devwithjevClassifies Menu Dishes with Jev
Florian tests TypeSafe’s jev-1.13 on 3,373 menu dishes to identify their category, achieving 84.5% agreement. The run takes 9.7 seconds and costs $0.11.
Florian | FlowGrapher@Fl0𝕏
Test de Jev (TypeSafe, jev-1.13) sur 3 373 plats à la carte. Objectif : trouver le type (viande, pâtes, pizza, etc.) Résultat : 9,7 s, 2,56 M tokens in, 0,11 $, 84,5 % d’accord (406 désaccords, 116 incertains) Deux ratés parlants : Tagliata → pâtes Fondue vigneronne → fromage Ça va très vite et ça reste typé. Ça n'hallucine pas un label hors liste. Les ratés viennent des noms proches et des mots p
The test records 406 disagreements and 116 uncertain cases. Examples include Tagliata classified as pasta and fondue vigneronne as cheese; the author says the model does not hallucinate labels outside the list.
Also filed under Triage & routing
- Jev-Assisted Codex Model and Effort Routing
The project provides Jev-assisted model and reasoning-effort routing for local Codex Desktop and CLI.
- Plain-English Gmail Filter Judge
The project uses Jev (TypeSafe) to judge Gmail filters written in plain English.
- Framework-Agnostic Jev Intent and Tool Router
Huzeyfe_dev built a lightweight, framework-agnostic intent and tool router using TypeSafe’s Jev decision model to speed up agent loops.
- MCP Tool Router with Jev and MCTS
Harness-router is a lightweight tool router that uses Jev for fast tool selection and Monte Carlo Tree Search for harder multi-step decisions. It uses MCP as the tool interface and falls back when routing is uncertain.