spb/forge-studio Public
The Instruments of LLM training — a native macOS cockpit for Forge. Train language models from scratch on Apple Silicon without a terminal.
Swift 95.7%
Shell 4.3%
1# CLAUDE.md — Forge Studio23<!--4Author: Simon-Pierre Boucher5Contact: contact@spboucher.ai6-->78> **Instruction to Claude:** Every source file created in this project (Swift, scripts,9> plists, Metal, Makefiles) must begin with a header comment containing:10> `Author: Simon-Pierre Boucher — contact@spboucher.ai`1112> **Quality bar:** This app must feel like a first-party Apple pro tool — the13> "Instruments of LLM training." Every screen, every chart, every error message is14> judged against that standard. If a feature would ship half-working, it doesn't ship;15> it gets fixed. No placeholder UI, no TODO left in released code, no chart that16> jitters, no run state that can be lost.1718(Spec complète fournie par l'utilisateur — voir la conversation d'origine ; ce fichier19reprend les points de contrat. L'app est le compagnon GUI de Forge :20https://github.com/spboucher-ai/forge — datasets, config, runs, dashboard de métriques21légendaire, checkpoints/génération/éval, DMG signé/notarisé.)2223- macOS 14+, Apple Silicon, Swift 5.10+/SwiftUI/Swift Concurrency, Swift Charts.24- Build canonique : SwiftPM (pattern zyquo-term) ; l'app pilote le binaire `forge`25 en sous-processus ; métriques via `log.csv` (flush par step, voir RESEARCH.md).26- Signing/notarization : identité et profil extraits de zyquo-term (voir27 scripts/notarize.sh) — Team 3YM54G49SN, profil keychain "MacLustr-Notarize".28