ROBOMAVS

Inspiration We were tired of choosing between staying yoked (hitting our daily protein goals) and staying solvent (not breaking the bank on college meals). Most macro trackers require tedious manual input, and none of them address the core problem: How do I find a cheap, high-protein meal near me right now, and how do I know the macros of the food I just put on my plate?

Inspired by the visual chaos and multidimensional precision of Spider-Man: Across the Spider-Verse, we built the Yoke Coach: an AI that crosses the "nutrition dimension" to deliver highly personalized, hyper-efficient, and budget-conscious coaching—all with a motivational voice.

What It Does The Yoke Coach is a comprehensive macro-tracking and meal-planning PWA that uses multimodal AI to automate goal-setting, food identification, and local meal finding.

AI Goal Setting (Personalization): Users input their profile (Height, Weight, Goal, Activity Level). The app uses the Gemini API to act as a certified nutritionist, calculating a precise, personalized daily Protein, Carb, Fat, and Calorie goal (TDEE/BMR).

Multimodal Food Logging: Users snap a photo of their meal. Gemini Vision analyzes the image, identifies ingredients, estimates portion sizes, and returns a macro breakdown in structured JSON. This logs the meal instantly and removes manual entry.

Local Macro Finder: Users enter a craving, their budget, and their ZIP code. Gemini searches for a high-protein, cost-effective meal nearby (e.g., "Grilled Chicken Bowl under $15") and provides a map link to the restaurant.

Multilingual Coaching: The entire interface is made accessible to all through a built-in Language Switcher (Español, Papi!).

Voice Feedback: The app delivers real-time, motivational status reports using the ElevenLabs API to keep the user hyped and accountable.

Secure Profile: Features are securely enabled only after a successful login via Auth0.

How We Built It The entire application runs as a Single Page Application (SPA) written exclusively in HTML, CSS (Tailwind), and JavaScript, avoiding any complex server-side Node.js or Python backend logic.

Gemini API (The Brain): We leveraged Gemini in three key ways:

Vision: Used for image analysis of food photos (analyzeMeal).

Goal Calculation: Used for complex formula processing based on personal stats (calculatePersonalGoals).

Structured Output: Used JSON Schema to force the model to return predictable, clean data (e.g., {"protein_grams": 55, "estimated_cost_usd": 12.50}), which is essential for accurate tracking.

ElevenLabs (The Voice): The API generates high-quality, motivational speech to replace boring text alerts.

Auth0 (The Security): We integrated the Auth0 JavaScript SDK to handle user login and session management, securing access to the personalized goal features.

Theme: Custom CSS was written to create the segmented, high-contrast, neon glitch aesthetic inspired by the Spider-Verse art style.

Challenges We Ran Into Auth0 Race Condition: We faced a persistent ReferenceError: createAuth0Client is not defined because the external Auth0 SDK script was not loaded before our custom JavaScript tried to call it. This required implementing a final, synchronous execution block to guarantee the load order.

Multilingual Widget Integration: The Google Translate widget often clashed with our custom dark/neon CSS, requiring extensive CSS overrides (!important tags) to ensure the dropdown selector was visible and usable.

Vision Accuracy: The biggest technical hurdle was ensuring the Gemini Vision model could accurately estimate macros based purely on an image, which required careful prompt engineering to guide the AI's nutritional reasoning.

Accomplishments That We're Proud Of Full API Integration: Successfully integrating Gemini, ElevenLabs, and Auth0 into a single, cohesive, client-side application structure.

Multimodal Utility: Creating a genuinely useful tool that leverages the power of Gemini Vision for a practical, real-world task (macro tracking).

Unique Theme: Delivering a stunning, memorable user experience with the custom Spider-Verse Glitch Theme.

Accessibility: Making the app multilingual (Spanish, French, etc.) via the Google Translate Widget, expanding its potential reach.

What We Learned We gained deep expertise in managing synchronous dependencies in client-side applications and how to effectively use JSON Schema to bridge the gap between large language models and structured application data. We learned that for image analysis, careful constraints and explicit output formatting are key to reliable results.

What's Next for ROBOMAVS Data Persistence (Firebase/Firestore): Move user logs and goals from temporary memory (browser session) to a secured database (like Firebase) to save history permanently across sessions.

Workout Vision: Integrate Gemini Vision to analyze workout videos or photos to count reps or critique form.

Goal Optimization: Use the macro history to generate dynamic, automated adjustments to the user's protein goals over time.