ROUTEAgents & automation
AgentOS Pilot Router Jev Strategy
AgentOS@useAgentOS
The AgentOS Pilot Router now includes Jev, a routing strategy by TypeSafe AI, which processes state and typed questions to return decisions with probability distributions and confidence scores.
TakeYou send it state plus typed questions (choice / score / noul) and it returns a decision with a calibrated probability distribution and a confidence score.
Pattern✦ Software function call decisions
1/3 AgentOS Pilot Router now has a third routing strategy: Jev, by TypeSafe AI.
Jev is @typesafeai "System One" model. It doesn't generate text. You send it state plus typed questions (choice / score / noul) and it returns a decision with a calibrated probability distribution and a confidence score. One request, ~0.9 s, $0.042 per million input tokens.
---
Quick context. AgentOS is an open source Local-First Agentic Trading AI Agent, Token-Efficient by Design. Pilot Router is the layer that picks a model tier for every turn, c0 for trivial chat up to c3 for deep work, so you don't pay frontier prices for "thanks".
It had two strategies: a local MiniLM+ONNX classifier (default, nothing leaves your machine) and an LLM judge. With Jev, Pilot Router is now more powerful and accurate.
Related builds
