Ludus is an AI-native decision operating system that turns complex questions into traceable evidence, structured analysis, causal simulations, and human-signed, reviewable decisions.
- Development phase: Gate 0 validated; Task 2/19A implementation pending
- Repository visibility: Public
- License: PolyForm Noncommercial 1.0.0 (Source Available; noncommercial use permitted, commercial use requires a written license)
- Capacity clock: Not started. Automated Gate 0 passed; the 6-agent/72-hour clock remains conditional on accepting the proven system-Chrome Playwright fallback or acquiring bundled Chromium.
- Production readiness: Not ready
See LICENSING.md and COPYRIGHT before copying, distributing, relicensing, or publishing any part of this repository. This repository is licensed under the PolyForm Noncommercial License 1.0.0: noncommercial use (learning, research, teaching, evaluation) is permitted; commercial use requires a separate written license from the rights holder.
The canonical product, architecture, contract, execution, and acceptance plan is now versioned inside this repository. Visual reference and disposable validation artifacts remain external workspace inputs:
| Purpose | Path |
|---|---|
| Product, architecture, contract, execution, and acceptance plan | docs/product-plan |
| Final static frontend design reference | ../look |
| Optional Gate 0 validation slice | ../decision-lab-G0 |
| Final implementation repository | . |
../look is a static design reference, not a runtime dependency. Production code must not load its app.js, and implementation changes must land in this repository rather than modifying the reference directory.
Gate 0 establishes a reproducible monorepo baseline and verifies, without exposing credentials:
- Python 3.12 with
uv; - Node.js 22 and pnpm;
- Docker daemon, Docker Compose, and PostgreSQL 16;
- canonical contract validation and OpenAPI-to-TypeScript drift checks;
- Ways
hardtech-market-direction/1.1.0and fixture boundaries; - Look V7 source snapshot integrity and frontend build;
- secure environment configuration;
- real provider probes for text, thinking/reasoning, strict tool calls, and structured output;
- the selected 3/4/6-agent capacity profile.
Offline bootstrap may continue while external services or credentials are unavailable, but it must not report PREFLIGHT_OK or start the capacity clock.
apps/web/ Next.js frontend (Task 1W)
packages/contracts/ Generated OpenAPI/TypeScript contract package
services/api/ FastAPI service
scripts/ Preflight, provider probe, contract generation, validation, QA teardown
ways/ Versioned decision-method packages (noncommercial core by default)
HEAD Current work record
HISTORY Append-only completed/prior work history
Do not install global tools, install project dependencies, create .venv, start persistent containers, or install browser runtimes without explicit product-owner approval. Secrets belong only in ignored local environment files or the host secret store; never place them in Git, logs, HEAD, or HISTORY.
QA/gate runs that start a compose stack must use a run-scoped project name (docker compose -p ludus-qa-<run-id> ... up -d) and must tear that stack down at the end of the run, pass or fail, via scripts/qa_teardown.ps1 -Project ludus-qa-<run-id>. Run scripts/qa_teardown.ps1 -Inventory first (read-only docker ps) and confirm with the product owner before any cleanup; stopping or removing containers not started by the current run requires separate authorization. A complete gate run must leave no QA containers behind in docker ps.
HEAD describes the current work. Before starting a different work item and when completing the current item, append the existing HEAD record to HISTORY; then write the new active or idle state to HEAD. Existing HISTORY entries are append-only.