What it is
Classroom demos that keep Jev on typed decisions and Python/TS thresholds on actions.
What it does: Classifies Jira tickets, pre-scans functions before Claude explains them, and routes PRs to rules / Haiku / Sonnet from Jev complexity and money-touch questions.
Why it belongs in the showcase: End-to-end System One client with real integrations and explicit cost comparison, not a stub README.
Caveats: No license file; Spanish docs; paid live APIs.
Evidence
Public TypeScript repo (5★ at intake, size 106) with shared/jev.ts posting to https://api.typesafe.ai/v1/systemone (choice, noul, score via askJev), per-demo questions.ts, thresholds with decide()/flagsFrom()/routeFor(), Claude Code skills (/triage, /lote, /pre-vuelo, /router, /comparar, /costos), and scripts/costs.ts comparing LLM-only vs Jev. README states every command hits live services with no mock mode. Requires Node 22.18+, TYPESAFE_API_KEY, Claude Code session, Jira Cloud and gh for demo 3.
Caveats
- No license file at intake
- Spanish primary docs; live demos spend real Jev and Claude tokens
- Needs Jira Cloud and GitHub CLI for the full three-demo path
- No published labelled accuracy study beyond author cost scripts
Unofficial case page. Stars / list position observed for heat; do not treat heat as a quality SLA. Review the upstream README before you depend on it.