Show HN: Jev routing coding tasks to Grok Build or Codex Astra

Category: ai-ml

Tags: model-routing, llm-agents, coding-agents, task-classification, prototype

Score: 4.3/10 (Innovation: 5, Technical: 4, Documentation: 5, Utility: 3)

Jev Model Router Demo is a local prototype that uses the Jev LLM to classify coding tasks into one of two execution harnesses: Grok Build for mechanical work or Codex Astra for complex or high-consequence tasks. It bundles three typed questions (route choice, complexity score, planning flag) into a single classifier call and optionally launches the chosen agent in a read-only sandbox. Interesting as a small experiment in model-level task routing, but it is a throwaway demo with zero stars and no license.

Target audience: AI agent developers and tooling engineers experimenting with LLM-based task routing

Repository: https://github.com/jcpsimmons/jev-model-router-demo · JavaScript

View on Hacker News