Inferoa

About

Inferoa is an agent framework designed for loop engineering with a focus on inference-native token optimization. The name Inferoa derives from "Infer" for its inference-native foundation, "o" representing tokenmaxxing loop engineering, and "a" for the agent harness. Centered on the vLLM ecosystem, it enables agents to co-design loop engineering alongside tokenmaxxing primitives rather than treating inference as an opaque process. Key components include prefix-cache discipline, context optimization, intelligent routing via the vLLM Semantic Router, and serving through vLLM, vLLM Omni, and RTK/CodeGraph context optimization.