Momus is the coordination repository for AI-video engineering work, beginning with the Patrick prompt-engineering pilot and the carryover from the local MomusStudiofree / waoowaoo workflow.
This repository is intentionally evidence-first. It does not pretend that the local video application source code is already present here. Instead, it preserves the engineering ledger, current state, executable sprint boundaries, proof requirements, and handoff rules so a fresh agent can continue without rediscovering the work.
- Read
AGENTS.md. - Read
docs/AI_ENGINEERING_LEDGER.mdfor the canonical plan and provenance. - Read
docs/CURRENT_STATE.mdfor the live handoff point. - Pick exactly one bounded sprint from
docs/SPRINTS.md. - Review supporting evidence under
docs/evidence/. - Run
python scripts/validate_repo.pybefore handing off or merging documentation changes.
The first program is a vertical proof of the AI-video workflow for Patrick:
idea/prompt -> authenticated local app -> API configuration -> text model -> image model -> video model -> representative 9:16 shot -> generation/export proof -> reproducibility
The source ledger records a local waoowaoo instance at http://localhost:13000, work performed via OpenCode, and a test-account browser flow. Credentials and provider keys are not repository artifacts and must never be committed.
- Canonical ledger:
docs/AI_ENGINEERING_LEDGER.md - Live state:
docs/CURRENT_STATE.md - Sprint registry:
docs/SPRINTS.md - Evidence:
docs/evidence/ - Harness manifest:
harness/manifest.v1.json - Validation:
scripts/validate_repo.py
The Google Drive document named AI Engineering Ledger remains the human-readable source ledger; the Markdown ledger in this repository is the agent-executable representation.