Jevstiller Further Cuts Jev's Decision Costs and Latencies
Techstrong.ai, Monday, October 5th, 2026
Open source Jevstiller answers routine Jev classification calls locally in 15ms, skipping API calls with a statistical match guarantee.
Techstrong.ai reports that developer Tomer Glick released Jevstiller, an open source system that sits in front of calls to TypeSafe's Jev decision model, which returns classification labels in about 300 milliseconds.
Jevstiller classifies routine requests locally in about 15 milliseconds and skips the Jev call when it can statistically guarantee its answer will match Jev's at a preset rate, calibrating against a 'test slice' of Jev responses and testing thresholds in strict sequence.
It targets high-volume tasks such as log parsing, security alert testing, rapid-fire agentic decisions and strict SLAs.