タグ：cost-optimization

Per-turn router for agent harnesses: one 350ms Jev call picks the model tier, effort, tools and skill, behind a hard deadline with a regex fallback. Claude Agent SDK adapter included.

Shift every LLM call to the cheapest model that can handle it. Routing decided by TypeSafe Jev in ~180 ms. No training data. Policy in plain YAML. TypeScript and Python.

Find the LLM calls you should move to JEV — scan logs & code for decision-shaped calls, price the migration, calibrate decision models, generate the migration PR. Zero deps, fully local.

Use Jev instead of an LLM for RAG decisions: relevance grading, query routing, hallucination checks and LLM-as-a-judge evals. Faster, cheaper, measured first.

An LSP server whose ambient pass runs the rules a repository states in `.jev/rules/*.json`: each candidate line is sent to Jev, a classifier that answers one typed question with a value and a probability, never prose.

Should I use JEV?