Jev playbook
This playbook is jev.pro’s operate pillar: checklists you can run in a review, not a story-shaped “how to build with System One” essay. TypeSafe’s official narrative remains how to build. We are independent and unofficial. We do not issue API keys.
This page is for builders crossing from curious to production-minded Jev use. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Confirm facts on docs.typesafe.ai.
Open-cluster pages are independent field-guide notes. Replicas and third-party interfaces mentioned anywhere on jev.pro are not Jev and not endorsed. This is not a TypeSafe playbook.
Path
| Stage | Page | Exit gate |
|---|---|---|
| Design | Design checklist | Action space + atomic questions written |
| Implement | Implementation checklist | Pinned model, traces, error map |
| Launch | Launch checklist | Harness + fail-closed + labeled claims |
| Compose | Intent routing, guardrails, RAG filter | Code owns side effects |
| Compare | vs classifiers, vs agents | You know when not to use Jev |
If you already have a mega-prompt or a frozen classifier, insert the migration guide between design and implement: dual-run actions in shadow mode before you let Jev auto-act. Do not big-bang irreversible side effects.
Design starts with an action space that includes abstain. Implement persists a decision trace. Scope test: when to use Jev. Named architectures that fight System One: anti-patterns.
Operate loop (weekly)
- Re-fetch the channel matrix (model ids, promos, 32k vs 64k cards).
- Diff
response.modelin logs — catch silent alias moves if someone shippedjev-latest. - Sample handoff traces; add misses to the gold file.
- Re-read TypeSafe jaggedness if you changed criteria.
First-party evaluation is still POST /v1/systemone (and GET /v1/models). Do not invent extra public evaluation routes. Compose atomic questions in code; one POST, many questions. Choice is unordered (max 255; add other / none). Score is 2–10 ordered levels — keep arithmetic in code; do not treat score as dollars. Noul is P(yes); 0.5 is uncertain, not medium. Confidence is peakedness of Choice/Score; Noul has none.
Production gates we will not skip (SEO-GEO)
- English public pages; unofficial disclaimer; outbound https://docs.typesafe.ai
- One intent per URL
- Vendor 40×–200× / $0.042 figures labeled
- No key CTAs
- Cover rival intents, do not clone their slugs or voice
Those gates are this site’s editorial bar. Fork them into your wiki if they help; do not ship jev.pro copy as if TypeSafe signed it.
What this playbook is not
Not official TypeSafe. Not a paid “Pro” tier. “Pro” here means production craft (see NARRATIVE.md in the repo). Not a substitute for docs.typesafe.ai.
Hub: Open. Official: docs.typesafe.ai.
FAQ
Is this TypeSafe’s official playbook?
No. Official how-to-build stays on their docs. This is an independent operate path.
Where does migration fit?
After you can write the action space, before you flip auto-act. See the migration guide.
What is the weekly loop for?
Catch alias drift, refresh door facts, and grow gold from real handoffs — not to invent metrics.
Disclaimer
This is an independent unofficial site and is not affiliated with TypeSafe AI; official documentation is available at https://docs.typesafe.ai. Never treat jev.pro as TypeSafe official documentation. We do not sell, issue, or proxy API keys.