python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install -e .
python -m unittest discover -s tests -v
python -m compileall -q src tests examples
node --check examples\refund_gate.mjsExpected: 21 tests pass, compilation is silent, and the Node syntax check exits 0.
Run LoopGrid Core 0.8.1-design-partner (or a compatible REST deployment) and use a workspace service key with ingest,read. Add review only if a live Jev result routes to human review and you choose to exercise the sandbox review action.
$env:LOOPGRID_BASE_URL="http://127.0.0.1:8000"
$env:LOOPGRID_WORKSPACE_ID="default"
$env:LOOPGRID_API_KEY="<workspace service key>"
$env:JEV_PROVIDER="vercel"
$env:AI_GATEWAY_API_KEY="<secret>"
python examples\refund_gate.pyThe v0.1.0 release gate was satisfied on 2026-09-25 with the exact source being published. The live run:
- received a real
typesafe-ai/jevresponse through Vercel AI Gateway; - resolved the final provider as
typesafe-aiwith provider HTTP 200; - recorded
model.provider = typesafe-aiandcontext.jev.access_path = vercel; - preserved Jev answers, usage/provider metadata and
evaluation_sha256; - wrote native
model_completedandpolicy_evaluatedevents; - recorded delegated authority;
- recorded sandbox
tool_executedandoutcome_observedon the auto route; - reached
evidence_completewith 100% evidence coverage; - finished with
valid: trueand no verification failures; and - created a new decision rather than replaying the earlier unrelated decision.
Do not print or commit provider or LoopGrid API keys.
- Push this exact source to
loopgridio/loopgrid-jev. - Confirm GitHub Actions is green.
- Tag
v0.1.0and create a GitHub Release. - Deploy the prepared
/integrations/jev/website page. - Record the 35–45 second live Jev → policy → sandbox action/outcome → Verify demo.
- Publish launch copy and contact TypeSafe/Vercel as a community interoperability example.
Do not claim official partnership, certification, or endorsement unless separately confirmed.