Jev AI use cases
Routing & decisions
Send each request to the right model, handler or tool, and know when to hand it to a person.
Routing & decisions
Send each request to the right model, handler or tool, and know when to hand it to a person.
About routing & decisions
Routing decides where work goes before anyone does it: which model tier, which specialist handler, which tool, or which person. A good router saves cost on easy requests and protects quality on hard ones.
Jev returns the routing decision as a typed choice with a probability for every route and a confidence value. Your application then forwards the request. Because routes are written descriptions, changing policy means editing text, not retraining.
Which page fits your task
- LLM RouterYou want to route requests between model tiers, support handlers or agent tools, with a fallback for uncertain cases.
Common mistakes
- Overlapping route descriptions, which split probability between two destinations.
- Having no default route for low-confidence decisions.
- Embedding prices or quotas in route text, where they go stale.
Routing & decisions FAQ
Does Jev send the request to the chosen model?
No. Jev returns the decision; your application or gateway makes the call. Provider keys and traffic stay in your own stack.
How many routes can I use?
The playground allows up to 12 options per choice. The API accepts many more, but keep routes distinct: fewer, clearer routes are easier to measure and tune.
Further reading · reviewed 2026-09-23
- Intent routing to the right handlerTypeSafe documentation · docs.typesafe.ai/patterns/intent-routing
- Confidence-gated routingTypeSafe documentation · docs.typesafe.ai/patterns/confidence-routing
One decision model. Your own rules.
Use a Noul for a yes/no probability, a Choice for a fixed set of outcomes, or a Score for an ordered rubric. Ask focused questions about the same input and combine the results in code.