laguagu/jev-skills
Practical Jev skills, decision examples, and a small guide to useful integrations.
Helps build applications with TypeSafe Jev by choosing decision patterns, wording and debugging typed questions, setting up API credentials, selecting SDKs or integrations, and designing fallbacks and evaluations. Use when adding Jev routing, ranking, classification, tool selection, evidence checks, or context selection, when a Jev question answers wrong or with low confidence, or when finding Jev skills and example projects. Complements the official typesafe-ai skill, which carries current API detail; measuring a finished workflow belongs to the jev-evidence-eval skill.
Measures a TypeSafe Jev evidence or classification workflow for accuracy on accepted cases, review rate and coverage, abstention, latency, usage and cost, with source receipts, missing evidence and contradictions in the test set. Use when checking whether a Jev decision is supported by supplied text, choosing or sweeping a confidence threshold, or comparing a Jev workflow against an existing classifier or LLM.