📼 ByteSized // The 2082 Fitness Protocol

"Wake up, space-cadet! The year is 2082, and your Meat-Mech is running on fumes!"

ByteSized is an autonomous, agent-driven fitness OS built for the year 2082. Standard health trackers fail because manual data entry is boring. ByteSized solves this by acting as an intelligent, multimodal backend controller that parses your natural language and images, calculates your macros, and gamifies your survival—all wrapped in a 1980s VHS Synthwave aesthetic.

Built solo in 6 hours for the Qoder + Insforge Hackathon.

---

🚀 The Tech Stack

This project was architected to push the absolute limits of the Qoder Agentic IDE and the Insforge Agent-Native Backend.

• Frontend: Next.js 14 (App Router), Tailwind CSS, Web Speech API.
• Backend Infrastructure: Insforge PostgreSQL, Insforge Auth (Google Provider).
• The "Edge Brain": Insforge Deno Edge Functions + Insforge Model Gateway.
• Realtime State: Insforge Realtime (Postgres Change Data Capture).
• AI Architect: Qoder (Using MCP to read the Insforge schema and generate context-aware UI/UX).

✨ Core Features

• Multimodal Ingestion Engine: Don't type your meals. Use the Web Speech API to tell ByteSized what you ate, or snap a photo of your food. The Insforge Edge Function securely routes the payload through the Model Gateway to extract calories, macros, and XP.
• Agentic Personalities: Choose your trainer. Whether it's RAD.os (the 80s aerobics daemon), CYC.L3 (the intense spin instructor), or SGT.1R0N (the tactical boot camp AI), the Edge Function dynamically alters its state-mutation logic and dialogue based on your selected persona.
• Realtime CDC UI: No loading spinners. When you log an action, the Next.js frontend subscribes directly to the Insforge vhs_logs table via WebSockets. The UI updates instantly as the Edge Function mutates the database.
• The "Eject Tape" Grader: At the end of the day, click [ EJECT TAPE ]. A secondary Edge Function aggregates your daily macros, passes the summary to the LLM, and generates a final A-F Report Card.

🧠 How Qoder & Insforge Built This in 6 Hours

1. Schema-to-UI Telepathy: By connecting Qoder to the Insforge Remote MCP, Qoder autonomously read the Postgres schema and generated perfectly typed Next.js hooks and Tailwind components.
2. Secure Edge Routing: Instead of exposing LLM API keys on the client, Qoder wrote Deno Edge Functions deployed on Insforge. These functions act as the secure "Brain," handling the Model Gateway calls and executing Postgres RPCs to update user stats.
3. Context-Aware Styling: Qoder's .qoder/rules engine was used to enforce a strict "2082 Cyberpunk meets 1980s VHS" design system across the entire codebase, generating complex Tailwind arbitrary values for neon glows and CRT scanlines without manual CSS tweaking.

🛠️ Local Setup

Want to initialize your own Meat-Mech?

1. Clone the repo & install dependencies:

   git clone https://github.com/voxelle-io/ByteSized.git
   cd ByteSized
   npm install

2. Set up Insforge:

   • Create a new Insforge project.
   • Run the SQL migrations (found in insforge/migrations/) to create the users, vhs_logs, and daily_summaries tables with RLS.
   • Deploy the Edge Functions using the Insforge CLI.

3. Environment Variables:

   • Create .env.local and add your Insforge URL and Anon Key.

   NEXT_PUBLIC_INSFORGE_URL=your_project_url
   NEXT_PUBLIC_INSFORGE_ANON_KEY=your_anon_key
   # Only needed for local seed scripts, do not expose to frontend
   INSFORGE_SERVICE_ROLE_KEY=your_service_role_key
   

4. Run the OS:

   npm run dev

🛡️ Security Note

This repository utilizes a strict .gitignore to prevent the leakage of Insforge Service Role keys and local Qoder chat states, while preserving the .qoder/rules directory to maintain architectural context for future agentic development.