{
  "description": "OWG conformance corpus. The numbered W-documents run as a regression suite. The named project documents demonstrate the multi-document registry pattern: documents sharing a project_id reference a registry document's participants, organizations, infrastructure and contexts, so validate them together rather than in isolation.",
  "owg_version": "0.9",
  "schema": "https://openworkflowgraph.org/schemas/core/v0.9.json",
  "count": 22,
  "documents": [
    {
      "file": "europa-project.owg.json",
      "url": "https://openworkflowgraph.org/corpus/europa-project.owg.json",
      "kind": "project",
      "tier": null,
      "id": "europa-project",
      "project_id": "europa",
      "description": "MovieLabs 2030 Greenlight — 'Europa' shared entity registry. 17-min virtual-production short (ETC × USC). Cross-organization production: ETC/USC, Bind Studio (VFX), Skywalker Sound, Filmlight/color, Cree8 + AWS infrastructure. Registry of organizations, participants (the fmam roles/participants graph), OMC contexts (characters, locations, narrative + production scenes) and the picture-locked standing assets that Scenario 3's post-picture-lock change ripples through.",
      "tasks": 0,
      "assets": 8,
      "participants": 17
    },
    {
      "file": "europa-sc3-postlock.owg.json",
      "url": "https://openworkflowgraph.org/corpus/europa-sc3-postlock.owg.json",
      "kind": "project",
      "tier": null,
      "id": "europa-sc3-postlock",
      "project_id": "europa",
      "description": "MovieLabs 2030 Greenlight — Scenario 3 (post-picture-lock change), modeled as an OWG. After picture lock the director changes the porthole view in scene 1J; the change ripples to VFX, sound, color and editorial. OWG resolves the blast radius by graph traversal (no separate fmam database), surfaces cost/time impact BEFORE re-initiating work, notifies only the affected participants/services (publish/notify, not broadcast), and re-works non-destructively. Demonstrates 2030 principles 2, 3, 6, 7, 8, 9, 10.",
      "tasks": 9,
      "assets": 6,
      "participants": 0
    },
    {
      "file": "helios-concept-dev.owg.json",
      "url": "https://openworkflowgraph.org/corpus/helios-concept-dev.owg.json",
      "kind": "project",
      "tier": null,
      "id": "helios-concept-dev",
      "project_id": "helios",
      "description": "Helios — Character 'Aria' concept development: train character LoRA, generate variations, VFX supervisor review.",
      "tasks": 6,
      "assets": 5,
      "participants": 3
    },
    {
      "file": "helios-project.owg.json",
      "url": "https://openworkflowgraph.org/corpus/helios-project.owg.json",
      "kind": "project",
      "tier": null,
      "id": "helios-project",
      "project_id": "helios",
      "description": "Starfall Pictures — Feature film 'Helios'. Shared entity registry: organization, participants, OMC contexts (characters, locations, scenes), and standing assets.",
      "tasks": 0,
      "assets": 6,
      "participants": 10
    },
    {
      "file": "helios-vfx-s001.owg.json",
      "url": "https://openworkflowgraph.org/corpus/helios-vfx-s001.owg.json",
      "kind": "project",
      "tier": null,
      "id": "helios-vfx-s001",
      "project_id": "helios",
      "description": "Helios — VFX for Scene 001 (Aria arrives at Mars Base). Composite approved Aria concept elements over plate, make review proxy, deliver.",
      "tasks": 4,
      "assets": 5,
      "participants": 4
    },
    {
      "file": "hybrid-genai-pipeline.owg.json",
      "url": "https://openworkflowgraph.org/corpus/hybrid-genai-pipeline.owg.json",
      "kind": "project",
      "tier": null,
      "id": "hybrid-genai-pipeline",
      "project_id": "hybrid-genai-pipeline",
      "description": "Hero demo: one OWG workflow that spans the generative / AI-node world (ComfyUI-class steps an agent assembles — script understanding, prompt generation, 2D concept, 3D turntable) and the traditional video pipeline (annotate, ingest to MAM, human review/approval). One graph, one language, both worlds — authored by an agent, viewed by humans through Samsyn.",
      "tasks": 8,
      "assets": 7,
      "participants": 7
    },
    {
      "file": "tams-live-sports.owg.json",
      "url": "https://openworkflowgraph.org/corpus/tams-live-sports.owg.json",
      "kind": "project",
      "tier": null,
      "id": "tams-live-sports",
      "project_id": "tams-live-sports",
      "description": "BBC TAMS binding — the graph of a live match while the match is on. A growing, time-addressable source feed (essence in a TAMS store) carries QC and AI-highlight tasks that hang off it in near-real time; a highlight is a bounded timerange of the growing source — no copy, no transcode — cut, reviewed by a producer, and delivered, with full provenance across the live seam.",
      "tasks": 5,
      "assets": 4,
      "participants": 4
