TrainSage

A native iOS strength coach built around a hybrid planning model: deterministic training rules own structure and safety, while provider-agnostic AI personalizes exercise selection, loading, sequencing, and coaching notes.

Built FitnessCore as a tested Swift package for domain models, exercise cataloguing, volume landmarks, plan generation, and response validation, with a deterministic rule-engine fallback when an AI response is missing or unsafe.

TrainSage adapts the week as training happens: daily energy and time check-ins, exercise swaps for occupied equipment, logged weights and feedback, and proactive reminders and coach insights.

Kept the product private and provider-agnostic with SwiftData persistence, Keychain-stored provider keys, pluggable OpenAI-compatible, Gemini, and Apple on-device providers, plus a per-call AI cost ledger.