Glossary: Ranked route
Ranked route is an ordered list of handlers produced by sorting a Choice probabilities map (or several Noul values), then applying your confidence floor before you execute rank 1. Jev (TypeSafe’s System One decision model) returns a distribution on POST /v1/systemone; it does not return a travel itinerary or a next-hop object. Rank is something your code computes. Rank 2 is a fallback only if policy says so.
This glossary entry is for builders who already heard of Jev / System One and need a citable definition before they wire model routing. jev.pro is an independent unofficial field guide — not affiliated with TypeSafe AI. Confirm field names against docs.typesafe.ai. We do not sell or issue API keys.
How it shows up
Put type: "choice" on an intent question, or score candidate models with independent Nouls. The official intent-routing pattern uses a Choice plus optional Score complexity. You still own the sort, the floor, and the side effect. A typical production loop: (1) read probabilities, (2) sort descending, (3) if confidence is below your confidence threshold, abstain instead of taking rank 1, (4) else dispatch the mapped handler.
Low confidence on rank 1 should usually mean a cheaper model, a deterministic default, or a human — not a silent second guess that pretends rank 2 was “almost as good.” Close races (two options near 0.45 / 0.40) are why you log the full map, not only choice. Question IDs are for your code; TypeSafe says they are not used in inference.
Pin jev-1.13.0 (or the current versioned id) once you fit floors. Changing jev-latest under a fitted rank policy is a silent policy change.
Contrast
Ranked route vs intent. Intent is the closed label you asked for (“billing”). A ranked route is the ordered handler list you derived after the answer arrived.
Ranked route vs fallback. Fallback is the safe path when Jev is unsure, rate-limited, or down. Rank 2 can be that path, but HTTP 429/529 are not “use rank 2.” Official errors are 401, 422, 429, and 529 — none of those mean “try the runner-up.”
Ranked route vs action space. The action space is the finite set of side effects you allow. Ranking does not enlarge that set. Jev cannot invent page_ceo if you never mapped it.
Ranked route vs argmax. choice is already the argmax. Ranking without a confidence floor is just executing the winner. The useful extra work is the floor and the documented runner-up policy.
Common mistakes
- Always executing rank 2 on a miss. Runner-up mass is not permission.
- Treating rank as a TypeSafe reserved word or a response field. It is unofficial.
- Copying a Noul τ onto a Choice sort. Noul has no
confidence; Choice uses peakedness. - Asking Jev to “return the routing plan as JSON.” That fights the product contract. Keep the plan in code.
- Forgetting that schema-safe ≠ factually correct: a valid winner can still be the wrong team.
See also
FAQ
Is ranked route a TypeSafe reserved word?
No. Unofficial working term. Building blocks are Choice + confidence (or several Nouls) plus your sort.
Should I always execute rank 2 on a miss?
Only if the second handler is safe. Do not treat runner-up mass as permission.
Can I rank Score levels?
You can sort legend by probabilities, but Score is an ordered rubric. Threshold “past level 1” is usually clearer than inventing a route from mid-rubric mass.
Does Jev return rank indexes?
No. You compute order from probabilities. Log the map and the branch you took.
Limits
This is a jev.pro working definition unless TypeSafe uses the same word as an API field. We do not invent search volume. Schema-safe answers are not automatically factually correct.
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.