    },
    {
      "file": "usd-scene-composition.owg.json",
      "url": "https://openworkflowgraph.org/corpus/usd-scene-composition.owg.json",
      "kind": "project",
      "tier": null,
      "id": "usd-scene-composition",
      "project_id": "usd-scene-composition",
      "description": "OpenUSD binding — USD composes the scene, OWG governs the assets. Two shot assemblies compose the same published component assets (Hero character v23, the set, a hero prop) via USD reference/payload arcs. The composition graph is read deterministically from the entry layers, so the graph can answer what USD files encode but nothing upstream can query: which shots reference Hero v23, and what breaks if it's retired — impact analysis with per-component approvals attached.",
      "tasks": 6,
      "assets": 5,
      "participants": 4
    },
    {
      "file": "W01.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W01.owg.json",
      "kind": "conformance",
      "tier": "tier-0",
      "id": "W01-transcode-mov-mp4",
      "project_id": null,
      "description": "The simplest possible OWG workflow: transcode one source file MOV→MP4. One task, one participant, one input, one output.",
      "tasks": 1,
      "assets": 2,
      "participants": 1
    },
    {
      "file": "W02.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W02.owg.json",
      "kind": "conformance",
      "tier": "tier-0",
      "id": "W02",
      "project_id": "standalone",
      "description": "W02 (T0 Atomic) — Convert Markdown to DOCX. One task, one actor, document conversion. Stresses F1 (atomic task), F7 (botverse MCP executor), F18 (cloud storage).",
      "tasks": 1,
      "assets": 2,
      "participants": 1
    },
    {
      "file": "W03.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W03.owg.json",
      "kind": "conformance",
      "tier": "tier-0",
      "id": "W03",
      "project_id": "generative-ai",
      "description": "W03 (T0 Atomic) — Single text→image via one ComfyUI graph. One generative task on cloud GPU, producing a C2PA-credentialed asset. Stresses F1 (atomic task), F6 (sub-graph/nesting — the ComfyUI interior is zoomable), F9 (comfyui_graph executor), F17 (C2PA content credentials), F22 (cloud GPU compute).",
      "tasks": 1,
      "assets": 2,
      "participants": 1
    },
    {
      "file": "W04.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W04.owg.json",
      "kind": "conformance",
      "tier": "tier-1",
      "id": "W04",
      "project_id": "audio-workflows",
      "description": "W04 (T1 Linear Chain) — Record raw podcast audio (human), edit in DAW (local_app), normalise levels (Botverse), publish to RSS host (SaaS API). First workflow to exercise F2 (linear chain), F10 (local_app), F12 (human), F8 (saas_api), F13 (CONTINUE failure mode on normalisation).",
      "tasks": 4,
      "assets": 4,
      "participants": 4
    },
    {
      "file": "W05.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W05.owg.json",
      "kind": "conformance",
      "tier": "tier-1",
      "id": "W05",
      "project_id": "video-workflows",
      "description": "W05 (T1 Linear + Conditional) — Import UGC footage (local Premiere), colour grade and edit, conditionally generate captions via AI agent (when: add_captions param), thumbnail review by human, then publish to YouTube. Stresses F2 (chain), F4 (conditional step via when:), F10 (local_app), F11 (agent), F12 (human approval gate), F8 (saas_api).",
      "tasks": 5,
      "assets": 5,
      "participants": 4
    },
    {
      "file": "W06.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W06.owg.json",
      "kind": "conformance",
      "tier": "tier-1",
      "id": "W06",
      "project_id": "photography-workflows",
      "description": "W06 (T1 Linear Chain) — Photographer culls selects in Lightroom (human), retouches selects (local_app), exports full-res JPEGs (local_app), delivers to client portal (saas_api). Stresses F2 (chain), F10 (local_app), F12 (human cull gate), F8 (saas_api delivery), F5 (retry on delivery).",
      "tasks": 5,
      "assets": 5,
      "participants": 3
    },
    {
      "file": "W07.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W07.owg.json",
      "kind": "conformance",
      "tier": "tier-1",
      "id": "W07",
      "project_id": "vfx-workflows",
      "description": "W07 (T1 Parallel Branches) — Build After Effects comp (local_app), then colour_comp and audio_mix run in parallel (both depend only on comp), render combines both branches, QC validates (Botverse), then deliver. First workflow to exercise F3 (parallel branches — colour_comp and audio_mix run concurrently). Also stresses F10 (local_app), F7 (botverse_mcp QC), F2 (fan-in at render).",
      "tasks": 6,
      "assets": 6,
      "participants": 4
    },
    {
      "file": "W08.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W08.owg.json",
