the schema rather than bolted on.

OWG is published by Entertainment Technologists Inc. It is an open
specification, not a ratified industry standard. v0.92 is a draft: expect
field-level change, and pin the schema you validate against.

Everything here is public. The schema is served at
https://openworkflowgraph.org/schemas/core/v0.92.json and the JSON-LD context
at /context/v0.9.jsonld. No authentication on any URL below.

## Start here

- [Full documentation, single file](https://openworkflowgraph.org/llms-full.txt): every page concatenated as markdown — fetch this if you want everything at once.
- [Agentic workflows and certainty](https://openworkflowgraph.org/raw/agentic-workflows.md): why an agent-authored workflow can be trusted to execute — the four properties that must be established before a run.
- [Conformance corpus index](https://openworkflowgraph.org/corpus/index.json): 22 complete, validated OWG documents as downloadable JSON, with task/asset/participant counts. Reference data — start here rather than transcribing examples out of the prose.
- [Core schema](https://openworkflowgraph.org/schemas/core/v0.92.json): validate against this. Every object is closed (additionalProperties: false); extend via `metadata`, an `x-` prefix, or (on context/asset/relationships) a namespaced `profiles.<id>` key.
- [For agents](https://openworkflowgraph.org/raw/for-agents.md): machine-oriented implementation guide, authoring rules, and a pre-flight checklist.
- [Versioning and stability](https://openworkflowgraph.org/raw/versioning-and-stability.md): what v0.92 defines, and how stable each part is.
- [Implementation considerations](https://openworkflowgraph.org/raw/implementation-considerations.md): decisions the specification leaves to implementers, with recommended approaches.

## Introduction

- [Overview](https://openworkflowgraph.org/raw/overview.md): What the Open Workflow Graph is, the problem it solves, and how it relates to the standards already in your pipeline.
- [Quickstart](https://openworkflowgraph.org/raw/quickstart.md): Write a valid OWG document, validate it, and understand what the graph now knows — in about ten minutes.
- [Core concepts](https://openworkflowgraph.org/raw/concepts.md): The five node types, the provenance backbone beneath them, and the two design choices that make a workflow both a plan and a record.
- [Agentic workflows and certainty](https://openworkflowgraph.org/raw/agentic-workflows.md): What changes when workflows are written by machines rather than people, the four things you must be able to establish about a workflow before running it, and how OWG makes each of them checkable.

## Specification

- [Document structure](https://openworkflowgraph.org/raw/specification.md): The OWG v0.92 document — every top-level field, the entity registries, parameters, and governance ceilings.
- [Tasks and executors](https://openworkflowgraph.org/raw/tasks.md): The task object in full — executors, provenance edges, dependencies, failure and compensation, retry, gates and re-routes.
- [Assets and versioning](https://openworkflowgraph.org/raw/assets-and-versioning.md): Asset identity, storage locators, the five typed version relations, lifecycle states, and how composition and external identifiers ride along.
- [Participants and authority](https://openworkflowgraph.org/raw/participants-and-authority.md): How humans, agents, and services are declared, and how an agent's authority chains to a responsible human.
- [Relationships and assurance](https://openworkflowgraph.org/raw/relationships.md): Explicit edges, the three assurance levels, and exactly where a machine proposal's confidence score and method live in the payload.
- [Subgraphs and turnovers](https://openworkflowgraph.org/raw/subgraphs-and-turnovers.md): How a task nests an interior graph, and how that same mechanism scopes a cross-company hand-off to exactly what a vendor needs.
- [Reference syntax](https://openworkflowgraph.org/raw/reference-syntax.md): The formal grammar for $. references, string interpolation, and when expressions — with the run-state vocabulary they evaluate against.
- [Profiles](https://openworkflowgraph.org/raw/profiles.md): How profiles declare the executors, tool vocabularies, and registry extensions a workflow may use, keeping domain and vendor detail out of the core specification.
- [Run records](https://openworkflowgraph.org/raw/runs.md): The run document — how one execution of a definition is recorded, with per-attempt status, cost, and the actual infrastructure used.

## Implementing

- [Validation](https://openworkflowgraph.org/raw/validation.md): Every check the validator performs — structural, referential, graph, and expression — and the error codes it emits.
- [Conformance corpus](https://openworkflowgraph.org/raw/conformance.md): The reference workflows, what each tier proves, how the regression harness runs them, and the true inventory.
- [Persistence requirements](https://openworkflowgraph.org/raw/persistence.md): What an implementation must be able to answer about an OWG graph — stated as required capabilities, not as a prescribed storage design.
- [For agents](https://openworkflowgraph.org/raw/for-agents.md): A machine-oriented implementation guide — canonical endpoints, the minimum viable document, a pre-flight checklist, and what an agent-authored workflow must satisfy to be trusted.
- [Examples](https://openworkflowgraph.org/raw/examples.md): Complete annotated workflows from the conformance corpus — atomic, generative, linear chain, and conditional with a human gate.
- [Implementation considerations](https://openworkflowgraph.org/raw/implementation-considerations.md): Decisions the specification deliberately leaves to implementers, with recommended approaches and the trade-offs behind them.
- [Failure-handling best practices](https://openworkflowgraph.org/raw/failure-handling.md): Four ways a task's failure can behave, when to reach for each one, and how they compose with fan-out and conditional merges.

## Ecosystem

- [Standards bindings](https://openworkflowgraph.org/raw/standards-bindings.md): How OWG references OpenUSD, BBC TAMS, OpenTimelineIO, C2PA, SOM, and OMC without absorbing their internals — and the field-level hooks each binding uses.
- [Security model](https://openworkflowgraph.org/raw/security-model.md): Why permissioning is a property of the graph rather than a system wrapped around it — and how scope- and time-bounded access falls out of the schema.
- [AI disclosure and compliance](https://openworkflowgraph.org/raw/compliance.md): How EU AI Act Article 50 disclosure becomes a lineage query, and the six schema fields that make a per-asset disclosure manifest computable.
- [Versioning and stability](https://openworkflowgraph.org/raw/versioning-and-stability.md): What v0.92 defines, how the specification versions, where it deliberately leaves room to extend, and what that means for building against it.
- [Changelog](https://openworkflowgraph.org/raw/changelog.md): What changed in each version of the specification, and when.

## Notes for implementers

- Any page is available as raw markdown at /raw/<slug>.md
- A document is entity registries (organizations, participants, infrastructure, assets, contexts) plus tasks[]. Required: `owg_version` ("0.92", "0.91", or "0.9") and `id`.
- Every defined object sets `additionalProperties: false`. Unknown keys are validation errors; use `metadata`, an `x-` prefix, or (v0.92, on context/asset/relationships) a namespaced `profiles.<id>` key to extend.
- Reference syntax (`$.params.x`, `$.tasks.y.outputs.z`) and `when` expressions have formal EBNF grammars — see /raw/reference-syntax.md.
- Validation runs in four passes: structural, referential, graph, expression. The reference implementation currently covers pass 1 fully and passes 2-3 partially, so verify referential integrity yourself.
- An `inferred` assurance requires `confidence` (0-1) and `method`. See /raw/relationships.md.