Symfony app that streams a public French support-request dataset into SQLite and qualifies each request in background batches: Choice for intent, Score for urgency, Noul for whether it is a bug to forward.
yoanbernabeu/demo-symfony-typesafe
User repository on GitHub
Démo : trier des demandes de support avec Jev (TypeSafe) et Symfony AI. Messenger, Live Components, Turbo, kit shadcn de UX Toolkit.
Repo stats from the GitHub API, cached Oct 2026.
Une application Symfony qui lit de vraies demandes adressées aux services publics français et laisse Jev, le modèle « System One » de TypeSafe, les qualifier : ce que veut la personne, à quel point c'est urgent, et s'il s'agit d'un bug à transmettre aux développeurs. Jev ne génère pas de texte. On lui donne un contenu et des questions typées, il répond par des probabilités, en quelques centaines de millisecondes. La démo utilise ses trois types de questions, posées ensemble, en un seul appel par demande :
Quoted from the project's README. Read the full README.
Watch out
The author reports 1000 requests qualified in 25 seconds at 428 ms mean latency for roughly 820,000 tokens, measured on one laptop, so treat those as an order of magnitude rather than a promise. app:triage:limit defaults to 1000 requests sent to Jev to cap the bill and 0 removes the cap, with since date, batch size and requests-per-minute as the other three knobs. The dataset is about 260 MB and is not in the repository; app:dataset:download fetches it from data.gouv.fr under Licence Ouverte 2.0, and only the request text is kept. These are real testimonies and the README says so, sometimes describing real distress. Requires PHP 8.4 with pdo_sqlite, ctype and iconv, and the Symfony CLI is recommended but optional. Copy reviewed against the repository README on 2026-10-04, not from running it.
iamaamir/system-oneTypeScript
wundercorp/lokiPython
doronp/jevcTypeScript
3clyp50/a0-typesafe-aiPython
typakon4/jev-layerJavaScriptPosts from builders shipping with Jev right now.
