Lightweight AI Decision Router

Offload task routing, classification, and risk checks from large models to a tiny engine

Problem

People use GPT/Claude for task routing, content classification, and risk judgment, but it's slow and expensive. A V2EX share highlights TypeSafe Jev, a model built for structured decisions that returns option scores and probabilities (e.g., PASS 0.92 / FAIL 0.06) at up to 250K tokens/s, showing this niche is real and general LLMs are overkill.

Solution

A lightweight decision-routing API or local library where callers define options and rules, and it returns scores and probabilities, with built-in templates for agent routing, content moderation, and ticket classification. No large model required; fast and cheap via small classifiers or rule engines.

Users

Indie developers and small teams building AI agents and automation workflows.

MVP

API for defining options and scores; built-in routing/classification/moderation templates; REST and local library; basic dashboard; free trial credits.

Revenue

Usage-based API pricing plus self-hosted license.

Why now

Agent and workflow automation is exploding; using large LLMs for simple decisions is slow and costly; niche small models and lightweight engines have an opening.