Tagline: MLflow versions your models. CRI versions your reasoning.
Position: CRI-CORE is the runtime that enforces AWO’s reproducibility rules as executable infrastructure.
AWO defines the rules. CRI enforces them.
CRI-CORE turns the Aurora Workflow Orchestration (AWO) method into a runnable, verifiable workflow system:
- Validates workflow manifests
- Executes a minimal run pipeline
- Emits an immutable
runs/<id>/record with hashes and a run manifest - Stages human gates/checklists for audit
It is not a UI, team platform, or ledger—those live in the CRI-Enterprise layer.
- Method origin: Aurora Workflow Orchestration (AWO) v1.2 — finalized under Waveframe Labs
- AWO concept DOI (authoritative): https://doi.org/10.5281/zenodo.17013612
- Role split: AWO = methodology (frozen); CRI-CORE = runtime (evolves)
See also: docs/AWO_ORIGIN.md (provenance note).
Repo uses a root CLI file (
cli.py). We’ll add a package entrypoint later if desired.
python3 cli.py validate workflows/minimal.json
python3 cli.py run workflows/minimal.json
# copy the printed run_id, then:
python3 cli.py commit <RUN_ID>What you get:
runs/<RUN_ID>/run_manifest.jsonruns/<RUN_ID>/SHA256SUMS.txtruns/<RUN_ID>/gates/{scope_checklist.md, merge_checklist.md}
Minimal JSON schema is included; strict validation and signing come next.
scope → (optional) fanout → audit → commit
- scope: load workflow; bind claims/IDs (MVP: structural checks)
- audit: basic gates/checklists; fail-closed design (MVP: staged templates)
- commit: write immutable artifacts (manifest + hashes)
See: docs/CRI_SPEC.md and docs/RUNTIME_OVERVIEW.md.
cli.py # root CLI (current entrypoint)
__init__.py # version stub (optional)
docs/
CRI_SPEC.md # MVP spec (fill as we evolve)
figures/
Demo/ # placeholder assets
schemas/
run_manifest.schema.json
workflow.schema.json
templates/
gates/
merge_checklist.md
scope_checklist.md
workflows/
minimal.json # minimal runnable example
cri-validate.yml # CI
release*.yml # release workflows
Runtime artifacts (local):
runs/(git-ignored).
v0.2.0 — AWO enforcement (strict)
- JSON Schema validation in CLI (
validate) - Tighten run manifest fields and failure semantics
v0.3.0 — Plugin manager (alpha)
- Declarative
plugins_active.yaml - Load/execute a simple metric plugin (e.g., compression/entropy proxy)
v0.4.0 — Attestation & evidence
- Signing pipeline (cosign/OIDC) optional
- Evidence registry draft (
evidence.yaml)
v0.5.0 — Enterprise scaffold
- UI dashboard (read-only)
- Auth hooks & provenance ledger adapter (file → API)
- Keep CRI-CORE strictly runtime; method changes belong in AWO (frozen).
- README and CI must agree on the same CLI invocation.
- Prefer small, auditable PRs: docs → schema → CLI → plugins.
- Code: Apache-2.0 (
LICENSE) - Enterprise: see
LICENSE-ENTERPRISE.md - Cite:
CITATION.cff
- AWO (method, finalized): https://github.com/Waveframe-Labs/Aurora-Workflow-Orchestration
- AWO Concept DOI: https://doi.org/10.5281/zenodo.17013612
© 2025 Waveframe Labs · Independent Open-Science Research Entity