DEV Community

Cover image for I Built an Entire Workflow Automation Platform Using the GitHub Copilot CLI 🤖✨
Hrishi Mehta
Hrishi Mehta

Posted on

I Built an Entire Workflow Automation Platform Using the GitHub Copilot CLI 🤖✨

#devchallenge #githubchallenge #cli #githubcopilot

GitHub Copilot CLI Challenge Submission

This is a submission for the GitHub Copilot CLI Challenge

What I Built

I built Sluice, a visual, node-based automation platform designed for developers who are tired of repetitive manual tasks.

We've all been there: "Can you summarize these PDF to text, md, etc ?" or "Can you change image type without losing quality of uploaded image ?" or "Just quickly verify this CSV data?" or "Suggestions to improve this code file?".

Instead of writing a new throwaway script every time, Sluice lets you drag and drop nodes to create reusable workflows.

Video Executing one flow - https://res.cloudinary.com/dfzqhhywm/video/upload/v1700000000/1f96fdaa-3673-4021-bbb5-35fb5d429b55_rknz56.mp4

Key Features:

  • 🧠 AI Nodes: "Explain Content", "Analyze Code", and "Describe Scene". I integrated LLMs to not just process files but understand them.
  • 📄 PDF Power: Watermark, Encrypt, Highlight Signatures, and Compress PDFs visually.
  • 🔊 Audio Analysis: Detect emotion in audio files (because why not?).
  • 💻 Developer Tools: A "Code Refactor" node that cleans up messy code and a "JSON Repair" node that fixes broken data structures.

It uses a Next.js frontend with React Flow for the canvas, and a robust FastAPI (Python) backend to handle heavy file processing.

My Experience with GitHub Copilot CLI

Wait, did I actually write code?

Okay, here's the kicker: I built this entire full-stack application using GitHub Copilot in the CLI.

I used the CLI agent for everything. You literally just prompt it, and it writes the code for you. It's not just "autocomplete" anymore; it's a full-on coding partner.

Why it's a Must-Try:

  1. Prompt & It Codes: I didn't have to manually wire up the complex React Flow nodes or write the API endpoints. I just told the CLI: "Create a Properties Panel that dynamically updates based on the selected node string" and it scaffolded the entire component, imports and all.

  2. MCP Servers are Insane: The Model Context Protocol (MCP) server integration is the real game-changer. It allows the CLI to interact with my local file system, run terminal commands, and even search the web contextually. It feels like the AI has "hands" inside my project.

  3. Easy Model Switching: One minute I'm using a fast model for quick CSS fixes, and the next I'm switching to a reasoning-heavy model to debug a race condition in the backend. The ability to fluidly switch models within the same CLI session meant I always had the right brain for the job.

  4. Complex Logic? No Problem:

    • I had a messy engine.py file handling 20 different file operations.
    • I asked the CLI to refactor the if/elif blocks for AI processing (explain, analyze, suggest).
    • It rewrote the entire conditional structure to use a clean, dynamic prompting system, handling file type detection and fallbacks for binary files automatically.

Building this project felt less like coding and more like directing a highly skilled engineer. If you haven't tried the new CLI agent workflow yet, you're working too hard. 🚀

Top comments (0)