Read-only Hugging Face mirror

Build workflow Cases that can be reset, run, and scored.

A local architecture preview for producing traceable workflow environments with known negative and positive states, fresh Runs, and objective Scores.

Release / status: 0.2.1 technical preview
Source snapshot: f0df52096bd0
Runtime: Static architecture tour; the FastAPI/SQLite/Docker product stays local

What the project owns

The Factory owns Case generation, reset, isolated Runs, and scoring. It shares only versioned RunCase files with other products.

Binds source, baseline, known-correct state, transformation, and positive/negative gates.

Each Run receives a new workspace, isolated Git state, and simulator snapshot.

Agent execution status remains distinct from verifier and task outcomes.

The sequence is explanatory. Install and execute the actual project locally from its canonical repository.

Select exact baseline and known-correct revisions or a confirmed workflow recording.

Prove reset consistency and discriminate known wrong from known correct.

Create isolated workspace, Git state, simulator, and restricted profile.

Unsupported platform isolation fails closed as an environment error.

Freeze changes, run the verifier, and emit a versioned Score.

Synthetic or presentation-only images from the canonical repository.

This page preserves the project's published limits instead of turning engineering checks into outcome claims.

The UI, database, source repositories, and Docker execution remain local.

Hosted and native isolation gaps remain unsupported, not passing evidence.

One Score is one attempt; it is not a repeatability or model-quality claim.

Local files below are curated public mirrors. GitHub remains authoritative for development, releases, and issue history.