scalinity.bio

Production-ready macOS SwiftUI app + Python ML pipeline to:

• Simulate a public-data-backed synthetic multiomic cohort (1,000+ individuals)
• Train a biological-age regressor and export to CoreML
• Run on-device inference and generate evidence-backed anti-aging regimens (RAG + OpenRouter)

Requirements

• macOS: 13+ (Ventura)
• Xcode: 15+ recommended (Swift 5.9)
• Python: 3.11+ recommended for the ML pipeline

Environment variables

Cursor blocks .env* files in this workspace, so use ENV.example as your template and export vars in your shell (or set in Xcode scheme env).

Required:

export OPENROUTER_API_KEY="..."
export ENTREZ_API_KEY="..."
export DATABASE_ENCRYPTION_KEY="..."   # 32+ chars; see App/Services/EnvironmentConfig.swift

Quick start (ML)

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Optional realism extras (BioPython + RDKit):
# pip install -r requirements-optional.txt

# Generate a 1k+ synthetic cohort:
python -m ML.data_simulation --out Data/synthetic_cohort.csv --n 1200

# Train + export CoreML:
python -m ML.model_training --data Data/synthetic_cohort.csv --outdir ML/artifacts
python -m ML.coreml_export --artifacts ML/artifacts --out Resources/Models/BioAgeRegressor.mlmodeldata --target bio_age_delta

Tests

# Python
source .venv/bin/activate
pytest -q

# Swift
swift test

Quick start (macOS app)

Open the repo in Xcode (File → Open… → select the folder). Run the executable product scalinity.bio.

The app reads:

• Data/synthetic_cohort.csv for reference distributions and KNN imputation
• Resources/FeatureStats.json and Resources/ModelSchema.json for preprocessing
• Resources/Models/BioAgeRegressor.mlmodeldata for inference (compiled at runtime)

Privacy model (high level)

• All stored data is written to a local SQLite DB, with payload encryption (AES-GCM via CryptoKit).
• Any identifiers are hashed (SHA-256 with per-install salt).
• Any analytics used for retraining requires explicit consent (see Settings → Privacy).

Weekly retraining (optional)

There are two supported options:

• Manual: run:

./ML/retrain_weekly.sh

• Scheduled (launchd):
  • Edit [ML/launchd/com.scalinitybio.retrain.plist](ML/launchd/com.scalinitybio.retrain.plist) and replace the placeholder path with your absolute repo path.
  • Load it:

launchctl bootstrap gui/$(id -u) ML/launchd/com.scalinitybio.retrain.plist
launchctl list | grep com.scalinitybio.retrain