Wheel Strat

Agentic options mentor powered by Gemini 3. Wheel Strat scans markets, explains trades grounded with real-time search, and backtests strategies on decades of historical data all from your pocket.

Comment

What It Does

Wheel Strat is a companion app for options traders with two core Gemini 3 integrations:

🧠 Strategy Explainer (On-Demand)

Tap any trade to get instant AI analysis. Gemini 3 evaluates the contract's Greeks, IV rank, and market context to explain the Goal, Risk, and Outcome in plain English. Perfect for learning or validating your thesis.

🏃 Marathon Agent (Scheduled)

An autonomous agent that scans the market at 9:30am and 3:30pm EST. It uses Gemini 3 to identify high-conviction opportunities, analyzes them against your risk profile, and pushes actionable alerts to your phone.

Additional Features

  • Portfolio Dashboard — Real-time Net Liquidity, Greeks, and P&L tracking
  • Scenario Backtesting — Historical win probability using 20 years of price data
  • Paper Trading — Execute against IBKR paper accounts with full reconciliation
  • Demo Mode — Pre-seeded "Mag7" portfolio

How We Built It

Gemini 3 Integration

Strategy Explainer Flow:

User taps "Explain" → Cloud Function → Gemini 3 Flash → Structured JSON → UI Modal
  • Prompts include full trade context: symbol, strike, DTE, Greeks, IV rank
  • Gemini 3's structured output ensures reliable parsing for UI rendering
  • Sub-second latency enables real-time interaction

Marathon Agent Flow:

Cloud Scheduler (9:30am/3:30pm) → Agent Function → IBKR Option Chains → Gemini 3 Analysis → Firestore → Push Notification
  • Agent fetches live option chains from IBKR bridge
  • Gemini 3 evaluates each contract against Wheel criteria (delta -0.25 to -0.35, IV Rank >50%, 30-45 DTE)
  • Results written to Firestore, synced to device via TinyBase

Architecture Highlights

  • Local-First: TinyBase + SQLite ensures the app works offline and syncs seamlessly
  • Production IBKR Bridge: Python Flask server on GCP connects to real brokerage data
  • Glassmorphic UI: Custom design system with 60fps Reanimated 4 animations

Challenges We Ran Into

1. IBKR Bridge Reliability (Infrastructure)

Running IB Gateway (Java) on GCP's free-tier e2-micro (1GB RAM) required extensive JVM tuning:

  • Reduced heap to 256MB with SerialGC
  • Added 1GB swap file for memory pressure spikes
  • Implemented clean crash-on-OOM for Docker restart recovery

2. Prompt Engineering for Finance (AI)

Balancing concise responses with educational depth for complex financial topics. We iterated on prompts to ensure Gemini 3 explains why a trade is risky, not just that it's risky.

3. Offline-First Sync (Mobile)

Ensuring TinyBase ↔ Firestore sync remained conflict-free across sessions, especially when trades execute while the app is backgrounded.

Accomplishments We're Proud Of

🏆 Production-Ready AI Integration

Both Gemini 3 features work end-to-end in production:

  • Strategy Explainer analyzes real trades in <1 second
  • Marathon Agent has generated 50+ daily reports with actionable opportunities

🎨 Premium Mobile Experience

  • Glassmorphic UI with consistent 60fps animations
  • Interactive Skia charts for price/volatility visualization
  • Haptic feedback on all user interactions

📱 TestFlight Deployment

The app is publicly available today: https://testflight.apple.com/join/vK1pNwbs

📊 20-Year Historical Backtesting

Pattern matching engine analyzes GOOGL data back to 2005 to calculate asymmetric win probabilities.

What We Learned

Gemini 3's Structured Output is a Game-Changer

The ability to request reliable JSON extraction means we can confidently render AI responses in production UI without parsing failures.

Local-First is Worth the Investment

TinyBase's reactive architecture made the app feel instantaneous. Users never wait for network—data is always there.

Financial AI Requires Guardrails

Options trading involves real money. We implemented multiple safety layers: paper trading only, confirmation modals, and clear risk disclosure.

What's Next for Wheel Strat

Phase 1: Expanded Universe

Support for additional tickers beyond the Mag7 focus for Pro plan.

Phase 2: Live Account Trading

Graduate from paper to real execution with OAuth broker integration (IBKR, Public.com).

Phase 3: Agentic Trading Mode

Agentic auto-suggest bot trading mode for Diamond plan.

Built With

Languages & Frameworks

  • TypeScript — Primary language for app and Cloud Functions
  • Python — IBKR Bridge server (Flask + ib_insync)
  • React Native (Expo SDK 54) — Cross-platform mobile app

AI & Cloud

  • Gemini 3 Flash — Core reasoning engine via Google AI Studio
  • Firebase — Cloud Functions (Gen 2), Firestore, Cloud Scheduler
  • Google Cloud Platform — Compute Engine for IBKR Bridge

Data & State

  • TinyBase 7.0 — Offline-first reactive store
  • SQLite — Local persistence via expo-sqlite
  • Cloud SQL (PostgreSQL) — 20-year historical market data

Visualization & UX

  • Victory Native (Skia) — Hardware-accelerated charts
  • React Native Reanimated 4 — 60fps animations
  • Expo Haptics — Tactile feedback

Infrastructure

  • Docker + Caddy — Containerized bridge with TLS termination
  • ngrok — Secure tunnel for local development
  • Interactive Brokers TWS API — Real-time market data and execution

Built With

Share this project:

Updates