Community · E2 · artifact verified
Route messages to specialists with one Jev call
Two Python apps on the official typesafe-sdk: a router sends each message to every specialist it needs in one Jev call - per-agent relevance Nouls plus urgency escalation to a human - and a triage answering three questions per failed pipeline task.
01 · Role in the system
What Jev does here
The router keeps multi-agent dispatch honest: one Jev call answers a relevance Noul per specialist agent (whose descriptions Jev sees) and an urgency Noul, so an agent runs only above the 0.5 relevance threshold and a human escalates at 0.7 urgency - thresholds as named constants, not vibes. The sibling triage app asks three questions per failed data-pipeline task in one call and code maps the calibrated answers to an action. Both apps read TYPESAFE_API_KEY from the environment, never store it, and the repo is a compact showcase of routing and triage on the official SDK.
02 · Control boundary
Where Jev sits
Jev as the relevance and urgency judge of multi-agent routing: one call scores every specialist per message, thresholds decide which run and when a human steps in, and the same calibrated pattern powers pipeline-failure triage.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- Small demo scope: two apps, thresholds hand-picked.
- 0 stars; created 2026-10-11, single-maintainer.
- Specialist agents are stubs that report their scope; real work is left to the caller.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- mani-bellan ↗Community · github · public · checked 2026-10-12