Processing Benchmark Suite

A web application for comparing algorithm outputs and metrics.

Workflow Architecture

graph TD
    A[Define System Protocol] --> B(Select Engines)
    B --> C(Tune Hyperparameters)
    C --> D[Execute & Monitor Stream]
    D --> E{Visual Diff Engine}
    E --> F(Analyze Word-Level Differences)
    F --> G[Save to Chronicle & Refine]

Loading

Directory Structure

/
├── app/               # Next.js App Router & Server Components
├── components/        # Metrics display, Diff views, Editor UI
├── lib/               # Diffing engine, SDK integration, Benchmarking logic
├── public/            # Static Assets & Screenshots
└── package.json       # Dependencies & Scripts

Metrics Collection

Captures TTFT (Time To First Token) latency during execution. Uses the diff algorithm to display word-level differences between outputs.

Features

• Diffing: Compare text outputs side-by-side.
• Scroll Synchronization: Scroll multiple columns simultaneously.
• Configuration: Adjust Temperature, Top-P, and Max Tokens per column.
• Session Storage: Saves history to localStorage.

Tech Stack

Layer	Technology
Framework	Next.js 16
State	React 19
Styling	Tailwind CSS 4

Getting Started

git clone https://github.com/GaneshArwan/MultiModelAI.git
cd MultiModelAI
npm install
npm run dev

Deployment

Requires client-side API keys. Keys are stored locally in the browser and passed directly to the provider endpoints.

License

MIT License © 2026 GaneshArwan