Research preview — not a validated paper-positioning service. Real-manuscript diagnostics and paid use remain unavailable until Gate 1 is completed.
This directory contains the first executable Gate 0 evidence for the open Workbench. It intentionally contains no private protocol, venue data, prompts, benchmark answers, or user manuscripts.
See PUBLIC_RESEARCH_PREVIEW.md for the exact public status, limits, and offline synthetic demo.
Included in v1:
- JSON Schema 2020-12 contracts for task, result, failure, cancellation, timeout, task event, and settlement event payloads;
- synthetic positive and negative fixtures;
- a server-controlled diagnostic state machine;
- an idempotent one-authorization/one-capture settlement ledger;
- contract, transition, replay, over-capture, and safe-failure tests.
Run the evidence suite:
python3 -m venv .venv
.venv/bin/python -m pip install 'jsonschema>=4.23,<5' 'pypdf>=5,<7'
.venv/bin/python -m unittest discover -s tests -vRun the entirely synthetic, offline product-shape preview:
.venv/bin/python scripts/run_public_preview.pyThe parser isolation tests require macOS sandbox-exec and must run outside an
already nested application sandbox. A Linux CI run must replace this policy
with an isolated runner or container that also enforces a hard memory limit.
The contracts are frozen as contract_version: 1.0. A breaking change requires
a new contract version; it must not overwrite v1 or prior run evidence.
Gate 1 currently needs one independent manual paper-positioning operator. See MANUAL_OPERATOR_INTEREST.md for the public eligibility and conflict rules. The private protocol, evaluation cases, prompts, profiles, thresholds, calibration records, and judge materials are not part of this repository or the public qualification intake.
The public Workbench software and documentation are licensed under the Apache License 2.0. Private protocols, prompts, profiles, benchmark materials, calibration assets, and user manuscripts are not included and are not licensed by this repository.