Labs / Fleet governance
Dispatch resolver
playground.
Paste a fictional task brief and see which illustrative crew dispatch profile Jev would pick, with calibrated probabilities and confidence gates - a public mirror of confidence-gated automation without private fleet data.
Task brief
How it works
[Your brief]
|
v
+-------------+
| Jev choice |
+-------------+
one choice question (rules + default)
|
v
+----------------------+
| Confidence floors |
+----------------------+
global 0.6 and per-rule minimums
|
+--> clear profile or escalate
The model never sees quota, secrets, or real fleet paths. Deterministic code applies thresholds after the single choice answer returns.
Illustrative rules
- rule_1 - The task is a small copy or styling change on the public marketing site with no new routes or data stores.
- rule_2 - The task adds a new authenticated workflow, persistence, or payment path that needs careful review.
- rule_3 - The task is exploratory research or a written report with no production deploy.
- rule_4 - The task is a focused bug fix with a clear reproduction and an existing test suite.
- default - No listed rule applies to this task.