AI Trainer

Unified iOS app + Python backend for workout videos, local gyms, and an AI coach.

Repository Layout

• AITrainer/: iOS app (SwiftUI)
• agent/: unified FastAPI backend (videos + gyms + coach) and agent logic
• data/: SQLite data used by the agent
• sources/: RAG source documents

Requirements

• Python 3.12+
• Xcode (for iOS app)

Environment Variables

Create a .env in the repo root (or agent/.env). See env.example.

Required:

• OPENAI_API_KEY (AI coach)
• GEMINI_API_KEY (meal photo scan)
• GEMINI_MODEL (optional override; auto-detected if empty)
• YOUTUBE_API_KEY
• GOOGLE_PLACES_API_KEY
• TAVILY_API_KEY (online recipe search)

Optional:

• REDIS_URL or UPSTASH_REDIS_REST_URL + UPSTASH_REDIS_REST_TOKEN

Run the Backend

From the repo root:

cd agent
./start.sh

Backend will be available at:

• http://localhost:8000/health
• http://localhost:8000/docs

Run the Agent CLI

From the repo root:

python -m agent.cli

iOS App

Open AITrainer.xcodeproj in Xcode and run on a simulator.

The app calls the backend here:

• AITrainer/Models/Services/YouTubeService.swift
• AITrainer/Models/Services/GymFinderService.swift
• AITrainer/Models/Services/AICoachService.swift
• AITrainer/Models/Services/FoodScanService.swift (meal photo analysis)
• AITrainer/Models/Services/RecipeImageService.swift (AI recipe images)

If you change the backend host, update the baseURL in those files. Backend host is centrally configured via Config/Info.plist (BackendBaseURL) and AITrainer/Models/Services/BackendConfig.swift.

Notes

• The AI coach endpoints are served by the agent logic via FastAPI:
  • POST /coach/chat
  • POST /coach/feedback
• RAG index is loaded from data/faiss_index when present.