🛰️ Fit-Live-Frame-AI: Neural Vault Edition

Official Submission for the Gemini Agent AI Challenge

Fit-Live-Frame-AI is a high-performance, biomechanical auditing agent that utilizes Gemini 1.5 Flash and a multi-tier Google Cloud Architecture to provide real-time coaching and massive-scale performance archival (7GB+).

Designed for high-fidelity fitness tracking and professional athletic audits, the system eliminates traditional browser-based synchronization limits via the proprietary Neural Tunnel protocol.

---

🌐 Live Infrastructure

• Live Deployment: fit-neural-vault.uc.r.appspot.com
• Official Repository: github.com/deepthi261/Fit-Live-Frame-AI
• Frontend: Google App Engine (Node.js 20)
• Neural Bridge: Google Cloud Functions (V4 Resumable Handshake)
• Vault: Google Cloud Storage (Bucket: fit-live-frame-ai)

---

🧠 Core Innovations: The "Neural Vault" Suite

1. 🧬 Neural Link (Gemini 1.5 Flash)

The "Brain" of the system. It processes live frames using a Universal Protocol compatible with stable v1 endpoints. It performs:

• Biomechanical Analysis: Real-time pose verification using MediaPipe pose landmarks.
• Rep-Counter Consensus: A hybrid AI/Heuristic engine that ensures 100% precision in movement tracking.
• Dietary & Coaching Insights: Context-aware metabolic recovery advice delivered via high-speed neural TTS.

2. ⚡ Infinity Protocol (Unbreakable 7GB Sync)

Traditional browser uploads fail for large video files. We implemented the Infinity Sync Protocol for multi-gigabyte archival:

• Handshake Resume: Automatically queries GCS for the last "caught" byte after a network flicker.
• Binary Streaming: Uses V4 Signed Resumable Tunnels to stream performance video (even 7GB+ .mov files) directly from the browser to the GCS Vault.
• Fault-Tolerance: Detects net::ERR_NETWORK_CHANGED and resumes without losing a single frame.

3. 🌉 Zero-CORB Bridge (Secure Sheets Proxy)

To eliminate Cross-Origin Read Blocking (CORB) errors when logging to Google Sheets/AppSheet, we built a server-to-server bridge:

• Telemetry Proxy: Routes sensitive workout data through a Cloud Function.
• Encryption-in-Transit: Ensures workout details (Reps, Calories, Precision) are verified by the bridge before reaching the final report.

---

🏗️ System Architecture

graph TD
    A[Browser / MediaPipe] -->|Neural Stream| B(Gemini 1.5 Flash)
    B -->|JSON Metadata| C{Neural Resolver}
    C -->|Stats HUD| D[User Dashboard]
    C -->|Zero-CORB Bridge| E[Cloud Function Proxy]
    E -->|Verified Sync| F[Google Sheets Dashboard]
    A -->|Infinity Protocol| G[GCS Media Vault]
    G -->|V4 Resumable Handshake| H[Cloud Function Tunnel]

Loading

---

🛠️ Technical Stack

• AI Backend: Google Generative AI (Gemini 1.5/3.0 Flash) with Gemini API key in built.
• Cloud Infrastructure: Google Cloud Platform (App Engine, Cloud Functions, GCS)
• Frontend Architecture: React 19, TypeScript, Vite
• Motion & UI: Framer Motion, Lucide React, Tailwind CSS
• Vision Layer: MediaPipe (Pose Landmarker)
• Database: Google Sheets (via Apps Script Bridge)

---

🚀 Getting Started & How to Test for Judges

To evaluate the Fit-Live-Frame-AI Agent, please follow this high-fidelity workflow:

1. Launch the Hub: Open the Live Application. The Neural Link is pre-configured and will activate automatically.
2. Neural Calibration: Grant camera permissions. You will see the Shadow Mentor V3.0 initialize on the right-side HUD.
3. Perform Movements: Stand back so your full body is visible. Perform a few Squats, Bicep Curls, or Pushups.
   • Watch the Predicted Activity HUD update in real-time as the Gemini 1.5 Flash agent classifies your movement.
   • Monitor the Precision Meter as it audits your form symmetry.
   • Listen for Voice Feedback providing metabolic and postural coaching.
4. Audit the Vault: Once finished, click "Finish Session & Submit".
   • This triggers the Infinity Protocol, handshaking with our Google Cloud Function to vault your performance data securely to Google Cloud Storage.
   • You will receive a Shadow Report summarizing your biomechanical precision and session metrics.

---

---

Local Development & Automated Deployment

# Install Dependencies
npm install

# Launch Development Server
npm run dev

# 🚀 AUTOMATED DEPLOYMENT (GEMINI CHALLENGE BONUS)
# The project includes a full 'Infrastructure-as-Code' deployment script
# to automate the build, GAE deployment, and Cloud Function sync.
bash deploy.sh

---

🏗️ Automated Deployment Proof

The following file demonstrates the automated, one-touch deployment of our infrastructure:

• Automation Logic: deploy.sh - This script handles the automated build of production artifacts and their synchronized deployment to Google App Engine and Cloud Functions.

---

📜 Repository Compliance & Formalities

This repository contains the full implementation of the Fit-Live-Frame-AI Agent, including the GCS Proxy Function (gcs_proxy_function.js) and the Neural Persistence Client (src/lib/cloudStorageClient.ts).

All code is provided under the MIT License for the Gemini Agent AI Challenge.

---

Developed by Deepthi Thotakura - Powered by Gemini AI & Google Cloud.