      "kind": "conformance",
      "tier": "tier-1",
      "id": "W08",
      "project_id": "generative-ai",
      "description": "W08 (T1 Autonomous + Parallel) — Fully autonomous AI pipeline: agent generates script from brief, then TTS and b-roll image generation run in parallel (both depend on script), agent assembles final video, cleanup runs as COMPENSATE if assemble fails. Stresses F2 (chain), F3 (parallel: tts + broll), F11 (agent), F8 (saas_api TTS), F9 (comfyui_graph), F17 (C2PA on generated imagery), F16 (COMPENSATE failure mode with compensation step).",
      "tasks": 5,
      "assets": 6,
      "participants": 6
    },
    {
      "file": "W09.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W09.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W09",
      "project_id": "sports-workflows",
      "description": "W09 (T2 Sports Live) — Detect a live sports highlight from a growing, time-addressable ingest source (BBC TAMS binding), reference the highlight as a bounded timerange of that source, transcode a deliverable clip, enrich metadata, and publish to platform. Exercises F2 (chain), F11 (agent for event detection + metadata), F8 (saas_api ingest + publish), F13 (CONTINUE on metadata enrich so publish proceeds even if tagging partial), plus the v1.1 TAMS binding (tams storage provider, opaque timerange, growing-source state, tams.source identifier pass-through).",
      "tasks": 5,
      "assets": 5,
      "participants": 6
    },
    {
      "file": "W10.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W10.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W10",
      "project_id": "sports-workflows",
      "description": "W10 (T2 Sports Graphics) — Pull live team stats, render lower-third and scoreboard graphics in CasparCG, composite over programme feed, deliver package to playout. Exercises F2 (chain), F3 (parallel: lower-third + scoreboard render), F8 (saas_api stats + playout), F10 (local_app CasparCG).",
      "tasks": 5,
      "assets": 5,
      "participants": 4
    },
    {
      "file": "W11.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W11.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W11",
      "project_id": "broadcast-workflows",
      "description": "W11 (T2 News QC) — Ingest news package, run loudness and artifact QC checks, agent aggregates report, human producer signs off, encode to broadcast spec, deliver to TX. Exercises F2 (chain), F3 (parallel: loudness + artifact checks), F12 (human sign-off gate), F11 (agent QC aggregation), F14 (SKIP_DEPENDENTS on critical artifact fail).",
      "tasks": 7,
      "assets": 7,
      "participants": 4
    },
    {
      "file": "W12.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W12.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W12",
      "project_id": "vfx-workflows",
      "description": "W12 (T2 VFX DI) — Ingest DPX plates in Resolve, apply CDL grade and LUT conform, export EXR sequence, composite VFX elements in Nuke, encode ProRes dailies preview, deliver to VFX house. COMPENSATE on composite failure to roll back partial EXR writes. Exercises F2 (chain), F10 (local_app Resolve + Nuke), F16 (COMPENSATE on composite), F8 (saas_api final delivery).",
      "tasks": 7,
      "assets": 7,
      "participants": 4
    },
    {
      "file": "W13.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W13.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W13",
      "project_id": "localization-workflows",
      "description": "W13 (T2 AI Dubbing) — Extract original dialogue and timecodes, fan out to four simultaneous translations, run TTS for each language in parallel, mix all tracks and deliver localised package. Exercises F2 (chain), F3 (wide parallel fan-out: 4 translate + 4 TTS branches), F11 (agent dialog extraction), F8 (saas_api translation + TTS), CONTINUE on individual TTS failures so remaining languages still deliver.",
      "tasks": 11,
      "assets": 12,
      "participants": 6
    },
    {
      "file": "W14.owg.json",
      "url": "https://openworkflowgraph.org/corpus/W14.owg.json",
      "kind": "conformance",
      "tier": "tier-2",
      "id": "W14",
      "project_id": "vfx-workflows",
      "description": "W14 (T2 Generative VFX — HEADLINE) — Full generative character asset pipeline: character artist ingests reference images, Botverse preprocesses for LoRA training, ComfyUI trains LoRA on cloud GPU, generates 20 character variations, artist reviews and selects approved set, VFX supervisor signs off, agent applies C2PA manifest to all approved assets, package and deliver to production asset management. Exercises F2 (chain), F9 (comfyui_graph: LoRA train + generate), F12 (two human gates: artist review + supervisor sign-off), F11 (agent C2PA signing), F17 (C2PA provenance on all generative outputs), F8 (saas_api delivery).",
      "tasks": 10,
      "assets": 10,
      "participants": 7
    }
  ]
}
